Mechatronics and Robotics

Mechatronics and Robotics

Edited by:
Zoran Gacovski

ARCLER
P r e s s

www.arclerpress.com
Mechatronics and Robotics
Zoran Gacovski

Arcler Press
224 Shoreacres Road
Burlington, ON L7L 2H2
Canada
www.arclerpress.com
Email: [email protected]

e-book Edition 2021

ISBN: 978-1-77407-978-2 (e-book)

This book contains information obtained from highly regarded resources. Reprinted
material sources are indicated. Copyright for individual articles remains with the au-
thors as indicated and published under Creative Commons License. A Wide variety of
references are listed. Reasonable efforts have been made to publish reliable data and
views articulated in the chapters are those of the individual contributors, and not neces-
sarily those of the editors or publishers. Editors or publishers are not responsible for
the accuracy of the information in the published chapters or consequences of their use.
The publisher assumes no responsibility for any damage or grievance to the persons or
property arising out of the use of any materials, instructions, methods or thoughts in the
book. The editors and the publisher have attempted to trace the copyright holders of all
material reproduced in this publication and apologize to copyright holders if permission
has not been obtained. If any copyright holder has not been acknowledged, please write
to us so we may rectify.

Notice: Registered trademark of products or corporate names are used only for explana-
tion and identification without intent of infringement.

© 2021 Arcler Press

ISBN: 978-1-77407-775-7 (Hardcover)

Arcler Press publishes wide variety of books and eBooks. For more information about
Arcler Press and its products, visit our website at www.arclerpress.com
DECLARATION

Some content or chapters in this book are open access copyright free
published research work, which is published under Creative Commons
License and are indicated with the citation. We are thankful to the
publishers and authors of the content and chapters as without them this
book wouldn’t have been possible.
 ABOUT THE EDITOR

Dr. Zoran Gacovski has earned his PhD degree at Faculty of Electrical
engineering, Skopje. His research interests include Intelligent systems and
Software engineering, fuzzy systems, graphical models (Petri, Neural and
Bayesian networks), and IT security. He has published over 50 journal and
conference papers, and he has been reviewer of renowned Journals. Currently,
he is a professor in Computer Engineering at European University, Skopje,
Macedonia.
 TABLE OF CONTENTS

List of Contributors........................................................................................xv
List of Abbreviations..................................................................................... xxi
Preface................................................................................................... ....xxiii

Chapter 1 A Brief History of Industrial Robotics in the 20th Century......................... 1

Abstract...................................................................................................... 1
Introduction................................................................................................ 2
The First Generation of Industrial Robots (1950-1967)................................ 3
The Second Generation of Industrial Robots (1968-1977)........................... 6
The Third Generation of Industrial Robots (1978-1999)............................... 9
Conclusion............................................................................................... 13
Conflicts of Interest................................................................................... 14
References................................................................................................ 15

Chapter 2 From Mechatronic Components to Industrial Automation Things: An

IoT Model for Cyber-Physical Manufacturing Systems............................. 17
Abstract.................................................................................................... 17
Introduction.............................................................................................. 18
Related Work............................................................................................ 21
Towards An Iot Model For Manufacturing................................................. 23
A Contiki Based Industrial Automation Thing............................................ 28
Automating The Generation Process Of IAT.............................................. 34
Conclusion............................................................................................... 37
References................................................................................................ 38

Chapter 3 The 3+1 SysML View-Model in Model Integrated Mechatronics.............. 43

Abstract.................................................................................................... 43
Introduction.............................................................................................. 44
Background And Related Work................................................................. 46
 The 3+1 Architectural View Model............................................................ 50
Using Sysml To Model The Mim Artifacts.................................................. 52
The Mts Development Process.................................................................. 55
Future Developments And Research Challenges....................................... 59
Concluding Remarks................................................................................. 61
Acknowledgments.................................................................................... 61
References................................................................................................ 62

Chapter 4 Design and Development of Mechatronic Application in

Agricultural Irrigation Device.................................................................. 65
Abstract.................................................................................................... 65
Introduction.............................................................................................. 66
Literature Review...................................................................................... 69
Hardware Design Of PLC......................................................................... 70
Materials And Methodology...................................................................... 73
Result ....................................................................................................... 76
Discussion................................................................................................ 77
Conclusion............................................................................................... 78
References................................................................................................ 79

Chapter 5 Design and Development of a

Competitive Low-Cost Robot Arm with Four Degrees of Freedom........... 81
Abstract.................................................................................................... 81
Introduction.............................................................................................. 82
Mechanical Design................................................................................... 83
Robot Arm Inverse Kinematics.................................................................. 88
End-Effector Selection............................................................................... 90
Robot Arm Control................................................................................... 91
Testing And Validation.............................................................................. 94
Results And Discussions........................................................................... 96
Conclusions.............................................................................................. 98
References.............................................................................................. 100

Chapter 6 An Adaptive Robust Approach to Modeling and Control of Flexible

Arm Robots............................................................................................ 101
Abstract.................................................................................................. 101
Introduction............................................................................................ 102

x
 Dynamic Equations................................................................................. 104
Control Design....................................................................................... 106
Simulation Results.................................................................................. 111
Conclusion............................................................................................. 115
References.............................................................................................. 116

Chapter 7 Interactive Heuristic D* Path Planning Solution Based on PSO

for Two-Link Robotic Arm in Dynamic Environment............................. 119
Abstract.................................................................................................. 119
Introduction............................................................................................ 120
Related Research Work........................................................................... 121
Two-Link Robot Arm Kinematics............................................................. 123
Free Cartesian Space Analysis................................................................. 125
Proposed Method: D* Algorithm Based On PSO..................................... 126
Simulation Result.................................................................................... 128
Conclusion............................................................................................. 135
References.............................................................................................. 138

Chapter 8 Optimal Task Placement of a Serial Robot Manipulator for

Manipulability and Mechanical Power Optimization............................. 141
Abstract.................................................................................................. 141
Introduction............................................................................................ 142
Manipulability........................................................................................ 145
Mechanical Power.................................................................................. 146
Task Specification................................................................................... 147
Multicriteria Programming...................................................................... 150
Optimization Strategy............................................................................. 151
Numerical Simulations........................................................................... 152
Conclusions............................................................................................ 163
Acknowledgements................................................................................ 164
References.............................................................................................. 165

Chapter 9 Kinematical Analysis and Simulation of High-Speed Plate

Carrying Manipulator Based on Matlab.................................................. 169
Abstract.................................................................................................. 169
Introduction............................................................................................ 170
The Structure Design And Link Parameters Of High-Speed Plate

xi
 Carrying Manipulator.................................................................... 171
The Kinematics Simulation Algorithm of Manipulator............................. 176
The Kinematics Simulation of Manipulator.............................................. 177
Conclusions............................................................................................ 181
References.............................................................................................. 182

Chapter 10 Flexible Impact Force Sensor................................................................. 183

Abstract.................................................................................................. 183
Introduction............................................................................................ 184
Measurement Of Force On Soft Material Surface By Using A
Sheet-Type Sensor......................................................................... 185
Mechanical Issues In Impact Force Measurement On Flexible
Material Surface............................................................................ 189
Flexible Sensor Made From Plural Sensor Elements Arranged In
A Matrix........................................................................................ 193
Experiment............................................................................................. 196
Improvement Of The Sensor Flexibility................................................... 200
Conclusions............................................................................................ 201
Acknowledgements................................................................................ 202
References.............................................................................................. 203

Chapter 11 A Sensing and Robot Navigation of Hybrid Sensor Network.................. 205

Abstract.................................................................................................. 205
Introduction............................................................................................ 206
Related Work.......................................................................................... 207
The Deployment Of Static Sensor Nodes And Communication
Algorithm...................................................................................... 208
The Mobile Sensor Nodes....................................................................... 210
Navigation Of Robot............................................................................... 212
Conclusions............................................................................................ 219
Acknowledgment.................................................................................... 219
References.............................................................................................. 220

Chapter 12 A Study on Vehicle Detection and Tracking Using Wireless Sensor

Networks................................................................................................ 223
Abstract.................................................................................................. 223
Introduction............................................................................................ 224
Classification Of Sensors Used In Vehicle Detection............................... 227

xii
 Signals For Vehicle Detection And Tracking............................................. 234
Algorithms For Vehicle Detection And Tracking....................................... 235
Analysis Of Algorithms........................................................................... 244
Summary Of Algorithms......................................................................... 248
Conclusions And Future Directions......................................................... 248
Acknowledgment.................................................................................... 249
References.............................................................................................. 250

Chapter 13 Deployment of Pre-Industrial Autonomous Microbe Sensor in

Saudi Arabia’s Injection Seawater System.............................................. 253
Abstract.................................................................................................. 253
Introduction............................................................................................ 254
Technical Improvements Of Pre-Industrial Ams....................................... 255
Design And Construction Of Pre-Industrial Ams...................................... 256
Performance Assessment......................................................................... 258
Lessons Learned...................................................................................... 265
Conclusion And Recommendations........................................................ 269
Acknowledgments.................................................................................. 270
References.............................................................................................. 271

Chapter 14 A Survey of Modeling and Control of Piezoelectric Actuators............... 273

Abstract.................................................................................................. 273
Introduction............................................................................................ 274
Piezoelectric Actuators And Their Behaviors........................................... 275
Modeling Of Piezoelectric Actuators...................................................... 277
Control Of Peas...................................................................................... 289
Emerging Issues In Pea Modeling And Control........................................ 295
Conclusions............................................................................................ 300
References.............................................................................................. 301

Chapter 15 Experimental Investigation of Photostrictive Materials for MEMS

Application............................................................................................. 319
Abstract.................................................................................................. 319
Introduction............................................................................................ 320
Methodology.......................................................................................... 327
Results.................................................................................................... 332
Conclusions............................................................................................ 340

xiii
 References.............................................................................................. 342

Chapter 16 Theory and Simulation Analysis of the Mode Shape and Normal
Shape Actuators and Sensors.................................................................. 345
Abstract.................................................................................................. 345
Introduction............................................................................................ 346
Actuator Equation................................................................................... 347
Sensor Equation...................................................................................... 356
Case Study: Theory And Simulation Analysis........................................... 358
Results And Discussion........................................................................... 360
Conclusion............................................................................................. 362
Acknowledgements................................................................................ 367
References.............................................................................................. 368

Chapter 17 Experimental Study of the Response of Transonic Diffuser Flow to a

Piezoceramic Actuator at Diffuser Throat.............................................. 371
Abstract.................................................................................................. 371
Introduction............................................................................................ 372
Experimental Apparatus And Procedure.................................................. 373
Experimental Results And Discussion...................................................... 375
Conclusions............................................................................................ 384
References.............................................................................................. 385

Chapter 18 An ARX-Based PID-Sliding Mode Control on Velocity Tracking

Control of a Stick-Slip Piezoelectric-Driven Actuator............................ 387
Abstract.................................................................................................. 387
Introduction............................................................................................ 388
ARX-Based PID-SMC And Its Design....................................................... 391
Experiments............................................................................................ 393
Conclusion............................................................................................. 399
Acknowledgements................................................................................ 400
References.............................................................................................. 401

Index...................................................................................................... 403

xiv
 LIST OF CONTRIBUTORS

A. Gasparetto
Polytechnic Department of Engineering and Architecture, University of Udine, Udine,
Italy

L. Scalera
Polytechnic Department of Engineering and Architecture, University of Udine, Udine,
Italy

Theodoros Foradis
Electrical and Computer Engineering, University of Patras, Patras, Greece

Kleanthis Thramboulidis
Electrical and Computer Engineering, University of Patras, Patras, Greece

Kleanthis Thramboulidis
Visiting Professor Helsinki University of Technology, Electrical & Computer
Engineering, University of Patras, Patras, Greece.

Mohd Hudzari Razali

Faculty of Bioresources & Food Industry, Universiti Sultan Zainal Abidin, Tembila
Campus, Besut, Terengganu Darul Iman, Malaysia

Syazili Roslan
Faculty of Bioresources & Food Industry, Universiti Sultan Zainal Abidin, Tembila
Campus, Besut, Terengganu Darul Iman, Malaysia

Abdul Ssomad Mohd Abd Halim

Faculty of Bioresources & Food Industry, Universiti Sultan Zainal Abidin, Tembila
Campus, Besut, Terengganu Darul Iman, Malaysia

Hayan Basit
Faculty of Bioresources & Food Industry, Universiti Sultan Zainal Abidin, Tembila
Campus, Besut, Terengganu Darul Iman, Malaysia

Ashraf Elfasakhany
Department of Mechanical Engineering, Faculty of Engineering, Taif University, Al-

xv
Haweiah, Saudi Arabia
Tecnológico de Monterrey, Campus Ciudad Juárez, Ciudad Juarez, Mexico

Eduardo Yanez
Tecnológico de Monterrey, Campus Ciudad Juárez, Ciudad Juarez, Mexico

Karen Baylon
Tecnológico de Monterrey, Campus Ciudad Juárez, Ciudad Juarez, Mexico

Ricardo Salgado
Tecnológico de Monterrey, Campus Ciudad Juárez, Ciudad Juarez, Mexico

Pedram Bagheri
Department of Electrical and Computer Engineering, University of Calgary, Calgary,
Alberta, Canada

Laleh Behjat
Department of Electrical and Computer Engineering, University of Calgary, Calgary,
Alberta, Canada

Qiao Sun
Department of Mechanical and Manufacturing Engineering, University of Calgary,
Calgary, Alberta, Canada

Firas A. Raheem
Control and Systems Department, University of Technology, Baghdad, Iraq

Umniah I. Hameed
Control and Systems Department, University of Technology, Baghdad, Iraq

Rogério Rodrigues dos Santos

School of Mechanical Engineering, Federal University of Uberlândia, Uberlândia,
Brazil

Valder Steffen, Jr
School of Mechanical Engineering, Federal University of Uberlândia, Uberlândia,
Brazil

Sezimária de Fátima Pereira Saramago

Faculty of Mathematics, Federal University of Uberlândia, Uberlândia, Brazil

Ke Wang, Jiping Zhou

School of Mechanical Engineering, University of Yangzhou, Yangzhou, China

xvi
Chen Liu
Department of Transportation and Environmental Engineering, Hiroshima University,
Higashi-Hiroshima, Japan

Yukio Fujimoto
Department of Transportation and Environmental Engineering, Hiroshima University,
Higashi-Hiroshima, Japan

Yoshikazu Tanaka
Department of Transportation and Environmental Engineering, Hiroshima University,
Higashi-Hiroshima, Japan

Shuncai Yao
Department of Electric Engineering, North University of China, Taiyuan, China

Jindong Tan
Department of Electrical Computer Engineering, Michigan Technological University,
Houghton, USA

Hongxia Pan
Department of Electric Engineering, North University of China, Taiyuan, China

G. Padmavathi
Department of Computer Science, Avinashilingam University for Women, Coimbatore,
Tamil Nadu, India

D. Shanmugapriya
Department of Information Technology, Avinashilingam University for Women,
Coimbatore, Tamil Nadu, India

M. Kalaivani
Department of Computer Science, Avinashilingam University for Women, Coimbatore,
Tamil Nadu, India

Mohammed A. Al-Moniee
Saudi Arabian Oil Company (Saudi Aramco), Dhahran, Saudi Arabia

Xiangyang Zhu
Saudi Arabian Oil Company (Saudi Aramco), Dhahran, Saudi Arabia

Rikke Markfoged
Danish Technological Institute (DTI), Aarhus, Denmark

xvii
Aabdullah H. Al-Wadei
Saudi Arabian Oil Company (Saudi Aramco), Dhahran, Saudi Arabia

Poul L. Pedersen
Danish Technological Institute (DTI), Aarhus, Denmark

Anders K. Tuxen
Danish Technological Institute (DTI), Aarhus, Denmark

Fuad I. Al-Nuwaiser
Saudi Arabian Oil Company (Saudi Aramco), Dhahran, Saudi Arabia

Lone Tang
Danish Technological Institute (DTI), Aarhus, Denmark

Tinna Staghøj Roesen

Danish Technological Institute (DTI), Aarhus, Denmark

Thomas Lundgaard
Danish Technological Institute (DTI), Aarhus, Denmark

Jingyang Peng
Department of Mechanical Engineering, University of Saskatchewan, Saskatoon,
Canada

Xiongbiao Chen
Department of Mechanical Engineering, University of Saskatchewan, Saskatoon,
Canada

Mosfequr Rahman
Georgia Southern University, Statesboro, GA, USA

Mahbub Ahmed
Southern Arkansas University, Magnolia, AR, USA

Masud Nawaz
Georgia Southern University, Statesboro, GA, USA

Gustavo Molina
Georgia Southern University, Statesboro, GA, USA

Abdur Rahman
Edinboro University, Edinboro, PA, USA

xviii
Jwo Ming Jou
Department of Mechanical Engineering, Cheng Shiu University, Kaohsiung City,
Taiwan

Minoru Yaga
Department of Mechanical Systems, Faculty of Engineering, University of the Ryukyus,
Okinawa, Japan

Yusuke Uechi
Mie Metal Industry Co. Ltd., Mie, Japan

Hiroaki Ozono
Graduate School of Engineering and Science, University of the Ryukyus, Okinawa,
Japan

Masaaki Ishikawa
Department of Mechanical Systems, Faculty of Engineering, University of the Ryukyus,
Okinawa, Japan

Isao Teruya
Department of Mechanical Systems, Faculty of Engineering, University of the Ryukyus,
Okinawa, Japan

Y. Cao, X. B. Chen
Department of Mechanical Engineering, University of Saskatchewan, Saskatoon,
Canada

xix
 LIST OF ABBREVIATIONS

AMR Anisotropic Magneto Resistive

AFM Atomic Force Microscope
AMS Autonomous microbe sensor
ARX Auto-regressive
ARX Autoregressive exogenous
COTS Commercial off-the-shelf
DTI Danish Technological Institute
DOF Degrees-of-freedom
EPFL Ecole Poly-technique Fédérale de Lausanne
EDM Electrical Discharge Machining
TEKES Finnish Funding Agency for Technology and Innovation
AMS Fluorescence count
FAO Food and Agriculture Organization
4 DOF Four degree of freedom
FCS Free Cartesian space
IAS Industrial automation systems
IAT Industrial Automation Thing
IC Integrated circuits
IoMT Internet of manufacturing things
IoT Internet of Things
MRC Maxwell resistive capacitor
MIC Microbiologically influenced corrosion
MEMS Micro-Electro-Mechanical-Systems
MIM Model Integrated Mechatronics
MTSs Model-driven development of complex mechatronic systems
NC Normally closed
NO Normally open
OO Object- oriented
PSO Particle Swarm Optimization
PFE Photo-ferroelectric
PEA Piezoelectric actuator
PVDF Piezoelectric film
PIM Platform Independent Model
PI PrandtlIshlinskii
PLC Programmable Logic Controllers
PD Proportional-Derivative
PID Proportional-integral- derive
QoS Quality of Service
RRT Rapidly-Exploring Random Tree
SOA Service-oriented architecture
SMC Sliding mode control
SMOs Sliding mode-based observers
SMC Sliding model control
SRB Sulfate-reducing bacteria
SysML Systems Modeling Language
TBD Track Before Detect
UV Ultraviolet

xxii
 PREFACE

Mechatronics is a synergy of mechanical engineering, electrical engineering and

information and communication technologies (computer science, informatics).
Highly automated systems are all around us, from music and video devices to
factory facilities. Due to the continuous advancement of the technology, where
there are no longer any sharp boundaries between individual fields of expertise,
there is a need for experts who will be able to design, operate and maintain such
complex systems.
Mechatronics is not a new technical branch, but a new approach in the field
of technical sciences in an effort to educate engineers who are able to fully
understand the functioning and design of modern sophisticated products
(devices) and systems.
Mechatronics is presented - not as a new engineering branch, but as a new
concept, a new approach to engineering that requires interaction and integration
between different engineering disciplines. The afore-mentioned technical areas
are taught as much as is necessary for understanding and servicing individual
systems, which require further specialization. The engineer should be able to
understand the principle of system operation, the role of individual components
and diagnose system operation. He is the link between the technicians who are
specialized, (he is the team leader).
Mechatronics is the process of embedding intelligence into physical systems.
The characteristics of mechatronic products and systems are:
• functional interaction between mechanical, electronic and information
technologies;
• spatial integration of subsystems into one unit; intelligence related to the
control functions of the mechatronic system;
• adaptability, the convenience with which - mechatronic products can be
adapted to changing tasks and situations;
• multifunctionality related to microprocessor functions determined by a
computer program;
• hidden functions performed by microelectronics, difficult to see and
understand by the end consumers.
Mechatronics is an interdisciplinary field that integrates knowledge in
electronics, mechanical engineering and computer science. Mechatronics
specialists will be the most flexible, the most versatile and the most modern
trained professionals of the future.
This edition covers different topics from mechatronics and robotics, including
mechatronics basics, robotics arms and manipulators, sensors and actuators in
mechatronics.
Section 1 focuses on mechatronics basics, describing a brief history of industrial
robotics in the 20th century; an IoT model for cyber-physical manufacturing
systems; 3+1 SysML view-model in model integrated mechatronics; design and
development of mechatronic application in agricultural irrigation device.
Section 2 focuses on robotics arms and manipulators, describing design
and development of a competitive low-cost robot arm with four degrees of
freedom; an adaptive robust approach to modeling and control of flexible arm
robots; interactive heuristic d* path planning solution based on PSO for two-
link robotic arm in dynamic environment; optimal task placement of a serial
robot manipulator for manipulability and mechanical power optimization;
kinematical analysis and simulation of high-speed plate carrying manipulator
based on Matlab.
Section 3 focuses on sensors in mechatronics, describing flexible impact force
sensor; a sensing and robot navigation of hybrid sensor network; a study on
vehicle detection and tracking using wireless sensor networks; deployment of
pre-industrial autonomous microbe sensor in Saudi Arabia’s injection seawater
system.
Section 4 focuses on actuators, describing a survey of modeling and control of
piezoelectric actuators; experimental investigation of photostrictive materials
for MEMS application; theory and simulation analysis of the mode shape and
normal shape actuators and sensors; experimental study of the response of
transonic diffuser flow to a piezoceramic actuator at diffuser throat; an ARX-
based PID-sliding mode control on velocity tracking control of a stick-slip Pi-
ezoelectric-driven actuator.
 CHAPTER

1
A Brief History of Industrial Robotics
in the 20th Century

A. Gasparetto, L. Scalera
Polytechnic Department of Engineering and Architecture, University of
Udine, Udine, Italy

ABSTRACT
Industrial robotics is a branch of robotics that gained paramount importance
in the last century. The presence of robots totally revolutionized the
industrial environment in just a few decades. In this paper, a brief history
of industrial robotics in the 20th century will be presented, and a proposal
for classifying the evolution of industrial robots into four generations is
set forward. The characteristics of the robots belonging to each generation
are mentioned, and the evolution of their features is described. The most

Citation:- Gasparetto, A. and Scalera, L. (2019), “A Brief History of Industrial Ro-

botics in the 20th Century”. Advances in Historical Studies, 8, 24-35. doi: 10.4236/
ahs.2019.81002.
Copyright: © 2019 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
2 Mechatronics and Robotics

significant milestones of the history of industrial robots, from the 1950’s

to the end of the century, are mentioned, together with a description of the
most representative industrial robots that were designed and manufactured
in those decades.
Keywords: Industrial Robots, History, Unimate, Stanford Arm, Delta
Robot

INTRODUCTION
The idea to design and build some sort of beings, or devices that could
carry out repetitive or heavy tasks, thus relieving men from this burden,
dates back to ancient times. Since the Greek-Hellenistic age some of these
devices, which were named automata, have been designed and created by
ingenious inventors, belonging to several different civilizations through the
centuries. The term “automata” mainly refers to human-like devices, while
the term “robot” has a more general meaning.
The origin of the term “robot” is placed in more recent times: namely,
it comes from the Czech word “robota”, meaning “heavy work” or “forced
labor”. The introduction of this term is due to the Czech writer Karel Čapek
(1890-1938), who used it for the first time in 1920 in his novel “R.U.R.:
Rossum’s Universal Robots” (Čapek, 2004) .
On the other hand, the word “Robotics” was employed for the first time
by Isaac Asimov (1920-1992) in his novel “Runaround” (1942), contained in
the famous series “I, Robot”. In that novel he defined three rules concerning
the behavior of robots and the interaction with humans: these rules would
later be named the three Laws of Robotics (Asimov, 1942).
In the literature, there are not so many works dealing with Robotics
from a historical perspective. For instance, Ceccarelli dealt with this
topic in (Ceccarelli, 2001) and (Ceccarelli, 2004), while Gasparetto
(2016) presents a historical outline of robotics from ancient times until the
Industrial Revolution.
Industrial applications of Robotics gained a paramount importance in the
last century. The beginning of “Industrial Robotics”, as we currently define
it, can be dated back to the 1950’s, although some kinds of automatization in
the industrial environment started to appear since the times of the Industrial
Revolution.
In this paper, the main milestones of the history of industrial robotics,
from its beginning (in the 1950’s and even earlier) to the end of the 20th
 A Brief History of Industrial Robotics in the 20th Century 3

century, will be mentioned and described. The evolution of industrial robots

can be subdivided in four categories, as in (Zamalloa, 2017), the first three
covering the timespan from the 1950’s to the end of the 1990’s. The robots
of the fourth generation (which ranges from 2000 to nowadays), that are
characterized by high-level “intelligent” features (such as the capability
of performing advanced computations, logical reasoning, deep learning,
complex strategies, collaborative behavior), are not analyzed in this paper.
In the scientific literature, not so many works on the history of industrial
robots are present. In some books, such as in (Ceccarelli, 2004), (Mathia,
2010), (Siciliano & Khatib, 2008), and (Westerlund, 2000), there are some
sketches on this topic, which could also be found in some internal reports,
e.g. in (Lagerberg & Jonson, 2016) and (Wallen, 2008) .

THE FIRST GENERATION OF INDUSTRIAL ROBOTS

(1950-1967)
Some Authors (Zamalloa, 2017) proposed a chronological categorization of
industrial robots, by defining four “generations”.
The first generation of industrial robot spans from 1950 to 1967. The
robots of this generation were basically programmable machines that did not
have the ability to really control the modality of task execution; moreover,
they had no communication with the external environment. With respect
to the hardware, the first generation robots were provided with low-tech
equipment, and servo-controllers were not present (Wallen, 2008). A peculiar
feature of these robots is the strong noise they produced, when their arms
collided with the mechanical stops built to limit the movement of the axes.
With respect to the actuators, almost all the first generation robots employed
pneumatic actuators and were controlled by a sort of “logic gates” acting
as automatic regulators. Such “logic gates” were basically cams which
activated pneumatic valves, or relays which controlled solenoid valves.
Finally, the tasks that the first generation robots were capable to perform
were necessarily quite trivial, such as loading-unloading or simple material
handling operations.
The history of industrial robotics is conventionally set in the 1950’s,
although some developments in automation had taken place before: namely,
a “programmable” paint-sprayer device invented by Pollard and Roselund in
1938 (Koetsier, 2019), and a tele-operated “manipulator” invented by Goertz
in 1949, for instance. However, the turning point for industrial robotics was
4 Mechatronics and Robotics

due to the genius of George Devol, who designed in 1954 a “Programmable

Article Transfer” (this was the name given when the patent request (Devol,
1954) was filed). Such a device was the base for the development of Unimate
that is considered the first “true” industrial robot in history.
Devol (Malone, 2011) , who was basically a scientist, needed an
entrepreneurial mate with whom set up in a more concrete way his idea of a
robotic manipulator which could be used to automatize industrial processes.
In 1954 he met Joseph Engel Berger, an engineer employed in the spatial
industry, and they agreed on setting up a company which could manufacture
robots for industrial applications. This led to the foundation of a company
named Unimaton, which produced in 1961 the first Unimate robot (Figure
1). Unimate, which was hydraulically actuated, was immediately installed
in an automotive company, namely in the General Motors factory located
in Trenton (New Jersey, USA), where it was employed to extract parts
from a die-casting machine (as mentioned earlier, it was used for a single
task, because it was very complicated to reprogram it). In the following
years, several other Unimates were installed in automotive factories,
where were mainly employed for spot-welding of cars and for handling of
workpieces (Wallen, 2008).
In the same years, several robot manufacturing companies were born,
due to the fact that many entrepreneurs understood that this kind of devices
had great potentialities, especially in the automotive sector. Companies like
Ford and General Motors started to consider the automatization of their
productive plants and needed device such as the new robot to achieve this
goal. Thus, there was a sudden increase in the orders of robotic devices,
which gave the opportunity to thrive to many robot manufacturers.
One of such companies was, for instance, AMF Corporation: in 1962
they manufactured a new robot that was called Versatran (i.e. “versatile
transfer”). It was a cylindrical robot (Figure 2) that was ordered by Ford
for its production plants in Canton (Ohio, USA), thus enjoying a good
popularity (Birnie, 1974). The Versatran was also the first robot to be
installed in a productive site in Japan (1967). In 1969, the presence of
robots in Japanese companies had a significant increase, also due to the fact
that Unimation allowed to Kawasaki Heavy Industries Ltd. the license to
produce robots for the Japanese and the Asian market. In particular, this led
to the development of the Kawasaki-Unimate 2000, the first industrial robot
ever built in Japan.
 A Brief History of Industrial Robotics in the 20th Century 5

Figure 1. Joseph Engelberger and George Devol (left); the Unimate robot
(right).

Figure 2. The Versatran robot.

Since then, Japan became one of the leading countries in the world in
the field of industrial robotics. It was in Japan that the first national robotic
association in the world was founded in 1971, namely JIRA (Japanese
Robotics Association).
The diffusion of robots in Europe took place more or less at the
same time as in Japan, namely in the second half of the 1960’s. The first
robotic installation in Europe appeared in 1967 in the company Svenska
Metallverken located in Upplands Väsby (Sweden): it consisted of robots
carrying out simple pick-and-place tasks (Wallen, 2008). Two years later, in
Norway, the first painting robot was manufactured by the Tralffa Company
and installed in a plant for wheelbarrow painting.
The appearance of robots allowed to automatize important production
tasks such as welding. Unimation was the first company to manufacture
welding robots, while General Motors was the first company to install this
6 Mechatronics and Robotics

kind of robotic devices at their automotive plants in Lordstown (Ohio, USA)

in 1969. Three years after, FIAT installed the first welding robots in Europe
at their plants in Turin (Italy).

THE SECOND GENERATION OF INDUSTRIAL RO-

BOTS (1968-1977)
The industrial robots of the second generation (conventionally ranging from
1968 to 1977) were basic programmable machines with limited possibilities
of self-adaptive behavior and elementary capabilities to recognize the
external environment (Zamalloa, 2017). These robots used servo-controllers,
which enabled them to perform both point-to-point motion, and continuous
paths as well. Their control system consisted of microprocessors or of
Programmable Logic Controllers (PLC), and they could be also programmed
by an operator by means of a teach box. With respect to those belonging to
the first generation, these robots could carry out more complex tasks (e.g.
control of work centers). However, their level of versatility was not very
high, because each robot had its own software, which was dedicated to
a specific task. Hence, these robots turned out to be application-specific
devices, meaning that it was very difficult to employ the same robot for
different tasks, since this would require a substantial modification of the
controller, and a thorough reprogramming of the operating software. With
respect to diagnostics, the robots of the second generation were not very per
formant, since the only diagnostic reports they could produce were those
related to failures, which were reported by means of indicator lights, without
any hint related to the cause of the failure that was left to the operator to
trace.
At the beginning of the history of industrial robotics, the robots had
hydraulic actuators. The shift from hydraulic to electric actuators took place
in the 1970’s, when the electronic components needed to govern a robot
reached the full technical maturity. As a matter of fact, microprocessors
and other components started to be widely used at that time: this allowed
the robot manufacturers to dispose of powerful and cost-effective devices
which could be employed to implement control systems able to deal with a
complex and computationally expensive task such as the control of a robot.
The economic and geopolitical situation, at the international level, also
pushed the trend towards electrically driven robots: for instance, the price
of crude oil suddenly grew after the oil crisis following the Kippur war
(October 1973). The companies were thus forced to find more efficient
 A Brief History of Industrial Robotics in the 20th Century 7

ways of production: robot (and in particular electrically driven robots) was

consistent with the goals of reducing the costs and increasing the productivity.
This gave a boost to the installations of industrial robots, which increased
more than 30% per year in the second half of the 1970’s (Karlsson, 1991).
From the scientific point of view, a significant base for the development
of electrically driven robots was the merit of Victor Scheinman (Scheinman,
1973). Scheinman was a mechanical engineering student at Stanford
University who in 1969 designed and built the famous Stanford Arm (Figure
3). This robot was the first prototype of a robot actuated by electric motors
(6 DC motors) and controlled by a PDP-6 microprocessor. The Stanford
Arm had 5 revolute and one prismatic joint, for a total of 6 DOFs, and its
kinematic chain was made of harmonic drives and spur gear reducers. Its
inverse kinematics could be analytically solved in a closed form, which
allowed a fast trajectory execution. Moreover, some sensors (tachometers
and potentiometers) were mounted on the robot, in order to measure position
and velocity of the robot joints.

Figure 3. The Stanford Arm.

Four years later (1973), Scheinman designed another electrical robot, named
Vicarm that was smaller and lighter than the industrial robots of that time.
This made Vicarm particularly suitable for use in tasks, such as assembly
of parts, where the robot was not required to lift and carry heavy loads.
These features of Vicarm were so appreciated, that Unimation bought the
company that produced Vicarm and exploited its know-how to design and
manufacture (in 1978) the famous PUMA robot (Figure 4).
8 Mechatronics and Robotics

The PUMA (an acronym for Programmable Universal Machine for Assembly)
was considered for many decades the archetype of the anthropomorphic
robots, and its kinematics is still taken as an example in several robotic
books at the undergraduate as well at the graduate level.
In the meanwhile, other companies developed and manufactured other
types of industrial robots. For instance, KUKA developed in 1973 the
Famulus robot, whose name in Latin means “servant”. In 1974 Cincinnati
Milacron, a major machine tool manufacturer, developed a robot called T3
(an acronym for “The Tomorrow Tool”), which was installed in several
automotive plants, and especially in the Volvo plants in Sweden. T3 was
the first commercially available minicomputer-controlled industrial robot
(Figure 5).
In 1974, the Swedish company ASEA (now ABB) started the production
of the robots of the famous and successful IRB series, well known worldwide
also for their typical orange color. The first robot of this series, that was
issued for more than 20 years, was the IRB-6, which was largely employed
in productive sites for complex tasks (machining, arc-welding), for its ability
to move smoothly along continuous paths (Figure 6).
In the same year, the Japanese company Hitachi developed the robot HI-
T-HAND Expert, that is relevant in the history of industrial robotics for the
precision it reached in insertion operations (it could insert mechanical parts
with a clearance of about 10 micrometers). It was also provided with a force
feed-back control system and a flexible wrist mechanism.

Figure 4. The PUMA robot.

 A Brief History of Industrial Robotics in the 20th Century 9

Figure 5. The Cincinnati Milacron T3 robot.

Figure 6. The “legendary” ABB IRB robot.

THE THIRD GENERATION OF INDUSTRIAL RO-

BOTS (1978-1999)
The industrial robots of the third generation (conventionally ranging from
1978 to 1999) were characterized by a larger extent of interaction with both
the operator and the environment, through some kind of complex interfaces
(such as vision or voice). They also had some self-programming capabilities,
and could reprogram themselves, although by a little amount, in order to
execute different tasks (Zamalloa, 2017). These robots were provided with
servo controls, and could execute complex tasks, by moving either from
point to point or along continuous paths. They could be programmed either
10 Mechatronics and Robotics

on-line (the operator could use a teach box with a keyboard) or off-line, being
connected to a PLC or a PC, which allowed to use a high-level language for
motion programming and enabled the robots to be interfaced with a CAD
or a database. The possibility of high-level, off-line programming enlarged
the operational potential of the robots: for in-stance, they could elaborate
data from sensor reading, in order to adjust the robot movements taking into
account changes in the environment (e.g. changes in position and orientation
of the work pieces). Moreover, the diagnostic capabilities could be greatly
enhanced: these robots could produce not only an indication of failure
detection, but also a report on the location and on the type of the failure.
In addition, some sort of “intelligence” was present in the robots of the
third generation, with some (although limited) adaptive capabilities. These
capabilities could be employed in some more complex tasks (such as tactile
inspection, assembly operations, arc welding), by using the data coming
from vision or perception systems to locate the objects and the work pieces
and guide the joint movements according to the task to be performed, taking
into account the possibility of small changes in the position of the objects.
Between the end of the 1970’s and the beginning of the 1980’s, other
scientific and technical improvements contributed to the diffusion of robots.
In 1978, a novel kinematic structure was proposed by the Japanese
scientist Hiroshi Makino from Yamanashi University. Such a structure was
made of three revolute joints with parallel axes and a prismatic joint lying at
the end of the kinematic chain (Figure 7). The robot with this structure was
named SCARA (an acronym from “Selective Compliance Assembly Robot
Arm”), since its compliance in the horizontal direction resulted lower than
the compliance in the vertical direction. For this reason, as well as for the
lightness of the kinematic chain (that allowed a simpler and faster controller),
this robot was suitable to be employed in tasks such as the assembly of small
objects (Makino & Furuya, 1980).
Another relevant technical improvement in industrial robotics was the
appearance of direct drive actuated robot. The first prototype of this kind was
the CMU Direct Drive Arm (Asada & Kanade, 1983), developed in 1981 by
Kanade and Asaka at Carnegie Mellon University (Pittsburgh, USA). This
kind of robot featured higher accuracy and faster operations because the
motors connected directly to the arms eliminating the need for intermediate
gear or chain systems.
Both the aforementioned findings were employed in the AdeptOne (Figure
8), the first commercially available direct-driven SCARA robot (1984).
 A Brief History of Industrial Robotics in the 20th Century 11

Figure 7. One of the first prototypes of SCARA robot, designed by Hiroshi

Makino.

Figure 8. Examples of AdeptOne SCARA robots.

These scientific and technical findings contributed to the diffusion of
robots in applications different from those in the classical automotive sector,
in particular in the production of consumer goods, especially those in the
electronic market. SCARA robots were employed in these types of assembly
12 Mechatronics and Robotics

lines, especially in Japan. For this reason, Japan became in the 1980’s not
only the world leader in robot manufacturing (with more than 40 companies
that built and sold robots worldwide), but also in the production of electronic
consumer goods.
Robotics in the 1980’s was a rising star, not only in Japan but in all the
developed countries. It appeared as a promising field that drew the interest
of journalist, scientists, policy makers and also common people.
From the scientific and technical point of view, this is the time when the
robots became even more versatile, by exploiting important improvements
both with respect to the hardware and the software. Concerning the former
aspect, robots started to be provided with advanced sensors (e.g. cameras,
force sensors, laser scanners); concerning the latter aspects, the control
software became more “intelligent” by introducing come techniques related
to Artificial Intelligence. Both these aspects increased the versatility and
the flexibility of the robots, which could be employed in more and more
complex tasks.
Despite the significant progress undergone in the 1980’s, the need for
robots that could carry out task at high speed pushed the scientific research
to design innovative kinematic structures. The idea of employing parallel
kinematic chains instead of the classical serial kinematic chains was put
forward and led to a type of lightweight robot featuring the capability of
moving at high speed. The archetype of this kind of robots was the Delta
robot (that appeared in 1992), conceived by the Swiss scientist Reymond
Clavel at the Ecole Poly-technique Fédérale de Lausanne (EPFL). This
type of robot, designed by Clavel in his PhD thesis, had three translational
DOFs and one rotational DOF (Clavel, 1991). With respect to serial robots,
parallel robots featured a smaller work-space, but the capability of operating
at much higher speed. The kinematic architecture of the Delta robot was
copied in many parallel manipulators, devoted to high speed pick-and-place
operations.
The first application of Delta robots was developed by the Swiss
company Demaurex in 1992: six Delta robots were operating inside a work
cell for loading pretzels into trays (Figure 9). Some years afterwards (1998)
ABB developed the Flex-Picker, the world’s fastest picking robot, based on
the structure of the Delta robot (Figure 10).
In this decade, the first attempts to synchronize robots took place:
in 1994 Motoman developed MRC, the first robot control system which
provided the synchronized control of two robots (up to 21 axes).
 A Brief History of Industrial Robotics in the 20th Century 13

The end of the third generation is conventionally set to the end of the century;
beginning from the year 2000, the industrial robots are considered to belong
to the fourth generation (which extends up to the current days). Such robots
feature high-level “intelligent” capabilities (such as performing advanced
computations, logical reasoning, deep learning, complex strategies,
collaborative behavior).

Figure 9. Delta robots in the Demaurex work cells for the packaging of pretzels.

Figure 10. The ABB Flex-Pitcher robot.

CONCLUSION
In this paper, a brief history of industrial robotics in the 20th century
was presented. The evolution of the industrial robots was conventionally
categorized into four generations, of which the first three cover the timespan
from the 1950’s to the end of the century.
14 Mechatronics and Robotics

In this historical sketch, not only the scientific and technical evolution
was taken into account, but some considerations about the economic and
geopolitical issues that determined the diffusion of industrial robots, were
also done.
The evolution of industrial robotics is not over, but is still developing in
the current days: innovative ideas and novel hardware devices, together with
some new programming techniques connected with Artificial Intelligence,
are revolutionizing the concept of industrial automation and giving a new
youth to the factory environment.

CONFLICTS OF INTEREST
The authors declare no conflicts of interest regarding the publication of this
paper.
 A Brief History of Industrial Robotics in the 20th Century 15

REFERENCES
1. Asada, H., & Kanade, T. (1983). Design of Direct-Drive Mechanical
Arms. Journal of Vibration, Acoustics, Stress, and Reliability, 105,
312-316. https://doi.org/10.1115/1.3269106
2. Asimov, I. (1942). I, Robot. In Runaround. New York: Spectra Books.
3. Birnie, J. V. (1974). Practical Implications of Programmable
Manipulators. Industrial Robot: An International Journal, 1, 122-128.
https://doi.org/10.1108/eb004720
4. Capek, K. (2004). R.U.R. Rossum’s Universal Robots. Fayetteville:
Penguin Ed.
5. Ceccarelli, M. (2001). A Historical Perspective of Robotics toward the
Future. Fuji International Journal of Robotics and Mechatronics, 13,
299-313. https://doi.org/10.20965/jrm.2001.p0299
6. Ceccarelli, M. (2004). Fundamentals of Mechanics of Robotic
Manipulation. Dordrecht: Kluwer/Springer. https://doi.
org/10.1007/978-1-4020-2110-7
7. Clavel, R., (1991). Conception d’un robot parallèle rapide à 4 degrés
de liberté. Ph.D. Thesis, Lausanne, Switzerland: EPFL.
8. Devol, G. (1954). Programmable Article Transfer. https://patents.
google.com/patent/US2988237
9. Gasparetto, A. (2016). Robots in History: Legends and Prototypes
from Ancient Times to the Industrial Revolution. In Explorations in
the History of Machines and Mechanisms. History of Mechanism and
Machine Science (vol. 32, pp. 39-49). Berlin/Heidelberg: Springer.
https://doi.org/10.1007/978-3-319-31184-5_5
10. Karlsson, J. M. (1991). A Decade of Robotics; Analysis of the
Diffusion of Industrial Robots in the 1980s by Countries, Application
Areas, Industrial Branches and Types of Robots. Stockholm, Sweden:
Mekanforbundets Forlag.
11. Koetsier, T. (2019). Towards the Global Intelligent Machine. In The
Ascent of GIM, the Global Intelligent Machine. History of Mechanism
and Machine Science (vol 36, pp. 293-317). Berlin/Heidelberg:
Springer. https://doi.org/10.1007/978-3-319-96547-5
12. Lagerberg, T., & Jonson, J. (2016). Une bio du robot. Petite histoire de
la robotique industrielle. ABB Review 3/16. http://new.abb.com/about/
technology/abb-review
16 Mechatronics and Robotics

13. Makino, H., & Furuya, N. (1980). Selective Compliance Assembly

Robot Arm. Proceedings of 1st International Conference on Assembly
Automation (ICAA), Brighton, March 25-27, 77-86.
14. Malone, B. (2011). George Devol: A Life Devoted to Invention,
and Robots. IEEE Spectrum Magazine. https://spectrum.ieee.org/
automaton/robotics/industrial-robots/george-devol-a-life-devoted-to-
invention-and-robots
15. Mathia, K. (2010). Robotics for Electronics Manufacturing. Principles
and Applications in Clean-room Automation. Cambridge: Cambridge
University Press. https://doi.org/10.1017/CBO9780511712173
16. Scheinman, V. (1973). Design of a Computer Controlled Manipulator.
PhD Thesis, Stanford, CA: Stanford University. http://www.dtic.mil/
docs/citations/AD0708069
17. Siciliano, B., & Khatib, O. (2008). Springer Handbook of Robotics.
Berlin Heidelberg: Springer-Verlag.
18. Wallen, J. (2008). The History of the Industrial Robot. Technical
Report, Linkopings: Automatic Control at Linkopings Universitet.
http://www.control.isy.liu.se/publications
19. Westerlund, L. (2000). The Extended Arm of Man—A History of the
Industrial Robot. Stockholm: Informationsforlaget.
20. Zamalloa, I., Kojcev, R., Hernandez, A., Muguruza, I., Usategui, L.,
Bilbao, A., & Mayoral, V. (2017). Dissecting Robotics—Historical
Overview and Future Perspectives.
 CHAPTER

2
From Mechatronic Components to
Industrial Automation Things: An
IoT Model for Cyber-Physical
Manufacturing Systems

Theodoros Foradis, Kleanthis Thramboulidis

Electrical and Computer Engineering, University of Patras, Patras, Greece

ABSTRACT
IoT is considered as one of the key enabling technologies for the fourth
industrial revolution that is known as Industry 4.0. In this paper, we consider
the mechatronic component as the lowest level in the system composition
hierarchy that tightly integrates mechanics with the electronics and software
required to convert the mechanics to intelligent (smart) object offering well
defined services to its environment. For this mechatronic component to be

Citation:- Foradis, T. and Thramboulidis, K. (2017), “From Mechatronic Components

to Industrial Automation Things: An IoT Model for Cyber-Physical Manufacturing Sys-
tems”. Journal of Software Engineering and Applications, 10, 734-753. doi: 10.4236/
jsea.2017.108040.
Copyright: © 2017 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
18 Mechatronics and Robotics

integrated in the IoT-based industrial automation environment, a software

layer is required on top of it to convert its conventional interface to an
IoT compliant one. This layer, which we call IoT wrapper, transforms the
conventional mechatronic component to an Industrial Automation Thing
(IAT). The IAT is the key element of an IoT model specifically developed
in the context of this work for the manufacturing domain. The model is
compared to existing IoT models and its main differences are discussed.
A model-to-model transformer is presented to automatically transform the
legacy mechatronic component to an IAT ready to be integrated in the IoT-
based industrial automation environment. The UML4IoT profile is used
in the form of a Domain Specific Modelling Language to automate this
transformation. A prototype implementation of an Industrial Automation
Thing using C and the Contiki operating system demonstrates the
effectiveness of the proposed approach.
Keywords: Mechatronics, Cyber-Physical Systems, Internet of Things
(IoT), Contiki, UML4IoT Profile

INTRODUCTION
Based on one of the most commonly used definitions, the term Mechatronics
emphasizes on the synergistic integration of the three discipline areas, i.e.,
mechanical engineering, electronics and intelligent computer control, in the
design and manufacture of products and processes [1] , i.e., it emphasizes
on synergy. What is not clear by this definition is the level at which this
integration should be performed, i.e., at the system level, which is the
traditional approach, or at the subsystem or even at the mechanical unit
(component) level.
The latter is proposed in Model Integrated Mechatronics [2] and
refined with the 3+1SysML-view model [3] [4]. This approach defines
the Mechatronic component as the main building block that abstracts the
mechanical object to the software level, and transforms it to a smart object
by adding additional functionality to the one offered by the mechanical part.
The so constructed mechatronic components are integrated with cyber
components and humans to construct the industrial automation system. This
approach slightly finds its road to production in the context of Industry 4.0,
e.g., [5], since it greatly reduces the coupling between the system components
compared to the traditional one, which considers the integration of the three
disciplines at the system integration level.
 From Mechatronic Components to Industrial Automation ... 19

A substantial number of communication mechanisms and middleware are

used for the integration of the constituent components of industrial automation
systems (IAS). DPWS [6], one of the most recent, is the result of extensive
research based on the service-oriented architecture (SOA). It is based on
SOAP which was the dominating technology used in SOA implementations.
However, the REST paradigm is gaining more attention in manufacturing
the past few years [7] [8]. A comparison of REST with SOAP in the context
of the manufacturing domain can be found in [7]. Cloud computing has
also attracted the interest of researchers. Cloud manufacturing defines a
new service-oriented manufacturing model where cloud computing and IoT
deeply influence the development process of manufacturing systems [9].
IoT technologies greatly reduce the time for decision-making that is very
critical in modern manufacturing environments [10]. Both, cloud computing
and IoT, have been moving from buzzwords and hypes to tangible practical
technologies that subtly influence and change our world [11]. A state-of-the-
art survey in cloud manufacturing is given in [9], which considers IoT as a
technology that deeply influences the development of cloud manufacturing.
IoT is aligned well with the architecture of a manufacturing enterprise and
it is able to provide “vital solutions to planning, scheduling, and controlling
of manufacturing systems at all levels” [12]. Several approaches consider IoT
as a technology that can be utilized as an integration mechanism to be used
down to the sensor and actuator level of the industrial automation system.
Others consider IoT as the new logical transition from the automation and
connectivity concepts that exist in the IAS domain for many years. Bradley
et al. [13] , in an article with title “The Internet of Things―The future
or the end of mechatronics”, argue that many of the smart components
associated with the IoT will be essentially mechatronic in nature, and will
be constructed as far as it regards their interaction with the physical world
on the conventional hierarchical model. This model considers the controller
of the mechatronic component in the loop with the controlled physical unit
through sensors and actuators.
For the conventional mechatronic component to be integrated in the
IoT- based industrial environment a software layer is required on top of
it to convert its conventional interface to an IoT-compliant one. Thus, the
adoption of the IoT as integration technology for the system, transforms
the conventional mechatronic component to an Industrial Automation Thing
(IAT). This transformation is more likely, as authors also argue in [13] , to
bring significant changes to the way mechatronic, and related, systems are
designed and configured. There is already an increasing complexity in the
20 Mechatronics and Robotics

job of the industrial engineer in the task of transferring the functionality of

the physical world in the software world in the level of the IAT. To this, the
complexity of adding an extra layer to transform the conventional mechatronic
component to an IAT is added. New protocols, languages, environments and
architectural paradigms should be used and successfully integrated with
the already used conventional architectures and this complicates the job of
the industrial engineer. To this direction, authors in [14] present UML4IoT
with focus on the modeling of the IAT. UML4IoT is a UML-based approach
that realizes the model driven engineering paradigm to exploit IoT in the
manufacturing domain.
In this paper, 1) we extend the model of the IoT introduced in [14] , and
2) define a model-to-model transformer to automate the construction of IATs
based on the Contiki OS [15] and the C language. In the extended IoT model,
the IAT is still the key artifact for the adoption of the IoT infrastructure in
the manufacturing domain. Based on this model the manufacturing system is
considered as a composition of cyber-physical and cyber components along
with humans [16]. All these components are considered as Things, either
permanent or on demand that collaborate exploiting an IoT communication
infrastructure to realize a higher level of behavior, i.e., the one of the system
level.
The presented approach is discussed in comparison with other
approaches and mainly the IoT-A reference architecture that has been
adopted by the Papyrus for IoT project [17], which is building a platform
for the design of IoT systems in general. This project has many similarities
with our project; both projects use UML and SysML as modeling languages,
and Papyrus as tool to provide a modeling solution for IoT. Our approach
focuses on manufacturing systems. More specifically, it focuses on the case
that a high-level design specification for the mechatronic component is not
available. Specific annotations were defined to annotate the C source code
specification of the mechatronic component so as to automatically transform
the mechatronic component to an IAT. The approach is presented using as
case study the Liqueur production laboratory system. The LWM2M IoT
application protocol [18] running on top of CoAP [19] is used as the IoT
protocol stack.
The remainder of this paper is structured as follows. Section 2 positions
this research against related work. In Section 3, the proposed extension
to the IoT model which is used in the UML4IoT approach is presented
and discussed in comparison with existing IoT models. In Section 4, the
 From Mechatronic Components to Industrial Automation ... 21

Contiki based Industrial Automation Thing is presented along with the case
study. The model-to-model transformer for the C language and the Contiki
operating system is presented in section 5 and the paper is concluded in the
last section.

RELATED WORK
IoT offers new levels of connectivity in the industrial domain that may
lead to higher efficiency, flexibility, and interoperability among industries
[20]. However, not only many definitions exist for the IoT but also several
models. These models, e.g., ETSI, IETF, SENSEI, have been developed to
capture the key concepts of the domain and provide the infrastructure to
develop frameworks and architectures for the systems based on IoT. Fei
et al. [21] claim that even though IoT has been used in various application
domains there is still no clear and uniform definition and architecture
about it. Therefore, the IoT domain suffers from an inconsistent usage and
understanding of the meaning of several key terms, as also claimed in [22].
The definition of an IoT domain model is a prerequisite for defining IoT
Architectures. A detailed discussion on the IoT models can be found in
[23], where the IoT-A reference model for IoT is presented and validated.
Authors in [24] present and discuss applications of IoT in Manufacturing
and describe a five-layer architecture for manufacturing based on IoT.
Authors in [25] focus on the device nature of Thing and consider sensors,
actuators and controllers as IoT devices, i.e., things. They focus on the data
field structures and evaluate the benefit of using an IP smart gateway as the
decentralized peripheral to integrated sensors, actuators and the controller
and claim that this may improve the performance of IoT devices. We do not
agree with the use of sensors and actuators as first-class model elements
in the high-level design specification of the system. Sensors and actuators
are just technology artefacts used to integrate the physical with the cyber
world so they have no place in the high-level design spec of the system.
In our approach, the IAT, which encapsulates sensors, actuators and the
controller, plays the role of the Thing and represents the key construct in the
IoT manufacturing environment.
MDD becomes more and more popular in the development of embedded
software systems and various reports refer efficiency gains, from up to 50%,
for example, in the development in the car industry [26], with high error
reductions and a rapid increase of the maturity level of developed products.
MDD is considered as a promising solution to address the complexity of
22 Mechatronics and Robotics

software development in IoT [27] and improve quality characteristics of the

produced software. Malavolta and Muccini [28] argue that MDD is the right
tools to address the complexity of wireless Sensor Networks development
exploiting abstraction, reuse, separation of concerns and automation. They
present a framework to systematically study, classify and compare existing
MDA approaches in this area [28]. Several works publish results that
exploit the MDD paradigm in IoT based systems to improve their quality
characteristics but also the ones of their development process.
Thang et al. [27] present FRASAD, a framework based on MDA to
manage the complexity of IoT applications. They present a rule based model
and a domain specific language to describe the application, using the sensor
node as key concept. The primary objective is to model the sensor node
software. Contiki is also supported among other OSs by this framework.
Authors assume that the application logic of the sensor node program is
captured in a Platform Independent Model (PIM), which is constructed
using a set of rules they have defined to describe behavior of the sensor node
programs. This PIM is next mapped through a Domain Specific Language
to the specific platform where it is indented to be executed. However, the
approach focuses more on the message dissemination compared to the
processing which is considered as an optional part of a sensor node. Our
approach focuses on the interface of the Industrial Automation Thing; the
behaviour which is very complicated compared to one of a sensor node is
defined using another DSML we have defined for structuring the cyber-
physical component.
Authors in [29] present the software architecture of a platform developed
to address issues, among which the lack of development toolkits, that limit
the diffusion of IoT within industrial environments. They also describe an
innovative, IoT oriented, model-driven development toolkit that focuses on
the seamless integration of heterogeneous industrial devices and sensors,
into existing legacy systems by transforming them into web services. The
proposed toolkit allows inexperienced developers to discover and compose
distributed devices and services into mashups using a modelling tool. Thus,
the use of the MDD approach is mainly on the generation of the mashups and
does not focus on the modelling of a mechatronic component as is the case
of our approach. Furthermore, authors do not refer or describe the domain
modelling language that they use in their MDD approach.
A very early approach to model complex IoT systems with UML and
then generate RESTful interfaces from these models is presented by Prehofer
 From Mechatronic Components to Industrial Automation ... 23

[30]. Authors do not define any DSML but they construct class diagrams
and state charts using only primitive UML model elements. Authors in [31]
describe an approach to define a visual DSML for the IoT based on UML.
They model the Thing, which they consider as key construct for building
an IoT system, using the UML component construct and its interface using
provided and required interfaces. Authors do not address the mapping of the
conventional object- oriented (OO) interfaces of the Thing with the ones of
the REST paradigm.
Yingfeng Zhang et al. [32] present a real-time information capturing
and integration architecture of the internet of manufacturing things (IoMT)
to provide a new paradigm by extending the techniques of IoT to the
manufacturing field.
They use the term manufacturing thing but they do not focus on its
structure and its development process. Instead, they focus on the overall
architecture for a manufacturing system and describe a framework with
focus on real-time tracking and tracing for the dynamic monitoring of the
manufacturing process.
To the best of our knowledge there is no other work that focuses on the
automation of the transformation process of the conventional mechatronic
component to an IoT compliant one, i.e., to an Industrial Automation Thing
ready to be integrated into the IoT-based industrial environment.

TOWARDS AN IOT MODEL FOR MANUFACTURING

The IoT-A Reference Model

De et al. [33] describe the key concepts of the IoT-A reference architecture
that is a result of an EU funded IoT project. These concepts and their
interrelations are depicted in Figure 1.
Based on this, device is attached to entity, which is associated with
resource that is accessed through service. In more detail, authors consider
the entity as the “thing” in the IoT, i.e., the focus of interactions by humans
and/or software agents.
The device represents the hardware component that is either attached
to an entity or it exists in its environment and monitors it. The resource is
the actual software component that provides information on the entity or
enables the controlling of the device. A service exposes the functionality of
a device by accessing its hosted resources.
24 Mechatronics and Robotics

The Proposed IoT Model

Figure 2 captures the high-level key concepts of our IoT model. Based
on this, the IoT is defined as a composition of Things and a processing
and communication (IPV6-based) infrastructure. Any artifact that is able
to communicate with other Things using the processing and IPV6-based
communication infrastructure (Proc & Comn Infr) is considered as Thing.
Things collaborate to achieve higher level of behavior compared to the one
offered by each one of the collaborating Things. Collaborating Things form
a new Thing of type System As Thing that represents a system of Things.

Figure 1. Example key concepts and interaction in the IoT-A model (De et al.
2011).

Figure 2. High level key concepts in IoT.

A Thing may be either a system of Things (System As Thing) or a
component (Component As Thing). A Component As Thing is a Thing that
does not utilize IoT for the integration of its constituent components. This
type of Thing is used to represent in the IoT world conventional systems or
components that have been transformed to Things (Component As Thing)
by adding on top of their conventional interface an IoT-compliant one. The
smallest Thing of this type is a sensor or actuator. The Proc&ComnInfr
is a composition of: 1) processing nodes and 2) communication devices,
 From Mechatronic Components to Industrial Automation ... 25

i.e., gateways, bridges, switches, routers, etc. Proc&ComnInfr is modelled

as a composition of Things (Component As Thing) assuming that these
devices are IoT enabled; this will be the case in the near future. The Cloud
is considered part of the Proc&Comn In fr.

The Model of Thing in Manufacturing

Figure 3 presents the proposed model for the Thing. Based on this, a Thing is
either real, cyber or virtual. A Real Thing is either permanent or on demand
Thing. A permanent Thing is a composition of a cyber-physical object (Cp
Object) and a cyber IoT enabler (Cyber Iot Enabler). An On Demand Thing
is an aggregation of a Physical Object and a cyber-physical IoT enabler (Cp
Iot Enabler). As cyber-physical IoT enabler we model any device such as
laptop, tablet, mobile phone, wearable, RFID reader that provides an IoT
like interface and is able to interact with a physical object. As physical object
we consider a human, an inanimate object or even animal with an embedded
or attached tag. A human interacts with an app, i.e., application specific IoT
wrapper and is temporarily transformed to a Thing.

Figure 3. The model of thing in IoT.

Physical objects of type animals or inanimate objects with an embedded or
attached tag interact with an RFID reader with IoT like interface for the same
reason. A cyber-physical object is any physical object that: 1) implements
a kind of functionality, i.e., material and/or energy transformations and
2) has been transformed to a smart object by appending on it information
processing functionality. A mechatronic component is an example of cyber-
physical object. A cyber IoT enabler integrated with virtual objects and
deployed on a processing unit constructs a virtual Thing (virtual Thing). The
Io Twrapper of the UML4IoT approach is an example of cyber IoT enabler.
26 Mechatronics and Robotics

A cyber Thing is defined as a composition of one Cp Iot Enabler and one-

to-many cyber Objects. This allows the developer to optionally group cyber
objects of the system design model and map these to one cyber Thing.
A Thing may expose to its environment: 1) part of its structure in terms
of properties and/or 2) part of its processing or storage functionalities.
These functionalities are exposed as services. Services are discriminated to:
a) IoT infrastructure services (Iot Service), b) application domain services
and c) application specific services. All types of services may be managed
(activated, configured, updated, etc.) through the IoT communication
infrastructure. Communication infrastructure services may be considered
similar to the Industrial Automation Thing services with the remark that
Industrial Automation Things perform material, energy and information
processing while communication infrastructure Things perform only energy
and information processing. For example, device management services,
defined by the OMA LWM2M [18] are Iot Services. Native services are
services that would be defined for a specific application domain, e.g., home
automation, manufacturing, or system specific services, such as the generate
Liqueur Type A service of the Liqueur Plant laboratory production system
[14]. Device management is implemented by the management interface of
the LWM2M. On the other side, Thing management is domain or application
specific and should be implemented by specific cyber components on top of
the device management and service interface, e.g., the one of LWM2M.
The IoT processing and communication infrastructure (Proc&ComnInfr)
should provide an environment for a service-based collaboration of
Things. Each Thing implements functionalities offered to the environment
as services with negotiated QoS that should be discovered and exploited
by other Things. It may also utilize services of other Things to realize its
behavior. The cyber components of the manufacturing system model are
deployed during the deployment time on Things of type Component As
Thing (see Figure 2). UML/SysML design models of the system are marked
with the IoT model elements using the UML4IoT profile to automatically
transform the system to an IoT-compliant one.

Discussion on the Proposed IoT Model

Our definition for the Thing is different from a widely accepted one, described
by De et al. [33], which defines the thing and its relations to devices,
resources and services. Device, Resource, Service and Thing are also model
elements in the IoT model presented by Stephan [34]. In our approach, the
 From Mechatronic Components to Industrial Automation ... 27

Component As Thing encapsulates and hides how sensing of its physical

part as well as actuation are realized, since this is an implementation issue.
Based on this, we do not consider sensors and actuators first class model
elements in our model and we do not capture these artifacts in the design
model of the system. Thus, actors of the OO approach or terminators of
the SA approach are modeled either as Industrial Automation Things or as
human Things.
This definition of Thing satisfies the requirement set by Zhuming et
al. [12] according to which all interactions among the system constituent
components, which may be humans, machines and products should be
performed under the same umbrella. This allows the developers to focus on
the system’s functionality and not worry about interactions, thus increasing
the productivity.
In the design model of the system we do not capture resources. A
resource is a technology artifact used: 1) to represent the exposed properties
and services of a Thing, and 2) to access these through a well-defined set
of operations to achieve low coupling between collaborating Things. The
LWM2M defines a set of such operations, i.e., READ, EXECUTE, WRITE,
etc., implemented on top of the http operations.
The RESTful as well as the SOAP paradigms can be utilized for
accessing the services offered by Things. Thus, we adopt a different meaning
for service from the widely accepted and described, e.g., [18], where access
to resources from the outside world finally happens through services. IoT-A
[23] which is a result of the IoT-A EU project, consider the service as an
entity that accesses a Resource which is associated with an Entity that has
attached a Device. It should also be noted that while a resource is defined
in IoT-A as the core software component that represents an entity in the
digital world, a Device is attached to a Resource. We do not adopt this model
because it is technology driven. Our model focuses on the system modeling
level and its objective is to offer a platform independent modeling of the
target system. In our model, a Thing has structural (attributes) and behavioral
properties (functions/methods). Those properties that are accessible from its
environment are represented as resources. The REST ful paradigm is adopted
for accessing the resources. In this context and to exploit the benefits of
IoT, the networking entities of the IoT Proc&ComnInfr are also considered
as Things (IoT-Thing Networking Thing) that provide their own set of
information processing services required to establish the communication
infrastructure of the IoT.
28 Mechatronics and Robotics

Plant processes as well as other functionalities of the plant are assigned to

cyber objects of the system’s design model. These cyber objects may be
marked as cyber Things. Alternatively, cyber objects may be deployed on
other cyberThings or on Things of the Proc&ComunInfr. In both cases the
corresponding services are mapped to resources of the corresponding Thing.
Our approach differs from the one of the ebb its platform [20] that
identifies the following four layers which are considered required to bind
the physical world with software services:
1) Physical-world layer, where devices, sensors and physical-objects
are captured,
2) The IoT layer,
3) The internet-of-services layer and,
4) The business system mediation and product life cycle layer.
Pramudianto et al. [35] capture sensors and actuators as first-class model
elements in their IoT meta model and use the term virtual object to refer
to the software entity that acts as a proxy of the real-world object. In our
model, the meaning of the virtual Thing is completely different. We use the
term software representative (SR) to refer to what authors in [35] call virtual
object.

A CONTIKI BASED INDUSTRIAL AUTOMATION

THING

The Liqueur Production Laboratory System

The Liqueur production mechatronic system, used as case study in this
work, is composed of the following mechatronic components: smartSilo1,
smartSilo2, smartSilo3, smartSilo4 and smart Pipe (see Figure 4). The
system is based on the case study initially used by Basile et al. [36] and
then extended by Thram boulidis [16] to be compliant with the mechatronic
component concept. The smart Silo mechatronic components are reserved
in couples to produce specific types of liqueurs. Smart Silos 1 and 4 form
one couple; smartSilos 2 and 3 form the other couple. A mechatronic
component has a well-defined interface through which exposes its behavior
to be used by the liqueur production processes. This interface exposes the
functionalities offered by the silo such as fill, empty, mix and heat. Using
the common pipe at the same time for liquid transfer among the silos is
not allowed. Moreover, mixing the liquid in two silos at the same time is
 From Mechatronic Components to Industrial Automation ... 29

not permitted due to a constraint in power consumption. Implementation

issues regarding the physical silo are encapsulated and hidden from the
mechatronic component’s environment.
Our intention is to integrate the components of this conventional
mechatronic system using IoT and gain from the low coupling that this
technology introduces among the interacting components. The use of the
IoT will also enable the system to exploit the benefits of this technology
regarding the user interaction by allowing end users to produce custom
types of liqueur. The end user would be able to define, through an app (my
Liqueur App), the production parameters of the desired type of liqueur, as
shown in Figure 4.

The Cyber-Physical Component

The legacy smart Silo Mechatronic/cyber-physical component is composed
of the physical silo (physical part), a processing, storage and communication
unit and the low-level control software (cyber part) required for the smart
Silo to provide a higher level of abstraction functionality compared to the
one provided by the physical silo.
As shown in Figure 5, which presents the high-level architecture of the
cyber-physical component, the software part is composed of two main parts.
The first part is the software representation of the physical object, i.e., the
mechanical unit, into the software domain.

Figure 4. The Liqueur production system used as case study.

30 Mechatronics and Robotics

Figure 5. The architecture of the cyber- physical component.

This part does not add any extra functionality; it only encapsulates the
details of the integration of the physical world with the cyber world. On top
of this, another part, the controller in Figure 5, transforms the physical object
to a smart one adding extra functionality. This part encapsulates the low-
level control of the physical object required to transform the physical world
object into a smart cyber-physical component that provides its functionality
through a well-defined interface.
We use the Interface construct of UML to specify the cyber interface
of a cyber-physical component. The Interface is used in UML to declare
a set of public features and obligations that together constitute a coherent
service [37]. In this sense, an Interface specifies a contract that any instance
of the mechatronic component shall fulfil. The UML class diagram of Figure
6 presents the interface of the smart Silo Mechatronic component in terms
of provided and required interfaces. The Smart Silo Usage If represents the
provided interface while the Smart Silo User If represents the required one.
In the required interface, we show how to model the interaction between
SmartSilo and its client with the Signal and Reception constructs of UML in
order to represent the possibly asynchronous nature of this interaction.

Figure 6. The cyber interface of the SmartSilo mechatronic component.

 From Mechatronic Components to Industrial Automation ... 31

Thus, the heating Completed and mixing Completed signals sent by the
Smart Silo will trigger an asynchronous without a reply reaction to the
Smart Silo client, e.g., the type A liqueur generation process, through the
corresponding Receptions captured in the Smart Silo User If.

Towards a Contiki-Based Industrial Automation Thing

The 6LoWPAN IoT gateway of Weptech electronic Gmbh running the
Contiki operating system is used to host the controller of the smartSilo
mechatronic component. The 6LoWPAN IoT gateway, which is based on
an ARM®Cortex®? M3 SoC with 512 kB Flash and 32 kB RAM, functions
as a border router in a 6LoWPAN network. It connects a wireless IPv6
network, over an 802.15.4 compliant radio interface in the 2.4 GHz band, to
the Internet via a 10BASE-T Ethernet interface.
Contiki is a lightweight operating system ported to various microcontroller
architectures on resource constraint devices. It was selected mainly for its
event-driven kernel that guaranties fast response times to events and to its
ability for dynamic loading and replacement of individual programs and
services that leads to very flexible Mechatronic components whose behavior
may be modified during run-time. The interfacing of the cyber part with the
physical one, has been developed using the event-driven handling mechanism
of Contiki to get a better response time compared to the traditional scan
cycle approach mainly used in industry. Sensor signals generate interrupts
which are handled by Contiki and transformed to asynchronous software
events. These Events are broadcasted and captured by the corresponding
event handling routines. Thus, the high-level sensor signal is transformed to
the high Level Reached asynchronous event. This event is handled by the
corresponding event handling routine, which is responsible to implement
the sensor data handling algorithm. The data handling algorithm sends
among others a close signal to the in Valve and activates the sending of a
filling Completed event to the client of the mechatronic component, i.e., the
liqueur generation process.
The response of the system from the time that the sensor generates
the signal to the time that the signal arrives to the in Valve actuator has an
average value of 39.20 μs. Listing 1 presents a part of the object-based C
implementation of the smart silo cyber part that is related with the interface
of the component with its environment. This implementation is for the case
that the required interface will be modelled by call back functions instead
of signals and receptions. It is evident that both alternatives, i.e., signals
32 Mechatronics and Robotics

or call back functions, imply a tight coupling among the smartSilo and the
components that use its behaviour, in the sense that these interfaces should
be known in advance for the development of the component’s clients.
A model-to-model transformer is required to automate the process of
generating the IAT. This transformer will use as input the properly annotated
with the DSML conventional mechatronic component. One approach is to
mark the UML design specification of the mechatronic component with the
stereotypes defined by the UML4IoT profile. If a UML design is not available
then the source code of the cyber part of the mechatronic component is
properly annotated with specific annotations that have been defined based on
the UML profile. An example of annotated code with Java-like annotations
is given in Listing 2, where objects and object types of the cyber-physical
component as well as their properties that should be exported to the IoT
environment are properly annotated.

Listing 1. Part of the C object-based implementation of the cyber part of the

mechatronic component.

Listing 2. Part of the C object-based implementation of the cyber part of the

mechatronic component annotated with the UML4IoT java-like annotations.
In Figure 7, the Contiki-based silo industrial automation thing developed
with the proposed approach is shown. In the current implementation, the
Weptech embedded board is used as processing unit while a hardware
simulator is used for the silo. The Raspberry Pi and the XDK of Robert
Bosch are alternative supported platforms.
 From Mechatronic Components to Industrial Automation ... 33

The Interfaces of the Industrial Automation Thing

Adopting the OMA LWM2M application protocol, the interface of the
Industrial Automation Thing is well defined and independent of the behavior
that is implemented by the component. This interface is defined using UML
provided and required interfaces as shown in Figure 8. Based on this figure
the IAT has three provided interfaces and three required that are independent
of the nature of the component. This feature combined with the ability of
Contiki to dynamically load and replace individual services, results to a
completely flexible component regarding its behavior. New or replaced
behavior can be activated by the same well-defined REST interface of the
Industrial Automation Thing.

Figure 7. The silo industrial automation thing based on a hardware silo simula-
tor.

Figure 8. Industrial automation thing interfaces (provided and

required).
A comparison with Figure 6 that captures the conventional mechatronic
component interface points out the flexibility of the AIT compared to the
conventional one. Through this REST interface, resources may be created
34 Mechatronics and Robotics

and used on demand based on requirements assuming that the physical part
supports the requested new behavior. Resources, Resource Instances, Objects,
Object Instances which are exposed by the IAT as well as their attributes,
are accessed by the clients of the IAT through the device management and
service enablement interface (DM&SE If). IPSO smart objects [38] have
been adopted in this work to satisfy the requirement for interoperability.

AUTOMATING THE GENERATION PROCESS OF IAT

For the automation of the generation process of the IAT a model-to-model
transformer has been implemented. This transformer accepts as input the
annotated legacy source code and applies a set of transformation rules
properly defined to get the source code of the IAT.

The Model-to-Model Transformer

Α prototype model-to-model transformer was developed based on Autogen
and GNU Guile. Autogen (https://www.gnu.org/software/autogen/) is a
GNU tool that supports creation and maintenance of source code. As shown
in Figure 9, Autogen takes as input the annotated legacy code of the smart
object, a template file and the definitions file. The template file implements
the set of transformation rules. It actually defines the structure of the textual
output of Autogen, i.e., the structure of the IAT source code, making use of
autogen’s macro-type format and embedded scheme code. The definitions
file provides the information required to instantiate the template file. More
specifically, it includes the LWM2M object/resource/instance properties and
other source-level information. The definitions file can be generated by the
appropriate scheme procedures of the c-parser Guile module that uses as
input the annotated legacy source code, as shown in Figure 9.

Figure 9. Transformation process of the IAT.

 From Mechatronic Components to Industrial Automation ... 35

The Guile module that implements the c-parser scheme procedures was
developed as a high-level recursive descent parser, based on the guile-parser-
combinators module (https://git.dthompson.us/guile-parser-combinators.
git). GNU Guile is an implementation of the Scheme programming language
(https://www.gnu.org/software/guile/). Autogen collaborates with the Guile
VM through guile procedure calls embedded in the template file. The output
of the transformation process is the source code of the IAT in a format ready
to be compiled with the Contiki operating system and then deployed.

Transformation Rules
The following rules that apply for each object of the lwm2m client have been defined.
Rule 1:Create wrapper functions for annotated behaviors.
For each function with the BehaviorResource annotation create a
wrapper function with input parameters:
lwm2m_context_t *ctx, const uint8_t *arg, size_targsize, uint8_t
*outbuf, size_t outsize
E.g., Source:static int fill(void);
Target: static intfill(lwm2m_context_t *ctx, const uint8_t *arg, size_
targsize, uint8_t *outbuf, size_t outsize);
Rule 2:Create getters/setters functions for each property annotated with
the PrimitiveRes annotation.
For each attribute annotated with the PrimitiveRes annotation create the
corresponding read and write function depending on the applied operations
on the attribute defined in the annotation.
E.g., for the silo_state property
staticintget_silo_state(lwm2m_context_t *ctx, uint8_t *outbuf, size_t
outsize) {
char *value;
value = get_silo_state_inString(silo->state);
returnctx->writer->write_string(ctx, outbuf, outsize, value,
strlen(value));}
Rule 3: Construct the Resource model.
For each annotated attribute or function create an entry using the
LWM2M_RESOURCE_CALLBACK macro of the lwm2m implementation
which is integrated into the Contiki OS. E.g., for the silo_state
36 Mechatronics and Robotics

LWM2M_RESOURCE_CALLBACK(0,{get_silo_state, NULL, NULL}),

Append this entry to a list of resources, i.e., silo_resources, using the
LWM2M_RESOURCES macro.
Create the corresponding object instance using the LWM2M_INSTANCE
macro and register the resources. E.g., LWM2M_INSTANCE(0, silo_
resources)
Append it to the list of silo instances using the LWM2M_INSTANCES.
E.g.,LWM2M_INSTANCES(silo_instances, …. );
Create the Object and registers its instances using the LWM2M_
OBJECT macro. E.g.,
LWM2M_OBJECT(silo_obj, 1663, silo_instances);
Rule 4: Modify setter functions.
For each attribute annotated with the ObservableResource annotation,
modify its setter function (set_ ())by appending a call to the lwm2m_object_
notify_observersfunction.
Assumption: For each observable attribute a setter function exists. E..g.,
Source: void set_filling_completed(){
silo->filling_completed = 1;}
Target: void set_filling_completed(){
silo->filling_completed = 1;
lwm2m_object_notify_observers(&silo_obj, “/0/7”); }
Rule 5: Generate and handle the initialize function for the object.
5.1 Generate an initialize function to initialize the legacy object and
register it to lwm2m by a call to thelwm2m_ engine_register_
object function.
The legacy initialize function of the object should be properly annotated.
E.g.,
void ipso_ silo_init(void) {
silo_init(); // legacy object initialization function
lwm2m_engine_register_object(&silo_obj); }
5.2 Append the initialize function prototype to theipso-objects.hfile.
E.g.
void ipso_silo_init(void);
 From Mechatronic Components to Industrial Automation ... 37

5.3 Append a call statement to the initialize function of each object to

theipso_objects_init()function body of theipso-objects.cfile. E.g.,
void ipso_objects_init(void) {
ipso_silo_init();

CONCLUSION
In this paper, we consider the tight integration of the physical world with the
cyber one at the mechatronic component level. A mechatronic component
offers its functionality through well-defined mechanical, electrical and
software interfaces. In this sense the industrial automation system is a
composition of mechatronic components along with cyber components and
humans. IoT is adopted for the integration of these components to exploit
the benefit of this technology and UML4IoT is utilized to automatically
transform the conventional mechatronic component into an IoT compliant
cyber-physical one, i.e., to an Industrial Automation Thing. The IoT model
used in the UML4IoT approach is extended towards a complete IoT model
for the manufacturing domain. The transformation rules required for the
development of the model-to-model transformer have been developed and
validated through a prototype implementation of the liqueur production
laboratory system. The prototype implementation of the silo industrial
automation thing based on a Contiki enabled embedded board and the C
language is used to demonstrate the applicability and the effectiveness of
the proposed approach.
38 Mechatronics and Robotics

REFERENCES
1. UNESCO Chair (2016) Home Page. On Mechatronics and Mechatronics
Research and Application Center. http://mecha.ee.boun.edu.tr/
2. Thramboulidis, K. (2005) Model Integrated Mechatronics—Towards
a new Paradigm in the Development of Manufacturing Systems.
IEEE Transactions on Industrial Informatics, 1, 54-61. https://doi.
org/10.1109/TII.2005.844427
3. Thramboulidis, K. and Scholz, S. (2010) Integrating the 3+1 SysML
View Model with Safety Engineering. 2010 IEEE Conference on
Emerging Technologies and Factory Automation (ETFA), Bilbao, Spain,
13-16 September 2010, 1-8.https://doi.org/10.1109/etfa.2010.5641353
4. Thramboulidis, K. (2012) Overcoming Mechatronic Design
Challenges: The 3+1 SysML-View Model. The Computing Science
and Technology International Journal, 2, 6-14.
5. FESTO (2016) Industry 4.0: Efficient Engineering Processes
with “OPAK”. https://www.festo.com/net/en_corp/SupportPortal/
MobilePressDetails.aspx?documentId=368146&q=
6. OASIS (2016) Devices Profile for Web Services (DPWS) Specification.
http://docs.oasis-open.org/ws-dd/ns/dpws/2009/01
7. Angulo, P., Guzmán, C.C., Jiménez, G. and Romero, D. (2016) A
Service-Oriented Architecture and Its ICT-Infrastructure to Support
Eco-Efficiency Performance Monitoring in Manufacturing Enterprises.
International Journal of Computer Integrated Manufacturing, 30, 202-
214. https://doi.org/10.1080/0951192X.2016.1145810
8. Morgan, J. and O’Donnell, G.E. (2015) Enabling a Ubiquitous
and Cloud Manufacturing Foundation with Field-Level Service-
Oriented Architecture. International Journal of Computer Integrated
Manufacturing, 30, 442-458.
9. He, W. and Xu, L. (2015) A State-of-the-Art Survey of Cloud
Manufacturing. International Journal of Computer Integrated
Manufacturing, 28, 239-250.https://doi.org/10.1080/095119
2X.2013.874595
10. Ghimire, S., Luis-Ferreira, F., Nodehi, T. and Jardim-Goncalves,
R. (2016) IoT Based Situational Awareness Framework for Real-
Time Project Management. International Journal of Computer
Integrated Manufacturing, 30, 74-83.https://doi.org/10.1080/095119
2X.2015.1130242
 From Mechatronic Components to Industrial Automation ... 39

11. Ren, L., Zhang, L., Wang, L., Tao, F. and Chai, X. (2014) Cloud
Manufacturing: Key Characteristics and Applications. International
Journal of Computer Integrated Manufacturing, 30, 501-515. https://
doi.org/10.1080/0951192X.2014.902105
12. Bi, Z., Xu, L.D. and Wang, C. (2014) Internet of Things for Enterprise
Systems of Modern Manufacturing. IEEE Transactions on Industrial
Informatics, 10, 1537-1546. https://doi.org/10.1109/TII.2014.2300338
13. Bradley, D., Russell, D., Ferguson, I., Isaacs, J., MacLeod, A. and
White, R. (2015) The Internet of Things—The Future or the End of
Mechatronics. Mechatronics, 27, 57-74.
14. Thramboulidis, K. and Christoulakis, F. (2016) UML4IoT—A UML
Profile to Exploit IoT in Cyber-Physical Manufacturing Systems.
Computers in Industry, 82, 259-272.
15. Dunkels, A., Gronvall, B. and Voigt, T. (2004) Contiki—A Lightweight
and Flexible Operating System for Tiny Networked Sensors. 29th
Annual IEEE International Conference on Local Computer Networks,
16-18 November 2004, 455-462. https://doi.org/10.1109/lcn.2004.38
16. Thramboulidis, K. (2015) A Cyber-Physical System-Based Approach
for Industrial Automation Systems. Computers in Industry, 72, 92-102.
17. Eclipse (2016) Papyrus for IoT—A Modeling Solution for IoT. https://
www.eclipse.org/community/eclipse_newsletter/2016/april/article3.
php
18. Open Mobile Alliance (OMA) (2015) Lightweight Machine to
Machine Technical Specification. OMA-TS-LightweightM2M-V1_0-
20151214-C, Candidate Version 1.0. 14 December 2015.
19. Kovatsch, M., Duquennoy, S. and Dunkels, A. (2011) A Low-Power
CoAP for Contiki. 8th IEEE International Conference on Mobile Ad-
Hoc and Sensor Systems, Valencia, Spain, 17-21 October 2011, 855-
860. https://doi.org/10.1109/MASS.2011.100
20. Khaleel, H., Conzon, D., Kasinathan, P., Brizzi, P., Pastrone, C., et al.
(2015) Heterogeneous Applications, Tools, and Methodologies in the
Car Manufacturing Industry through an IoT Approach. IEEE Systems
Journal, PP, 1. https://doi.org/10.1109/JSYST.2015.2469681
21. Tao, F., Cheng, Y., Xu, L.D., Zhang, L. and Li, B.H. (2014) CCIoT-
CMfg: Cloud Computing and Internet of Things-Based Cloud
Manufacturing Service System. IEEE Transactions on Industrial
Informatics, 10, 1435-1442.https://doi.org/10.1109/TII.2014.2306383
40 Mechatronics and Robotics

22. Bauer, M., Bui, N., De Loof, J., Magerkurth, C., Nettstrter, A., Stefa, J.
and Walewski, J. (2013) Iot Reference Model. In: Bassi, A., et al., Eds.,
Enabling Things to Talk, Springer, Berlin, Heidelberg, 113-162.https://
doi.org/10.1007/978-3-642-40403-0_7
23. IoT-A (2016) Final Architectural Reference Model for the IoT. http://
www.meet-iot.eu/deliverables-IOTA/D1_5.pdf
24. Tao, F., Zuo, Y., Xu, L.D. and Zhang, L. (2014) IoT-Based Intelligent
Perception and Access of Manufacturing Resource toward Cloud
Manufacturing. IEEE Transactions on Industrial Informatics, 10, 1547-
1557.https://doi.org/10.1109/TII.2014.2306397
25. Diaz-Cacho, M., Delgado, E., Falcon, P. and Barreiro, A. (2015) IoT
Integration on Industrial Environments. 2015 IEEE World Conference
on Factory Communication Systems (WFCS), Palma de Mallorca, 27-
29 May 2015, 1-7. https://doi.org/10.1109/wfcs.2015.7160553
26. Broy, M., Kirstan, S., Krcmar, H., Schatz, B. and Zimmermann, J.
(2013) What Is the Benefit of a Model-Based Design of Embedded
Software Systems in the Car Industry? In: Software Design and
Development: Concepts, Methodologies, Tools, and Applications, IGI
Global, 310-334.
27. Nguyen, X.T., Tran, T., Baraki, H. and Geihs, K. (2015) FRASAD:
A Framework for Model-Driven IoT Application Development.
2015 IEEE 2nd World Forum on Internet of Things (WF-IoT),
Milan, 14-16 December 2015, 387-392. https://doi.org/10.1109/WF-
IoT.2015.7389085
28. Malavolta, I. and Muccini, H. (2014) A Study on MDE Approaches
for Engineering Wireless Sensor Networks. 40th EUROMICRO
Conference on Software Engineering and Advanced Applications
(SEAA), Verona, 27-29 August 2014, 149-157.https://doi.org/10.1109/
seaa.2014.61
29. Conzon, D., Brizzi, P., Kasinathan, P., Pastrone, C., Pramudianto,
F. and Cultrona, P. (2015) Industrial Application Development
Exploiting IoT Vision and Model Driven Programming. 2015 18th
International Conference on Intelligence in Next Generation Networks
(ICIN), Paris, 17-19 February 2015, 168-175. https://doi.org/10.1109/
icin.2015.7073828
30. Prehofer, C. (2015) Models at REST or Modelling RESTful Interfaces
for the Internet of Things. IEEE 2nd World Forum on Internet of Things
(WF-IoT), Milan, 14-16 December 2015, 251-255.
 From Mechatronic Components to Industrial Automation ... 41

31. Eterovic, T., Kaljic, E., Donko, D., Salihbegovic, A. and Ribic, S. (2015)
An Internet of Things Visual Domain Specific Modeling Language
Based on UML. 2015 XXV International Conference on Information,
Communication and Automation Technologies (ICAT), Sarajevo, 29-
31 October 2015, 1-5. https://doi.org/10.1109/icat.2015.7340537
32. Zhang, Y., Zhang, G., Wang, J., Sun, S., Si, S. and Yang, T. (2015)
Real-Time Information Capturing and Integration Framework of the
Internet of Manufacturing Things. International Journal of Computer
Integrated Manufacturing, 28, 811-822. https://doi.org/10.1080/09511
92X.2014.900874
33. De, S., Barnaghi, P., Bauer, M. and Meissner, S. (2011) Service
Modelling for the Internet of Things. 2011 Federated Conference on
Computer Science and Information Systems (FedCSIS), Szczecin, 18-
21 September 2011, 949-955.
34. Haller, S. (2010) The Things in the Internet of Things. Internet of
Things Conference 2010, Tokyo, 29 November-1 December 2010.
35. Ferry, P., et al. (2014) IoTLink: An Internet of Things Prototyping
Toolkit. IEEE International Conference on Ubiquitous Intelligence and
Computing, Bali, 9-12 December 2014, 1-9.
36. Basile, F., Chiacchio, P. and Gerbasio, D. (2013) On the Implementation
of Industrial Automation Systems Based on PLC. IEEE Transactions
on Automation Science and Engineering, 10, 990-1003. https://doi.
org/10.1109/TASE.2012.2226578
37. OMG (2015) Object Management Group, OMG Unified Modeling
Language (OMG UML), Version 2.5, OMG Document Number
Formal/2015-03-01. http://www.omg.org/spec/UML/2.5
38. Internet Protocol for Smart Objects (IPSO) Alliance (2014) IPSO
Smart Object Guideline. IPSO Smart Object Committee, 21 September.
https://www.ipso-alliance.org/ipso-community/resources/smart-
objects-interoperability/
 CHAPTER

3
The 3+1 SysML View-Model
in Model Integrated
Mechatronics

Kleanthis Thramboulidis
Visiting Professor Helsinki University of Technology, Electrical & Computer
Engineering, University of Patras, Patras, Greece.

ABSTRACT
Software is becoming the driving force in today’s mechatronic systems. It
does not only realize a significant part of their functionality but it is also
used to realize their most competitive advantages. However, the traditional
development process is wholly inappropriate for the development of these
systems that impose a tighter coupling of software with electronics and

Citation:- K. Thramboulidis, “The 3+1 SysML View-Model in Model Integrated Me-

chatronics,” Journal of Software Engineering and Applications, Vol. 3 No. 2, 2010, pp.
109-118. doi: 10.4236/jsea.2010.32014.
Copyright: © 2010 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
44 Mechatronics and Robotics

mechanics. In this paper, a synergistic integration of the constituent parts of

mechatronic systems, i.e. mechanical, electronic and software is proposed
though the 3+1 SysML view-model. SysML is used to specify the central
view-model of the mechatronic system while the other three views are for
the different disciplines involved. The widely used in software engineering
V-model is extended to address the requirements set by the 3+1 SysML view-
model and the Model Integrated Mechatronics (MIM) paradigm. A SysML
profile is described to facilitate the application of the proposed view-model
in the development of mechatronic systems.
Keywords: Systems Engineering, System Modeling, Mechatronic
Component, Model Driven Development, Model Integrated Mechatronics,
SysML Profile, V-Model, IEC61499

INTRODUCTION
Software does not only implement a significant part of the functionality of
today’s mechatronic systems, but it is also used to realize their most competitive
advantages. It is the evolving driver for innovations in many mechatronic
systems and in general it is considered as the driving force in improving this
kind of systems. However, the traditional development process is wholly
inappropriate for the development of systems characterized by complexity,
dynamics and uncertainty as is the case with today’s mechatronic systems
[1]. According to the traditional development process the constituent parts
of the mechatronic system, i.e. the mechanical, electronic and software, that
constitute the system are developed independently and then are integrated to
compose the final system.
The software development starts when the development of electronic
and mechanical is already at a stage where any change in these parts is
expensive and time consuming. This is why the mechanical and electronic
properties impose several constrains and narrow the solution space for
software development. Moreover, as claimed in [2] “the actual cooperation
during the construction is less developed. There is no joint development
process, no joint tool usage, no joint modeling formalism and no joint
analysis. Every discipline has its own approaches”. As a result of this, the
current process that is traditionally divided into software, electronics and
mechanics, emphasizes on domain-spanning design methods and tools and
is unable to address the demand for synergetic mechatronic dependability
predictions; this is why many products suffer from severe dependability
problems [3].
 The 3+1 SysML View-Model in Model Integrated Mechatronics 45

An integrated framework for the construction of mechatronic systems is

missing [2]. Such a framework should address current challenges in the
development of mechatronic systems that include among others, synergistic
modeling and integration, design synchronization, as well as model execution
and analysis. It should provide the infrastructure required for applying a tight
integration of mechanics with electronics and software in order to replace
conventionally designed mechanical and electromechanical systems into
smart ones where significant part of functionality will be implemented by
software. It is also expected to result in massive improvements in system’s
Quality of Service (QoS) characteristics and allow a smooth integration of
dependability predictions during the early development phases.
Model Integrated Mechatronics (MIM) [4] is a paradigm that was
proposed to address the need for an integrated development in mechatronic
systems. MIM supports the model-driven development of complex
mechatronic systems (MTSs) through the evolution of models that have
as primary construct the mechatronic component (MTC). The concept
of component has already been adopted in the development process of
manufacturing systems by other researchers too. However, they mostly
focus on the software part of the component; they do not address the
whole development process; and they do not provide an architecture for
the concurrent engineering of all constituent components, i.e., mechanical,
electronic and software.
In this paper, the Systems Modeling Language (SysML) [5] is adopted
for the system’s modeling process in the MIM paradigm. An architectural
view-model, the one called 3+1 SysML is proposed to address the
synergistic integration of the constituent parts of mechatronic systems. The
main view is the SysML view that corresponds to the mechatronic layer
of the MIM Architecture. This view captures the system model that is the
one constructed by the MTS developer. Each of the three views is used to
describe the system from the perspective of the corresponding discipline.
Specific tools of every discipline may be exploited for the model execution
and analysis of the SysML models.
SysML is used to represent the models of the mechatronic system that
are proposed by the MIM paradigm. A SysML profile was defined using
Papyrus, an open source tool for graphical UML 2/SysML modeling. This
profile supports not only the modeling of mechatronic systems using the
concept of mechatronic component but also a hybrid development process
that integrates the traditional approach with the MTC-based one, to allow
46 Mechatronics and Robotics

the reuse of legacy systems. It provides a SysML based implementation of

the MIM architecture that will allow the MIM paradigm to be exploited
in the development process of real world mechatronic systems. An
effective development process that exploits the 3+1 SysML view-model
and the SysML profile is described extending the well-known in software
engineering V-model [6].
The remainder of this paper is organized as follows. In the next section,
a brief introduction to the MIM architecture and the SysML is given and the
related work is briefly discussed. Section 3 presents the proposed 3+1 SysML
view-model that emphasizes the importance of the common system-level
model. In Section 4, a SysML profile that represents the artifacts used in the
MIM architecture is described. Section 5 presents the proposed modifications
to the widely used in software systems V-model to address the needs of the
MTS development process. Future developments and research challenges are
discussed in Section 6 and the paper is concluded in the last section.

BACKGROUND AND RELATED WORK

A Brief Introduction to MIM

The upper layer of the MIM architecture that is shown in
Figure 1, i.e. the mechatronic layer, was defined to systematically
address complexities in the model-driven development process of
component-based mechatronic systems. This layer is projected into three
dimensions representing the application, the resource, and the mechanical
process respectively. The controlling application software is modeled in the
application layer, while the hardware, i.e., computing and communication,
as well as the software resources that constitute the infrastructure required
for the execution of the controlling application software, are modeled in the
resource layer. Mechanical, hydraulic and pneumatic parts are modeled in
the mechanical layer.
Mechatronic system integrators work horizontally in the model
evolution dimension of the MIM architecture. They interactively compose
the MTS using already defined MTCs without worrying on lower layers’
implementation details. They go through a model-driven development
process to build the MTS using descriptions of already existing MTCs. They
only have to capture the application logic in application layer components,
as well as to identify their required QoS characteristics from the resource
layer infrastructure.
 The 3+1 SysML View-Model in Model Integrated Mechatronics 47

MTC builders work in the model integration dimension and apply an

information integration process that crosses the boundaries between
mechanical, electronic, and computer science fields. They work horizontally
and vertically, either top-down or bottom-up, in the lower three layers of
the architecture in a concurrent way. Constructed MTCs are stored in MTC
repositories to be discovered and used by mechatronic system integrators [7].
MIM is a new paradigm that promotes model integration not only of
implementation space artifacts but also of early analysis and design phase
ones. It promotes reuse at the mechatronic level and significantly decreases
development and validation time of the system.

Figure 1. The Model Integrated Mechatronics (MIM) architecture [4].

Figure 2. SysML diagrams and its comparison with UML 2 [5].

MIM addresses the need for synergistic integration at the model and
process levels; it facilitates the integration between the design processes
of the different disciplines which is the approach considered as the most
effective to improve the development process of mechatronic systems. The
other two approaches being: implementation time integration and design
time integration at data and model level [8].
48 Mechatronics and Robotics

Systems Modeling Language (SysML)

SysML was developed to support the specification, analysis, design,
verification and validation of a broad range of complex systems [5]. These
systems may include hardware, software, information, processes, personnel,
and facilities. The objective of SysML is to unify the diverse modeling
languages currently used by system engineers. SysML reuses a subset of
UML 2 [9] and provides additional extensions needed to address system
engineering aspects not covered by UML 2. It includes diagrams that can
be used to specify system requirements, behavior, structure and parametric
relationships. Requirements diagram and parametric diagrams are the new
diagram types proposed by SysML, as shown in Figure 2 which presents the
diagrams that are used by SysML.
SysML provides modeling constructs to represent text-based
requirements and relate them to other modeling elements. The requirement
diagram can be used to represent many of the relationships that exist between
requirements and visualize them. It provides a bridge between traditional
requirements management tools and the other SysML models. It can be used
to depict the requirements in graphical, tabular, or tree structure format and
highlight their relationships, as well as to capture the relationships between
requirements and other model elements that satisfy or verify them.
The other new diagram, i.e. the parametric diagram, is used to describe
the constraints among the properties associated with blocks. It allows the
specification of continuous components by parametric constraints on class
attribute values expressing corresponding differential equations. However, the
syntax and the semantics of behavioral descriptions captured in parametric
diagrams have not been defined to allow the integration with other simulation
and analysis modeling techniques for the proper execution of the models. So,
the parametric diagram is used to integrate the system descriptive behavior and
structure models expressed in SysML with other simulation and engineering
analysis models such as performance, reliability, and mass property models.
The fact that SysML is based on UML 2, will allow system engineers
modeling with SysML and software engineers modeling with UML 2 to
collaborate on models of the mechatronic system. This will improve
communication among the various stakeholders who participate in the
mechatronic systems development process and will promote interoperability
among modeling tools in different disciplines. All the above mentioned
characteristics make SysML ideal for the representation of models used in
the MIM paradigm.
 The 3+1 SysML View-Model in Model Integrated Mechatronics 49

Related Work
Several researchers are already working in the direction of improving the
effectiveness of the development process of mechatronic systems. Schafer
and Wehrheim [2] survey on current developments in mechatronics and
present the architecture of their mechatronic rail system that seems to provide
an excellent platform for studding and analyzing future developments and
research chalenges in mechatronic systems. They identify the need of an
integrated framework for the construction of mechatronic systems and
they discuss future trends in mechatronics especially from the software
engineering point of view. Habib [10] argues on the urgent need for theories,
models, and tools that should facilitate modeling, analysis, synthesis,
simulation, and prototyping of mechatronic systems. He emphasizes the
argument that the approach based on optimization within each domain
separately will not result in the optimum system design, and he proposes
a data and model integration approach to address the integration problem.
Burmester et al. [11] claim that in today’s mechatronic systems most of
the control and reconfiguration functionality is realized in software. They
present mechatronic UML to exploit the Model Driven Architecture
approach for the design of hybrid mechatronic real-time systems that have
to fulfill safety-critical requirements. “Mechatronic UML” is defined as an
extension of UML to built platform independent models for mechatronic
systems. Various UML models have been extended to cover the requirements
of modeling the structural view as well as the behavioral view of the system.
However, the proposed extension is used to model only the software part of
the mechatronic system. Authors in [12] briefly refer to a process model of
Robert Bosch GmbH for the development of mechatronic systems in Motor
Vehicles to support aspects such as reuse, exchangeability, scalability and
distributed development. They use the concept of mechatronic component,
even though not well defined, as the basic construct of their process. They
argue: a) on the need of a clear specification of component interfaces;
and b) the great contribution of re-use to increase the quality properties
of mechatronic systems and decrease development time. Nordmann [13]
is using the concept of mechatronic component and presents an example
of using Active Magnetic Bearings to increase performance, reliability,
reusability and longer lifetime. Authors in [14] propose for the development
of multidisciplinary systems, such as mechatronics, the integration of the
various domain-specific tools. They mainly focus on the integration of used
data and models and not on a process level integration. Moreover, none
of these approaches provide a high level architecture for an integrated,
50 Mechatronics and Robotics

synergistic development process for mechatronic systems and they do not

describe a systems level development process based on the mechatronic
component and the emerging standard in the domain that is SysML.

THE 3+1 ARCHITECTURAL VIEW MODEL

Each of the three lower layers of the MIM architecture provides a specific view
of the central models that are captured in the upper layer, i.e. the mechatronic
layer. Each view is used to describe the system from the perspective of the
corresponding discipline. The software view (s-view), for example, provides
the models of the software part of the MTS and allows for software specific
tools to be used to elaborate and further refine these models. The IEC61499
function block model is an example of such a domain specific model that can
be used to further refine the s-view [15]. Figure 3 depicts the 3+1 SysML
view-model that is proposed for the development of mechatronic systems.
The MTS model is the heart of this architecture and is depicted in the center
of the picture. It is surrounded by 4 views which correspond to the roles that
engineers play during the development process of mechatronic systems.
The main view is the MTS-view that corresponds to the mechatronic
layer of the MIM Architecture. This is the view that is used by the MTS
developer. The other 3 views correspond to the 3 lower layers of the MIM
architecture. The m-View, for example, corresponds to the Mechanical layer
of the MIM architecture and captures all the mechanics, hydraulics and
pneumatics of the MTS model. These models are generated by projecting the
MTS models to the mechatronic layer and are fully synchronized with the
MTS models. Any modifications imposed by the mechanical engineer to the
models of this view directly affect the corresponding central MTS models.
Moreover, modifications done by the MTS developer on the central models
directly affect the corresponding m-view models. The m-view is mainly
used during the primitive MTC development process where a concurrent,
synergistic engineering on the three views is adopted at the primitive MTC
level as a more effective process. It is also used to have a whole view of the
mechanical system model and perform optimization and analysis activities on
this.Figure 4 presents the structure of the primitive MTC and its interfaces
to the environment. A primitive MTC may expose to the environment
mechanical, electronic and software interfaces through the corresponding
ports. Interfaces between its constituent parts are also shown. Sensors and
actuators are used to realize the interactions between the mechanical part
and the electronic one.
 The 3+1 SysML View-Model in Model Integrated Mechatronics 51

Figure 3. The 3+1 SysML view-model for mechatronic systems development.

Figure 4. The structure of the primitive mechatronic component.

This means that sensors and actuators are fully encapsulated by the
MTC construct. The mechanical part interacts with the environment only
through mechanical ports that are mainly in the form of mounting ports or
ports for flow of material or energy. A detailed description of mechanical
connections that can be discriminated into fixed and moveable arrangements
is given in [16]. The sOperator Panel, i.e. the software part of the Operator
Panel primitive MTC, exposes its functionality along with the corresponding
QoS through a provided interface of the sPort. A hosting functionality for
application specific components will be optionally provided by the ePort.
Currently there is no tool to execute the MTS models, not even to analyze
their behavior. Discipline specific tools may be exploited for the execution
and analysis of MTS models. This is obtained by the proper integration and
coordination of specific model execution and analysis tools of the three
views. The tool of each view is used to execute the primitive MTC model of
52 Mechatronics and Robotics

its perspective so it has to provide specific interfaces to the tools of the other
views, in order to implement the interactions of its own part to the other
parts of the primitive MTC. The arrows that cross the boundaries between
the three views in Figure 3 represent the interactions of the corresponding
models and have to be implemented by specific interfaces of the model
refinement, analysis and execution tools of the three disciplines. The AP233
or more formally the ISO 10303-233 standard for systems engineering [17],
that provides a data exchange format for the reliable interchange of data
between software tools may be exploited to effectively implement these
interactions. The execution and analysis of the primitive MTC is obtained
through a collaboration of the corresponding tools of the three views. It is
clear that the contribution of the three views’ specific tools is restricted at
the primitive MTC internal level while the execution and analysis of MTC
models is done at the MTS level with the coordination and synchronization
between MTCs carried by this level. This makes the tool integration a major
challenge in the domain of mechatronic systems.
The 3+1 SysML view-model when used with the MTS V-Model that is
described in one of the following sections, promote the synergistic integration
of the three constituent parts of the mechatronic system and emphasizes the
importance of a common model for the system. However, this model can
also be used with the traditional development process that is based on the
independent development of constituent parts of the mechatronic system and
their subsequent integration. Even in this case the existence of a common
model for the system greatly improves the effectiveness of the development
process.
It should be noted that in each view the corresponding discipline’s
specific architectures and tools may be exploited, as for example the 4+1
architectural view [18] that may be exploited in the context of the e-view by
the software engineer or the MTC developer/integrator.

USING SYSML TO MODEL THE MIM ARTIFACTS

The Systems Modeling Language can be used to represent the artifacts of
the mechatronic systems development process that correspond to the system
level activities. These include requirements specifications for the MTS and
MTC levels, as well as architectural specifications for the MTS and MTC
levels. UML 2.0 will be used for the modeling of the software part of the
primitive MTC and corresponding tools from the electronics and mechanics
domain will be used for the modeling of the other two constituent parts of
 The 3+1 SysML View-Model in Model Integrated Mechatronics 53

the MTC. In this section the modeling of the MTS and MTC levels using
SysML is considered. It is evident that specific interfaces have to be defined
for the integration of the different views and these interfaces have to be
realized by the tools used in the various disciplines to create a completely
integrated tool chain to support the MTS development process. The SysML
to AP233 mapping [19] is towards this direction.

Modeling of the Mechatronic Component

The MTS stereotype that is shown in Figure 5, which presents part of the
SysML4MIM profile, is considered as a composition of MTS Components
and MTS Connectors (not shown in the figure). The abstract stereotype MTS
Componet was defined to provide more flexibility in system modeling. It
allows the definition of the different disciplines’ components in any level
of the system’s decomposition hierarchy; it also allows the application of
the traditional approach where the system is considered as consisting of
mechanical, electronic and software components. An MTS Component
that is abstract is specialized to the MTC abstract stereotype and the m
Compoment, eComponent and s Component stereotypes. The m Component
stereotype is used to represent in the model any mechanical component of
the mechatronic system.

Figure 5. SysML4MIM profile (part); the MTS stereotype.

This way of modeling allows the profile to be used also in the traditional
development process of mechatronic systems since the MTS may be
considered as a composition of m, e and sComponents. The MTC stereotype
which is also abstract is specialized to the Composite MTC and the Primitive
MTC. This allows a hierarchical decomposition scheme for the MTS up
to the level of primitive MTC that is considered as composition of m, e
and sComponents. An m, e and sComponent may be further decomposed
54 Mechatronics and Robotics

in corresponding Components allowing a component based synthesis in

each one of the three disciplines. The SysML4UML profile that was created
using Papyrus, an open source tool for graphical UML 2 modeling (http://
www.papyrusum-l.org/), imports the UML4SysML profile that is already
supported by Papyrus. This allows the MTS stereotype to extend the Block
stereotype of the UML4SysML profile. All the other components also extend
the Block stereotype even not shown in figure. The proposed SysML4MIM
profile allows the synergistic integration in the development of mechatronic
systems to any level of granularity down to the primitive MTC component
which is the one that is not decided or it is not possible to be decomposed
into lower layer MTC components.

Modeling of the Mechatronic Port

Allowable inputs and outputs of an MTC are defined using the concept of
the port. This allows the design of modular reusable MTCs, with clearly
defined interaction points and interfaces with the environment. The construct
of Mechatronic port (MTPort) was defined as an extension of the UML port
to fulfill this requirement. SysML provides standard ports which support
client-server communication and FlowPorts that define flows in and/or out
of a block. An MTC may own MTPorts, as shown in Figure 6, which allows
the MTC to declare the items it may exchange with its environment and
the interaction points through which this exchange is made. Furthermore,
MTPorts allow the MTC to declare the provided to the environment services
but also the services that the MTC expects from it. An MTPort is defined as
an aggregation of mPort, ePort and sPort. Each port is used to represent the
interaction point of the corresponding part of the primitive MTC with the
environment (see Figure 4). All these ports extend the SysML port; mPort
and ePort extend it through the SysML flow port while sPort extends it
directly. So, a sPort is characterized by provided and required interfaces.
The specification of what can flow in or out of an mPort or ePort is achieved
by typing them with a specification of the things that flow in and/or out. It
should be noted that an mPort may accept or transmit energy or material
but may also accept or transmit information that has been decided to be
transferred by mechanical means. Of course the same information may be
transferred by electronic signals using an ePort or by software messages
using a sPort. The support of several alternatives through configuration,
results in increased reuse potential for the MTC. The specification of the
services of the sPort is achieved by typing it with the provided and/or
required interfaces.
 The 3+1 SysML View-Model in Model Integrated Mechatronics 55

Figure 6. The MTPort stereotype in the SysML4MIM profile.

Flows of mPorts and ePorts may be atomic or non-atomic; an atomic
flow is specified with a single type representing the items that flow in or out.
A non-atomic flow is specified with a flow specification which lists the items
that constitute the flow. A sPort accepts software signals, i.e. packages of
information, which usually need a more complicated specification supported
by UML 2.

THE MTS DEVELOPMENT PROCESS

The MIM development process adopts the V-model as basis and updates it
to address the needs of the mechatronic systems domain. Figure 7 presents
the proposed MTS V-Model. A system modeling process is applied down
to the primitive MTC level, as shown in the left-hand part of the V-model.
For primitive MTCs that have to be constructed, a concurrent engineering
process of all three constituent parts, i.e., mechanics, electronics and software
is adopted, as depicted in the bottom of the V-Model. The system integration
and verification process is depicted by the right-hand side of the V-model.
MTS-level requirements are captured using the SysML requirements
diagram. Essential use cases, which are used to capture the functional
requirements at this level, are defined in abstract, simplified, and independent
of technology or implementation way. They are written as “an abstract
dialog representing user intentions and system responsibilities, and they
are typically small and focused on a highly specific user goal, yielding a
finegrained model of user activity” [20]. After the definition of the essential
use cases there are two alternatives to proceed in the system’s architecture
definition phase:
56 Mechatronics and Robotics

1. Use cases are decomposed in sub-use cases.

2. Responsibilities of the system are identified.
In the first case the decomposition of use cases to sub-use cases allows:
a) the reuse of existing components on the basis of their requirement
specifications that should have been defined in terms of use cases, and b)
modularity and reuse in requirements specification artifacts. In the second
case, activity diagrams are defined for each use case in order to identify
the activities/responsibilities of the actors and the ones that are required by
the system in the context of the specific use case. After this step the list of
abstract activities (functions) that have to be performed by the system is
available. In other words the responsibilities of the system in the context of
the specific use case are defined.

Figure 7. The mechatronic system V-Model (MTS V-Model).

The term responsibility is used to emphasize the fact that only the
abstract definition of the activity/function is provided at this time and not
its implementation.
As a next step, for both alternatives, a system composition should be
proposed to satisfy either the system use-cases or the system responsibilities.
Some use cases or system responsibilities may be directly supported by
existing mechatronic components. However, it is common that for a required
system level use case or responsibility, a collaboration of MTCs has to
be defined in order to achieve it. The system level use cases and sub-use
cases or the system responsibilities/functions correspondingly are mapped
to system’s components. An analogous mapping applies also for the non-
functional requirements.
The definition of system’s structure in terms of MTCs is a design process
and results in the selection of system components and the definition of their
 The 3+1 SysML View-Model in Model Integrated Mechatronics 57

collaboration. Furthermore each MTC has assigned responsibilities that are

handled in the subsequent phases as its required responsibilities. The result
in both cases is a system architecture that is comprised of:
• Class or component diagrams to specify the structure of the
system;
• Sequence or activity diagrams (or even state charts) to specify
the components’ collaborations to provide the higher layer
functionality.
Domain analysis is used to capture the domain key concepts and provide
the information required to create the first architectural model of the system.
SysML diagrams are used to specify the proposed architecture. Block
definition diagrams (bdd) are used to capture the structure of the system
and internal block diagrams (ibd) are used to capture the components’
interactions, all expressed using the SysML4MIM profile.
During the architecture definition the developer has to assign the system
required responsibilities to the system’s components. This assignment
results to an architectural diagram that represents the system components,
their responsibilities and the components interactions.
The allocation relationship of SysML provides an effective means
to capture this assignment and allow the navigation between the system
models by establishing cross-cutting relationships among them. There are
two alternatives to proceed in the definition of the architecture:
1. The bottom-up approach (synthesis).
2. The top-down (decomposition).
According to the bottom-up approach, for every system-level
responsibility a set of commercial off-the-shelf (COTS) MTCs is selected.
We assume that each COTS MTC has its own provided functions that are well
defined by the developer of the MTC [7]. These provided functions are part
of the MTC package that specifies the real world MTC. QoS characteristics
are also included in the MTC package and can be used to examine if the QoS
aspects of the proposed collaboration scheme satisfy the required system-
level QoS aspects for the specific responsibility. If the QoS characteristics
of this specific collaboration meet the QoS requirements of the system-level
required responsibility, the design is accepted. Either wise corrective actions
should be proposed and analyzed. Corrective actions may include: a) re-
engineering of the collaboration scenario, b) the use of components with
better QoS characteristics than the ones used in the previous design, or c) a
58 Mechatronics and Robotics

combination of the above. It is assumed that the MTC developer has already
performed a QoS analysis for the MTC. All this information comprises the
offered QoS characteristics of the MTC [4]. It should also be stated that the
MTC developer does not know during the MTC’s development time all the
systems that this MTC will be used in the future.
According to the top-down approach, for every system-level
responsibility that has not been assigned to a single MTC, a set of abstract
MTCs is defined and the required MTC responsibilities are specified
along with the required collaboration scheme. Required system-level QoS
characteristics are decomposed to derive the component-level required QoS
characteristics.
This process results to the definition of the required QoS characteristics
at the level of constituent MTCs. The process of deriving MTC-level QoS
characteristics from system-level ones is a complex process and has to be
defined. At this time the engineer has well defined required specifications
(functional and nonfunctional, including QoS characteristics) for every
abstract MTC. Using these required QoS characteristics the engineer is able
to select from the market or his components repository the ones that their
offered QoS characteristics meet the required ones [8]. If such MTC’s do not
exist they have to be further analyzed in order to be developed.
Advantages and disadvantages for the above approaches that result in
the definition of the architecture of the MTS of its composite MTCs may
be identified but it is expected that in the real MTS development process
a combination of both approaches will be used resulting to a hybrid more
efficient approach.
The above process, bottom-up or top-down, is again applied to every
composite MTC that has to be developed. It is applied iteratively down to
the primitive-MTC level; the identification of primitive MTCs signals the
end of this iteration. For each composite MTC the system modeling process
as defined by the left-hand side part of the V-Model is followed. Analysis
is applied and its architecture is defined in terms of constituent components
(composite and/or primitive). Sequence diagrams are defined to realize use
cases of the MTC and identify the activities that are involved in the specific
use case.
This is not the case for primitive MTCs that have to bypass the system
process and follow a synergistic integration of the three constituent parts,
i.e. mechanic, electronic and software (MTC synthesis) as shown in the
bottom of the V-model in Figure 7.
 The 3+1 SysML View-Model in Model Integrated Mechatronics 59

For each primitive MTC, verification follows its integration as shown

in the right-hand side of the MTS V-Model. Each composite MTC is
integrated according to its MTC architecture and then it is verified against
its requirements. After the integration and verification of the MTCs of the
system, the MTS integration test is performed and the MTS is verified
against its requirements.
It should be noted that the system analysis phase is followed by a system
architecture design phase as shown in the proposed MTS V-model. This is
also the case for the V-Model in software engineering. After this point the
proposed V-model is completely differentiated from the traditional software
engineering V-model.
After the system architectural design, repetitions of analysis followed
by architecture design for every composite MTC are applied following the
system modeling process up to the primitive MTC level.
This is the point where the system development process is terminated
and the synergistic integration of constituent parts of the primitive MTC
is performed working independently but in a synergistic way in the three
disciplines. For every primitive MTCs that has complex software constituent
part, a software V-model can be applied for its development, as shown in the
bottom of the MTS V-Model.
Figure 8 presents two real world MTCs and the Parallel Kinematic
Machine evolium MTS, that were developed based on the basic principles
of the MIM architecture by a high-tech Italian company. Each axis of the
Parallel Kinematic Machine has its own intelligence, so there is no need of
an external entity to control the motion trajectory.

FUTURE DEVELOPMENTS AND RESEARCH CHAL-

LENGES
Mechatronic systems development is a very complicated process imposing
many challenges. In this section we refer to the ones that are of higher
priority considering the 3+1 SysML view-model. The identification of the
mechanical discipline information that has to be captured in the system level
models is one of the challenges.
60 Mechatronics and Robotics

Figure 8. Real-world MTS and MTCs

There are two possible approaches: a) exploit SysML constructs to
represent as much of the mechanical discipline information, including
component interfaces and behavior; and b) extend SysML constructs with
new ones with the objective of creating a complete SysML model of the
mechanical component. In the first case specific tools will be used for
further refinement of the mechanical models and their subsequent execution
and analysis. In the latter, the SysML models have to be automatically
transformed to models of the specific tools for execution and model analysis.
The integration of SysML with the Modelica language (www.modelica.org/)
is towards this direction. This challenge is greatly related to the one that
concerns the model execution, analysis and assessment of models on the
MTS-view level.
Fully automated generation of the three views from the MTS view, as
well as automatic update of the MTS-view with changes in the discipline
views are important challenges that have to be addressed to improve the
effectiveness of the process. This also imposes the challenge of integration
of existing mechanical and electronic domain design tools.
The identification and definition of reusable MTCs is another major
challenge in mechatronic systems. The Workpartner [21], a mobile service
robot, is planned to be used in the context of a TEKES (Finnish Funding
Agency for Technology and Innovation) funded project as a case study for
the application of MIM, but also as a case study for the identification of
reusable MTCs.
Since many of the MTSs are from the safety critical domain, the
integration of the MTS V-model development process and the 3+1 SysML
view-model with safety engineering is another major challenge for the MIM
paradigm to be effectively exploited in safety critical mechatronic systems.
 The 3+1 SysML View-Model in Model Integrated Mechatronics 61

CONCLUDING REMARKS
The traditional approach in the development of mechatronic systems is unable
to address the needs of today’s complex mechatronic systems. An integrated
framework for the construction of mechatronic systems is missing. The work
presented in this paper attempts to contribute to this direction by: a) using
SysML to define the artifacts of the MIM paradigm; b) proposing the 3+1
SysML view-model imposed by the MIM architecture; and c) extending the
well accepted and widely used in the software domain V-model to address
the demands of the mechatronic system development process. However,
the challenges for a fully automated MTS development process crosses the
boundaries of the three disciplines of mechatronic systems and impose a
joint effort and collaboration between computer science, electronics and
mechanics. The current status of discipline isolation imposed in many cases
by the existing structure of engineering degree programs makes the task
even more complicated.

ACKNOWLEDGMENTS
Part of this work has been funded by TIKOSU, a project belonging to
the Digital Product Process -program of the Finnish Funding Agency for
Technology and Innovation (TEKES). The author wishes to thank the
partners and especially Jarmo Alanen, Kari Koskinen, and Seppo Sierla for
fruitful comments on these ideas. Thanks are also due to Piero Larizza, Eric
Coatanea and Jussi Suomela for discussions on these concepts.
62 Mechatronics and Robotics

REFERENCES
1. G. Rzevski, “On conceptual design of intelligent mecha- tronic
systems,” Mechatronics, 2003.
2. W. Schafer and H. Wehrheim, “The challenges of build- ing advanced
mechatronic systems,” Future of Software Engineering, International
Conference on Software Engi- neering, IEEE Computer Society, 2007.
3. Philipp Limbourg, “Dependability modelling under un- certainty: An
imprecise probabilistic approach,” Springer, 2008.
4. K. Thramboulidis, “Model integrated mechatronics: To- wards a new
paradigm in the development of manufac- turing systems,” IEEE
Transactions on Industrial Infor- matics, Vol. 1, No. 1, February 2005.
5. OMG, “OMG Systems Modeling Language (OMG SysML™),” V1.0,
September 2007.
6. GD250, “Lifecycle process model ‘V-Model’,” Available online: http://
www.informatik.uni-bremen.de/gd-pa/vmodel/ vm1. htm#application
7. K. Thramboulidis, G. Doukas, and G. Koumoutsos, “A SOA-
based embedded systems development environment for industrial
automation,” EURASIP Journal on Embed- ded Systems, Article ID
312671, pp. 15, 2008.
8. K. Thramboulidis, “Challenges in the development of mechatronic
systems: The mechatronic component,” 13th IEEE International
Conference on Emerging Technolo- gies and Factory Automation
(ETFA), Hamburg, Ger- many, September 2008.
9. OMG, “Unified modeling language: Superstructure,” version 2.1.1,
formal/2007-02-03.
10. M. Habib, “Mechatronics,” IEEE Industrial Electronics Magazine,
Vol. 1, No. 2, Summer 2007.
11. S. Burmester, H. Giese, and M. Tichy, “Model-driven development of
reconfigurable mechatronic systems with mechatronic ‘UML’ in model
driven architecture,” Springer Berlin/Heidelberg, Vol. 3599, 2005.
12. K. Knorr, A. Lapp, P. Torre Flores, J. Schirmer, D. Kraft
13. J. Petersen, M. Bourhaleb, and T. Bertram “A process model for
distributed development of networked mecha- tronic components
in motor vehicles,” Proceedings of the IEEE Joint International
Conference on Requirements Engineering (RE’02), 2002
14. R. Nordmann, “Use of mechatronic components in rotat- ing machinery,”
 The 3+1 SysML View-Model in Model Integrated Mechatronics 63

Book: Vibration Problems ICOVP 2005, Springer Netherlands, Vol.

111, January 20, 2007.
15. J. El-khoury, O. Redell, and M. Torngren, “A tool integra- tion platform
for multi-disciplinary development,” Pro- ceedings of the 2005 31st
EUROMICRO Conference on Software Engineering and Advanced
Applications, 2005.
16. G. Doukas and K. Thramboulidis, “A real-time Linux based framework
for model-driven engineering in control and automation,” IEEE
Transactions on Industrial Elec- tronics (forthcoming).
17. G. Pahl, W. Beitz, J. Feldhusen, and K. H. Grote, “Engi- neering design:
A systematic approach,” Third Edition, Springer-Verlag London, 2007.
18. R. Eckert, W. Mansel, and G. Specht, “Model transfer among CASE
tools in systems engineering,” Systems En- gineering, Vol. 8, No 1, pp.
41–50, March 2005.
19. P. Kruchten, “The 4+1 view model of architecture,” IEEE Software,
Vol. 12, No. 6, pp. 42–50, November 1995.
20. OMG, “SysML and AP233 mapping activity,” OMG SysML portal,
http://www.omgwiki.org/OMGSysML/ doku. php?id=sysml-
ap233:mapping_between_sysml_and_ap233.
21. L. Constantine, “Activity modeling: Toward a pragmatic integration of
activity theory with usage-centered design,” Technical Paper, Available
on-line: http://www. foru se. com/ articles/activitymodeling.pdf
22. “The workpartner mobile service robot,” http://automa- tion.tkk.fi/
 CHAPTER

4
Design and Development of
Mechatronic Application in Agricultural
Irrigation Device

Mohd Hudzari Razali, Syazili Roslan, Abdul Ssomad Mohd Abd

Halim, Hayan Basit
Faculty of Bioresources & Food Industry, Universiti Sultan Zainal Abidin,
Tembila Campus, Besut, Terengganu Darul Iman, Malaysia

ABSTRACT
In order to help the small-scale farmer, an automatic irrigation control
system was proposed. This system will provide an irrigation system that
will ease the burden of the citizen to take care of the plant. This system will
run automatically by referring to the time set by the user. As the name itself

Citation:- Razali, M. , Roslan, S. , Halim, A. and Basit, H. (2016), “Design and Devel-
opment of Mechatronic Application in Agricultural Irrigation Device”. World Journal
of Engineering and Technology, 4, 450-459. doi: 10.4236/wjet.2016.43045.
Copyright: © 2016 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
66 Mechatronics and Robotics

is a water control system, this system will only start irrigating when the time
set triggered the water control level for the plant to grow healthily. It will
automatically stop when the timer is off (1 hour). The brain of the system is
the PLC (Programmable Logic Controller). This is the place where all the
activities are done. The irrigation will be provided by a pump that is also
connected to the microcontroller. The pump will be activated until the timer
has reached its time set. This system will continue running until the user
presses the OFF button.
Keywords: Irrigation System, Water Control System, PLC
(Programmable Logic Controller)

INTRODUCTION
As technology undergoes rapid advancement, complicated task such as
control system is accomplished with a highly automated control system.
Example of the system may be in the form of Programmable Logic Controller
(PLC) and possibly a host computer and accompanied with signal interfacing
to the field devices such as opera- tor panel, motors, sensors, switches, and
solenoid valves. Network communication is capable of providing a large
scale implementation and process coordination besides enabling greater
versatility in understanding distributed control system. Every single integral
component module in a control system plays a major role regardless of size.
Generally, PLC can be explained as a digital electronic device that uses a
programmable memory to store direction and to accomplish functions such
as counting, logic sequencing, arithmetic and timing in order to control
machine and processes. The term logic was used because the programming
is basically concerned with implementing logic and switching operations
[1]. Among the most important implementation for PLCs in Southeast Asia
is the oil & gas industry. For example, Malaysia is world famous for its
palm oil, accounting for a large portion of total global production. Here,
many mid and high-end PLCs are used in the production of palm oil, fed
mostly by government funding. Meanwhile, large PLCs are used in the oil
and natural gas industry in Indonesia and Malaysia. Another important for
PLCs implementations in Southeast Asia is the automotive manufacturing
industry. Thailand is world-famous for its production of automotive
electronics; with industry requirements for PLCs multiply after the country
was hit by critical flooding in 2011. Malaysia has a big automotive industry
markets in the area among the locals, where a very large population will
soon requesting more favorable modes of transport. A third application is
 Design and Development of Mechatronic Application in ... 67

in agriculture industry. Thailand and Indonesia are the main contributors to

the expansion of this sector, and advanced PLCs will be required to produce
innovative food processing and packaging ingredient, aimed at local use,
consumption and exports alike. The use of PLCs and other similar devices
in the agricultural mechanization industry is growing widespread. Example
applications include food processing, building environmental control, grain
drying, aquaculture production, and tractor and machinery systems. As we
know that Malaysia government policy in terms of making industrializing in
agriculture activity sector, so in this project, we introduce the application of
PLC that can be used in agriculture activity. It includes how to set up the PLC
programmer, integration between the software and hardware, construction
of the circuit, and run monitoring.

Problem Statement
The world’s population is estimated to reach 9.2 billion by 2050.The UN
Food and Agriculture Organization (FAO) has projected that farmers will
need to increase their production by 70% more food than in 2006 to meet
this demand.
Inability to do so will causing in food scarcity and poorer healthcare
in developing countries, with damaging impact for development and
the potential for conflict within and between nations. While agriculture
productivity has been increasing, production capacity is stagnant, and food
security remains a major issue in many nations due to hiking prices as well
as availability. Continuing increase costs of input of agriculture, commodity
speculation and competition with other uses for crops. In poorer nations, this
causing the number of people considered to be ‘food insecure’ increasing
because they are not able to afford enough food even if it were available
to buy. Both productivity and absolute production need to increase if this
issue is to be addressed. In Malaysia, small farms remain at the Centre of
agriculture and rural development. However, one of the main causes for the
low agricultural productivity in most developing countries in the region is
the lack of appropriate machineries that cater to and suit the requirements
of small-scale farms. For this reason, many small farms are deemed as
unproductive and inefficient. Farm mechanization plays a significant role
in every nation’s economy. However, it is often misconstrued to mean
modernization, beneficial only to industrialized countries with highly
mechanized agriculture. Developing countries often have to rely on a
variety of imported farm machines, which are seldom appropriate for small
farms [2]. One of the issues that need to be address is the lack of usage of
68 Mechatronics and Robotics

automation such as PLC especially in Malaysia. Many small, medium, and

even large-scale farm having less production capacity that can be prevented
if the owner using an automation for their respective farm. This Project
will give me an overview of how PLC works and insight so more advance
automation can be carried out in the future.

Significant of Study
Automation and robotic can play a large role in society to meet the
agricultural production needs. For decades automation and robots have
played a core role in multiplying the efficiency and lowering the cost of
industrial production and products. In the past several years, a similar shift
has started to take place in agriculture industry, with GPS- and vision-based
such as self-guided tractors and harvesters already being available in the
market commercially.
More recently, developers have started to test with autonomous systems
that combine operations such as thinning, pruning and harvesting, as well
as sowing, spraying, mowing and weed removal. In the fruit production
industry, for example, robotic platforms rode by workers have shown to be
doubling the efficiency compared to workers using ladders. Advancement
in sensors technology and control systems allow for optimum resource and
integrated pest and disease management.
This will be a revolution in the way that food nowadays is grown,
tended, and harvested. The level of input of engineering technologies into
agriculture is generally still low [3]. Automation component such as PLC is
an important module in agricultural automation.
It can be found used in irrigation systems, food processing, building
environmental control, grain drying, aquaculture production, and tractor and
machinery systems. Using automation such as PLC enables us to increase the
number of production by reducing error such as human and environmental
factor (damaged produce).

Objective of Study
• To develop the hardware and software by using PLC
(Programmable Logic Controller) as a main controller.
• To interface PLC module with input and output component.
• To design prototype system for motion control.
 Design and Development of Mechatronic Application in ... 69

LITERATURE REVIEW

Programming Logic Control

Programmable Logic Controller is often found in the field close to the
processing unit. Tiny and operator interface of a PLC may be simple as
button switch. In later generation PLC manufacturer have added analogue
to digital conversion proficiency and provided enough logic to adjust simple
control loops [4]. Nowadays there are at least two recognized PLC sizes:
Small sized PLC, which is basically a relay replacement and provides a
reliable control to stand-alone section of process of PLC. Medium sized
PLC that can performs all the relay replacement tasks expected of it, and
also performs functions like timing, counting, and complex mathematical
applications.
There are five core components in a PLC system:
1) The PLC processor or controller.
2) I/O (Input/Output) modules.
3) Chassis or backplane.
4) Power supply.
5) Programming software that runs in a PC.
Several advantages of PLC: Increased reliability; once a selected
program has been drafted and installed, other PLCs can download it without
any difficulties. Since almost all the logic pattern is contained in the PLC’s
internal memory, there is lower chance of it making a logic wiring error.
More Flexibility; It is easier to design and alter a program in a PLC than
to wire and rewired a circuit board. Equipment manufacturers can provide
system upgrade by simply broadcast out a new program.
Lower Cost; PLC were originally made and designed to replace relay
control logic, and the cost reduction have been so exceptional that relay
control is becoming non-existent and obsolete except used for power
application.Communications Capability; PLC can interact with other
controllers or computer equipment to perform such functions as monitoring
devices, process parameters, supervisory control, data gathering, and also
download and upload of programs.
Faster Response Time; PLCs are designed for real-time and high-speed
applications. The PLC operates in real time; causing event taking place in
the field will result in the carried out of an operation or output.
70 Mechatronics and Robotics

Easier to troubleshoot; PLCs have resident diagnostics and override

functions that allow users to easily trace and correct software and hardware
problems [5].
Many types of PLC are available such as Omron, Mitsubishi, Siemen,
Nais and many more. In this project, I am using Omron PLC model CP1E-
N30DR-D.This PLC is suitable to control the switching of valve and speed
of motor pump (Figure 1).

HARDWARE DESIGN OF PLC

Input Device
Intelligence of an automated system is greatly depending on the ability of
a PLC to read in the signal from various types of automatic sensing and
manually input field devices. Toggle switches, push buttons, and keypad,
which from the basic man machine interface, are types of manual input
device. On the other hand, for detection of workplace, monitoring of moving
mechanism, checking on pressure and or liquid level and many others, the

Figure 1. Omron PLC.

PLC will have to tap the signal from the specific automatic sensing
device like proximity switch, limit switch, photoelectric sensor, and level
sensor and so on (Figure 2). Types of input signal to be connected to the PLC
would be analogue or ON/OFF logic. These input signals are interfaced to
PLC through various types of PLC input module.

Output Devices
An automatic system is incomplete and the PLC system is virtually paralyzed
without means of interface to the field output devices. Some of the most
commonly controlled device is relay indicator, motors, solenoids, buzzers
 Design and Development of Mechatronic Application in ... 71

and so on (Figure 3). Through activation of motors and solenoids the PLC
can control from simple pick and place system to a much complex servo
positioning system. These type of output devices are the mechanism of an
automated system and so its direct effect on the system performance.

Software Design of PLC Programming

Ladder Logic
To ease the use of PLC’s we programmed it using ladder logic format. Ladder
logic is a visual representation of a set of inputs and outputs. Ladder Logic
format resembles familiar hardware systems and it doesn’t require extra
training for technicians, engineers and also those who without backgrounds
in these hardware systems. The name is derived from the fact that the diagram
upon completion resembles a ladder. A Ladder Logic diagram consists of a
vertical line on the left hand side, known as the hot rail, and a vertical line
on the right hand side, known as the neutral rail [6] . They are connected
by lines, known as rungs; with several different symbols each represents an
input or output. The logic is following the path and determining if the rung
is true. For the rung to be true it is necessary to be able to flow across closed
contacts to the opposite rail. If the rung is true the output is then considered
true. Figure 4 is sample ladder logic diagram.

Relay
Relay is one of the switch types which are electrically operated. Most
of the relay uses an electromagnet to operate a switching mechanism
mechanically. It is used to control a circuit by a low-power signal (with
complete electrical isolation between control and controlled circuits), where
several circuits must be controlled with one signal. A contactor is one of
the relay types that can handle a high power that required controlling an
electric motor electrical circuit from overload or faults; in modern electric
power systems these functions are performed by digital instruments still
called “protective relays”. In general, components in relay are inductor coil,
a spring (not shown in figure), swing terminal, and two high power contacts
named as normally closed (NC) and normally open (NO). Relay uses an
Electromagnet to move swing terminal between two contacts (NO and NC).
When there is no power applied to the inductor coil (Relay is OFF), the
spring holds the swing terminal is attached to NC contact. Figure 5 shows
the pin diagram for relay:
72 Mechatronics and Robotics

The relay will start operating when the required power is applied to the
inductor coil will generates a magnetic field which will move the swing
terminal from normally close contact to normally open contact. When the
power is OFF, the spring will move back to normally close contact.

Water Pump Motor

A stepper motor is electromechanical actuators, changing electrical pulses
into mechanical actions. When the signals are applied to the system with the
right sequence pulses, the shaft of a stepper motor rotates in distinct step.

Figure 2. Type of input.

Figure 3. Type of output.

Figure 4. Ladder logic programming.

 Design and Development of Mechatronic Application in ... 73

The stepper motors revolution has various direct relationships to applied

input pulses. The speed of the shafts rotation is connected to the frequency
of the input pulses and the length of rotation is associated to the amount of
input pulses applied. Meanwhile the direction of the shaft rotation is depends
on sequences of the input pulses. Not like other type of motor, the stepper
motor has no contacts or brushes. It is a synchronous motor in order to
rotate armature magnet through the magnetic field switching. The essential
function of a stepper motor is to translate switching excitation changes into
precisely 15 defined increments of rotor position. Generally the stepper
motor capable works as an electric motor when the drive running without
commutated. The rotor of the stepper motor can be permanent magnet or
variable reluctance motor which has a toothed block of some magnetically
soft material.

Figure 5. Relay general pin diagram.

Typically, all the windings in the stepper motor are part of the stator. A
stepping motor control system consists of three basic elements; controller,
driver and motor [7].

MATERIALS AND METHODOLOGY

Materials
1) PLC training kit
• ・ PLC microcontroller (CP1E-N30DR-D)
• ・ CX one PLC programmer
• ・ Stepper motor
• ・ Touch-screen teach pendant
74 Mechatronics and Robotics

2) S8VK-G01505 Omron Automation

3) Omron MY2N 24VDC Relay
4) Wire cutter
5) Power source
6) Connecting Wire
7) Water pump motor
8) Push switch

Methodology
Figure 6 showed the methodology or work flow of the project that has been
used as the guideline in order to do the project. In started with investigate
the topic and objectives with supervisor. After doing literature review,
equipment that needed was investigated like relay, water pump motor, PLC
and AC/DC converter.

Pilot Experiment
Pilot experiment was done first so we will learn first on how to write a
ladder logic system and install it into PLC.
This experiment was done using PLC training kit as a prototype that
already had been assembled so it can be used as a learning tool before actual
experiment.
The programming that we are going to use is CX one programmer that
can be used to write ladder logic system. First, we need to connect PLC to a
computer so we will be able to install the programing into our PLC. We put
the logic:
When push button is ON, turn the stepper motor ON. After that, we
write this logic into a ladder logic system (CX one).
Then we load this program into the PLC. We connect the sensor input to
the PLC, the one we specified in our program (push button).
Lastly, we connect the PLC External Output Terminal (specified by our
program) to the stepper motor. Now, we execute the logic program on the
PLC.
 Design and Development of Mechatronic Application in ... 75

Figure 6. Flowchart of construction of model.

Construction of Model
Firstly, we need to connect our PLC to a computer using an adapter. Then
we run the program (CX One programmer shown in Figure 7) and write the
intended logic sequence for our experiment that is:
1) When switch is ON, turn the Timer A for 100 seconds.
2) During Timer A ON, the water pump motor is ON until Timer A
is OFF.
3) When Timer A is OFF, turn ON Timer B (100 seconds).
4) When Timer B is OFF, switch ON the water pump motor and
Timer C (100 seconds).
5) When Timer c is OFF, turn ON Timer D (100 seconds).
6) When Timer D is OFF, the counter resets the main switch and the
cycle continue.
Next, we load the program that has been written into the PLC. After
that, connect the push switch to the PLC by using connecting wire, the one
we specified in our programming earlier. Then, connect the PLC External
Output Terminal to the water pump motor (output). Now, we execute the
logic program on the PLC. After the model has been completed, we do
several test run for our model to make sure the PLC programming work as it
intended. Any failure will be observed and documented so we will be able to
do troubleshooting for our model. Figure 8 shows the model of construction
software for irrigation control.
76 Mechatronics and Robotics

RESULT

Software Development PLC Programming

Programming instructions:

Figure 7. CX one programmer.

Figure 8. Construction software model for irrigation control.

1) When switch is ON, turn the Timer A for 1 hour.
2) During Timer A ON, the water pump motor is ON until Timer A
is OFF.
3) When Timer A is OFF, turn ON Timer B (8 hours).
4) When Timer B is OFF, switch ON the water pump motor and
Timer C (1 hour).
5) When Timer C is OFF, turn ON Timer D (14 hours).
6) When Timer D is OFF, the counter RESET the main switch and
the cycle continue.
 Design and Development of Mechatronic Application in ... 77

Hardware Development
Figure 9 shows an overview of work table in Farm Mechanization laboratory
in Universiti Sultan Zainal Abidin campus. Figure 10 shows constructed
model for irrigation control.

Simulation
Figure 11 shows an algorithm for automatic irrigation control.

DISCUSSION

Introduction
Both hardware and software part are developed and built successfully. The
system is complete when both of them are connected together. This chapter
explains all the results related to both part.

Hardware
The microcontroller is connected directly to AC to DC adaptor that is
functioning as power supply to switch on the control unit. The moisture
sensor will start measuring and the data obtain will be sent to control unit
in voltage form that will be processed by PIC microcontroller. Since the
microcontroller has the capability of 10-bit analog to digital converter, it
will convert and manipulate the data obtained to get the actual value of
humidity. After that, these values are displayed on the LCD. The control unit
also will instruct the pump to start working.

Figure 9. Overview of work table in Farm Mechanization Lab.

78 Mechatronics and Robotics

Figure 10. Constructed model for irrigation control.

Figure 11. Algorithm for automatic irrigation control.

CONCLUSION
The project involved the design activity on both software and hardware
development of PLC in agriculture mini hydroponic system. The design on
circuitry system is involved for connecting between input and output system
including push button device, microcontroller device, 240 volts power
supply and water motor pump. The software development is created from
basic concept on timer and counter application. By using this technique, this
project was successful to automate basic irrigation system for hydroponic.
The system will automatically circulate the fertilizer mix water throughout
the system on selected period of morning and afternoon daily with less
intervention of the user.
 Design and Development of Mechatronic Application in ... 79

REFERENCES
1. Selvaraj, K. (2010) Development of a “Programmable Logic Controller
Circuitry” for Optimal Power Distribution in a Manufacturing
Industries. Journal of Computer Science, 6, 250-252.http://dx.doi.
org/10.3844/jcssp.2010.250.252
2. (2005) Small Farm Mechanization Systems Development, Adoption
and Utilization. Report, FFTC Annual Parts, Sri Lanka, 13-17.
3. Kamaruddin, R., Rukunuddin, I.H. and Seng, O.H. (2007) Research and
Development of Agricultural Engineering in Malaysia. Paper, Country
Asian, United Nations Centre, Pacific Engineering, Agricultural, 1-11.
4. Halim, M.Z.B. (2007) The Temperature Control System Using PLC.
5. Ismail, M.H. (2008) Rain Water PLC Based Detector and Valve
Switcher.
6. Schumann, D., Fietsam, J., Rochel, E., Dixon, K., Imbayan, M. and
Ibayan, M. (n.d.) Literature Review for the Design of a SCADA
System.
7. Shah, C. (2004) Sensor Less Control of Stepper Motor Using Kalman
Filter. Department of Electrical and Computer Engineering, Cleveland
State University, Master of Science in Electrical Engineering.
 CHAPTER

5
Design and Development of a
Competitive Low-Cost Robot Arm with
Four Degrees of Freedom

Ashraf Elfasakhany1,2, Eduardo Yanez2, Karen Baylon2, Ricardo

Salgado2
1
Department of Mechanical Engineering, Faculty of Engineering, Taif
University, Al-Haweiah, Saudi Arabia
2
Tecnológico de Monterrey, Campus Ciudad Juárez, Ciudad Juarez, Mexico

ABSTRACT
The main focus of this work was to design, develop and implementation of
competitively robot arm with enhanced control and stumpy cost. The robot
arm was designed with four degrees of freedom and talented to accomplish

Citation:- A. Elfasakhany, E. Yanez, K. Baylon and R. Salgado, “Design and Develop-

ment of a Competitive Low-Cost Robot Arm with Four Degrees of Freedom,” Modern
Mechanical Engineering, Vol. 1 No. 2, 2011, pp. 47-55. doi: 10.4236/mme.2011.12007.
Copyright: © 2011 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
82 Mechatronics and Robotics

accurately simple tasks, such as light material handling, which will be

integrated into a mobile platform that serves as an assistant for industrial
workforce. The robot arm is equipped with several servo motors which do
links between arms and perform arm movements. The servo motors include
encoder so that no controller was implemented. To control the robot we used
Labview, which performs inverse kinematic calculations and communicates
the proper angles serially to a microcontroller that drives the servo motors
with the capability of modifying position, speed and acceleration. Testing
and validation of the robot arm was carried out and results shows that it
work properly.
Keywords: Robot Arm, Low-Cost, Design, Validation, Four Degrees of
Freedom, Servo Motors, Arduino Robot Control, Labview Robot Control

INTRODUCTION
The term robotics is practically defined as the study, design and use of robot
systems for manufacturing [1]. Robots are generally used to perform unsafe,
hazardous, highly repetitive, and unpleasant tasks. They have many different
functions such as material handling, assembly, arc welding, resistance
welding, machine tool load and unload functions, painting, spraying, etc.
There are mainly two different kinds of robots: a service robot and
an industrial robotic. Service robot is a robot that operates semi or fully
autonomously to perform services useful to the well-being of humans and
equipment, excluding manufacturing operations [2]. Industrial robot, on
the other hand, is officially defined by ISO as an automatically controlled
and multipurpose manipulator programmable in three or more axis [1].
Industrial robots are designed to move material, parts, tools, or specialized
devices through variable programmed motions to perform a variety of tasks.
An industrial robot system includes not only industrial robots but also any
devices and/or sensors required for the robot to perform its tasks as well as
sequencing or monitoring communication interfaces.
In 2007 the world market grew by 3% with approximately 114,000 new
installed industrial robots. At the end of 2007 there were around one million
industrial robots in use, compared with an estimated 50,000 service robots
for industrial use [3].
Due to increase using of industrial robot arms, an evolution to that topic
began trying to imitate human movements in a detail mode. For example a
group of students in Korea made a design of innovations that robotic arm
 Design and Development of a Competitive Low-Cost Robot Arm with ... 83

take account of dancing hand, weight lifting, Chinese calligraphy writing

and color classification [4]. Another group of engineers at USA develop eight
degrees of freedom robot arm. This robot is able to grasp many objects with
a lot of shapes from a pen to a ball and simulating also the hand of human
being [5]. In space, the Space Shuttle Remote Manipulator System, known
as SSRMS or Canadarm, and its successor is example of multi degree of
freedom robot arms that have been used to perform a variety of tasks such
as inspections of the space shuttle using a specially deployed boom with
cameras and sensors attached at the end effector and satellite deployment
and retrieval manoeuvres from the cargo bay of the space shuttle [6].
In Mexico, Scientists are on track to design and develop many robot
arms, and the Mexican government estimates that in Mexico there are about
11,000 robotic arms used in different industrial applications. However, the
experts think that the apogee of the robot arms is not only of higher quality,
but also accurately, repeatability, and stumpy cost.
Most robots are set up for an operation by the teachand-repeat technique.
In this mode, a trained operator (programmer) typically uses a portable control
device (a teach pendant) to teach a robot its task manually. Robot speeds
during these programming sessions are slow.The present work is part of a two-
phase project, which requires a mobile robot to be able to transport the tools
from the storage room to the industrial cell. In this phase in the project, which
carried out at Monterrey University of Technology, Mexico, the main focus
was to design, development and implementation of an industrial robotic arm
with stumpy cost, accurate and superior control. This robot arm was designed
with four degrees of freedom and talented to accomplish simple tasks, such as
light material handling, which will be integrated into a mobile platform that
serves as an assistant for industrial workforce.

MECHANICAL DESIGN
The mechanical design of the robot arm is based on a robot manipulator
with similar functions to a human arm [6-8]. The links of such a manipulator
are connected by joints allowing rotational motion and the links of the
manipulator is considered to form a kinematic chain. The business end of
the kinematic chain of the manipulator is called the end effector or end-of-
arm-tooling and it is analogous to the human hand. Figure 1 shows the Free
Body Diagram for mechanical design of the robotic arm. As shown, the
end effector is not included in the design because a commercially available
gripper is used.
84 Mechatronics and Robotics

Figure 1. Free body diagram of the robot arm.

This is because that the end effector is one of the most complex parts of
the system and, in turn, it is much easier and economical to use a commercial
one than build it.
Figure 2 shows the work region of the robotic arm. This is the typical
workspace of a robot arm with four degree of freedom (4 DOF). The
mechanical design was limited to 4 DOF mainly because that such a design
allows most of the necessary movements and keeps the costs and the
complexity of the robot competitively. Accordingly, rotational motion of the
joints is restricted where rotation is done around two axis in the shoulder and
around only one in the elbow and the wrist, see Figure 1.
The robot arm joints are typically actuated by electrical motors. The
servo motors were chosen, since they include encoders which automatically
provide feedback to the motors and adjust the position accordingly. However,
the disadvantage of these motors is that rotation range is less than 180˚ span,
which greatly decreases the region reached by the arm and the possible
positions [9]. The qualifications of servo motors were selected based on the
maximum torque required by the structure and possible loads. In the current
study, the material used for the structure was acrylic.
Figure 3 shows the force diagram used for load calculations. The
calculations were carried out only for the joints that have the largest loads,
since the other joints would have the same motor, i.e. the motor can move
the links without problems. The calculations considered the weight of the
motors, about 50 grams, except for the weight of motor at joint B, since it is
carried out by link BA. Figure 4 shows the force diagram on link CB, which
contains the joints (B and C) with the highest load (carry the links DC and
ED) and the calculations are carried out as follows.
 Design and Development of a Competitive Low-Cost Robot Arm with ... 85

Figure 2. Work region of the robotic arm.

Figure 3. Force diagram of robot arm.

Figure 4. Force diagram of link CB.

The values used for the torque calculations:
Wd = 0.011 kg (weight of link DE)
Wc = 0.030 kg (weight of link CD)
Wb = 0.030 kg (weight of link CB)
L = 1 kg (load)
86 Mechatronics and Robotics

Cm = Dm = 0.050 kg (weight of motor)

LBC = 0.14 m (length of link BC)
LCD = 0.14 m (length of link CD)
LDE = 0.05 m (length of link DE)
Performing the sum of forces in the Y axis, using the loads as shown
in Figure 4, and solving for CY and CB, see Equations (1)-(4). Similarly,
performing the sum of moments around point C, Equation (5), and point
B, Equation (6), to obtain the torque in C and B, Equations (7) and (8),
respectively.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)
The servo motor that was selected, based on the calculations, is the
Hextronik HX12K, which has a torque of 280 oz/in. This motor was
recommended because it is much cheaper than any other motor with same
specifications. Since we need more torque at joint B, see Equation (8),
we used two motors at point B to comply with the torque requirements;
however, one motor is enough for the other joints. Using two motors at joint
B is much cheaper than using one big motor with 560 oz/in. Other relevant
 Design and Development of a Competitive Low-Cost Robot Arm with ... 87

characteristics of the motors, which can be shown in Figure 5, are that they
can turn 60 degrees in 130 milliseconds and they have a weight of 47.9
grams each.
Once the initial dimensions for the robot arm and the motor were defined,
the design were carried out using the SolidWorks platform; design should
carefully take into account the thickness of the acrylic sheet and the way that
the pieces would be attached to each other. The acrylic sheet used to make
the robot is 1/8 thickness and that thin sheet was chosen because it easier for
machining and less weight with a good resistance.

Figure 5. Servo motor.

During design, we faced some difficulties due to the way of joining
thin acrylic parts strongly. It was needed tools to burn and join the acrylic
parts and that weren’t available and the team considered that a mechanical
junction based on screws and nuts would be much strong than other
alternatives, such as glue for example. In order to accomplish this, a small
feature was designed which allowed to fasten the bolts with the nuts without
having to screw in the thin acrylic layer. The result of this process was the
tridimensional design shown in Figure 6.
By end of design, each part was printed in full scale in cardboard paper
and then we verified all the dimensions and the interfaces of the assembly.
In turn, we built the first prototype of the robot arm. Next, parts of the robot
arm were machined from the acrylic sheet using a circular saw and Dermal
tools. The detailing on the parts was done in a professional workshop since
the parts of robot arm were too small and it is not an easy for accomplishing
such small and accurate cuts.
During assembling the robot parts with the motors, few problems pop up.
88 Mechatronics and Robotics

There were critical points that did not resist the fastening and, in turn, may
break down; hence, reinforcements in these points were considered. The
final result of the robot arm is shown in Figure 7.

ROBOT ARM INVERSE KINEMATICS

To validate the right positioning of the robotic arm, inverse kinematics
calculations are carried out. Such calculations are used to obtain the angle of
each motor from a position given by using the Cartesian coordinate system,
as shown in Figure 8.

Figure 6. Robot arm 3D model

Figure 7. Robot arm complete assembly.

Each motor will have a specific function: the motor located in the A union
 Design and Development of a Competitive Low-Cost Robot Arm with ... 89

positions the final element in the y axis, the motors B and C positions the
final element in the x and z axis.
The problem was simplified by using the xz plane, as shown in Figure 9.
In which the following known values were defined [9]:
LAB: the forearm length.
LBC: the arm length.
z: the position in the z axis.
x: the position in the x axis. y: the position in the y axis.
Using trigonometry relations, as shown in Figure 9, the motor angles θ2
and θ1 are obtained, as seen in Equations (9) and (10).

` (9)

(10)

(11)
The motor B is going to use θ1 and the motor C is going to use θ2. The
angle for the motor A is calculated as seen in Equation (11).

Figure 8. Coordinate system.

90 Mechatronics and Robotics

Figure 9. xz Plane.
With these calculations, the angles of servomotors are obtained and in
turn they take the action to move the whole structure to the specific position.

END-EFFECTOR SELECTION
The end effector is probably one of the most important and most complex
parts of the system. Wisely, it is much easier and economical to use a
commercial one than build it. The end effector varies mainly according to
the application and the task that the robot arm accomplishes for; it can be
pneumatic, electric or hydraulic. Since our robot arm is based in an electric
system, we may choose electric basis of end effector. Besides, the main
application of our system is handling, accordingly, the recommended type of
our end effector is a gripper, as shown in Figure 10. Please note that the end
effector is controlled by a servo motor and, in turn, the total servo motors
used for our robot arm will be 5 motors that move the structure.

Figure 10. Gripper with servo.

 Design and Development of a Competitive Low-Cost Robot Arm with ... 91

ROBOT ARM CONTROL

The robot arms can be autonomous or controlled manually. In manual mode,
a trained operator (programmer) typically uses a portable control device (a
teach pendant) to teach a robot to do its task manually. Robot speeds during
these programming sessions are slow. In the current work we enclosed the
both modes.
The control for the presented robot arm consists basically of three
levels: a microcontroller, a driver, and a computer-based user interface.
This system has unique characteristics that allow flexibility in programming
and controlling method, which was implemented using inverse kinematics;
besides it could also be implemented in a full manual mode. The electronic
design of control is shown in Figure 11.
The microcontroller used is an Atmega 368 which comes with a
development/programming board named “Arduino”, as shown in Figure 12.
The programming language is very similar to C but includes several libraryies
that help in the control of the I/O ports, timers, and serial communication.
This microcontroller was chosen because it has a low price, it is very easy to
reprogram, the programming language is simple, and interrupts are available
for this particular chip.
The driver used is a six-channel Micro Maestro servo controller
board. It supports three control methods: USB for direct connection to a
computer, TTL serial for use with embedded systems, such as the Arduino
microcontroller, and internal scripting for self-contained and host controller-
free applications. This controller, as shown in Figure 13, includes a 0.25 μs
resolution for position and built-in speed and acceleration control.

Figure 11. Electronic scheme of control.

92 Mechatronics and Robotics

Figure 12. Arduino microcontroller board.

Figure 13. Servo controller driver.

The user interface depends on the control method used, i.e., inverse
kinematics or a full manual mode. In the following, each interface is
described:

Inverse Kinematics Control

In this control method, the user inputs the coordinate system position where
the gripper should be. As consequence, interface is generated with Labview
through a visual user, as shown in Figure 14. The program automatically
performs the inverse kinematics calculations to obtain the angles that each
motor should have and then sends a command either to the microcontroller
or directly to the driver that will move the robot to the specified position.
Communication is performed with the RS-232 protocol. In the following,
you may see the Labview user interface inputs and output.
 Design and Development of a Competitive Low-Cost Robot Arm with ... 93

The Labview user interface inputs are:

x axis position.
y axis position.
z axis position.
Gripper opening.
Gripper attack angle.
Serial port.

The Labview user interface outputs are:

Motor A angle.
Motor B1 angle.
Motor B2 angle.
Motor C angle.
Attack angle.
Gripper angle
Such output variables are treated and sent by an appropriate way, so that
information can be interpreted in a correct manner. The outputs are sent via
the serial port which is communicated with the controller. When the button
“Move” is clicked, a process will take place, as shown in Figure 15. With
this action, the robotic arm will change its position according to the input
values. In addition, it has a standby button that stops the communication
controller.

Figure 14. Labview user interface.

94 Mechatronics and Robotics

Figure 15. Program process.

The main advantages of this approach are that it uses an efficient way
of moving and offers further capabilities that could be implemented, such
as position and sequence learning. A disadvantage, on the other hand, is that
the possible positions that have valid angles after the inverse kinematics
calculations are very limited because the servo motors have a restraint of
180˚.

Manual Control
This type of control is an extra option for our system that useful in specific
positions. In case of mandatory positions that the inverse kinematics mode
cannot calculate their valid angles, we may use the manual control instead.
Basically, manual control consists of a series of analog inputs, such as
potentiometers, that are connected with the microcontroller which will
interpret the values and send a command to the servo driver. In order to
implement this, a control board, as shown in Figure 16, should be built
to work as an interface with the user. Possible implementation includes a
teaching feature where the microcontroller stores positions in memory and
by a keypad or a series of switches we may recall these positions.

TESTING AND VALIDATION

Several tests were carried out to validate the robot arm and its components.
The testes covered both the particular elements and the overall system,
as shown in Figure 17. For the microcontroller, the tests are occurred by
sending different commands by the software to the microcontroller and
check changes on the output which was connected to a servo motor that
turned on or off depending on the command.
 Design and Development of a Competitive Low-Cost Robot Arm with ... 95

The servo motors were tested afterwards by sending different direct

pulses to each servomotor and verifying the response of moving to the
right position. We used a mark to know where the initial position was and
the final position of the motors is determined by sending a signal with the
microcontroller and, in turn, it is interpreted by the servo and compared to
the signal provided by the encoder, resulting in the rotation to the desired
position. During this test, the servo motor was inconsistence with the robot
arm system because of an incorrect polarization.

Figure 16. Potentiometer board.

Figure 17. Robot arm tests.

The servo motor driver was also tested using the Lab view software to
send commands to the microcontroller which sent the specific commands to
the driver which had one motor connected to change the position according
to the commend. It is important to notice that at the beginning of the project
a different servo motor driver was selected but several problems related to
the communication between them and the microcontroller were present. So
we choose a driver that allows the data to be sent directly from the computer
96 Mechatronics and Robotics

to it with only a USB wire, so the microcontroller would only be used in

case of the implementation of manual control.
Other tests were performed to verify the functionality of the whole
system, as shown in Figure 18.

Figure 18. Robot arm in action.

Those tests were occurred by introducing a specific position in the
Labview interface and measuring the distance between a reference point
and the final point in order to verify: the correct transformation from inverse
to direct kinematics, the relationship between the specified angles and the
rotation of the motors.
Testing and validation of the robot arm is one of the tasks that require
elongated time because several iterations are needed. During our tests,
many problems arise as: wrong angle calculations, wrong calibration of the
motors, problems with the physical angle and position measurements, and
one of the servo motors burned because of an overload that wasn’t expected.

RESULTS AND DISCUSSIONS

Results from the robot arm at different operating conditions are presented
as follows:

Servo Motors Movement Range

The limits of the servo motors were obtained since specification of this type
of motors contains that it has less than a 180 degree span. The real range for
all motors was found to be in the range 125 142 degrees, as shown in Table
 Design and Development of a Competitive Low-Cost Robot Arm with ... 97

1. This clearly demonstrated that real operation of robot arm is different

from the stander case.

Table 1. Motor angle ranges.

Current Consumption
The current consumption depends on the load and the type of motion
of the robotic arm. In the current study, there are 4 levels of current
consumptions:
• Low (from 0 to 200 mA). This consumption takes place when the
robot is at rest (not motion case).
• Normal (from 200 to 500 mA). This happened when the robot
arm is moving with capability to go to the target without needs
of great torque.
• High (from 500 mA to 900 mA). This range is reached at the
beginning of carrying loads. By overcoming the initial moment
of inertia for loads, the normal range takes a place.
Over current (more than 900 mA). The load is too heavy and the motor
cannot move at all. For being under this condition for more than one minute,
the motor will burn, i.e. it is not possible to be used any more.

Maximum Load
These results were obtained using different weights; a bag of corn was used
with a scale to determine bag’ weight. Results carried out by using the robot
arm to pick up the bag and move it to specific positions. Table 2 presents
the current consumption at different weights of bag of corn. From Table 2,
it can be seen that the robot can move without problems at loads lower than
50 grams. At loads 60 grams, the robot arm start having difficulties and
after passing 80 grams severe condition occurred where irreversible damage
could be happened in motors.
98 Mechatronics and Robotics

Final Position
Results show the precision of the robot arm to move different weight (<50
grams) is presented in Table 3. As shown, the robot arm is able to perform the
movement to the position specified. However, this movement is not smooth
and sometimes the motors do not have enough force, especially when the
load is heavy. In addition, some problems may appear due to synchronizing
the two bottom motors. The steps of the two motors were not coincidental
and that causes tension in the acrylic parts, which in case of being too much
will break the parts.

Table 2. Load vs. current consumption.

Table 3. Precision on all axis.

CONCLUSIONS
This paper presents the design, development and implementation of robot
arm, which has the talent to accomplish simple tasks, such as light material
handling. The robot arm was designed and built from acrylic material where
servo motors were used to perform links between arms and execute arm
movements. The servo motors include encoder so that no controller was
implemented; however, the rotation range of the motor is less than 180º
span, which greatly decreases the region reached by the arm and the possible
positions. The design of the robot arm was limited to four degrees of freedom
since this design allows most of the necessary movements and keeps the
 Design and Development of a Competitive Low-Cost Robot Arm with ... 99

costs and the complexity of the robot competitively. The end effector is not
included in the design because a commercially available gripper is used
since it is much easier and economical to use a commercial one than build it.
During design, we faced some difficulties due to the way of joining thin
acrylic parts strongly. A mechanical junction based on screws and nuts is
used and in order to accomplish that, a small feature was designed which
allowed fastening the bolts with the nuts without having to screw in the thin
acrylic layer.
To control the robot arm, three approaches are implemented: a
microcontroller, a driver, and a computer-based user interface. This system
has unique characteristics that allow flexibility in programming and
controlling method, which was implemented using inverse kinematics;
besides it could also be implemented in a full manual mode. This robotic
arm is contrast with others as being much cheaper than available robot
arms, also it can be controlled all of its movements from a computer, using
a Labview interface.
Several tests were carried out to validate the robot arm where the testes
covered both the particular elements and the overall system; results at
different operating conditions show trustful of the robot arm presented.
100 Mechatronics and Robotics

REFERENCES
1. Manipulating Industrial Robots—Vocabulary, International
Organization for Standardization Standard 8373, 1994.
2. Industrial and Service Robots, IFR International Federation of
Robotics, 2010. http://www.ifr.org/home
3. Case Studies and Profitability of Robot Investment, The IFR Statistical
Department, 2008. http://www.ifrstat.org/downloads/2008_Pressinfo_
english .pdf
4. R. J. Wang, J. W. Zhang, et al., “The Multiple-Function Intelligent
Robotic Arms,” FUZZ-IEEE Journal, Korea, 20-24 August 2009, pp.
1995-2000.
5. L. B. Duc, M. Syaifuddin, et al., “Designing 8 Degrees of Freedom
Humanoid Robotic Arm,” International Conference on Intelligent and
Advanced Systems, Kuala Lumpur, 25-28 November 2007, pp. 1069-
1074.
6. C. R. Carignan, G. G. Gefke and B. J. Roberts, “Intro to Space Mission
Design: Space Robotics,” Seminar of Space Robotics, University of
Maryland, Baltimore, 26 March 2002.
7. Occupational Safety and Health Administration Technical Manual,
OSHA 3167, United States Department of Labor, 1970.
8. B. Siciliano, L. Sciavicco, L. Villani and G. Oriolo, “Robotics,
Modelling, Planning and Control,” Springer, London, 2009.
9. M. P. Groover and M. Weiss, “Robotica Industrial, Technologia,
Programacion y Aplicaciones,” Mc-Graw Hills, Mexico D.F., 1989.
 CHAPTER

6
An Adaptive Robust Approach to
Modeling and Control of Flexible
Arm Robots

Pedram Bagheri1, Laleh Behjat1, Qiao Sun2

Department of Electrical and Computer Engineering, University of Calgary,
1

Calgary, Alberta, Canada

Department of Mechanical and Manufacturing Engineering, University of
2

Calgary, Calgary, Alberta, Canada

ABSTRACT
In this paper, a novel adaptive robust approach to modeling and control of
a class of flexible-arm robots subject to actuators unmodeled dynamics is
proposed. It is shown how real-time signals measured from a dynamical

Citation:- Bagheri, P., Behjat, L. and Sun, Q. (2018), “An Adaptive Robust Approach
to Modeling and Control of Flexible Arm Robots”. World Journal of Engineering and
Technology, 6, 282-295. doi: 10.4236/wjet.2018.62017.
Copyright: © 2018 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
102 Mechatronics and Robotics

system can be utilized to improve the accuracy of the mathematical

model of flexible robots. Given the elasticity of the robot’s arms, flexible
manipulators have both passive and active degrees of freedom. A nonlinear
robust controller is designed for the active degrees of freedom to enable the
robot to follow desired trajectories in the presence of actuators unmodeled
dynamics. Furthermore, it is shown that under some feasible conditions,
another nonlinear robust controller is designed for the passive degrees
of freedom. Moreover, to use the system response for model extraction,
two auxiliary signals are proposed to provide sufficient information
for improving the accuracy of the dynamics of the system numerically.
Additionally, two adaptive laws are proposed in each case to update the
two introduced auxiliary signals. As a result, the controller controls the
passive degrees of freedom after the active degrees of freedom converge
to their desired trajectories. Simultaneously, the information collected from
the system to update the auxiliary signals enhances the model accuracy. In
the end, simulation results are presented to verify the performance of the
proposed controller.
Keywords: Adaptive Control, Robust Control, Vibration Control

INTRODUCTION
The area of flexible-arm robots has attracted much attention during the last
few decades [1] [2]. This interest is due to the advantages that flexible arms
offer compared to their rigid counterparts. Weight reduction, lower energy
consumption, and faster system response are among several benefits utilized
in their numerous applications such as space missions [3].
There has been a great number of studies coping with controlling
flexible-arm robots, many of which investigate both theoretical and
experimental aspects in this field [4] [5]. On grounds of the flexibility of
the arms, along with a trajectory-tracking control problem, vibration control
should be also considered for such systems to improve the control system
performance [6]. As a result of the vibrations caused by the flexibility of
the arms, designing controllers for such systems becomes a challenging
task. There exist several research works in the literature addressing the
flexibility of the arm. Passive control methods are one way to deal with the
vibrations of elastic arms which require modification in physical parameters
of the system structure [7]. Due to these structural modifications, absorption
properties of the structure can be employed to increase damping properties
of the arms. Active control approaches have also been widely used to control
 An Adaptive Robust Approach to Modeling and Control of ... 103

flexible systems, in which actuation moments and forces are applied to

address the vibrations [8] [9]. However, the mere use of passive control
methods to reduce vibrations does not seem to suffice. Merely using active
control approaches is not sufficient as well. It is because vibration modes
with frequencies near actuators frequencies can lead to instability. Hence, a
combination of passive and active controllers can be employed as a suitable
solution to vibrations reduction [10] [11]. The boundary feedback scheme
is another way to dampen the vibrations of flexible manipulators [12]. Luo
et al. [13] controlled the vibrations of a class of flexible robots using a shear
feedback control method. Lyapunov-based control has also been vastly
utilized to cope with the challenges associated with controlling flexible
structures [14]. As another example of Lyapunov-based methods, Dadfarnia
et al. [15] used the Lyapunov stability theory to develop a piezoelectric
controller for flexible robots.
Infinite dimensionality is one of the most considerable challenges in
modeling flexible robots. The existence of flexibility in the system results
in dynamics governed by partial differential equations. Thus, techniques
such as modal truncation are employed to express the dynamics by a set of
ordinary differential equations [16]. As examples of such methods, Arts et
al. [17] proposed an adaptive model integration method as a model reduction
technique for planar flexible manipulators. Bruls et al. [18] used the global
modal parameterization technique to reduce model-order of flexible multi-
body dynamics. The procedure in both aforementioned methods is to divide
the motion into two parts of rigid and elastic.
On account of the elasticity of the arms, there are degrees of freedom
on which no actuation acts (Passive degrees of freedom). Consequently,
flexible-arm robots fall into the category of underactuated mechanical
systems. That is, systems with a lower number of control inputs compared to
the number of degrees of freedom. Designing controllers for underactuated
systems is an open problem. Control designs for such systems are dynamics-
dependent while dynamical characteristics of the system play an essential
role in the development of a control strategy. As an example of the existing
control methods for underactuated systems, Mahmut Reyhanoglu et al.
presented a theoretical scheme for modeling and control of underactuated
systems. As one of the requirements of their approach, the non-integrable
acceleration relations are required to be satisfied [19]. As another example,
Zhang and Tarn designed a hybrid switching control strategy for nonlinear
and underactuated mechanical systems [20].
104 Mechatronics and Robotics

Feedback linearization methods are wildly used for nonlinear systems. Nev-
ertheless, As a result of the underactuated nature of flexible arms, the exact
feedback linearization method [21] cannot be employed for such mechanical
systems. Hence, the partial feedback linearization method [22] is considered
as one the most suitable way to cope with underactuated systems. That is,
the entire system can be either linearized with respect to the active degrees
of freedom (collocated problem) or with respect to the passive degrees of
freedom (non-collocated problem). Be that as it may, such a control strategy
fails to handle uncertainties as control inputs to feedback-linearized systems
depend on the governing equations.
In this paper, a novel adaptive robust nonlinear controller is designed for
a class of multi-link flexible arms subject to uncertainties and unmodeled
dynamics of the actuators. The contribution of the current research work
is to utilize the system’s real-time responses to improve the accuracy of
the available mathematical model. Additionally, despite uncertainties and
underactuated nature of the system, the proposed controller is able to
track desired trajectories asymptotically. That is, despite the presence of
uncertainties and disturbance sources, the tracking error converges to zero.
To do such, first, an adaptive robust controller is designed for the active
degrees of freedom. Meanwhile, adaptive laws are proposed to estimate the
system uncertainties as well as approximating actuators uncertain dynamics.
Furthermore, another adaptive robust controller is designed to control the
passive degrees of freedom where the stabilities of the overall closed-loop
system, in both cases, in the presence of uncertainties are established. In
addition, due to the flexibility of the arms, the extraction of the system
dynamics is demanding and, for the most part, should be done analytically.
In the present approach, two adaptive signals are introduced and synthesized
such that the cumbersome analytical part of the model extraction can be
done numerically.

DYNAMIC EQUATIONS
In the current study, a multi-link flexible-arm robot is considered for
modeling and control purposes. The investigated system is assumed to
include m links containing l elastic modes. Therefore, the total degrees of
freedom come to n=m+l.
Equations governing a flexible arm using the Euler-Bernoulli beam can
be written as follows
 An Adaptive Robust Approach to Modeling and Control of ... 105

(1)

Where is the positive definite inertial matrix;

includes Coriolis and centrifugal terms; contains gravitational
terms; is the generalized coordinates vector where denotes
the passive degrees of freedom, and represents the active degrees of
freedom. is the torque generated by the actuators acting on the active
degrees of freedom. and G(q) can be calculated as [23]

(2)

(3)
where U denotes potential energy of the system. Extracting the dynamic
equations of flexible robot arms is time consuming and can be analytically
complicated. Due to its analytical nature, there exists a high chance of
computational errors as the equations are derived. The extraction of the
dynamic equations can be divided into two phases: 1) Extracting the portions
of the dynamics that are straightforward and demand less calculations such
as kinetic and potential energy. It is to be noted that the inertia matrix
M(q) can be arrived at directly from the rearrangement of the kinetic energy
of the system. 2) Obtaining the portions of the dynamics which require
lengthy analytical calculations such as G(q) and . It is worth noting
that calculating G(q) and require calculus of variations which can be
cumbersome for systems with a large degrees of freedom. The idea of this
paper is to propose a way to procure G(q) and numerically using the
system responses. For this purpose, (1) is rewritten as

(4)
where

(5)

(6)
106 Mechatronics and Robotics

and d(t) is an unknown vector added to the equations to account for uncertain
dynamics of the actuators. It is also assumed that where γ is an
unknown constant.
In order to complete the dynamic equations given in (4), and
should be computed. As it follows from (5) and (6), and
are expressed in terms of the matrix M, system states, and its potential
energy. Therefore, one can use the measurement of the states along with the
kinetic and potential energy of the system and use numerical computations to
obtain h1 and h2. Hence, the system dynamics is computed as it runs without
going through analytical calculations. However, such dynamic equations are
prone to numerical and measurement errors. Thus, a special measure should
be taken to improve the accuracy of such computations. To compensate for
such errors, two unknown auxiliary signals P and S are employed as follows

(7)

(8)
where are two unknown auxiliary signals
introduced to account for numerical and measurement errors;
are two diagonal matrices constructed as
and denotes the
jth element of the vector hi where i=1,2 . Given the physical and geometric
properties of flexible arm robots, it is not restrictive to assume that
both H1 and H2, as defined above, have full ranks. Therefore, their column
space can be utilized as a basis for the vectors h1 and h2 at any given time.
Thus, for any exact value of h1 and h2 at any given time, there exist two
unique vectors P and S such that h1=H1P and h2=H2S. It is clear that the
vectors P and S change in values over time. Hence, their values need to be
updated according to the system response. Therefore, an adaptive procedure
is developed for each of these vectors along with designing a controller in
Section 3. Consequently, as the system runs, its response is used to update
the values of P and S.

CONTROL DESIGN
In this section, an adaptive controller is designed for the considered system
in the presence of uncertainties. Alongside the adaptive controller, two
adaptive laws are derived for updating the values of P and S which are used
 An Adaptive Robust Approach to Modeling and Control of ... 107

to complete the dynamic equations. As it was shown in the previous section,

(7) and (8) exhibit the dynamic equations governing an n-degree of freedom
flexible-arm, with m active and l passive degrees of freedom. The first
objective is to design a controller for the active degrees of freedom.

Control Design for Active Degrees of Freedom

Employing (7), (8) can be rewritten as follows

(9)
Two matrices N and R are defined as .
Consequently, (9) comes to

(10)
As N is positive definite, (10) can be expressed as follows

(11)
where , and . Tracking error and its dynamics
can be defined as , respectively; q2d is the desired
trajectory for the active degrees of freedom. A measure of the tracking error
is also defined as

(12)
where is a positive definite matrix. The estimation errors of the
uncertainty vectors are shown by which are defined as and
are the estimations of the uncertainty vectors. Using the
definitions of (11) is rewritten as follows

(13)
Considering (12), ˙ is calculated as follows

(14)
To proceed with the control design, the following Lyapunov function is
proposed
108 Mechatronics and Robotics

(15)
where is the Euclidean norm of the estimation error of the
uncertain dynamics of the actuators. Taking time-derivative of (15) gives

(16)
where is decomposed into . Given the structure of (16), the
following control and adaptive laws are proposed

(17)

(18)
(19)
where is a positive definite matrix; (17) is the proposed control
law, and (18)-(19) are the adaptive laws to update the two auxiliary signals
introduced in (7)-(8). Substituting (17)-(19) into (16) yields

(20)
(20) can be rewritten as follows

(21)
where ∥ . ∥ stands for the Euclidean norm. Since N is a bounded positive
definite matrix, the adaptive law to update is proposed as

(22)
Employing (22), (21) becomes

(23)
To establish the asymptotic stability of the closed-loop system, the
second part of the control law is introduced as follows in (24). It is to be
noted that the second part of the control law compensates for the actuators
 An Adaptive Robust Approach to Modeling and Control of ... 109

uncertain dynamics.

(24)
where is an arbitrary signal where . By substituting (24)
into (23), the following is obtained

(25)
Since are positive, the following inequality is concluded

(26)
Hence, (25) is simplified as follows

(27)
Taking integral from both sides of (27) gives

(28)
Considering the fact that , the following is achieved

(29)
where a is a positive constant. It is observed that the left side of (29) is
non-negative, thus, it follows that (V V (0) − ∞( )) is bounded. As a result, it
is deduced that the right side of (29) remains bounded as well. Consequently,
QL∈2. It is also easy to check that . Therefore, it is concluded from
Barbalat’s lemma that which implies . Thus, it was proven
that adopting the extracted control laws (17), (24), and adaptive laws (18),
(19)) and (22) guarantees the asymptotic stability of the closed-loop system.
That is, the system is able to track desired trajectories asymptotically in the
presence of model and actuators uncertainties.
110 Mechatronics and Robotics

Control Design for Passive Degrees of Freedom

In this section, a robust nonlinear controller is designed for the passive
degrees of freedom alongside extracting adaptive laws to update uncertainties
vectors. To do such, (8) is rewritten as

(30)
Substituting (30) into (7) yields

(31)
where . It follows from the structure of M that
M22 is nonsingular. In addition, it can be easily shown that F is a positive
definite matrix as well. Therefore (31) can be expressed as follows

(32)
where , and J is a control signal yet to
be defined. The actual control input can be retrieved from J as follows
(33)
It is concluded from (33) that G should be nonsingular. Thus, according
to the definition of G, M12 needs to be invertible as well. Hence, in order to
extract a non-singular control law for the passive degrees of freedom, the
following assumption is made:
Assumption 1 It is assumed that the number of passive degrees of
freedom is less or equal than the number of active degrees of freedom.
The tracking error for the passive degrees of freedom and its dynamics
are defined as where q1d is the desired trajectory
for the passive degrees of freedom. Moreover, a measure of the tracking
error is defined as is a positive definite matrix.
Using the definitions of which are defined in the previous section, (32)
is expressed as follows

(34)
Based on the structure of (34), the following Lyapunov function is introduced

(35)
The time-derivative of (35) is calculated as
 An Adaptive Robust Approach to Modeling and Control of ... 111

(36)
where . Given (36), the following control and adaptive laws
are proposed

(37)
(38)

(39)
Where is a positive definite matrix? Substituting (37)-(39) into
(36) and taking the Euclidean norm form both sides lead to the following

(40)
To make (40) negative semi-definite, the following control and adaptive
laws are proposed

(41)

(42)
where σ(t) is a positive arbitrary signal belonging to L1 . Substituting
(40)-(42) into (40) gives

(43)
Similar to the stability analysis presented in the previous section, it is
concluded from (43) that , which implies that . Hence, it is
guaranteed that the tracking error converges to zero asymptotically in the
presence of uncertainties.

SIMULATION RESULTS
In this section, a simulation is conducted to validate the analytical results.
The simulated case study is a one-link flexible-arm robot with a payload at
the tip of the arm. It is assumed that the system is subject to uncertainties
112 Mechatronics and Robotics

and unmodeled dynamics in the actuators. The strategy is, first, to control the
active degrees of freedom to their desired values. Then, the control law and
adaptive laws switch to the non-collocated control to dampen the vibrations
of the flexible arm. The physical properties of the investigated arm are as
follows: Arm s Length = 1 m; Density = 7850 kg/m3; Beams section area
= 2e−4 m2; Elasticity = 2.07e11; Load mass = 0.1 kg. The actual values
uncertainty signals are S=8.2 and P=5.5.
It is shown in Figure 1 that the flexible arms hub angle is successfully
controlled to the desired angle. It is also observed from Figure 1 that the
hubs angle remains in its desired angle after the controller switches.
In Figure 2, the vibrations of the flexible arm are displayed, and its
zoomed-in version is depicted in Figure 3. It follows from Figure 3 that the
arm residual vibrations are dampened as the controller switches.
Figure 4 and Figure 5 correspond to the estimation of the introduced
auxiliary signals to compensate for uncertainties. As noticed from the
figures, the estimation of the uncertain parameters have converged to their
limits after about two seconds through the simulation.
Finally, the required actuator torque to control the arm is shown in Figure
6. It follows from the simulations results that the arm tip can be controlled
to its desired position more accurately. The reason stems from the fact that
the non-collocated control helps decrease the response time. It is because
that it would dramatically take a longer time for the residual vibrations to
dampen naturally.It was shown both analytically, in Section 3, and through
simulation that the proposed controller can achieve asymptotic tracking in
the presence of unknown and bounded disturbance sources.

Figure 1. Arm’s hub angle (rad).

 An Adaptive Robust Approach to Modeling and Control of ... 113

Figure 2. Flexible arm’s vibrations (m).

Figure 3. Zoomed-in version of the arm’s vibrations.

Figure 4. Auxiliary signal estimation (P).

114 Mechatronics and Robotics

Figure 5. Auxiliary signal estimation (S).

Figure 6. Control signal (Nm).

As a comparison to other recent robust approaches to the control problem
of such flexible manipulators, authors in [24] linearize the dynamics to
simplify the complexity of their model. However, the nonlinear dynamics
of the system is considered in the present study, and the controller is
designed accordingly. It is to be noted the asymptotic tracking in this study
is established on the postulate that no prior knowledge of the bound of the
disturbance is available. Be that as it may, existing methods in the literature
merely guarantee the roundedness of control and state signals in case of
unknown disturbance sources. That is, they cannot prove the output can
converge to desired trajectories asymptotically. As an example, authors
in [25] take a robust adaptive approach to control a flexible manipulator
using the Lyapunov stability theory. Their analytical analysis suggests
the roundedness of all signals in the presence of disturbance. Thus, given
 An Adaptive Robust Approach to Modeling and Control of ... 115

a desired trajectory, the output can stay in the vicinity of the trajectory.
However, the convergence of the output to the desired trajectory is not
guaranteed.

CONCLUSION
As presented in detail, an adaptive robust nonlinear controller was designed
for a class of flexible-arm robots based on the Lyapunov stability theory.
A new numerical technique was presented to facilitate the extraction
of the governing equations of flexible-arm robots. It was shown how
computational errors caused by numerical operations could be compensated
and approximated as uncertainty signals. In the control design process,
first, an adaptive controller was designed for the active degrees of freedom
to enable the system to follow desired trajectories in the presence of
uncertainties. Further, it was shown that under some feasible conditions,
another adaptive robust controller could be designed for the passive degrees
of freedom. Therefore, as the system was being controlled, its responses
were utilized to improve the accuracy of the mathematical model. Thus, the
presented method leads to improvements in the controller performance and
increases tracking accuracy. Furthermore, due to a numerical approach to
extracting the dynamics, the amount of analytical computations is reduced
dramatically.
116 Mechatronics and Robotics

REFERENCES
1. Giorgio, I., Della Corte, A. and Del Vescovo, D. (2017) Modelling Exible
Multi-Link Robots for Vibration Control: Numerical Simulations and
Real-Time Experiments. Mathematics and Mechanics of Solids, 21,
1234-1246.
2. Kiang, C.T., Spowage, A. and Yoong, C.K. (2015) Review of Control
and Sensor System of Exible Manipulator. Journal of Intelligent &
Robotic Systems, 77, 187-213. https://doi.org/10.1007/s10846-014-
0071-4
3. Dwivedy, S.K. and Eberhard, P. (2006) Dynamic Analysis of Exible
Manipulators, a Literature Review. Mechanism and Machine Theory,
41, 749-777. https://doi.org/10.1016/j.mechmachtheory.2006.01.014
4. He, W., Ouyang, Y. and Hong, J. (2017) Vibration Control of a
Exible Robotic Manipulator in the Presence of Input Deadzone.
IEEE Transactions on Industrial Informatics, 13, 48-59. https://doi.
org/10.1109/TII.2016.2608739
5. Kumar, A., Pathak, P.M. and Sukavanam, N. (2011) Reduced Model
Based Control of Two Link Exible Space Robot. Intelligent Control
and Automation, 2, 112-120. https://doi.org/10.4236/ica.2011.22013
6. Nani, V.-M. and Cires, I. (2014) Vibration Control for Two Air
Compressor Units Used in Tandem. World Journal of Engineering and
Technology, 2, 314-321. https://doi.org/10.4236/wjet.2014.24033
7. Sales, T., Rade, D. and De Souza, L. (2013) Passive Vibration Control of
Exible Spacecraft Using Shunted Piezoelectric Transducers. Aerospace
Science and Technology, 29, 403-412. https://doi.org/10.1016/j.
ast.2013.05.001
8. Bai, K., Luo, M., Li, T., Wu, J., Yang, L., Liu, M. and Jiang, G. (2016)
Active Vibration Adaptive Fuzzy Backstepping Control of a 7-Dof
Dual-Arm of Humanoid Robot with Input Saturation. Journal of
Intelligent & Fuzzy Systems, 31, 2949-2957. https://doi.org/10.3233/
JIFS-169179
9. Zhang, Q., Li, C., Zhang, J. and Zhang, J. (2017) Smooth Adaptive
Sliding Mode Vibration Control of a Exible Parallel Manipulator
with Multiple Smart Linkages in Modal Space. Journal of Sound and
Vibration, 411, 1-19. https://doi.org/10.1016/j.jsv.2017.08.052
10. Mottershead, J.E. and Ram, Y.M. (2006) Inverse Eigenvalue Problems
in Vibration Absorption: Passive Modification and Active Control.
 An Adaptive Robust Approach to Modeling and Control of ... 117

Mechanical Systems and Signal Processing, 20, 5-44. https://doi.

org/10.1016/j.ymssp.2005.05.006
11. Svensson, J.L., Andersson, P.B. and Kropp, W. (2010) On the Design
of Structural Junctions for the Purpose of Hybrid Passive Active
Vibration Control. Journal of Sound and Vibration, 329, 1274-1288.
https://doi.org/10.1016/j.jsv.2009.11.002
12. Coleman, M. and McSweeney, L. (2004) Analysis and Computation of
the Vibration Spectrum of the Cartesian Exible Manipulator. Journal
of Sound and Vibration, 274, 445-454. https://doi.org/10.1016/j.
jsv.2003.09.036
13. Luo, Z.-H., Kitamura, N. and Guo, B.-Z. (1995) Shear Force Feedback
Control of Exible Robot Arms. IEEE Transactions on Robotics and
Automation, 11, 760-765. https://doi.org/10.1109/70.466607
14. Herman, P. (2009) Strict Lyapunov Function for Sliding Mode
Control of Manipulator Using Quasi-Velocities. Mechanics
Research Communications, 36, 169-174. https://doi.org/10.1016/j.
mechrescom.2008.09.010
15. Dadfarnia, M., Jalili, N., Xian, B. and Dawson, D.M. (2004) A
Lyapunovbased Piezoelectric Controller for Exible Cartesian Robot
Manipulators. Journal of Dynamic Systems, Measurement, and
Control, 126, 347-358. https://doi.org/10.1115/1.1767854
16. Liu, X., Schauer, T., Goldack, A. and Schlaich, M. (2017) Multi-Modal
Active Vibration Control of a Lightweight Stress-Ribbon Footbridge
Based on Subspace Identication. IFAC-Papers Online, 50, 7058-7063.
https://doi.org/10.1016/j.ifacol.2017.08.1352
17. Aarts, R.G. and Jonker, J.B. (2002) Dynamic Simulation of Planar Exible
Link Manipulators Using Adaptive Modal Integration. Multibody
System Dynamics, 7, 31-50. https://doi.org/10.1023/A:1015271000518
18. Bruls, O., Duysinx, P. and Golinval, J.-C. (2007) The Global Modal
Parameterization for Non-Linear Model-Order Reduction in Exible
Multibody Dynamics. International Journal for Numerical Methods in
Engineering, 69, 948-977. https://doi.org/10.1002/nme.1795
19. Reyhanoglu, M., van der Schaft, A., McClamroch, N.H. and
Kolmanovsky, I. (1999) Dynamics and Control of a Class of
Underactuated Mechanical Systems. IEEE Transactions on Automatic
Control, 44, 1663-1671. https://doi.org/10.1109/9.788533
20. Zhang, M. and Tarn, T.-J. (2003) A Hybrid Switching Control
118 Mechatronics and Robotics

Strategy for Nonlinear and Underactuated Mechanical Systems.

IEEE Transactions on Automatic Control, 48, 1777-1782. https://doi.
org/10.1109/TAC.2003.817911
21. Chen, J.L. and Chang, W.-D. (2009) Feedback Linearization Control of
a Twolink Robot Using a Multi-Crossover Genetic Algorithm. Expert
Systems with Applications, 36, 4154-4159. https://doi.org/10.1016/j.
eswa.2008.01.048
22. Arisoy, A., Gokasan, M. and Bogosyan, O. (2006) Tip Position Control
of a Exible-Link Arm. 9th IEEE International Workshop on Advanced
Motion Control, 445-450.
23. Olfati-Saber, R. (2001) Nonlinear Control of Underactuated Mechanical
Systems with Application to Robotics and Aerospace Vehicles. PhD
Dissertation, Massachusetts Institute of Technology.
24. Mohamed, Z., Khairudin, M., Husain, A. and Subudhi, B. (2016) Linear
Matrix Inequality-Based Robust Proportional Derivative Control of
a Twolink Exible Manipulator. Journal of Vibration and Control, 22,
1244-1256. https://doi.org/10.1177/1077546314536427
25. He, X., He, W. and Sun, C. (2017) Robust Adaptive Vibration Control
for An uncertain Exible Timoshenko Robotic Manipulator with Input
and Output Constraints. International Journal of Systems Science, 48,
2860-2870. https://doi.org/10.1080/00207721.2017.1360963
 CHAPTER

7
Interactive Heuristic D* Path Planning
Solution Based on PSO for Two-Link
Robotic Arm in Dynamic Environment

Firas A. Raheem, Umniah I. Hameed

Control and Systems Department, University of Technology, Baghdad, Iraq

ABSTRACT
This paper is devoted to find an intelligent and safe path for two-link robotic
arm in dynamic environment. This paper focuses on computational part of
motion planning in completely changing dynamic environment at every
motion sample domains, since the local minima and sharp edges are the most
common problems in all path planning algorithms. In addition, finding a path
solution in a dynamic environment represents a challenge for the robotics

Citation:- Raheem, F. and Hameed, U. (2019), “Interactive Heuristic D* Path Planning

Solution Based on PSO for Two-Link Robotic Arm in Dynamic Environment”. World
Journal of Engineering and Technology, 7, 80-99. doi: 10.4236/wjet.2019.71005.
Copyright: © 2019 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
120 Mechatronics and Robotics

researchers, so in this paper, a proposed mixing approach was suggested

to overcome all these obstructions. The proposed approach methodology
for obtaining robot interactive path planning solution in known dynamic
environment utilizes the use of modified heuristic D-star (D*) algorithm
based on the full free Cartesian space analysis at each motion sample with
the Particle Swarm Optimization (PSO) technique. Also, a modification
on the D* algorithm has been done to match the dynamic environment
requirements by adding stop and return backward cases which is not included
in the original D* algorithm theory. The resultant interactive path solution
was computed by taking into consideration the time and position changes of
the moving obstacles. Furthermore, to insure the enhancement of the final
path length optimality, the PSO technique was used. The simulation results
are given to show the effectiveness of the proposed method.
Keywords: D* Algorithm, Particle Swarm Optimization (PSO), Path
Planning, Two-Link Arm, Known Dynamic Environment

INTRODUCTION
The words “path planning” refer to find a path solution in robotics where
the process is totally or sub automated. It is used to indicate the type of
computational process for moving a robot from one place to another with
respect to obstacles. In other words, it represents the search for initial
feasible path which is considered as the first step to solve the path planning
problem and includes finding a path for a robot that must move from initial
given start point to the goal point which is also given as the destination
position. The environment usually contains a fully de-fined set of obstacles.
Therefore, the robot does not collide with any of the obstacles [1]. Generally,
obstacles cannot always be static. When the obstacle changes its location
over time, it called dynamic. Dynamic environment represents a revolution
and important side in modern automation. It redounds in many important
diverse applications like air and sea traffic control negotiating, freeway
traffic, intelligent vehicles, automated assembly, and automated wheelchairs.
Because of existence the moving objects, the robot needs to make decision
quickly for avoiding possible collisions [2] [3]. There are three types of
dynamic environment. The first type is called a known dynamic environment
when all the information regarding the obstacles motions, sizes, shapes, and
locations is completely known. The second type is called a partially known
dynamic environment when not all information about the obstacles exists
at the planning time. In this state, it needs to calculate the robot motion
 Interactive Heuristic D* Path Planning Solution Based on ... 121

according to insufficient information about the environment. It is important

to know when the robot’s mission is to complete a map of the environment
by scanning the area to make sure if there is a new object found. Finally,
when there is nothing to know about the obstacles completely, it’s called the
totally unknown and dynamic environment and this needs some intelligent
methods for sensing and making a proper decision [3].
Due to the rapid increase in the use of robots in different life and industry
applications, path planning in dynamic environment is still considered as
an actual challenge and needed to be solved interactively for both mobile
and manipulator robots. The proposed method in this paper ensures finding
an intelligent, proper, safe and relatively optimal path depending on the
dynamic environment complexity. The structure of this paper contains
seven sections in addition to the introduction, Related Research work, two-
link robot arm kinematics, free Cartesian space analysis, proposed method,
simulation results, and conclusions.

RELATED RESEARCH WORK

To overcome the dynamic environment challenges, many path planning
algorithms have been used in the literature. Here we started with Anthony
Stentz [4] whom used The D* algorithm which is represented the dynamic
A*, it aims to create a short path in real-time by incrementally reforming paths
for the robot’s state as a new information is found. The D* extension which
concentrate the repairs of significantly reduction the whole time required to
subsequent and calculate the initial path re-planning operations which this
paper aims to, followed by Yueshi Shen and Knut Huper [5] who suggested
using a method to calculate the optimal joint trajectories for multi-DOF
manipulator robot performing constrained motion task. The path planning is
completed in two stages, firstly converting the variational calculus problem
into a finite dimensional optimization problem, using Newton’s method to
compute the joint trajectory’s inter-mediate points. Secondly, interpolating
these points to form the joint curve and adjust it to ensure that the motion
constraint will be satisfied throughout the entire trajectory. These calculations
are easier to apply for multi-DOF manipulators than the Tow-Link robot arm
because multi-DOF manipulators are more flexible.
Many other path planning algorithms were applied for solving dynamic
environment problems to find an optimal path for robots in an environment
that is only partially known and continuously changing. A method was
suggested for generating a collision-free near-optimal path for an autonomous
122 Mechatronics and Robotics

mobile robot in a dynamic environment containing moving and static

obstacles using the neural network and fuzzy logic with a genetic algorithm.
The neural network has a limited data size therefore, it cannot solve path in
bigger environment furthermore, the simulation results show that the method
is efficient and gives near-optimal and not optimal path reaching the target
position of the mobile robot [6]. An improved virtual force field (VFF)
algorithm approach was proposed for the real-time robot path planning, where
the environment is unknown and changing. An area ratio parameter with the
consideration of the robot size and obstacles was used. Furthermore, a fuzzy
control module was added to deal with the problem of obstacle avoidance in
dynamic environments. Due to the local minimum problem and the higher
computational complexity, the path could not reach optimality [7].
From the same point of view, an algorithm for specifying Cartesian
constraints with a single and dual arm operation is proposed based on
Rapidly-Exploring Random Tree (RRT). One parameter is used to be the
common in all RRT based on planers is the range. The range represents a
threshold for the samples drawn newly and their small distance values after
that those will add to the tree. Range value will be affective on the planning
time which means that it should be selected carefully since the higher values
causing short time planning which is advantage but as a result, it also cause
a tensive motions and the smaller values give a long time planning and it
composed more points in the tree. Also, more checking will be needed for
the collision that may occur in the space; therefore it needs extra collision
checking [8].
Most of the existing literature in dynamic environment considers only
the traditional path planning algorithms and there are only a few studies
that consider the heuristic algorithms like Anirudh Vemula [9] whom
uses a heuristic based graph search planning algorithm that uses adaptive
dimensionality consider time dimension corresponding to region where
potential dynamic obstacle collisions can occur. Their results show the
approach can only model simple interactions and fail to generalize for
complex crowded settings.
Inspired by the above discussion, the contribution of the proposed
method in this paper can be summarized in the following points:
1) Multi free Cartesian space analysis has been computed at each
motion sample to identify the feasible workspace area.
2) Modifying the heuristic D* algorithm by adding stop case and
return backward case for taking into consideration the obstacles
 Interactive Heuristic D* Path Planning Solution Based on ... 123

position and time which does not used before. These modification
cases not included in the original D* theory.
3) Finding a path solution utilizing the proposed method in this
paper that use the modified heuristic D* algorithm based on the
PSO optimization technique in known dynamic environment.
4) Applying the proposed method to simulate the path solution for
the two-link robot arm as case study, which is more difficult
than the mass-point case study usually used in path planning
researches.
5) The resulting path solution from this proposed method can solve
the problem of finding an optimal path in dangerous, harsh, toxic
environments and inaccessible areas.

TWO-LINK ROBOT ARM KINEMATICS

Kinematics is the science of motion that treats the subject without regard to
the forces that cause it. Two types of kinematics problems can be classified;
the first one is the forward kinematics problem which is concerned with the
relationship between the individual joints of the robot manipulator and the
position and orientation of the tool or end-effector [10].
The second is the inverse kinematics problem of the serial manipulators
has been studied for many decades. It is needed in the control of manipulators.
Solving the inverse kinematics is computationally expensive and generally
takes a very long time in the real-time control of manipulators. Tasks to be
performed by a manipulator are in the Cartesian space, whereas actuators
work in joint space. Cartesian space includes orientation matrix and a
position vector. However, joint space is represented by joint angles. The
conversion of the position and orientation of a manipulator end-effector
from Cartesian space to joint space is called an inverse kinematics problem
[11]. The planar manipulator is shown in Figure 1(a) to illustrate the
2-DOF. The planar manipulators have l1l1 and l2l2 as their link lengths
and θ1θ1 and θ2θ2 represent joint angles with x and y as task coordinates.
The forward kinematics equations are shown below:

(1)

(2)
124 Mechatronics and Robotics

In 2-DOF planner manipulator, the Geometric Solution Approach of

inverse kinematics is used for obtaining the joint variables for all the
points in Cartesian space regardless their orientation. The calculations
of θ1θ1 and θ2θ2 for the 2-DOF manipulator could be found by applying
the following equations.

Figure 1. (a) Two degrees of freedom manipulator; (b) The configurations of

two-link arm elbow up and elbow down.
Utilizing the cosines law, where the angle is given by:

(3)
Since sin θ2 is

(4)
To solve θ2, it has to use the cos−1 function. Although it is more benefical
to use tan−1tan−1 for numerical solution, in the software implementation,
we will use the function tan−1 (b/c) which is ATAN2 (b, c). This function
has a uniform accuracy over the range of its arguments, returns a unique
value for the angle depending on the signs of b and c, and gives the correct
solution if b and/or c is zero [11].
Subsequently, the two conceivable solutions for θ2 can be written as:

(5)
At last, θ1θ1 can be solved by:

(6)
where l1 and l2 are lengths of the first and the second links, respectively.
 Interactive Heuristic D* Path Planning Solution Based on ... 125

FREE CARTESIAN SPACE ANALYSIS

The result of inverse kinematics has been used to build and analyze the
workspace in this thesis. The specific points that the end-effector can reach,
can be defined as free Cartesian space (FCS). The angles (θ1 and θ2) which
were found by inverse kinematics are related to these specific points directly.
The shape and volume of the FCS in an environment, which contains many
obstacles, are changed depending on the number of obstacles, position
and size in addition to the joints limits. In this paper, Table 1 shows the
modeling and simulation. These restrictions impact and limit movement of
manipulator and additionally split up the workspace into a reachable region
and an inaccessible one.

Table 1. Theoritical range of the two-link robotic arm case study.

Link No. Joint 1 Joint 2

Ranges 0 ≤ θ1θ1 ≤ 360 − 9 0
≤ θ2θ2 ≤ 90
To find all possible solutions, we should compute the FCS and that
done was by analyzing the environment’s points. The center point and the
diameter of the obstacles are the basic issues upon which the checking
function depends.
Moreover, there are three possible states for each point in the whole
environment. Firstly and as shown in Figure 1(b), there are two solutions to
each point, the elbow up solution and the elbow down solution.
But in the second state, there is just one solution, either elbow up or
elbow down, and in the third one, there is no solution because the point lies
out of the reachable area of the manipulator or on an obstacle, or the one or
two links collide with an obstacle [12].
The resulted areas are separated into three parts, the first one contain the
points which can be reached by the elbow up solution, and the second area
contains the points which can be reached by the elbow down solution, and
the third one represents the unreachable points that contain points outside
the accessible range of manipulator workspace, points in obstacle’s body
and points that make collision with the obstacles.
The biggest factor that affects the free Cartesian is the length of the arm
link [12].
126 Mechatronics and Robotics

PROPOSED METHOD: D* ALGORITHM BASED ON

PSO
After applying the D* algorithm to find the non-interactive path solution in
known dynamic environment and studying the produced path, it has been
proposed to make the environment interactive considering the position
and time of the obstacles to enhance the path planning. Employing the D*
algorithm in the known obstacle position and time dynamic environment,
to find the robot path needs to build a map with its limitation, start and
target locations and the obstacles locations and dimensions. The proposed
planning for the two-link robot arm in dynamic environment is shown
in Figure 2. In the interactive path solution of known dynamic environment,
in each move of the robot from node to node in each motion sample the free
Cartesian space (FCS) will be calculated. As a result, the movement range
of the robot will be defined and described accurately. It is a more intelligent
behavior having multiple FCS with each motion sample the obstacles move.
The new FCS, which was calculated, gives an indication to find the feasible
area for the robot to move on. This process takes time but gives an accurate
decision for the environment and high accuracy to the robot movement in
the environment.
Since the process is off-line and after constructing the environment, the
algorithm will start its work when the robot firstly stands on the target node
and starts the initial expand to its 8-connected neighbor. Each node will have
an initial cost function where the robot will choose the least cost function
node to move on and put it in the closed list.

Figure 2. Proposed planning for the two-Link robotic arm in dynamic environ-
ment.
The other nodes will go to the open list with taking into consideration
the obstacles having a huge cost function and sorting the initial locations
in the closed list. With each motion samples, the closed list will be updated
to catch the new obstacle location. This means, the closed list is changing
dynamically over motion samples to save the obstacle nodes in the current
motion sample only, and the process complete like that until reaching the
start node. When the search is completed the robot will move from the start
 Interactive Heuristic D* Path Planning Solution Based on ... 127

node with respect to the closed list instantaneously to have information

about free node for moving on towards the target.
f = g + h (7)
where f is objective function and g is the estimated cost from start point
while ℎ is the cost to the goal.
The velocity of the moving obstacle is calculated by:

where Δx represents the change of distance in X-axis, Δy represents

the change of distance in Y-axis and Ts represents the sample time which
assumed to be 0.01 sec.
After finding the D* path, the PSO optimization technique is used for
enhancing the path, removing sharp edges and making it the shortest because
the D* path has stairs shape sometimes. The principle of PSO technique
depends on generated specified number of particles in random positions in
the specified workspace, also the velocity of those particles is nominated
randomly, each particle has a memory that stores all the best position have
been visited before, in addition to the fitness in that position which has been
improved over time [13].
In each iteration of PSO technique, the posij and veli,jt vectors of particle
i is modified in each dimension j in order to lead the particle i toward either
the personal best vector ( pbestijj ) or the swarm’s best vector ( gbestij )

(8)
The position of each particle (the point that is chosen from D* path) is
updated by using the new velocity for that particle, according to equation
below:

(9)
where c1 and c2 are the cognitive coefficients (c1+c2≤4 ), and r1,j and r2,j
are random real numbers between [0, 1], the inertia weight w controls the
particle momentum.
128 Mechatronics and Robotics

The value of velij is clamped to the range to reduce the probability

of leaving the search space by the particle. If the search space is defined by the
bounds , then the value of velmax is typically set so that
, where 0.1 ≤ k ≤ 1.0.
A large inertia weight (w) eases the local when its value is large and
eases the global search when its value is small [14].
The first step to operate the PSO is to set the parameters of PSO (c1 = 1, c2 =
3, w = 1) which are found by a trial and error technique to reach the best results.
Now, the D* path enters the PSO work, which will represent the D* path as the
search space and choose the optimal points from it taking into account the general
shape for the D* path and without distorting the original form. After determining
the PSO search space which is the D* path matrix, the search process begins with
defining the number of via points (the points that is chosen from D* path) that
search for the new points according to the search space dimension and the PSO
parameters. These via points are chosen randomly from the search area (D* path)
and will represent the position parameters in the PSO equation.
The cubic spline interpolation equation has been used to connect the
best chosen points from the PSO optimization technique to generate the
corresponding smoothed path. The cost function of PSO, which is the path
length, is computed. Then the elementary velocity is set, and updating the
local and global parameters are the next step. This procedure is reiterated
until the given points that are looked over the path are finished. The next
step is updating the velocity and position equations, for converging the via
points into the best cost, and also updating the local and global parameters.
The procedure is repeated for the number of specified populations and the
entire process is repeated for a number of specified iterations. The global
values that construct the shortest path are then allocated. The entire process
flowchart is represented in Figure 3. After getting the smoothed path, the
robot will move also interactively with position and time on the new PSO
path.

SIMULATION RESULT

First Environment
Here we have applied the two-Link robot arm as a case study. The arm length
is equal to the half of the environment with 50 cm for each link and the arm
has 360 degrees. The obstacle motion behavior is shown in Figure 4. Since
 Interactive Heuristic D* Path Planning Solution Based on ... 129

the multi FCS is applied as clarified in Figure 5, where the column (5(a))
represents the elbow down FCS analysis at some selected motion samples;
1, 4, 5, 7, and 8, while in the column (5(b)) and (5(c)) represents the elbow
up FCS analysis and the gathering between elbow up and elbow down FCS
analysis respectively at the same motion samples. In each FCS figure the
colored area (blue, red or green) is represented the free and safe area for
moving and the white area is represented the unreachable area. Therefore, it
has calculated the elbow up and elbow down FCS to enable the robot from
making decision which free space is the suitable for moving.

Figure 3. Proposed D* via Point flowchart for position and time consideration
dynamic environment.

Figure 4. Obstacle motion behavior.

After computing these calculations the robot arm could move more
freely because it could detect with each FCS the free area to move
130 Mechatronics and Robotics

on. Figure 6shows the two-link robot arm D* path environment where part
(a) represents the beginning of the planning while the end-effector is on the
start point, part (b) is the arm location after 8 motion sample from planning
process. At part (c) the arm is return backward at motion sample 15 to avoid
collision, part (d) and

Figure 5. FCS for the D* path and PSO path where (a) represents elbow down,
(b) represents elbow up and (c) represents gathering between elbow up and
elbow down.
 Interactive Heuristic D* Path Planning Solution Based on ... 131

Figure 6. The Two-link robot arm result for the D* path.

(e) showed the planning after 21 and 36 motion sample respectively and
after 38 motion sample the arm is returned backward to evade the obstacle
then it reached the goal point safely as shown in part (g) and (h). Where
the process is off-line, the PSO optimization technique is used to get rid
of sharp edge and shorten the path as shown in Figure 7(a), where the blue
line represents the D* path and the black line represents the PSO path.
Also, Figure 7(b) clarified that the D* path based on PSO decreased by
3.2846 cm long from the original path. Figure 8 shows the D* based on PSO
path planning process from the initial state in part (a) where the end-effector
is on the start location to the part (f) where the process has completed by
reaching goal location. At part (b) where the motion sample equal 7 the
arm has restarted backward to safe itself from collision. In addition, Figure
9 and Figure 10 represent the two link robot arm joint angles and change in
these angles for the optimized path respectively. The same procedure will be
done for the second environment.
These environments featured the interactive style motion where the
robot should make a decision and must be stopped in some situations if it
is needed, in this case, the stop state and the backward motion guarantee a
collision free path. Table 2 clarifies the environment information.

Second Environment
For more verification, a second more complicated environment with
four moving obstacles in such different way in comparing with the first
environment and more close to the robot links. Two from these obstacles
132 Mechatronics and Robotics

move in a square motion behavior as in Figure 4. Since the multi FCS is

applied as clarified in Figure 11, where the column (11(a)) represents the
elbow down FCS analysis at some selected motion samples; 1, 5, 7, 8, and
11, while in the columns (11(b)) and (11(c)) represents the elbow up FCS
analysis and the gathering between elbow up and elbow down FCS analysis
respectively at the same motion samples.

Figure 7. (a) Final D* and PSO path and (b) the path cost function.

Figure 8. Two-link robot arm result for the D* based on PSO path.
In each FCS figures the colored area (blue, red or green) indicates the
free and safe area for moving robot links while the white area indicates the
unreachable area.
 Interactive Heuristic D* Path Planning Solution Based on ... 133

Figure 9. The two-link robot arm joint angles for the optimized path.

Figure 10. The change in joint angles for the optimized path.

Table 2. Environments information.

Environment 1 2
Obstacles num- 3 4
ber
Population 50 50
iteration 100 100
D* path 72.4264 cm 78.2132 cm
PSO path 69.6865 cm 73.8408 cm
Therefore, the robot can take a proper decision for selecting the next
points in a free and intelligent path inside the permitted motion area. Figure
12 shows the two-link robot arm D* path environment where part (a)
represents the beginning of the planning while the end-effector is on the
start point, part (b) shows the arm location after 11 motion samples, part (c)
indicates that the arm is returning backward case at the motion sample 21 to
avoid collision, part (d) and
134 Mechatronics and Robotics

Figure 11. FCS for the D* path and PSO path for the second environment where
(a) represents elbow down, (b) represents elbow up and (c) represents gathering
between elbow up and elbow down.

Figure 12. The two-link robot arm result for the D* path for the second envi-
ronment.
 Interactive Heuristic D* Path Planning Solution Based on ... 135

part (e) show the planning after 33 and 42 motion samples respectively,
while after 46 motion samples, the arm is returned backward again to evade
the closest obstacle before reaching the goal point safely as shown in part
(g) and (h).Because of the known position and time motion behavior of the
obstacles, the process of the path planning can be made off-line, therefore the
PSO optimization technique can plays an important role for finalizing this
process and finding the best possible shortest and optimal path to get rid of the
heuristic sharp path edges. Both of the D* and PSO paths are shown in Figure
13(a), where the blue line represents the D* path and the black line represents
the PSO path. Also, Figure 13(b) represents the optimization cost function
progress and proves that the final D* based on PSO path length was shorter by
4.3724 cm than the original path. Figure 14 confirms the path safety validation
by showing clearly all the robot configurations during motion according to
the optimized D* based on PSO path from the initial point as in part (a) to
the goal point as in part (f). In addition, Figure 15 and Figure 16 represent
the two link robot arm joint angles and the changes in these angles during
the optimized path respectively. Also, Table 2 clarifies the environment
information.In comparison with the nearest research literature [4] that try to
solve the same problem for mobile robot using the D* heuristic method, our
proposed method has different results because of our original modification
that applied on D* algorithm which guarantees the interactivity and safety of
the path and because of finalizing the path using PSO optimization technique.

CONCLUSION
This paper proposed an optimized heuristic approach for solving the
problem of two-link robot arm path planning in known position and time
changes dynamic environment. This approach utilizes the use of heuristic
D* algorithm for finding the best path and the use of PSO optimization
technique for getting the final optimal path.

Figure 13. (a) Final D* and PSO path and (b) the path cost function for the
second environment.
136 Mechatronics and Robotics

Figure 14. The two-link robot arm result for the D* based on PSO path for the
second environment.

Figure 15. The two-link robot arm joint angles for the optimized path of the
second environment.
 Interactive Heuristic D* Path Planning Solution Based on ... 137

Figure 16. The change in joint angles for the optimized path of the second
environment
Also, a modification on the D* algorithm has been done by adding a
stop case and return backward case to ensure the path solution interactivity
response to the obstacles position and time continuous changes during robot
motion. From this approach, it was possible to get an intelligent path in
spite of the environmental difficulties where a completely collision free path
could be found at the expense of optimality. Therefore, the PSO optimization
technique has been used to enhance the final path by removing the sharp
edges until reaching path optimality. The result clearly proves the advantage
of using this proposed methodology since the robotic arm has preserved the
intelligence interactive behavior even after applying PSO Technique. The
length of the final path was smart, smooth, safe, and optimal, and traditional
D* algorithm could not find such a path. Our proposed method has a
new approach of solution that leads to different results since it keeps the
motion behavior of the robot to be intelligent and interactive with dynamic
environment changes. The future directions for developing this proposed
method include utilizing other intelligent methods and theories.]

Conflicts of Interest
The authors declare no conflicts of interest regarding the publication of this
paper.
138 Mechatronics and Robotics

REFERENCES
1. Miao, H., Tian, Y.-C. and Feng ,Y.M. (2013) Robot Path Planning in
Dynamic Environments using a Simulated Annealing Based Approach.
Faculty of Science and Technology, Queensland University of
Technology, Brisbane.
2. Raheem, F.A. and Bader, M.M. (2017) Development of Modified Path
Planning Algorithm Using Artificial Potential Field (APF) Based on
PSO for Factors Optimization. American Scientific Research Journal
for Engineering, Technology, and Science, 37, 316-328.
3. Fujimura, K. (1991) Motion Planning in Dynamic Environments.
Springer-Verlag, Tokyo. https://doi.org/10.1007/978-4-431-68165-6
4. Stentz, A. (1995) The Focused D* Algorithm for Real-Time
Replanning. Proceedings of the International Joint Conference on
Artificial Intelligence, Montreal, 20-25 August 1995, 1652-1659.
5. Shen, Y.S. and Huper, K. (2004) Optimal Joint Trajectory Planning for
Manipulator Robot Performing Constrained Motion Tasks. Department
of Information Engineering, Australian National University, National
ICT Australia Limited, Canberra.
6. Khelchandra, T. and Huang, J. (2015) Robot Motion Planning
with Neuro-Genetic- Fuzzy Approach in Dynamic Environment.
International Journal of Information Technology, 21, 1-23.
7. Ni, J.J., Wu, W.B., Shen, J.R. and Fan, X.N. (2014) An Improved
VFF Approach for Robot Path Planning in Unknown and Dynamic
Environments. Mathematical Problems in Engineering, 2014, Article
ID: 461237. https://doi.org/10.1155/2014/461237
8. Asadi, B. (2015) Single and Dual Arm Manipulator Motion Planning
Library. IEEE/RSJ International Conference on Intelligent Robots and
Systems (IROS), Hamburg, 28 September-2 October 2015.
9. Vemula, A. (2017) Safe and Efficient Navigation in Dynamic
Environments. Carnegie Mellon University, Pittsburgh.
10. Spong, M.W., Hutchinson, S. and Vidyasagar, M. (2005) Robot
Modeling and Control. 1st Edition, Wiley, Hoboken.
11. Craig, J.J. (2005) Introduction to Robotics Mechanics and Control. 3rd
Edition, Pearson Prentice Hall Pearson Education, Inc., Upper Saddle
River, NJ.
12. Sadiq, A.T., Raheem, F.A. and Abbas, N.A.F. (2017) Optimal Trajectory
Planning of 2-DOF Robot Arm Using the Integration of PSO Based on
 Interactive Heuristic D* Path Planning Solution Based on ... 139

D* Algorithm and Cubic Polynomial Equation. The First International

Conference for Engineering Researches, Baghdad, March 2017, 458-
467.
13. Boonyaritdachochai, P., et al. (2010) Optimal Congestion Management
in an Electricity Market Using Particle Swarm Optimization with
Time-Varying Acceleration Coefficients. Computers & Mathematics
with Applications, in press.
14. Poli, R., Kennedy, J. and Blackwell, T. (2007) Particle Swarm
Optimization: An Overview. Swarm Intelligence, 1, 33-57.
 CHAPTER

8
Optimal Task Placement of a Serial
Robot Manipulator for Manipulability
and Mechanical Power Optimization

Rogério Rodrigues dos Santos1, Valder Steffen, Jr.1, Sezimária de

Fátima Pereira Saramago2
School of Mechanical Engineering, Federal University of Uberlândia,
1

Uberlândia, Brazil
Faculty of Mathematics, Federal University of Uberlândia, Uberlândia,
2

Brazil

ABSTRACT
Power consumption and accuracy are main aspects to be taken into account
in the movement executed by high performance robots. The first aspect is

Citation:- R. Santos, V. Steffen and S. Saramago, “Optimal Task Placement of a Se-

rial Robot Manipulator for Manipulability and Mechanical Power Optimization,”
Intelligent Information Management, Vol. 2 No. 9, 2010, pp. 512-525. doi: 10.4236/
iim.2010.29061.
Copyright: © 2010 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
142 Mechatronics and Robotics

important from the economical point of view, while the second is requested to
satisfy technical specifications. Aiming at increasing the robot performance,
a strategy that maximizes the manipulator accuracy and minimizes the
mechanical power consumption is considered in this work. The end-effector
is constrained to follow a predefined path during the optimal task positioning.
The proposed strategy defines a relation between mechanical power and
manipulability as a key element of the manipulator analysis, establishing a
performance index for a rigid body transformation. This transformation is
used to compute the optimal task positioning through the optimization of a
multicriteria objective function. Numerical simulations regarding a serial
robot manipulator demonstrate the viability of the proposed methodology.
Keywords: Optimal task placement, Optimal robot path planning,
Multicriteria optimization

INTRODUCTION
Minimization of production costs and maximization of productivity are
some of the major objectives of industrial automation. In this scenario, serial
robot manipulators have been proved to be very useful tools.
Due to the augmenting use of serial robot manipulators to perform a
number of tasks in industry, requirements concerning higher precision,
improved productivity, reduced costs, and better manufacturing quality
become very important. To effectively explore all the potential of the robot,
its path planning is a subject of major concern.
There are several works in the literature dedicated to different approaches
concerning this subject. Regarding the off-line path planning approach, in
Saramago [1] a solution of moving a robot manipulator with minimum cost
along a specified geometric path in the presence of moving obstacles is
presented. The optimal traveling time and the minimum mechanical energy
of the actuators are considered together to form a multiobjective function to
be minimized along the process.
In Santos [2] a strategy to determine the trajectory for a defined movement
taking into account the requirements of torque, velocity, operation time and
robot positioning along the movement is proposed. The analysis is performed
as an off-line path planning through spline interpolation techniques in the
manipulator joint space. In the paper of Chiddarwar [3] an off-line path
planning for coordinated manipulation is proposed. The swept sphere
volume technique is used to model multiple robots and static obstacles. In
 Optimal Task Placement of a Serial Robot Manipulator for ... 143

Santos [4] the trajectory of two manipulators while manipulating a single

object in a collaborative task and the object placement are written as an
optimization problem. End-effector positioning and torque requirements
are considered together in an optimal control formulation. Accuracy and
energy consumption are improved during the path planning. Also, the
flexibility effects of manipulators working in a vertical workspace are taken
into account, and joint limits are considered in a box-constrained objective
function to ensure the movement feasibility at the optimal configuration.
In von Stryk [5] an optimal control path-planning strategy for which
the dynamics of the robot and the total traveling time are considered as
objectives to be optimized is proposed. In Santos [6] an optimal control
strategy that considers the presence of moving obstacles in the workspace
is presented.
The problem of time-optimal control along a specified path has been
investigated for several authors (Bobrow [7], Slotine [8]). In Constantinescu
[9] a smooth time optimal strategy for serial robot manipulators is presented.
The paper of Seraji [10] addressed a geometric approach to determine
the appropriate base location from which a robot can reach a target point.
The work of Zeghloul [11] presented some kinematic performance criteria
for the optimal placement of robots, and proposed a general optimization
method to determine the placement of manipulators automatically.
In Feddema [12], an algorithm for determining the optimal placement
of a robotic manipulator for minimum time coordinated motion is proposed.
In Park [13] a study that characterizes a set of desired goal positions and a
left-invariant distance metric parametrized by length scale is studied.
The paper of Chirikjian [14] proposed several metrics to be used to
generate interpolated sequences of motions of a solid body. In Martinez
[15] an analysis regarding the metric of a rigid displacement obtained from
the kinematic mapping is presented and other metrics for the set of spatial
and planar displacements are proposed. The work of Tabarah [16] proposed
an optimization strategy for determining the optimal base for a given
manipulator, and two cooperating manipulators, following a continuous path
with constant velocity. They considered the constant velocity as a constraint.
A strategy for velocity minimization is proposed by Zhang [17], based on
a neural network solver. A kinematic planning scheme is reformulated as a
quadratic problem that is solved by using a real-time algorithm, and applied
to a Puma 560 robotic arm.
144 Mechatronics and Robotics

In Abdel-Malek [18] a strategy regarding the placement of a robot

manipulator aiming at reaching a specified set of target points is described.
The placement of a serial manipulator in a given workspace is achieved
by defining the position and orientation of the manipulator’s base with
respect to a fixed reference frame. The strategy is based on characterizing
the robot placement by adjusting a constrained cost function that represents
the workspace to a set of target points.
With the aim of increasing productivity in the path following, industrial
robotic applications have been addressed in the literature by determining
path-constrained time-optimal motions, by taking into account the torque
limits of the actuators. In these formulations, the joint actuator torques are
considered as controlled inputs and the open loop control schemes result in
bang-bang or bang-singular-bang controls (Bobrow [7], Chen [19], Shiller
[20]). The paper of Xidias [21] proposed the generalization of the task
scheduling problem for articulated robots in a constrained 2D environment.
An algorithm for the optimum collision-free movement is proposed.
However, the increase of the speed may result (in some cases) in
increasing the robot end-effector vibration or decreasing its accuracy. This is
observed due to several factors, such as the bang-bang nature of the control,
joint flexibility, joint friction, gear arrangement, or even a combination of
them.
Vibration is not desired because it can both degrade the system
performance and decrease the actuator lifetime. A related and negative effect
of vibration is the poor accuracy of the system.
The present contribution discusses an alternative way to improve the
overall performance of serial robot manipulators. The use of manipulability
and mechanical power related indexes to achieve the best task placement is
proposed.
As the basic idea of manipulability consists in describing directions
in the task or joint space that maximizes the ratio between some measure
of effort in joint space and a measure of performance in task space, the
present contribution consists in the optimal positioning of the task to be
performed, which results in the maximization of the robot manipulability
and the minimization of the required mechanical power.
This will lead to accuracy maximization and power consumption
minimization, since manipulability can be understood as an ability
measurement (Fu [22]) and mechanical power is related to power
consumption (Saramago [1]).
 Optimal Task Placement of a Serial Robot Manipulator for ... 145

In the present paper both kinematics and dynamics features are considered
in the proposed task placement strategy. From the best of the authors’
knowledge, this problem was not previously addressed in the literature,
despite its importance and applicability.
The position of the end-effector is fixed according to a Cartesian
reference system, and described as a sequence of sucessive positions. The
optimization process applies a translational and rotational transformation
matrix with respect to a given reference frame. As a result, an optimal
sequence of kinematical robot positions is determined.
The outline of the paper is the following. In Section 2, a review about
manipulability is presented. In Section 3, the mechanical power concept is
discussed.
A strategy to describe the task and its positioning in the design space
is proposed in Section 4. Additionally in this section the design variables
are defined. In Section 5 the multicriteria programming formulation is
presented. In Section 6, the optimization strategy is established and the
numerical experiments are performed in Section 7. Finally, the conclusions
are drawn in Section 8.

MANIPULABILITY
Some factors should be taken into account when the posture of the
manipulator in the workspace is determined for performing a given task
during operation. An important factor is the ease of arbitrarily changing the
position and orientation of the end-effector at the tip of the manipulator.
As an approach for evaluating quantitatively the ability of manipulators
from the viewpoint of kinematics, the concepts of manipulability ellipsoid
and manipulability measure (Yoshikawa [23]) are presented.
Consider a manipulator with n degrees of freedom. The joint variables are
denoted by an n-dimensional vector, q = [q1, q2, ..., qn]T. An m-dimensional
vector r = [r1, r2, ..., rm]T, (m ≤ n) describes the position and orientation of
the end-effector.
The relation between the velocity vector v corresponding to r and the
joint velocity is given by:

(1)
where J(q) is the Jacobian matrix, computed from the kinematics
description of the robot.
146 Mechatronics and Robotics

The set of all end-effector velocities v that are realizable by joint velocities
such that the Euclidean norm of satisfies defines an ellipsoid in the
m-dimensional Euclidean space. In the direction of the major axis of the
ellipsoid, the end-effector can move at high speed, and in the direction of the
minor axis, the end-effector can move only at low speed. Also, the larger the
ellipsoid is, the faster the end-effector can move.
Since this ellipsoid represents an ability of manipulation, it is called the
manipulability ellipsoid.
The principal axes of the ellipsoid can be found by making use of the
singular-value decomposition of J(q).
One of the representative measures for the ability of manipulation
derived from the manipulability ellipsoid is the volume of the ellipsoid.
Since this volume is proportional to the eigenvalues of the Jacobian, it can
be seen as a representative measure.
Therefore, the manipulability measure, f1, for the manipulator
configuration, q, is defined as

(2)
Generally f1 ≥ 0 holds, and f1 = 0 if and only if rank J(q) < m.
There is a direct relation between singular configuration and
manipulability (through the Jacobian concept). According to this relationship,
it can be shown that the larger the manipulability measure, the larger the
ability of avoiding singular configurations.

MECHANICAL POWER
The consideration of the dynamics behavior of a serial robot manipulator is
of great importance for its path planning. This information allows a detailed
analysis and consequently the development of a precise control specification.
Many efficient schemes have been proposed to model the dynamics of
rigid multibody mechanical systems (Richard [24], Mata [25], Mata [26]).
The dynamics model can be obtained explicitly through algebraically
computation, or numerically through iterative computation.
The techniques based on the Newton-Euler method starts from the
dynamics of all individual parts of the system. They look at the instantaneous
or infinitesimal aspects of the motion, using vector quantities such as
Cartesian velocities and forces.
 Optimal Task Placement of a Serial Robot Manipulator for ... 147

Alternatively, the Euler-Lagrange based methods starts from the kinetic and
potential energy of the entire system, by considering the states of the system
during a finite time interval. This approach works with scalar quantities,
namely the energies.
Independently of the approach used, at the end of the process the
generalized forces are determined:

(3)
where and are the joint position, velocity and acceleration,
respectively. M(q) is the joint space mass matrix and is the vector
of Coriolis and centrifugal forces. Vector g(q) is the vector of gravitational
forces and τ is the generalized force vector.
The energy that is necessary to move the robot is an important design
issue, because in real applications energy supply is limited and any energy
reduction leads to smaller operational costs. Due to the relationship that
exists between energy and force, the minimal energy can be estimated from
the generalized force, , that is associated to each joint i at the time instant
t0 ≥ t ≥ tf.
In the present contribution, for a manipulator with n degrees of freedom,
the interval between the initial time t0 and the final time tf is represented by
a set of N points and the computation is numerically carried out through
a recursive Newton-Euler formulation (Craig [27]). Then, the mechanical
power is used for design purposes as defined by:

(4)
This expression is representative of the phenomenon under study
because it considers both the kinematics and the dynamic aspects of the
trajectory, simultaneously (Saramago [1]).

TASK SPECIFICATION
The task is specified as a set S of N Cartesian points (x, y, z) and the
respective time step. Hence,

(5)
148 Mechatronics and Robotics

In the case of non-redundant manipulators there exist a finite number of

configurations that satisfies the end-effector positioning specification.
This path-following problem is a highly constrained task, and the inverse
kinematics problem will have four solutions or less.
Considering the need of movement continuity together with a time-fixed
specification at each step and a relative small step size, the point-to-point
path planning strategy may fail due to the complexity of the constraints.
Therefore, a task positioning optimization is proposed as an alternative
to increase the robot performance without changing the position of the task
reference points with respect to each other. In other words, all the points will
be moved simultaneously through a rigid body transformation to find the
best position from the point of view of the ability of the manipulator.
To achieve this goal, the use of a homogeneous transformation A is
proposed, which in this case is defined as

(6)
where β is a rotation angle in the z axis (given in rad) and δx , δy and
δz are translations along the reference axes (given in meters), respectively.
If desired, it is possible to extend this formulation to consider the rotation
with respect to the reference axes.
The physical meaning of considering a rotation angle with respect to x or
y is that the robot’s base plane will not be parallel to the ground’s reference
plane. In other words, the robot’s z axis (its height) is not normal to the
ground’s reference plane (xy plane). As positioning two planes according to
a precise angle may not be possible in a number of practical situations (even
when it is possible, it is not an easy procedure to be done, in general terms),
the rotation along the z axis only is considered, maintaining the robot’s base
plane parallel to the task’s base plane.
After defining the set S that describes the task, the inverse kinematics
with respect to each Cartesian point is computed.
The inverse kinematics computation can be achieved algebraically, by
considering the analytical model of the manipulator and its geometry, or
numerically. With the aim of presenting a general procedure, in the present
contribution the inverse kinematics problem is solved numerically through
 Optimal Task Placement of a Serial Robot Manipulator for ... 149

an inverse problem formulation. This approach results an efficient and

general procedure that works satisfactorily for manipulators exhibiting
arbitrary complexity (Santos [28]).
The obtained path planning presents an end-effector positioning error,
while the robot executes the path. The interest is focused on errors due to
task positioning, as the inverse kinematics can generally be solved with a
sufficient precision.
Once the objective is specified, it is necessary to define the optimization
domain, i.e., the design space. As the design variables are defined in a
theoretical space (defining a rigid body transformation) and the positioning
constraints are defined in the Cartesian space (the workspace), a constraint
parameter is included in the formulation.
Defining the theoretical positioning error at each point as

(7)
where Pbase, Pend and are the robot base point, the end-effector
point, and the kinematics representation of the manipulator, respectively.
The penalty parameter used as a constraint in the following optimization
formulation is given by

(8)
and represents the sum of all end-effector positioning errors during the
movement.
Equation (7) means that the end-effector error is considered null when
the accuracy is better than the one specified by ε (given in meters). Otherwise,
this error corresponds to the distance from the end-effector specified position
to the position obtained from the direct kinematics calculation.
In some robotic implementations it is necessary to plan the layout of
the workspace, i.e., it is required to locate the robot base in such a way that
dexterity is maximized at (or around) given targets. It is worth mentioning
that such cases are also covered by the present formulation. Since the optimal
design ((β, δx, δy, δz) in Equation (6)) means a coordinate transformation
of the end-effector path regarding the fixed robot base, it is also possible
to adjust the location of the robot base regarding the fixed end-effector
reference through the relation (-β, -δx, -δy, -δz).
150 Mechatronics and Robotics

MULTICRITERIA PROGRAMMING
In problems with multiple criteria, one deals with a design variable vector x,
which satisfies all constraints and makes the scalar performance index that
is calculated from the m components of an objective function vector f(x)
as small as possible. This goal can be achieved by the vector optimization
problem:

(9)
where is the function domain (the design space).
An important feature of such multiple criteria optimization problem is
that the optimizer has to deal with objective conflicts (Eschenauer [29]).
Other authors discuss the same topic by defining compromise programming,
because there is no unique solution to the problem (Vanderplaats [30]).

The Proposed Formulation

To consider together the manipulability and the total mechanical power, the
following scalar objective function is proposed

(10)
where f1 and f2 are given by Equations (2) and (4), respectively. The
reference values are obtained from the computation of the
manipulability and mechanical power sum at the initial task positioning.
Therefore, in the ideal case the objective function value is lower than
2, where (the manipulability was increased) and (the
mechanical power was decreased).
The specification of a reference value to both the manipulability measure
and the mechanical power are of paramount importance in the construction
of the multicriteria objective function.
The inverse kinematics error is included in the problem as an equality
constraint in the optimization formulation, through the expression sE = 0
(Equation (8)).
It should be pointed out that the proposed formulation is able to handle
the case in which the task is outside the robot workspace. This configuration
reflects in the inverse kinematics error, which is taken into account as a
constraint (Equation (8)) of the optimization formulation.
 Optimal Task Placement of a Serial Robot Manipulator for ... 151

Additionally, initial and final velocity specifications are considered as

(11)

(12)
for each joint j in the initial and final times t0 and tf, respectively. To
represent a rest-to-rest motion, zero initial and final velocities are considered
above.

OPTIMIZATION STRATEGY
An important feature of the proposed analysis is the computation of highly
nonlinear equations. Despite the good performance of classical nonlinear
programming techniques, the global optimal design is hardly reached due to
the existence of local minima in the design space.
With the aim of improving the robustness of the optimization process, the
use of a global methodology is considered. The tunneling algorithm (Levy
[31,32]) is a heuristic methodology designed to find the global minimum of
a function. It is composed of a sequence of cycles, each cycle consists of two
phases: a minimization phase having the purpose of lowering the current
function value, and a tunneling phase that is devoted to finding a new initial
point (other than the last minimum found) for the next minimization phase.
This algorithm was first introduced by Levy [31], and the name derives from
its graphic interpretation.
In summary, the computation evolves according to the following phases:
a) Minimization phase: Given an initial point q0, the optimization
procedure computes a local minimum q* of f(q). At the end, it is
considered that a local minimum is found.
b) Tunneling phase: Given q* found above, since it is a local
minimum, there exists at least one , such that

(13)
In other words, there exists q0 Z = {q Ω – {q*} | f(q) ≤ f(q*) }. To
move from q* to q along the tunneling phase, it is defined a new initial point
0

q = q* + δ, q Ω that is used in the auxiliary function

(14)
152 Mechatronics and Robotics

which has a pole in q* , for a sufficient large value of η By computing

both phases iteratively, the sequence of local minima may lead to the desired
global minimum..
Different values for η are suggested for being used in Equation (14)
to avoid undesirable points and prevent the search algorithm to fail (Levy
[31,32]). In the present case the stop criterion is five consecutive iterations
without further improvement in the minimal objective value.
The minimization phase is performed by using a constrained
formulation, where the objective function is given by Equation (10) and the
constraint function is defined by Equation (8). The velocity specification
(Equations (11) and (12)) is achieved through the determination of a suitable
interpolation polynomial function using the joint position coordinates.
In Equation (14), during the tunneling phase, only the unconstrained
formulation given by Equation (10) is considered. This unconstrained
nonlinear programming problem is solved by using the BFGS variable metric
method (Luenberger [33], Vanderplaats [30]). The constrained formulation
of the minimization phase (Equations (10), (11) and (12)) is solved by using
the modified method of the feasible directions (Vanderplaats [34]).
The algorithms are implemented in FORTRAN by using the optimization
library DOT (Vanderplaats [35]).

NUMERICAL SIMULATIONS
The numerical simulations performed to illustrate the presented methodology
use two robot models, namely the Scara and the Puma 560 robot manipulators.
These two types of serial robot manipulators are classical configurations in
the field. The case studies have been proposed due to their complexity as
compared to those normally found in the literature. The following results
considers only optimal profiles whose sum of the end-effector error is zero
(Equation (8)), given ε = 0.001 m (Equation (7)). This means that the inverse
kinematics solution is found for all positions with the required precision.
Different tolerance values may result from different optimal profiles.
As different values for the weighting factor k lead to different
optimization problems (Equation (10)), the value of the global minimum is
normalized according to the following equation

(15)
 Optimal Task Placement of a Serial Robot Manipulator for ... 153

This performance index enables comparing the improvement of each

objective without the weighting effect, and is used in the subsequent analysis.
In summary, the global optimization for each weighting factor (k = 0, 0.1,
0.2, 0.3, ..., 1) is performed according to Equation (10) and the comparison
among the optimal designs obtained is performed by using Equation (15).
Thus, eleven global optimizations for each task have been carried out. The
maximum inverse kinematics error was defined as being lower than ε =
0.001 m in Equation (7).
A Scara manipulator was used in the first and second numerical
experiments. The third numerical experiment refers to a Puma 560
manipulator. By considering the ability of describing a complex path, the
data for a set of Cartesian specification points (Equation (5)) is determined
through a cubic spline interpolation of the parametric coordinates of the
desired path. Therefore, all the values are obtained from a suited set of
Cartesian points.
The path for each experiment is arbitrarily defined with the aim of
exploring the efficiency of the methodology in performing a complex task.
All the obtained results consider only configurations for which inverse
kinematics errors are lower than ε, i.e., sE = 0 (Equation (8)).
Each value k in Equation (10) defines a new optimization problem.
To compare the values given by different optimization formulations, the
normalization of the results, despite the weighting factor, is taken into
account as established by f4 (Equation (15)).

First Experiment
In this experiment, a subset of 7 reference points is defined: (0, 0.4, 0, –0.3),
(1, 0.4, –0.1, –0.3), (2, 0.2, –0.1, –0.3), (3, 0.2, –0.2, –0.3), (4, 0.4, –0.25,
–0.3), (5, 0.4, –0.2, –0.3) and (6, 0.2, –0.25, –0.3). Comparing these values
with Equation (5), the initial task is inside a rectangle defined by 0.2 ≤ x
≤ 0.4, –0.3 ≤ y ≤ –0.1 and z = –0.3. The corresponding traveling time is
defined by 0 ≤ t ≤ 6. These data are computed by interpolating cubic splines
for t, x, y and z, respectively. The interpolated data encompass a set of N
= 27 points (Equation (5)) that describes the path-following problem, as
presented in Figure 1.
At the original position, the values of the design variables are: β= 0 rad, δx =
0 m, δy = 0 m, and δz = 0 m. These values are used as an initial guess for the
optimization process (Equation (10)).
154 Mechatronics and Robotics

The global minimum values obtained are presented in Figure 2.

The optimal values of manipulability and mechanical power indexes
demonstrate that there is no clear dependence between the variables. Good
results are provided for k = 0.2, 0.5, 0.6, and 0.9.

Figure 1. The proposed path-following problem from two different perspec-

tives.
Details about the minimal performance index, k = 0.9, are presented
in Table 1. The overall performance was improved 28%, while the minimal
manipulability and the mechanical power were improved 62% and 18%,
respectively.
Some of the values provided by other weighting factors may result in
further improvement of the manipulability or mechanical power, separately.
However, the present configuration is a global minimum of the performance
index f4.
The improvement obtained during the tunneling process is shown
in Figure 3.
After the 6th iteration of the tunneling process, no further improvement
was achieved during the next five consecutive iterations. This behavior
defines the stop criteria for the global optimization search. The above result
highlights the good performance of the method, as satisfactory results are
obtained after a small number of iterations.
Figure 4 shows the robot performing the path following task and Figure
5 presents the robot optimal positioning, according to the optimal design
variable values given in Table 1.
 Optimal Task Placement of a Serial Robot Manipulator for ... 155

Figure 2. Optimal values obtained by using different weighting factors.

Figure 3. Performance index along the tunneling iterations (for k = 0.9).

Table 1. Initial and optimal parameter values (for k = 0.9).

Figure 4. Robot performing the path following task.

The joint coordinates at each time step are obtained from inverse kinematics
computation. Once the task is considered as a set of consecutive Cartesian
points with continuous transition between them, the corresponding joint
156 Mechatronics and Robotics

coordinates presents a smooth behavior. Figure 6 shows the initial and

optimal joint coordinates of the first three links and the corresponding
manipulability index.
Once the inverse kinematics is iteratively computed for each end-effector
positioning, it is possible to include the obstacle avoidance feature in the
above strategy. The corresponding procedure for the inverse kinematics
computation is addressed in Santos [28], for example.

Second Experiment
In this experiment a subset of 8 reference points is defined: (0, 0.2, 0,
–0.3), (1, 0.4, –0.1, –0.3), (2, 0.4, –0.05, –0.3), (3, 0.2, –0.1, –0.3), (4,
0.25, –0.15, –0.3), (5, 0.3, –0.17, –0.3), (6, 0.27, –0.13, –0.3) and (7, 0.26,
–0.2, –0.3). By comparing these values with Equation (10), the initial task
is inside a rectangle defined by 0.2 ≤ x ≤ 0.4, –0.2 ≤ y ≤ 0 and z = –0.3.
The corresponding traveling time is defined by 0 ≤ t ≤ 7. These data are
computed by interpolating cubic splines for t, x, y and z, respectively. The
interpolated data encompass a set of N = 54 points

Figure 5. Proposed (---) and optimal (––) paths from two different perspectives.
(Equation (5)) that describes the path-following problem, as presented
in Figure 7” target=”_self”> Figure 7.
 Optimal Task Placement of a Serial Robot Manipulator for ... 157

Figure 6. Joint coordinates and the corresponding manipulability index (Scara

manipulator).

Figure 7. The proposed path-following problem from two different perspec-

tives.
At the original position, the values of the design variables are: β = 0 rad, δx =
0 m, δy = 0 m, and δz = 0 m. These values are used as an initial guess for the
158 Mechatronics and Robotics

optimization process (Equation (10)). Global optimizations were performed

for eleven different values of the weighting factor k, (k = 0, 0.1, 0.2, ..., 1)
(see Equation (10)).
The minimum values obtained by the optimization process are presented
in Figure 8” target=”_self”> Figure 8.
In this experiment good results are provided by k = 0.3, 0.4, and 0.7.
Details about the minimal performance index, k = 0.7, are presented
in Table 2. The overall performance was improved 17%, while the minimal
manipulability and the mechanical power were improved 46% and 3%,
respectively.
Eventually, the values provided by other weighting factors may result in
further improvement either of the manipulability or the mechanical power,
separately. However, the above configuration is a global minimum of the
performance index f4.
The design improvement achieved by the tunneling process is shown
in Figure 9” target=”_self”> Figure 9.
After iteration 18 of the tunneling process, no further improvement was
achieved for five consecutive iterations, defining the stop criterion of the
global optimization search.
Figure 10 shows the robot performing the path-following task and Figure
11 presents the optimal positioning, according to the optimal design values
presented in Table 2.

Figure 8. Optimal values obtained by using different weighting factors.

 Optimal Task Placement of a Serial Robot Manipulator for ... 159

Figure 9. Performance index along the tunneling iterations (for k = 0.7).

Table 2. Initial and optimal parameter values (for k = 0.7).

Figure 10. Robot performing the path-following task.

The joint coordinates at each time step are obtained by inverse kinematics
computation. As the discrete set of points defining the task (Equation
(5)) is supposed to describe a continuous path, the transition between
joint coordinates is computed through cubic spline interpolation. Figure
12 shows the initial and optimal joint coordinates for the first three links and
the corresponding manipulability index.
For the examples presented above, the task was defined as being parallel
to the xy plane, i.e., the path is parallel to the ground.

Third Experiment
In this experiment, a subset of 7 reference points is defined: (0, 0.4, 0.1, 0), (1,
0.4, 0, 0), (2, 0.4, 0, –0.4), (3, 0.4, –0.3, –0.4), (4, 0.4, –0.4, 0), (5, 0.4, –0.3,
0), and (6, 0.4, –0.4, –0.4). By comparing these values with Equation (5), the
initial task is inside a rectangle defined by x = 0.4, –0.4 ≤ y ≤ 0.1, and –0.4
160 Mechatronics and Robotics

≤ z ≤ 0. The corresponding traveling time is defined by 0 ≤ t ≤ 6. These data

are computed by interpolating cubic splines for t, x, y and z, respectively.
The interpolated data encompass a set of N = 54 points (Equation (5)) that
describe the path-following problem, as presented in Figure 13.

Figure 11. Proposed (---) and optimal (––) paths from two different perspec-
tives.

Figure 12. Joint coordinates and the corresponding manipulability index (Scara
manipulator).
 Optimal Task Placement of a Serial Robot Manipulator for ... 161

Figure 13. The proposed path-following problem from two different perspec-
tives.
At the original position, the values of the design variables are: β = 0 rad,
δx = 0 m, δy = 0 m, and δz = 0 m. These values are used as an initial guess for
the optimization process (Equation (10)).
The optimal values of both the manipulability and the mechanical power
indexes demonstrate that there is no significant difference among most of
the results presented. The best values are provided by k = 0.4 and 0.5.
Details about the minimal performance index, k = 0.5, are presented
in Table 3” target=”_self”> Table 3. The overall performance was improved
40%, while the minimal manipulability and the mechanical power were
improved 64% and 41%, respectively.
Some of the values obtained by using other weighting factors may
result in further improvement of either the manipulability or the mechanical
power, separately. However, the above configuration is a global minimum
of the performance index f4.
The performance provided by the tunneling process is shown in Figure
14.
After the 5th iteration of the Tunneling process, no further improvement
was achieved for the next five conescutive iterations. This behavior defines
the stop criteria for the global optimization search. In this case, it is possible
to see that good results are obtained in a few iterations.
162 Mechatronics and Robotics

Figure 15 shows the robot performing the path following task and Figure
16 presents the optimal positioning, according to the optimal design variable
values given in Table 3.

Table 3. Initial and optimal parameter values (for k = 0.5).

Figure 14. Performance index along the tunneling iterations (for k = 0.5).

Figure 15. Robot performing the path-following task.

Figure 16. Proposed (---) and optimal (––) paths from two different perspec-
tives.
 Optimal Task Placement of a Serial Robot Manipulator for ... 163

Figure 17. Joint coordinates and the corresponding manipulability index (Puma
560 manipulator).
The joint coordinates at each time step are obtained by inverse kinematics
computation. As the task is considered as a set of consecutive Cartesian
points with continuous transition, the corresponding joint coordinates
presents a smooth behavior. Figure 17 shows the initial and optimal joint
coordinates for the first three links and the corresponding manipulability
index.
The previous example considered a task initially defined as being
parallel to the xz plane, i.e., the path is parallel to a vertical plane. It can be
noticed that the optimal design corresponds to rotating the task about 0.391
rad anticlockwise with respect to the reference frame.

CONCLUSIONS
In this work the path-following problem of robot manipulators was
addressed. An approach to increase the manipulability while decreasing
dynamics requirements was proposed by using optimization techniques.
The concept of manipulability was revisited. The basic idea behind
manipulability analysis consists in describing directions either in the task
space or the joint space that optimize the ratio between a measure of effort in
the joint space and a measure of performance in the task space. In this work
a classical formulation to describe the relation between the two spaces (task
and joint spaces) was considered.
Since the end-effector presents various performances due to a number of
factors (control specification, joint flexibility, joint friction, etc.), a general
methodology to reduce high torques and to increase manipulability was
proposed. The corresponding equation considers the overall performance
of the entire manipulator, as the determinant of the Jacobian is proportional
164 Mechatronics and Robotics

to the product of its singular values, and the mechanical power index is
obtained by a sum involving all the joints. The methodology also includes the
manipulator positioning error analysis, through the definition of constraint
functions in the optimization model.
The serial manipulator usually performs differently in different zones of
the workspace. Considering also the cases for which different joint velocity
and torque profiles may result different performance levels, a better position
into the workspace (in the sense that the performance indexes are optimized)
for performing a specified task was determined.The optimization problem
was based in a global search heuristic, improving the robustness of the process
with respect to the initial guess.Numerical experiments showed that the task
repositioning successfully increases the manipulability and reduces the total
mechanical power, simultaneously. As changes in the task positioning may
result different associated manipulability indexes and different joint velocity
and torque profiles, the methodology searches for an optimal way of using
the robot possibilities (as described by the performance indexes) by finding
an optimal position into the workspace to perform the task.
In some practical cases, the manipulated objects cannot be displaced
because of physical constraints such as weight or geometry. Even in such
cases the proposed strategy is suitable to be applied by changing the position
of the manipulator itself. The corresponding changes in the robot’s base are
given by the optimal profile by using the opposite sign of the optimal values,
i.e., (-β, -δx, -δy, -δz). It is also possible to adjust the analysis to consider
different objectives, as the total torque instead of the mechanical power, for
example. The steps to be followed remain the same.
In summary, the main contributions of this paper are the simultaneous
analysis of both the required mechanical power and the manipulability
(dynamics and kinematics aspects), together with a global deterministic
heuristic to achieve the optimal task placement.
As the proposed methodology is suitable for serial manipulators of
arbitrary complexity, the authors believe that the present methodology
represents a contribution concerning general optimal robot path planning
and design.

ACKNOWLEDGEMENTS
The authors are thankful to the agencies FAPEMIG and CNPq (INCT-EIE)
for partially funding this work.
 Optimal Task Placement of a Serial Robot Manipulator for ... 165

REFERENCES
1. S. F. P. Saramago and V. Steffen Jr., “Optimization of the Trajectory
Planning of Robot Manipulators Taking Into-Account the Dynamics
of the System,” Mechanism and Machine Theory, Vol. 33, 1998, pp.
883-894.
2. R. R. D. Santos, V. Steffen Jr. and S. F. P. Saramago, “Robot Path
Planning: Avoiding Obstacles,” In: 18th International Congress of
Mechanical Engineering, Ouro Preto-MG, Brazil, 2005.
3. S. S. Chiddarwar and N. R. Babu, “Offline Decoupled Path Planning
Approach for Effective Coordination of Multiple Robots,” Robotica,
Vol. 28, No. 4, 2010, pp. 477- 491.
4. R. R. D. Santos, V. Steffen Jr. and S. F. P. Saramago, “Optimal Path
Planning and Task Adjustment for Cooperative Flexible Manipulators,”
ABCM Symposium Series in Mechatronics, Associação Brasileira de
Engenharia e Ciências Mecânicas, ABCM, Vol. 3, 2008, pp. 236-245.
5. O. Von Stryk and M. Schlemmer, “Optimal Control of the Industrial
Robot Manutec r3,” Computational Optimal Control, International
Series of Numerical Mathematics, Vol. 115, 1994, pp. 367-382.
6. R. R. D. Santos, V. Steffen Jr. and S. F. P. Saramago, “Robot Path Planning
in a Constrained Workspace by Using Optimal Control Techniques,”
In: III European Conference on Computational Mechanics, Lisbon,
Portugal, 2006, pp. 159-177.
7. J. E. Bobrow, S. Dubowsky and J. S. Gibson, “Time-Optimal Control
of Robotic Manipulators along Specified Paths,” The Intemational
Journal of Robotics Research, Vol. 4, 1995, pp. 3-17.
8. J. E. Slotine and H. S. Yang, “Improving the Efficiency of Time-Optimal
Path-Following,” IEEE Transaction on Robotics and Automation, Vol.
5, No. 1, 1989, pp. 118- 124.
9. [9] D. Constantinescu and E. A. Croft, “Smooth and TimeOptimal
Trajectory Planning for Industrial Manipulators along Specified Paths,”
Journal of Robotic Systems, Vol. 17, 2000, pp. 233-249.
10. H. Seraji, “Reachability Analysis for Base Placement in Mobile
Manipulator,” Journal of Robotic Systems, Vol. 12, 1995, pp. 29-43.
11. S. Zeghloul and J. A. Pamanes, “Multi-Criteria Optimal Placement of
Robots in Constrained Environments,” Robotica, Vol. 11, No. 2, 1993,
pp. 105-110.
12. J. T. Feddema, “Kinematically Optimal Robot Placement for Minimum
166 Mechatronics and Robotics

Time Coordinated Motion,” Proceedings of the IEEE International

Conference of Robotics and Automation, Minneapolis, 1996, pp.
3395-3400.
13. F. C. Park, “Distance Metrics on the Rigid-Body Motions with
Applications to Mechanicsm Design,” ASME Journal of Mechanical
Design, Vol. 117, 1995, pp. 48-54.
14. G. S. Chirikjian and S. Zhou, “Metrics on Motion and Deformation of
Solid Models,” ASME Journal of Mechanical Design, Vol. 120, 1998,
pp. 252-261.
15. J. M. R. Martinez and J. Duffy, “On the Metrics of Rigid Body
Displacement for Infinite and Finite Bodies,” ASME Journal of
Mechanical Design, Vol. 117, No. 1, 1995, pp. 41-47.
16. B. Tabarah, B. Benhabib, R. Fenton and R. Cohen, “Cycle-Time
Optimization for Single-Arm and Two-Arm Robots Performing
Continuous Path Operation,” 21st Biennial Mechanisms Conference,
Chicago IL, 1990, pp. 401-406.
17. Y. Zhang, and K. Li, “Bi-Criteria Velocity Minimization of Robot
Manipulators Using LVI-Based Primal-Dual Neural Network and
Illustrated Via PUMA560 Robot Arm,” Robotica, Vol. 28, No. 4, 2010,
pp. 525-537.
18. K. Abdel-Malek and W. Yu, “On the Placement of Serial Manipulator,”
Proceedings of DETC00 2000 ASME Design Engineering Technical
Conferences, Baltimore, MD, 2000, pp. 1-8.
19. Y. Chen and A. A. Desrochers, “Structure of Minimum-Time Control
Law for Robotic Manipulators with Constrained Paths,” IEEE
International Conference on Robot and Automat, Scottsdale, Az, USA,
1989, pp. 971- 976.
20. Z. Shiller and H. H. Lu, “Computation of Path Constrained Time
Optimal Motions with Dynamic Singularities,” ASME Journal of
Dynamic Systems, Measurement, and Control, Vol. 114, 1992, pp. 34-
40.
21. E. K. Xidias, P. T. Zacharia and N. A. Aspragathos, “Time-Optimal Task
Scheduling for Articulated Manipulators in Environments Cluttered
with Obstacles,” Robotica, Vol. 28, No. 3, 2010, pp. 427-440.
22. K. S. Fu, R. C. Gonzales and C. S. G. Lee, “Robo-Tics: Control,
Sensing, Vision and Intelligence,” McGraw-Hill, New York, 1987.
23. T. Yoshikawa, “Manipulability of Robot Mechanisms,” The
 Optimal Task Placement of a Serial Robot Manipulator for ... 167

International Jouranl Robotics Research, Vol. 4, 1985, pp. 3-9.

24. M. J. Richard and C. M. Gosselin, “A Survey of Simulation Programs
for the Analysis of Mechanical Systems,” Mathematics and Computers
in Simulation, Vol. 35, No. 2, 1993, pp. 103-121.
25. V. Mata, F. Benimeli, N. Farhat and A. Valera, “Dynamic Parameter
Identification in Industrial Robots Considering Physical Feasibility,”
Journal of Advanced Robotics, Vol. 19, No. 1, pp. 101-120.
26. V. Mata, S. Provenzano, F. Valero and J. I. Cuadrado, “Serial Robot
Dynamics Algorithms for Moderately Large Numbers of Joints,”
Mechanism and Machine Theory, Vol. 37, No. 8, pp. 739-755.
27. J. J. Craig, “Introduction to Robotics: Mechanics & Control,” 2nd
Edition, Reading, MA: Addison-Wesley, 1989.
28. R. R. D. Santos, V. Steffen Jr. and S. F. P. Saramago, “Solving the Inverse
Kinematics Problem through Performance Index Optimization,” In:
XXVI Iberian Latin-American Congress on Computational Methods
in Engineering, Guarapari-ES, 1985.
29. H. Eschenauer, J. Koski and A. Osyczka, “Multicriteria Design
Optimization,” Springer-Verlag, Berlin, 1990
30. G. N. Vanderplaats, “Numerical Optimization Techniques for
Engineering Design,” 3rd Edition, VR & D Inc., 2001.
31. A. V. Levy and S. Gomez, “The Tunneling Method Applied to Global
Optimization. Numerical Optimization,” (Ed. P. T. Boggs) R. H.
Byrd and R. B. Schnabel, Eds., Society for Industrial and Applied
Mathematics, 1985, pp. 213-244.
32. A. V. Levy and A. Montalvo, “The Tunneling Algorithm for the Global
Minimization of Functions,” The SIAM Journal on Scientific and
Statistical Computing, Vol. 6, No. 1, 1985, pp. 15-29.
33. D. G. Luenberger, “Linear and Nonlinear Programming,” 2nd Edition,
Addison-Wesley, USA, 1984.
34. G. N. Vanderplaats, “An Efficient Feasible Direction Algorithm for
Design Synthesis,” American Institute of Aeronautics and Astronautics
Journal, Vol. 22, No. 11, 1984, pp. 1633-1640.
35. G. N. Vanderplaats, “DOT-Design Optimization Tools Program,”
Vanderplaats Research & Development, Inc., Colorado Springs, 1995.
 CHAPTER

9
Kinematical Analysis and Simulation of
High-Speed Plate Carrying Manipulator
Based on Matlab

Ke Wang, Jiping Zhou

School of Mechanical Engineering, University of Yangzhou, Yangzhou,
China

ABSTRACT
In order to construct the more effective kinematics method for industry, by
taking a high-speed plate handing robot as an example, the structure and
parameters of the robot linkages are analyzed, and the standard Denavit-
Hartenberg method is applied to establish the coordinates and the kinematic
equation of the linkages. Depending on the graphics and matrix calculation

Citation:- K. Wang and J. Zhou, “Kinematical Analysis and Simulation of High-Speed

Plate Carrying Manipulator Based on Matlab,” Engineering, Vol. 4 No. 12, 2012, pp.
850-856. doi: 10.4236/eng.2012.412108.
Copyright: © 2012 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
170 Mechatronics and Robotics

ability of Matlab especially including the Robotics Toolbox, the handling

robot has been modeled and its kinematics, inverse kinematics and the
trajectory planning have been simulated. Therefore, the correctness
of kinematic equation has been verified, meanwhile, the functions of
displacement, velocity, acceleration and trajectory of all the joints are also
obtained. In a further step, this has verified the validity of all the structure
parameters and provided a reliable basis for the theoretical research on the
design, dynamics analysis and trajectory planning of the manipulator control
system.
Keywords: High-Speed Plate Carrying Manipulator; D-H Method;
Robotics Toolbox; Trajectory Planning

INTRODUCTION
As a typical representative and a main technical means of information
technology and advanced manufacturing technology, the complete set of
automatic stamping processing line has become a high technology to which
the developed countries have paid much attention. And its development
level has become one of the most important standards to measure a nation’s
technical development. It has been extensively applied in the industry of
mechanical manufacturing, nuclear, aerospace, energy and transportation,
petroleum chemistry, building, electronic and etc. [1]. But, one key aspect of
the automatic stamping processing line is to develop a mechanical manipulator
with the characteristic of the high-speed and dynamic transmission.
At present, the sheet metal forming production line is operated at the
high productive rate of 15 - 25 SPM. Removing the time spent by the press
stamping operations, there is only about 2 s left to the robot manipulator. In
such a short period of time, the manipulator has to operate at a high speed
of 200 - 250 m/min, so that the operations such as loading, moving and
unloading can be fulfilled. And this has put forward higher requirements
to the mechanical structure, material friction characteristics, and structural
dynamic characteristics of the robot manipulator.
Taking a high-speed plate carrying manipulator as the research object,
this paper firstly analyzes the structure and connecting rod parameters; then
adopts the standard D-H method [2] to establish the kinematics equation;
and then discusses the positive and inverse kinematics algorithm, and the
trajectory planning problems; finally in the environment of Matlab, the
kinematics model is built to take kinematics simulation by using of Robotics
Toolbox [3]. In simulation process, we can not only directly observe the
 Kinematical Analysis and Simulation of High-Speed Plate Carrying ... 171

robot motion, but also get the required data in the graphic form. Therefore,
the virtual performance of the product can be tested in the conceptual design
stage, so as to improve the design performance, reduce design cost and
decrease product development time.

THE STRUCTURE DESIGN AND LINK PARAME-

TERS OF HIGH-SPEED PLATE CARRYING MANIPU-
LATOR

The Design Requirements of High-Speed Plate Carrying Ma-

nipulator

The Workplace of High-Speed Plate Carrying Manipulator

The traditional stamping processing method relies on a stand-alone
manual which is inefficient, inaccurate and insecurity. So it has already
become increasingly unsuited to the requirements of modern mass
production, especially when the requirements of the sheet metal processing
annual output exceeds thousands tons, this contradiction is more prominent.
The current stamping equipment is toward the trend of automation, sets
and online development. With the rapid development of China’s market
economy, especially in the coastal areas where the shortage of skilled workers
is severe, and labor costs rises sharply, many machinery manufacturers have
the urgent demand for manufacturing automation, and require machine and
equipment manufacturing industry to provide users with a complete set
of on-line technical services to improve production efficiency and reduce
labor costs. Therefore, the development of the technology of sheet metal
production line is even more important. One of the core technologies of
sheet metal stamping equipment line is the design of high-speed plate
carrying manipulator. As shown in Figure 1, it is high-speed plate carrying
manipulator in the application of stamping processing complete sets of
equipment on-line system.

The Operation Process of High-Speed Plate Carrying Manipula-

tor
According to the composition of the stamping process sets of on-line system
and the role of the high-speed plate carrying manipulator, the working cycle
of a manipulator contains nine action process, as shown in Figure 2:
172 Mechatronics and Robotics

1) From the point of origin, the left electromagnetic valve is

energized after pressing the start button, then the manipulator
moves to the left. It won’t stop until it encounters the left limit
switch.
2) Simultaneously the manipulator begins to drop after the decreased
electromagnetic valve opens, then, it won’t stop until it encounters
the lower limit switch.
3) At the same time the clamp electromagnetic valve is energized,
then the manipulator is clamped.
4) After the rise electromagnetic valve is energized, the manipulator
begins to rise, then it won’t stop until it encounters the rise limit
switch.
5) Simultaneously the manipulator moves to the right after the right
electromagnetic valve is energized, then, it won’t stop until it
encounters the right limit switch.
6) Simultaneously the manipulator begins to drop after the decreased
electromagnetic valve is energized, then, it won’t stop until it
encounters the lower limit switch.
7) At the same time the clamp electromagnetic valve is opened,
then, the manipulator is opened.
8) After the rise electromagnetic valve is energized, the manipulator
begins to rise, then, it won’t stop until it encounters the rise limit
switch.

Figure 1. The composition of the stamping processing sets of on-line system.

 Kinematical Analysis and Simulation of High-Speed Plate Carrying ... 173

Figure 2. The operation process of the manipulator.

9) Simultaneously the manipulator moves to the origin after the left
electromagnetic valve is energized, then, it won’t stop until it
encounters the left limit switch.
So far, the manipulator after 9 step action completes a cycle of its
movement, and then, it continues the cycle work.
Based on the above analysis, we have developed a three-dimensional
model of the manipulator as shown in Figure 3, which can be used to analyze
the dynamic characteristics of the manipulator.

The Design Parameters of High-Speed Plate Carrying Ma-

nipulator

D-H Transformation
In order to describe the translational and rotational relationships between
adjacent rods, Denavit and Hartenberg (1955) have proposed a matrix
method to establish the possessed coordinate system for each rod in the
linkage chain. This method is to establish a homogeneous transformation
matrix for each of the join bar coordinate, which represents the relationship
of the previous coordinate system of the bar, and the principle [4,5] is
detailed as follows:
OXYZ: A fixed reference coordinate system of the fixed coordinates, is
called the coordinate system.
OiXiYiZi: Fixed connected with the member bar of number I of the robot,
the origin of coordinate is at the center point of the joint of the I + 1th.
174 Mechatronics and Robotics

Identify and establish each coordinate system by following three rules.

1): Zi−1 axis along the motion shaft of the join of the i th.

Figure 3. The assembly model of manipulator, 1-body, 2- feeder, 3-beam,

4-slipway, 5-balance cylinder, 6-servo motor, 7-X axis transmission system.
2): Xi axis vertical the axis of Zi and Zi−1 and point to the direction of
away from the Zi−1 axis.
3): Yi axis according to the requirements of the right hand coordinate
system to establish.
Meanwhile the notation of D-H of the rigid bar depends on the four
parameters of the connecting rod.
The angle of two connecting rod : Xi−1→Xi around the corner of Zi−1.
The distance of two connecting rod : Xi−1→Xi along the distance of
Zi−1.
The length of the connecting rod : Zi−1→Zi along the distance of Xi−1.
The torsion angle of the connecting rod: : Zi−1→Zi around the corner
of Xi.
For rotational joints, is the joint variables, the rest are joint parameters
(remain unchanged); for mobile joints, is the joint variables, the rest are
joint parameters.
 Kinematical Analysis and Simulation of High-Speed Plate Carrying ... 175

The Design Coordinate of Manipulator

High-speed plate carrying manipulator is mainly composed of vertical pillars
(body), horizontal arm (beam), sliding table (Y axis transmission system)
and base. Horizontal arm mounted on the machine body level can move
around and can move up and down along the vertical pillars. By reference
to the high-speed plate carrying manipulator in Figure 3, we establish the
coordinate system of manipulator in Figure 4 based on the above analysis.
According to the D-H parameters method, four parameters are defined
for each link: the connecting rod angle , the distance of two connecting rod
, the length of the connecting rod , the connecting rod torsion Angle , the
D-H parameters of manipulator as is shown in Table 1.

Figure 4. The D-H coordinate of high-speed plate carrying manipulator.

Table 1. The D-H parameters and joint variables of manipulator.

176 Mechatronics and Robotics

THE KINEMATICS SIMULATION ALGORITHM OF

MANIPULATOR

The Kinematics of Manipulator

The so-called kinematics problem [4, 5] is given to the manipulator of each
joint variable, and then obtains the position and posture of the end of the
actuator, and its essence is to establish the kinematics equations. For the
solution of kinematics equation, this paper uses homogeneous transformation
matrix to describe the coordinate system of i relative to the position and
pose of the coordinate system of the i – 1, this is the general formula for
transformation matrix .

(1)
Now, putting the D-H parameters and joint variables of manipulator
substitution of the Table 1 in (1), we get the homogeneous transformation
matrix of the i coordinate system relative to the position and posture of the
base coordinate system, expressed as:
(2)
Especially, when , can be obtained, it determines the position
and posture of the end of the manipulator relative to the base coordinate
system, the matrix of T can be expressed as:

(3)
Note:

The Inverse Kinematics Issue of Manipulator

The inverse kinematics issue of manipulator is defined as follows: with the
known the position and posture of the end of the actuator, we need to solve
 Kinematical Analysis and Simulation of High-Speed Plate Carrying ... 177

the variables of each joint of the manipulator, here in our case the variables
are and .
The target point of the movement of the manipulator center is
. By using the transitional joint, the manipulator’s center
point can coincide with the target point, and the target point can be expressed
as: , then we can establish the following equation:

(4)
Combining (3) and (4) and solving the kinematics equation, we can get
the following equation.

(5)

(6)

THE KINEMATICS SIMULATION OF MANIPULA-

TOR

Use Link and Robot Functions to Establish the Manipulator’s

Object
1) Before the manipulators simulation, firstly input the manipulator’s
parameters by calling the function Link:

(note: represent the variables of , , θi and di respectively;

“sigma” represents the joint types: 0 for the rotational joint, and 1 for the
transitional joint; “standard” is for the standard D-H parameters. The function
robot is used to create a manipulator object by using the Link function in the
format of Robot (Link…). The commands for creating the high-speed plate
carrying manipulators is expressed as follows:

2) We can immediately see the three-dimensional view of the

manipulator which can be used in the form manually for driving
through the slider in the chart to drive the movement of the
178 Mechatronics and Robotics

manipulator, which is just like the actual control of the manipulator

[3]. As shown in Figure 5, it has brought great convenience for
manipulator’s teaching and training.

The Test of Kinematic Model

In the Figure 5, by moving each slide in the bar to move the manipulator,
the first joint movement can change the height of the manipulator in the
vertical direction. The second joint can change the length of the manipulator
in horizontal direction, So that we can obtain different position and posture.
By adjusting the slides and the joint variables, we can get the approximate
model corresponding to the actual structure in Figure 3, as shown in Figure
6.
In order to verify the correctness of the kinematics Equation (3), the
geometric parameters and the joint variables of each rod of the manipulator
are put into the kinematic equations to solve the setting position and posture
of the coordinate system of the end link rod in relation to the base coordinate
system. And then the corresponding coordinate values are input to the
manipulator’s trajectory planning, and the actual end position and posture
information are compared with the results from solving the equation. Two
groups of joint variable values are randomly chosen:

The error is very small through the comparative analysis of the actual
value and set point, which explains that the kinematics equation and the
model are reliable and consistent.

The Simulation of Trajectory Planning

According to the requirements of the task, trajectory planning will designate
in advance the robot’s operating procedure and action process. The
simulation method could describe in details the movement process of the
industrial robot [6-10]. Planning can be made in the space of both joint and
operation. We would introduce two main motions [4,5]: 1) PTP, Point-To-
Point motion; 2) CP, Continuous-Path. With regard to continuous-path, not
only the initial and final points of the manipulator need to be set, but also
some other points between the two points (called path points) which must
move along some specific path (path constraint) need to be indicated.
The design of the high-speed plate carrying manipulator adopts a point-to-
 Kinematical Analysis and Simulation of High-Speed Plate Carrying ... 179

point trajectory planning. Set A as the starting point, moving to B where

a task is finished, and then set B as a starting point, moving to C where a
certain task is completed, and then set C as a starting point, moving to point
C to complete the preset tasks, and continue to move, and so on. Here, the
move from point A to point B and from point B to point C does not exist any
point between them, and there is not any requirement for the known path of
the movement. Therefore, we can take this above planning as PTP planning.

Figure 5. The three-dimensional diagram and slide block control chart of the
high-speed plate carrying manipulator.

Figure 6. A state of high-speed plate carrying manipulator’s three-dimensional

diagram and slide block control chart.

The starting point is set to be ending point , and

between the two points the initial and final speeds are both zero, with the
time of movement time: , the relative program as follows:

By conducting the above program, we can observe the whole process of

the high-speed plate carrying manipulator moving from Figures 5 and 6. We
can also draw, by way of the function of
180 Mechatronics and Robotics

each joint’s displacement curve, as shown in Figure 7(a) and (b). (Note:
q represents the displacement, i represents the joint Numbers). Also we can
draw speed curve as shown in Figures 7(c)-(d), acceleration curve as shown
in Figures 7(e)-(f), through the corresponding functions:

The Analysis Based on the Results of Simulation

From Figures 7(a) and (b), it can be seen that the displacement of the
mobile joint 1 gradually changes from zero to 0.62832 m, movable joint 2
displacement from zero change to 1.2566 m; from Figures 7(c) and (d) we
(a) The displacement curve of joint 1
b) The displacement curve of joint 2
(c) The velocity curve of joint 1
(d) The velocity curve of joint 2
(e) The acceleration curve of joint 1
(f) The acceleration curve of joint 2
 Kinematical Analysis and Simulation of High-Speed Plate Carrying ... 181

Figure 7. Displacement, velocity and acceleration curves of high-speed plate

carrying manipulator.
can obtain that the initial and final speeds of the joints 1, 2 are zero, the
maximum velocity appearing at the intermediate time t = 1 s; from Figures
7(e)-(f), it can be seen, the initial and final acceleration speeds are zeros, two
maximum values in motion, with one positive and the other negative. The
manipulator’s displacement curve is smooth, and the curve of velocity and
acceleration is continuous. Therefore, it is concluded that: in the process, the
manipulator runs smoothly, and the whole body vibrates in a normal scope,
which thus explains the validity of the design.

CONCLUSIONS
Ke Wang, Jiping Zhou The paper carries out a kinematical analysis of a
highspeed plate carrying manipulator which is designed by using the module
function of Robotics Toolbox in the Matlab. Based on this, we have achieved
the following conclusions in three aspects: 1) the structure and parameters
of the plate carrying manipulator have been designed; 2) the kinematics
model has been established with the standard D-H method, and its positive
and inverse kinematics have been analyzed; 3) by using Robotics Toolbox
in Matlab, the kinematics model has been set up in order to make the study
of the motions of manipulator easier and more apparent, and to conduct a
kinematics simulation for the manipulator for obtaining the displacement,
velocity and acceleration curve of each manipulator joint. Thus, this proves
the validity of the design of manipulator structure parameters, as well
as provides the reliable basis for the theoretical analysis in terms of the
design, dynamics analysis, and trajectory planning of control system of the
manipulator.
182 Mechatronics and Robotics

REFERENCES
1. Q. Wang, X. J. Hu and L. Z. Li, “The Kinematic Analysis and Simulation
of the Humanoid Welding Manipulator Based on the MATLAB,”
Materials Science and Engineering College of Hefei University of
Technology, 2011.   
2. J. J. Craig, “Introduction to Robotics,” Mechanical Industry Press,
Beijing, 2006.   
3. P. I. Corke, “A Robotics Toolbox for MATLAB,” IEEE
Robotics and Automation Magazine, Vol. 3, No. 1, 1996, pp. 24-
32. doi:10.1109/100.486658   
4. X. S. Jiang, “Introduction to Robotics,” Liaoning Science and
Technology Press, Shenyang, 1994.
5. Z. X. Cai, “The Robotics,” Tsinghua University Press, Beijing, 2000.
6. J. X. Luo and G. Q. Hu, “The Kinematic Simulation Research of Robot
Based on the MATLAB,” Journal of Xiamen University, JCR Science
Edition, Vol. 44, No. 5, 2005, pp. 640-644.
7. J. Han and L. Hao, “Trajectory Planning and Simulation of Robot in
Joint Coordinate System,” Journal of Nanjing University of Science
and Technology, Vol. 24, No. 6, 2000, pp. 540-543.
8. Z. X. Wang and W. X. Fan, “The Kinematics Analysis and Simulation
of Industrial Robot Based on the Matlab,” Mechanical and Electrical
Engineering, Vol. 29, No. 1, 2012, pp. 33-37.
9. The MathWorks Inc., “Matlab the Language of Technical Computing
Version 6,” The MathWorks Inc., 2002.
10. P. I. Corke, “Robotics Toolbox for Matlab (Release 7.1) [EB/OL],”
2002. http://www.cat.csiro.au/cmst/staff/pic/robot
 CHAPTER

10
Flexible Impact Force
Sensor

Chen Liu, Yukio Fujimoto, Yoshikazu Tanaka

Department of Transportation and Environmental Engineering, Hiroshima
University, Higashi-Hiroshima, Japan

ABSTRACT
When human body is punched by boxing glove, both body surface and
boxing glove deform in a complex shape. The purpose of this study is to
develop a flexible sensor that can be used in such interface. Firstly, several
mechanical phenomena, which are the cause of the error signal of the sensor,
are discussed. These are the influences of out-of-plane bending deformation,

Citation:- Liu, C. , Fujimoto, Y. and Tanaka, Y. (2014), “Flexible Impact Force Sen-
sor”. Journal of Sensor Technology, 4, 66-80. doi: 10.4236/jst.2014.42008.
Copyright: © 2011 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
184 Mechatronics and Robotics

shear force caused by rubbing, shear force caused by the Poisson’s effect
of contact material, and the lateral compressive force caused by the
overhanging deformation of flexible material. Then, a prototype sensor
that can eliminate the error factors of these is developed. The sensor is
a distribution type impact sensor in which sixteen sensor elements are
arranged in a 4 × 4 matrix. Punching experiments using a boxing glove are
carried out by installing the sensor on the load cell, on the concrete wall and
on the sandbag. From the experiment, it is found that the impact force can
be measured with good accuracy by using the sensor. Despite the fact that
the sensor has inadequate distribution number of sensor elements, the sensor
structure meets mechanical requirements for the flexible impact sensor.
Keywords: Flexible Sensor, Impact Force, Boxing, Sandbag,
Distribution Type Sensor, Human Body

INTRODUCTION
When a boxer punches the human body, both body surface and boxing
glove deform in a complex shape. At present, the accurate measurement of
impact force in such interface is said to be difficult. The reason is because
the complex deformation and the complex force distribution are induced in
the flexible interface and they affect the sensor output. Up to now, many
kinds of tactile sensors such as resistance type [1], piezo-resistance type [2]
, capacitance type [3] and light intensity type [4] have been developed. Most
of them are assuming to mount the sensor on the robot members. However,
it has not been studied much about the influences of the deformation and
the associated force distribution at the flexible interface on the accuracy
of the sensor output. The reason of this is most of the tactile sensors are
intended to use under a small force level and a small deformation level on
the installation surface [5] [6].
By the way, it has been reported that peak impact force of about 4 kN
is generated when a strong boxer punches the face of crash test dummy [7].
When the hitting target is sandbag, the interface of glove and sandbag will
be deformed more greatly. A suitable sensor which can measure the impact
force in such condition is not seen at present. The purpose of this study
is to provide a sensor that can measure the impact force accurately at the
interface of large deformation and large force.
In this study, firstly, difficulty of force measurement in flexible interface
is discussed based on the experiments using a sheet type sensor (Chapter
2). Then, some mechanical factors necessary to achieve the flexible sensor
 Flexible Impact Force Sensor 185

are discussed (Chapter 3). These factors are the out-of-plane bending
deformation of sensor, the shear force caused by rubbing on sensor surface,
the shear force generated by contact friction, and the transverse compression
caused by the overhanging deformation of the flexible material.
In Chapter 4, as a sensor that can overcome the mechanical problems in
flexible interface, a flexible sensor of distribution type is developed. In Chapter
5, boxing punching tests are carried out using a prototype sensor by installing
the sensor on the sponge rubber sheet which is placed on the load cell or the
concrete wall. Further, the punching test is carried out by installing the sensor on
a sandbag. In particular, the present study is focused to achieve a sensor structure
that is hard to cause trouble under a big impact force like boxing punching.

MEASUREMENT OF FORCE ON SOFT MATERIAL

SURFACE BY USING A SHEET-TYPE SENSOR
Spherical Surface Pressing on the Flexible Material Surface
As a simple case, spherical pressing experiment of a sensor installed
on the flexible material surface was carried out. Figure 1(a) shows a sheet
type sensor made from piezoelectric film (PVDF) of which both surfaces are
covered by silicon rubber sheets [8]. When compressive force F is applied
as shown in Figure 1(b), the rubber sheets shrink in the thickness direction
and also be extended in the horizontal direction by the Poisson’s effect.
Since the elastic modulus of PVDF is much larger than that of rubber sheet,
the PVDF is stretched by the rubber elongation in the horizontal direction.
Then, the output signal proportional to the compressive force F is generated
in the sensor.Figure 2(a) shows a sheet type flexible sensor with slit notches
around the sensor edge. A single sheet type sensor is not possible to follow
the out-of-plane deformation. In order to make out-of-plane deformation
possible, one way is to process the slit notches around the sensor edge. The
slit notches are expected to get a similar result with “preventing wrinkles
of thin paper to wrap an apple”.Figure 2(b) is the sensor covered by doubly
stretch cloth on both surfaces. A polyethylene sheet is sandwiched in between
the doubly stretch cloth. When the sensor with slit notches is directly placed
on the flexible material and pressed by curved surface object, the output
signal of the sensor is influenced by the contact friction between the sensor
and the contact object. The stretch cloth and the polyethylene sheet play a
role to reduce the influence of contact friction by slide effect.
Figure 3 shows the condition of compression test. Thick acryl plate was
186 Mechatronics and Robotics

fixed on the top of a load cell. Sponge rubber plate with 40 mm thickness was
placed on the acryl plate. The sensor was put on it and pressed by spherical
body with 50, 100 or 150 mm diameter. The experiment was carried out by
changing the compressive force.
Figure 4 compares the relationship between the peak force of the load
cell and the sensor output. A good linear relationship is seen in the figure.
Despite the change in the diameter of spherical body, the output gradient does
not change much. The dashed line in the figure shows the output gradient
under the condition that the sensor is placed directly on the flat acryl plate.
The sensor detects the normal force on the curved surface, but the load cell
detects the vertical component. The difference in gradient of the two straight
lines is considered to be related with the detection characteristics.

Fist Pressing, Palm Pressing and Fist Punching on the Surface

of Flexible Material
Next experiment was carried out under the condition that the contact
interface deforms in a complex shape. Figure 5 shows the conditions of fist
pressing, palm pressing and fist punching. The experiments were conducted
by changing the compressive force. Figure 6 shows the relationship between
the peak force of the load cell and the peak output of the sensor during
compression. A quite large scatter of plots is seen in the figure.

Figure 1. Sensor cross-section and its detection principle of compressive force.

 Flexible Impact Force Sensor 187

Figure 2. Sheet type flexible sensor. (a) Main body of sensor with slits in outer
edges; (b) Sensor covered by stretch cloth.

Figure 3. Spherical press of the sensor on the sponge rubber.

Figure 4. Sensor output vs. load cell output. Dashed line is the output gradient
of the sensor on flat plane surface..
188 Mechatronics and Robotics

Figure 5. Fist press, palm press and fist punch of the sensor placedon the sponge
rubber plate.

Figure 6. Relationship between the output peak of sensor and the peak force of
load cell.
In the experiment of Section 2.1, the loading side was a hard material
with simple contact surface and the receiving side was a flexible material.
On the other hand, the contact surface of fist or palm is more complex and
it changes in every trial. Further, rubbing force and inclined force are also
working. The cause of the scatter of the output gradient can be considered as
the influences of complex force distribution and the excessive out-of-plane
deformation occurring in the sensor. This result means that a single sheet-
type sensor is not suitable to achieve the measurement of impact force on the
flexible material surface.
Figure 7 classifies the condition of impact force measurement into
four levels, in view of the difficulties of measurement. Figure 7(a) is the
condition that a hard object with simple contact geometry collides. Figure
7(b) is the condition that a soft object with simple contact geometry
collides. In accordance with the deformation of soft object, in this condition,
unexpected shear forces (rubbing force and contact friction) work at the
 Flexible Impact Force Sensor 189

interface. Figure 7(c) is the condition that a hard object with complex contact
geometry collides. In this case, the out of-plane bending deformation of the
sensor and the distributions of shear forces at the interface become more
complex. Figure 7(d) is the condition that a soft object with complex contact
geometry collides. In this case, stress distribution and deformation becomes
most complicated. Boxing punching on sandbag surface is close to the
condition of Figure 7(d).

MECHANICAL ISSUES IN IMPACT FORCE MEAS-

UREMENT ON FLEXIBLE MATERIAL SURFACE
The objective of this research is to develop a sensor which can be used even
when both objects of loading side and load receiving side are soft and the
contact geometry is complicated. After repeating several experiments, the
authors realized that the discussions about the error signals are necessary for
the flexible sensor. They will be described in the following.

Influence of Out-of-Plane Bending Deformation of the Sensor

Element
Figure 8 shows a sensor element used in the flexible sensor. A pair of silicon
rubber sheet is adhered on the both surfaces of piezoelectric film (PVDF).

Figure 7. Difficulty of force measurement at flexible interface is classified

into four levels depending on the shape of the impact object. (a) Hard object
with simple contact geometry; (b) Soft object with simple contact geometry;
(c) Hard object with complex contact geometry; (d) Soft object with complex
contact geometry.

Figure 8. Error signal caused by the out-of-plane bending deformation of sen-

sor element. (b) and (c) represent the shift of the neutral axis by bending defor-
mation.
190 Mechatronics and Robotics

The bending neutral axis comes to the center of PVDF thickness,

theoretically, due to the vertically symmetric structure. Therefore, the
sensor output should not occur almost by bending deformation. However,
in practice, by variation in the thickness of the adhesive layer and etc., the
neutral axis is shifted slightly from the center of PVDF thickness [9]. Thus
result in the occurrence of tensile or compressive strain in PVDF (Figure
8(b)). If a resin sheet like a flexible substrate is adhered on one side surface
of the silicon rubber, the bending neutral axis is further shifted, and results
in greater strain (Figure 8(c)). It is extremely difficult to control the strain of
PVDF caused by bending deformation.

Influence of Rubbing Force on the Sensor Surface (Shear

Force Caused by Rubbing)
Figure 9 shows a schematic view of a sensor element placed on the floor, on
which a compressive force F on the left side and a rubbing force S on the
right side are acting. The vertical reaction force RF occurs below the F, and
the shear force RS against S occurs caused by the friction between floor and
sensor element. As the result, tensile strain is produced in the region between
the points F and S (shown by red arrow). The rubbing force S sometimes is
generated as the horizontal component of inclined force which is applied to
the sensor surface.

Influence of Contact Friction on the Sensor Surface (Shear

Force Caused by the Poisson’s Effect)
Figure 10(a) shows a sensor element sandwiched in between the metal
plates. Since the elastic modulus of silicon rubber is smaller than those of
PVDF and aluminum plate, the silicon rubber deforms into barrel-shape
when compressive force F acts.
Figure 10(b) shows a state in which a pair of sponge rubber sheet is
sandwiched between the silicon rubber and aluminum plate additionally and
applied a compressive force F. Since the elastic modulus of sponge rubber
is further smaller than that of silicon rubber, the sponge rubber deforms
into bigger barrel-shape. Even if the same F is applied, the silicon rubber is
stretched more in this condition. This means that the sensor output is affected
by the Young’s modulus of the contact material as well as its thickness.
As an example, Figure 11 shows the magnitude of sensor output in
accordance with the change of contact material. The output is adjusted as
1.0 when the sensor element is sandwiched directly between the aluminum
 Flexible Impact Force Sensor 191

plates. A circular sensor element with 50 mm diameter is used. The silicon

rubber on both sides of PVDF is

Figure 9. Example of the tensile strain (red arrow) generated in the sensor ele-
ment caused by rubbing force. F is compressive force, S is rubbing force and
RF and RS are the corresponding reaction forces.

Figure 10. Error signal caused by the difference of contact friction.

Figure 11. Variability of sensor signal caused by the difference of contact mate-
rial.
1 mm thickness. Sponge rubber sheet (T = 5 mm, 10 mm thick) or natural
rubber sheet (T = 1 mm) or cork plate (T = 2 mm) was sandwiched between
the sensor element and the aluminum plate. As the contact material becomes
softer, the sensor output increases as shown in the figure. In the case of thick
sponge rubber (T = 10 mm), the magnitude becomes several times compared
with the case of the aluminum plate.
192 Mechatronics and Robotics

Influence of Lateral Compression Caused by the Overhanging

Deformation of Flexible Material
Figure 12 shows the state that a soft material in large size enough to cover the
sensor element collides with the element. When the force F acts on the sensor
surface, impact force (solid curve) in compressive direction is produced as
shown in Figure 12(a) at first. However, by the succeeding deformation of
soft material, a part exceeding the area of the sensor element may press the
element from the lateral direction. By the transverse compressive force from
the surrounding, the sensor generates a signal in opposite direction as shown
by dashed curve of Figure 12(b). As a result, the sensor signal becomes the
sum of solid and dashed curve as shown in Figure 12(c).

Figure 12. Lateral compression phenomenon caused bythe overhanging defor-

mation of flexible material.
Figure 13 shows an example of such waveforms by the free drop test of
soccer ball.
A circular sensor element with 50 mm diameter was placed on the load
cell, and the ball was fallen onto the sensor. Figure 13 compares the sensor
output (red curve) and the load cell output (black curve). When bottom of
the ball collide the center of sensor, the sensor output is generated on the
negative side.
After that, it turns into the opposite side in the halfway, since the
overhanging part of deformed ball presses the sensor laterally.
Therefore, the sensor waveform becomes quite different from the load
cell waveform. Such lateral compression occurs as well when the sensor
bottom is in buried condition in soft material.
 Flexible Impact Force Sensor 193

FLEXIBLE SENSOR MADE FROM PLURAL SENSOR

ELEMENTS ARRANGED IN A MATRIX
In order to eliminate the above influences, in this study, we develop a sensor
element of rigid plate type. The sensor elements are distributed in a matrix,
and the seam portion of each sensor element is connected flexibly in order
to ensure the flexibility. Further, a new manner to acquire the sensor signal
without using electrical wiring is employed.

Prototype of Sensor Element

Figure 14 shows the sensor element with 30 mm square and the twenty
elements made for this research. At first, silicon rubber sheet with 1 mm
thickness was adhered on both surfaces of PVDF. Next, a circular hole
was drilled in the center of the silicon rubber, and the hole was filled with
conductive rubber.
The conductive rubber was designed to contact the electrode of PVDF.
Furthermore, aluminum plate was adhered on the silicon rubber on each
side. The upside and downside electrode of PVDF are connected with the
upper and lower aluminum plate respectively via conductive rubber. In order
to assure the electric contact between the PVDF and the aluminum plate, the
conductive rubber was pressed by screw. The reason of using 5 mm thick
aluminum plate is that metal plate with certain thickness is necessary in
order to prevent the bending deformation.
Each sensor element was placed on load cell and punched by hand
hammer (with felt cap) several times. Figure 15 shows examples of the
impact force waveforms. Solid curve is the sensor output and dashed curve is
the load cell output. Eventually, sixteen elements among them were selected
and used for the flexible sensor.
Figure 16 shows the output gradient of the chosen sensor elements.
The horizontal axis is the peak impact force obtained by load cell and the
vertical axis is the peak output of the sensor element. As shown in the figure,
load-bearing capacity of a single sensor element is more than 1.5 kN. If the
impact force is distributed to several sensor elements, then the load bearing
capacity is more than 10 kN.

Prototype of Flexible Impact Force Sensor

Figure 17(a) shows a flexible sensor in which sixteen sensor elements are
arranged in a 4 × 4 matrix. The sensor is covered with artificial leather.
194 Mechatronics and Robotics

Figure 13. Impact force waveform in which vertical and lateral compression is
mixed.

Figure 14. Configuration of the prototype sensor elements.

Figure 15. Example of impact force waveform of sensorelement (solid curve)

and load cell (dashed curve).
Its surrounding and the vertical and horizontal column are finished by
sewing. The total size is about 230 mm square. Figure 17(b) shows the
sensor inserted into conductive cloth bag for static shielding. Green cable is
 Flexible Impact Force Sensor 195

the ground wire. Piezoelectric sensors are prone to be influenced by static

electricity [9] [10] . In order to prevent the signal drift by static electricity, it
is preferable to use in the state of electrostatic shielding. Figure 17(c) shows
the state in holding by bending the sensor.

Figure 16. Relationship between sensor output and load cell output obtained by
the hammer punching test.

Figure 17. Prototype sensor. (a) Flexible sensor consisting of 4 × 4 sensor ele-
ments; (b) Conductive cloth bag; (c) Hold the sensor bend.
Even though the sensor element itself is a rigid body, the connecting
portions of the sensor elements can keep the flexibility.
Figure 18 shows the cross section view of the flexible sensor. The
upper aluminum plates of all sensor elements are in contact with the upper
conductive cloth. Further, the lower aluminum plates of all sensor elements
are in contact with the lower conductive cloth. In order to prevent short-
circuiting of the upper and lower conductive cloth, bilayer deer skin is
sandwiched in the thickness center of the sensor as the insulation sheet.
After sewing the intermediate portion of the sensor element and element,
each sensor element is in a state put into a pocket.
The electrical wires are connected to the lower and upper conductive
cloth respectively. In other words, the output signal of the sensor element
is obtained from the conductive cloth via upper and lower aluminum plate.
Because there is no electrical wiring to connect each sensor element, there
is no risk of electric disconnection.
196 Mechatronics and Robotics

It is obvious that the flexibility of the sensor increases as the increase of the
number of sensor elements distributed. If the sensor element is assumed as
“a plate placed on the elastic floor”, the bending moment reduces along with
the sensor dimension becomes smaller. Therefore, it is possible to reduce the
sensor thickness as the size of sensor element is small.

EXPERIMENT

Comparison of Sensor Output and Load Cell Output

Figure 19 shows the photos of boxing punching test. First, thick acryl plate
was fixed on the top of a load cell with 10 kN capacity. Sponge rubber sheet
with 40 mm thickness was placed on it. Then, the flexible sensor inserted
in the conductive cloth bag was placed on the sponge rubber and a student
punched it by boxing glove. Since the sponge rubber is pretty thick, the
central portion of the sensor is deformed concave by the punching.

Figure 18. Schematic view of sensor cross-section. Cover sheet is artificial

leather. Insulation sheet is bilayer deer skin.

Figure 19. Impact test of flexible sensor on the load cell. (a) Sponge rubber (t =
40 mm) put on the load cell; (b) Sensor placed on the sponge rubber; (c) Punch-
ing test by boxing glove.
 Flexible Impact Force Sensor 197

Also, similar experiment was conducted under the condition that the flexible
sensor was placed directly on the acryl plate.
Figure 20 shows an example of impact force waveform measured in the
above punching test. Solid and dashed curves indicate the experiments on
the sponge rubber and the acryl plate respectively. The red curve is sensor
output and the black curve is load cell output. The impact force waveform
of the sensor is in good agreement with that of load cell.
Figure 21 shows the relationship between output peak of the sensor and
output peak of the load cell. Regardless of the presence or absence of sponge
rubber, the sensor output shows almost same gradient. This means that the
sensor can measure the impact force in the flexible interface.
Figure 22 shows sponge sheet with 40 mm thickness overlaid on the
sponge rubber. The total thickness of the bilayer is 80 mm. The sensor was
placed on the sponge sheet and punched by boxing glove. In the figure, two
examples of the impact force waveform are compared with that of the load
cell output. As the sponge is so soft, the sensor was greatly deformed in a
concave shape. There is somewhat difference in the waveform of sensor and
load cell.
Figure 23 shows the sensor placed in bent condition on the wood
semicircular column with diameter 150 mm. Some positions of the sensor
were fixed with gum tape. Then, the sensor was punched by boxing glove.
The relationship between the load cell output and the sensor output are
indicated by black diamond. The red circle in the figure is the output gradient
of the experiment shown in Figure 22. These two gradients are substantially
coincident. Moreover, this output gradient is in good agreement with that
in Figure 21.

Punching Test of the Sensor Installed on Concrete Wall

Figure 24 shows the punching test by boxing glove on concrete wall. Firstly,
sponge rubber sheet (40 mm thickness) was pasted on the concrete wall. The
sensor was placed on the sponge rubber sheet and the several parts are fixed
on the wall with gum tape.
198 Mechatronics and Robotics

Figure 20. Impact force waveforms of the sensor and the load cell. Red curve
is sensor, black curve is load cell.

Figure 21. Comparison of the peak output between sensor and load cell.

Figure 22. Impact force waveform of sensor and load cell. Sponge is over-
laid on top of the sponge rubber.
 Flexible Impact Force Sensor 199

Figure 23. Relationship between sensor output and load cell output. Black dia-
mond: semicircular column. Red circle: sponge & sponge rubber sheet.

Figure 24. Boxing punching of the sensor installed on the 40 mm thick sponge
rubber pasted on concrete wall.
Four impact force waveforms weakly punched by student “U” and two
waveforms strongly punched by student “F” are shown in the figure. The
peak impact force by the strong punching is around 2 kN. The contact time
is about 30 - 40 msec.

Punching Test of the Sensor Installed on Sandbag

Figure 25 shows the experiment of boxing punching sensor was placed on
the surface of a sandbag (soft sandbag) of gymnasium and fixed with gum
tape. Compared with the punching on the concrete wall, the sandbag is softer
and a concave deformation is induced on sandbag surface by punching.
Figure 26 shows the impact force waveforms. The peak impact force
ranges 0.7 kN and 0.9 kN. The contact time is about 70 msec and this is
longer than that on the wall punching. In the figure, it is seen that the gradient
of impact force becomes gentle from the middle of unloading. This reason
is considered as follows. The sandbag swings rearward by the punching.
200 Mechatronics and Robotics

Therefore, the forward speed of the glove persists for a while even after
passing the peak load. Since the pull-back of the glove is delayed, the contact
between glove and sandbag is continued for a while. The difference of the
gradient of unloading segment is clear when compared to the waveform of
wall punching in Figure 24.

IMPROVEMENT OF THE SENSOR FLEXIBILITY

Sensor elements described above is 30 mm square and 12 mm thickness. To
improve the flexibility of the sensor, it is necessary to increase the distribution
density by reducing size and thickness of the sensor element. Figure
27(a) shows slightly small sized sensor elements with 25 mm square and 8
mm thickness. Figure 27(b) is a flexible sensor made from the six elements.

Figure 25. Boxing punching of the sensor installed on the sandbag.

Figure 26. Impact force waveform of the sensor installed on the soft sandbag
surface and punched by boxing glove.
 Flexible Impact Force Sensor 201

Figure 28(a) shows the sensor fixed to the instep of soccer shoes with cords
and rubber bands. Figure 28(b) shows the impact force waveforms of instep
kick of soccer ball. The strength of the kick is the extent to path to nearby
players. The peak impact force is about 1 kN and the impact time is about 12
- 15 msec. This contact time is close to the value shown in the reference [11].
From the experience, it seems possible to miniaturize the element size to
15 mm square and 5 mm thickness. When the element becomes smaller than
that, however, reconsideration of the manufacturing process is necessary. In
spite of the rough distribution density of the sensor elements, the sandbag
sensor was not much discomfort at the time of use. On the other hand, in
the soccer shoe sensor, the sensor weight and the lack of flexibility were
felt uncomfortable. “Flexibility” and “light weight” are thought to be more
important for the sensors installed on human body side.

CONCLUSIONS
This study is aimed to develop a flexible sensor which can be used in the
measurement of impact force in flexible interface. The boxing punching is
chosen as the measurement object. The results obtained are as follows.

Figure 27. (a) Sensor elements of 25 mm square and 8 mm thickness; (b) Flex-
ible sensor made fromsix sensor elements.

Figure 28. (a) Set up of the sensor on soccer shoe; (b) Impact force wave-
forms of instep kick of soccer ball.
202 Mechatronics and Robotics

1) A flexible sensor in which sixteen sensor elements are arranged

in a 4 × 4 matrix is developed. Each sensor element is a rigid
plate structure. The sensor elements are arranged at regular
intervals. Both entire surfaces of the sensor elements are covered
with artificial leather, and its surrounding and the vertical and
horizontal columns are finished by sewing. Flexibility of the
sensor is achieved by the flexible leather connections between
adjacent sensor elements;
2) Sum of the output signal of each sensor element is collected
through the conductive cloth. Since there is no electrical wiring
inside the sensor, the sensor structure is hard to break by the
impact force which is realized;
3) Boxing punching experiments were carried out by installing the
sensor on concrete wall and sandbag. The flexible sensor showed
a good performance in the experiments;
4) The proposed sensor is inadequate distribution number of sensor
elements, but the proposed sensor includes mechanisms to
achieve a flexible sensor with high accuracy. By increasing the
distribution density of the sensor element, it can approach to the
practical sensor.
Flexibility, thinness, lightness, toughness and accuracy are the keys of
flexible impact force sensor. Overcoming of these problems is essential for
the realization of the sensor that can be installed on human body or robot.

ACKNOWLEDGEMENTS
Chen Liu, Yukio Fujimoto, Yoshikazu Tanaka The authors wish to thank
Tadashi Uesugi, Takahiro Namikawa, Atsushi Maeda and Takuya Okada
of the Graduate School of Engineering, Hiroshima University for their
assistance in experiments. This work was partially supported by the Venture
Business Laboratory of Hiroshima University.
 Flexible Impact Force Sensor 203

REFERENCES
1. Ishikawa, M. and Shimojo, M. (1982) A Method for Measuring the
Center Position of a Two Dimensional Distributed Load Using
Pressure-Conductive Rubber. Transactions of the Society of Instrument
and Control Engineers, 18, 730-735.  
2. Mei, T., Li, W.J., Ge, Y., Chen, Y., Ni, L. and Chan, M.H. (2000) An
Integrated MEMS Three-Dimensional Tactile Sensor with Large Force
Range. Sensors and Actuators, 80, 155-162. http://dx.doi.org/10.1016/
S0924-4247(99)00261-7  
3. Lee, H.K., Chang, S.I. and Yoon, E. (2006) A Flexible Polymer
Tactile Sensor: Fabrication and Modular Expandability for Large Area
Deployment. Journal of Micro Electromechanical Systems, 15, 1681-
1686. http://dx.doi.org/10.1109/JMEMS.2006.886021   
4. Donati, M., Vitiello, N., De Rossi, S.M.M., Lenzi, T., Crea, S.,
Persichetti, A., Giovacchini, F., Koopman, B., Porobnik, J., Munih, M.
and Carrozza, M.C. (2013) A Flexible Technology for the Distributed
Measurement of Interaction Pressure. Sensors, 13, 1021-1045. http://
dx.doi.org/10.3390/s130101021   
5. Lee, M.H. and Nicholls, H.R. (1999) Review Article Tactile sensing for
Mechatronics—A State of the Art Survey. Mechatronics, 9, 1-31. http://
dx.doi.org/10.1016/S0957-4158(98)00045-2   
6. Lumelsky, V.J., Shur, M.S. and Wagner, S. (2001) Sensitive Skin. IEEE
Sensors Journal, 1, 41-51. http://dx.doi.org/10.1109/JSEN.2001.923586  
7. Walilko, T.J., Viano, D.C. and Bir, C.A. (2005) Biomechanics of the
Head for Olympic Boxer Punches to the Face. British Journal of Sports
Medicine, 39, 710-719. http://dx.doi.org/10.1136/bjsm.2004.014126
8. Fujimoto, Y. and Setyanto, A.T. (2007) Sheet Type Impact Force Sensor
by the Use of Piezoelectric Film. Transactions of the Japan Society
of Mechanical Engineers, Series C, 73, 184-191. (in Japanese) http://
dx.doi.org/10.1299/kikaic.73.184   
9. Fujimoto, Y., Shintaku, E., Tanaka, Y. and Fujiyoshi, J. (2012) Pad
Type Compressive Force Sensor Suitable for High-Speed Impact
Force Measurement. Transactions of the Japan Society of Mechanical
Engineers, Series C, 78, 2438-2449. (in Japanese)   
10. Fujimoto, Y., Liu, C., Uesugi, T., Tanaka, Y. and Shintaku, E. (2013)
Pipe Surface Sensor for Impulsive Force Measurement. Transactions
of the Japan Society of Mechanical Engineers, Series C, 79, 113-123.
204 Mechatronics and Robotics

(in Japanese) http://dx.doi.org/10.1299/kikaic.79.1949   

11. Shinkai, H., Nunome, H., Ikegami, Y. and Isokawa, M. (2008) Ball-
Foot Interaction in Impact Phase of Instep Soccer Kick. In: Reilly, T.
and Korkusuz, F., Eds., Science and Football VI, The Proceedings of
the 6th World Congress on Science and Football, Routledge, Abingdon,
41-46.   
 CHAPTER

11
A Sensing and Robot Navigation of
Hybrid Sensor Network

Shuncai Yao1, Jindong Tan2, Hongxia Pan1

Department of Electric Engineering, North University of China, Taiyuan,
1

China
Department of Electrical Computer Engineering, Michigan Technological
2

University, Houghton, USA

ABSTRACT
Traditional sensor network and robot navigation are based on the map of
detecting fields available in advance. The optimal algorithms are explored to
solve the energy saving, shortest path problems, etc. However, in practical

Citation:- S. Yao, J. Tan and H. Pan, “A Sensing and Robot Navigation of Hybrid Sen-
sor Network,” Wireless Sensor Network, Vol. 2, No. 4, 2010, pp. 267-273. doi: 10.4236/
wsn.2010.24037.
Copyright: © 2010 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
206 Mechatronics and Robotics

environment, there are many fields, whose map is difficult to get, and need to
detect. This paper explores a kind of ad-hoc navigation algorithm based on
the hybrid sensor network without the prior map. The system of navigation
is composed of static nodes and mobile nodes. The static nodes monitor
events occurring and broadcast. In the system, a kind of cluster broadcast
method is adopted to determine the robot localization. The mobile nodes
detect the adversary or dangerous fields and broadcast warning message.
Robot gets the message and follows ad-hoc routine to arrive the events
occurring place. In the whole process, energy saving has taken into account.
The algorithms of nodes and robot are given in this paper. The simu- late and
practical results are available as well.
Keywords: Robot Navigation, Hybrid Sensor Network, Routine Search

INTRODUCTION
Sensor network is an advance technology in recent years. There have been
many papers in this field. Most of these are concerned in static sensor
network or mobile sensor network separately. The communication and
related technology have been studied. The static sensor network consists of
the fixed nodes and adopts the ad-hoc communication. This kind of system
can monitor the environment and send the messages to the base station.
Energy would be saved, but the system can hardly do any- thing better if
the whole condition is changed or the adversary available. In mobile sensor
network, the nodes can adjust their places according the algorithms, which
are based on the changing condition. However, at the same time, the energy
would be cost much. It is well know that is a serious problem for sensor
networks.
In navigation, most of these two systems always adopt a method that
is based on mapping. That is to say, the whole environment map is known
in advance. In spite that the optimal algorithms have been adopted in these
systems, it is not so suitable for the genuine conditions yet.
In this paper, we explore a hybrid sensor network. It is composed of both
static sensor nodes and mobile sensor nodes. The static sensor nodes are
deployed regularly, and detect the events of environment. While, the mobile
sensor nodes detect the adversary or dangerous field. When these fields are
found the mobile nodes will stop at the edge of adversary or dangerous fields
and send a message to robot. Finally, there will be more nodes at the edge
of adversary or dangerous fields to protect the robot not to approach. The
method is a compromise between the energy saving and flexible of whole
 A Sensing and Robot Navigation of Hybrid Sensor ... 207

system. Based on this scheme, non-mapping navigation comes into reality.

It is proved that it is a better scheme in energy saving and flexibility through
the simulate experiment at the end of this paper.

RELATED WORK
In a static sensor network, the quantity of nodes is an important problem.
The more nodes means the more coverage of whole area and more accurate
of navigation. Lots of research has focused on the sensor localization [1],
structural monitoring [2] or protocols [3]. Recently, researchers have begun
to explore the method, which low down the quantity. And, it is taken into
account that the adversary and dangerous fields available in environment.
Since these methods adopt the skeleton graphs and optimal algorithm,
the quantity of nodes and the price of communication have been slowed
down.
The quantity of nodes is not such a pivotal problem in mobile sensor
network because its flexibility. The problem of energy costs is tried to low
down by sorts of algorithm based on the passive algorithm [4, 5] or statistic-
learning algorithm. On the other hand, these algorithms are complicated.
They are not so easy to come into realities and the use more capacity of
nodes memory.
The sensor network of this paper develops the hybrid sensor network
and adopts a kind of cluster navigation. At first, the static sensor nodes are
deployed as regular as possible in the whole environment. That is to say, the
localization of static sensor nodes should be arranged at the correct place,
but we need not care it so much. These static sensor nodes may local in the
adversary or dangerous fields, but it does not matter. We can take measure to
solve the problem. The sensing program can be downloaded in these sensor
nodes. They can sense the change of environment area that we care. The
mobile sensor nodes, actually they are the small mobile robots, but they just
do the dangerous fields detecting work and nothing else. Navigated robots
are not these nodes. Actually, there is another robot being navigated. In
Figure 1, the static sensor nodes are deployed in the whole area. They have
sensing and communication abilities. Actually, these nodes are tmote Sky
[6] made by Moteiv Corporation. They are ultra-low power wireless module
for use in sensor networks, monitoring applications, and rapid application
prototyping. Tmote Sky leverages industry standards like USB and IEEE
802.15.4 to interoperate seamlessly with other devices.
208 Mechatronics and Robotics

Figure 1. The structure of system.

The mobile sensor nodes can gather around the edge of adversary or
dangerous. The robot is in a waiting status. The mobile nodes are four-wheel
cart, which carry the Mica nodes. The Mica wireless platform serves as a
foundation for the emerging possibilities. With sensing, communication,
and I/O capabilities, Mica can simultaneously act as a data router, sensor
interface, and control point. Mica’s flexible design serves as a building
block for creating efficient application-specific protocols. And Mica does
not require use of predefined protocols [7].
We have done many works both on the communication of static sensors
and moving of mobile sensor nodes. This is the base that we can explore the
scheme.

THE DEPLOYMENT OF STATIC SENSOR NODES

AND COMMUNICATION ALGORITHM
The static sensor nodes deploy all the detecting area. In view of the energy
saving, we adopt a kind of street map and skeleton graph [8]. This scheme
constructs a reduced graph with as few as possible working nodes instead of
the whole area covering static sensor nodes. All the nodes stand at the key
 A Sensing and Robot Navigation of Hybrid Sensor ... 209

points of the detecting area. They are the working nodes. That does not mean
we do not need large number of static nodes. If the quantity of static nodes
is not enough, the navigation can become very difficult, we discuss this
problem at navigation. The nodes, which are not on the key points, fall in
sleep. That means these sleeping nodes are working at the low power state.
When the system needs them, they will be taken up. These sleeping nodes
are in working state, and afford robot navigation. That is to say, there is large
number of nodes in system, but, the working nodes are not so many and
energy is saved sequentially. On the other hand, the flexibility and accuracy
of system can be maintained. Those nodes, which are not in adversary fields
and not on the point, sleep unless the robot navigation needs them (We will
discuss in Section 5.).
Except these non-key point nodes, there are still some nodes falls into
sleep. They are the nodes lie in the adversary or dangerous fields. They
can keep the sleeping status through the all process. Actually, they have
quitted events monitor. Therefore, the street map and skeleton graph can be
described as Figure 2.
Given such preliminary works, we modify the static nodes algorithm
to arrange the nodes places, monitor events and broadcast the message as
shown in Algorithm 1.
Algorithm 1. Deploy of static sensor nodes
1: while TRUE do
2: detect the location
3: if location! = dangerous fields & location! = key point then
4: send navigation requirement
5: else
6: node keeps sleep & quit monitor
7: end if
8: if navigation response! = TRUE then
9: node keeps sleep
10: else
11: broadcast navigation message
12: end if
13: end while
210 Mechatronics and Robotics

Figure 2. The diagram of on point nodes and the sleeping nodes. These nodes
which is in the dangerous fields are put into sleeping status.
Once there is an event occurs, the monitoring message can be sent by ad-
hoc module to find the robot. In order to save the energy, we set a threshold
for the event. Therefore, the working static nodes are not always at “working
status”.

THE MOBILE SENSOR NODES

The mobile sensor nodes afford the task that detect adversary or dangerous
for there is no prior knowledge about the field or map. Actually, mobile
sensor nodes are the little robots. The working style of mobile nodes adopts
a kind of active learning [9, 10].
Since the map of detected field is unknown, the mobile nodes scan the
whole detecting area in raster fashion.
A penalized estimate can be constructed. It is . When they find these
area or fields, they do the estimate of the area by using average smooth.
Then, in this section, the boundary of dangerous area can be located roughly.
Consequently, the mobile nodes calculate to smooth these detected regions
and get refine localization by average algorithm. A complexity estimate will
be got, just like the first coarse scan. It is r . Finally, the estimate of the
fields can be generated by coarse scan and refine localization.
The mean square error can be described as follow, by using Craig-
Bernstein inequality [11]:
 A Sensing and Robot Navigation of Hybrid Sensor ... 211

(1)

is calculated estimate value that conclude the coarse estimate and

refine estimate . K1 and K2 are the constants larger than 0.
Figure 3 shows the process about mobile sensor nodes detecting and
their final localization.
Once the localization of adversary or dangerous regions are estimated,
the mobile can communicated each other, move and stand around the edge
of these regions according average criterion. From then on, the mobile
sensor nodes work as static nodes and send warning message. Algorithm 2
has shown the mobile nodes motion and positioning method.
Algorithm 2 Mobile nodes define dangerous fields
1: while only received the static nodes’ signal do
2: Doing raster scan
3: Collect the sample
4: Launch estimate algorithm
5: if routine = BLOCKED then
6: if other mobile nodes! = stay then
7: stay as a static node
8: sending the warning message
9: else
10: record the sample
11: end if
12: else
13: raster scan
14: end if
15: end while
Thus, the mobile nodes can gather around the edge of adversary or
dangerous. According the adaptive statistic algorithm, the mobile nodes can
find the adversary fields as soon as possible. The path length detect can be
de- scribed as [12]:
212 Mechatronics and Robotics

Figure 3. The mobile nodes (a) Raster fashion moving; (b) Mobile nodes loca-
tion.

(2)
where l is the path length, ~ means the order of polynomial, and d is the
dimensions of function. And the time can be estimated as:

(3)
After the algorithms of static sensor nodes and mobile nodes are
designed, we can go into the next step. That is the navigation of robot.

NAVIGATION OF ROBOT
In navigation, the robot is Pioneer Ⅱ-AT four-wheel drive. The robot
also has a variety of expansion power and I/O ports for attachment and close
integration of additional sensors and other accessories. Expansion includes
an addressable I/O bus for up to 16 devices, two RS-232 serial ports, eight
digital I/O ports, five A/D ports, PSU controllers and more—all accessible
through a common application interface to the robot server software, P2OS
[13]. Therefore, we use the Mica as a communicate de- vice and serial ports
send dictate to the robot.
Most navigation is based on the detect fields map, which is available in
advance. So, the location of robot can be decided by coordinate of the map
 A Sensing and Robot Navigation of Hybrid Sensor ... 213

[14]. And the optimal navigation algorithm can be explored. There have
been many algorithms in these fields. But they can hardly adapt the non-
mapping condition. That is to say, if there is a novel environment or a region
that its condition is difficult to get to know, what these algorithms do is
detect the whole region and acquire the whole information of the region.
Only after the information of region is known, can the optimal algorithm
can be adopted. Therefore, the energy may be cost in acquiring information.
In order to solve the problem, we explore a kind of non-mapping navigation
algorithm to guide the robot to arrive the event occurring place.
Generally speaking, navigation can be described as 3 questions. That
is where I am, where I should go, and how to arrive [15]. We adopt a kind
of dynamic cluster navigation algorithm. It can decide the location of robot
and use both static nodes and mobile nodes to guide the robot to arrive the
event place.
1). Where I am (Decide the location of robot)
Since what we discuss is no prior field map in this paper; the first thing
is the location of robot should be decided. It is the important step.
In order to acquire the location of robot, we adopt a cluster-navigating
algorithm. It depends on the ad-hoc relay communication. The static nodes
work as event heads is 1, and so do as other nodes in the next cluster until
the robot receive the message [16]. Robot receives the hierarchy of the static
nodes. Therefore, the robot can decide approximately how long the distance
between event and itself not precise. It is enough. And the robot should
distinct the message of event or relay. It can decide itself whether arriving
the place or not. At the same time, robot should also distinct the static or
mobile nodes, so it can decide it is close to danger fields or not.
If the static nodes are too sparse to decide the robot in a field, it is difficult
for robot to make a decision its direction to move. The robot cannot decide
where to go. The system must take measure to make up. Those nodes, which
are not on the key point, should be taken up the sleeping nodes around them
and form a new cluster in order to inform the robot.
Figure 4 shows the process of detect the robot location. The detail
algorithm of finding robot location is showed as follow:
Algorithm 3 The Transmission of Event 1: while robot! = arrived do
2: select working static nodes as cluster
3: if the nodes is too sparse then
214 Mechatronics and Robotics

Figure 4. The process of finding robot by using cluster communication.

4: wake up the sleeping nodes
5: else
6: static nodes work as cluster
7: end if
8: if event = TRUE then
9: hierarchy = 0
10: broadcast message
11: else
12: if received message = TURE then
13: hierarchy < - hierarchy+1 14: broadcast
15: end if
16: end if
17: end while
2). Where should I go
In a large scope, robot can move to the direction whose hierarchy is
always lowing. When robot arrives in the range of cluster, the static nodes
should shut down the message broadcasting and its hierarchy minus one.
While in the cluster, robot always finds the leader of this cluster.
3). How to arrive
The robot can follow a routine, which the nodes’ hierarchy is lowing
unless it receives the warning message of the mobile nodes. When it receives
warning message of the mobile nodes, it have to use the smooth routine in
order to pass the danger field until the warning message vanish.
 A Sensing and Robot Navigation of Hybrid Sensor ... 215

Finally, the robot will arrive the place where events occur when it receive
the event’s source message.
The distance from robot to dangerous fields can be calculated by
potential [17]:

(4)
where R is the Euclidean distance from the edge to the mobile nods. In
this paper, it can be shown as:

(5)
(xe, ye) is the location of dangerous fields. The energy of mobile nodes
can be defined as follow:

(6)
Therefore, we can conclude the algorithm of robot navigation. There are
still some handicaps (they are the points not area) in the robot routine, but
there is sonar in the robot. The robot can avoid these handicap points. It is
the algorithm 4 shown as follow:
Algorithm 4 Navigation of Robot
1: While event message received = TURE Do
2: if arrived! = TURE then
3: if mobile warning message! = TURE then
4: use sonar (or communicate with the static nodes) follow navigation
routine
5: else
6: adopt the smooth routine 7: end if
8: else
9: send arrive message
10: inform the event node turn off requirement 11: end if
12: end while
Since there is no map in advance, what the robot does is receive the
message and do probing motion. Thus, the time to arrive the event is not
shortest and navigation routine is not the optimal routine. It is the cost of
216 Mechatronics and Robotics

none prior map available and the lower cost of both static and mobile nodes.
On the other hand, there are no so many working nodes in this system, the
energy cost of nodes can be saved.
4). How to wake up the sleeping nodes
When the static nodes, which are on the key point, is too sparse to
accomplish the navigation task, it is necessary to wake up the sleeping nodes.
At first, the cluster head, the static nodes on the key point, acquires the
message and sends the message to next closed cluster head. If the closed
cluster head can receive the message, it will send back a message to the
sending cluster head. That means that the statistic nodes are not too sparse
there. The message can relay to the next static nodes. If the closed cluster
head cannot receive the message, it cannot send back the message to the
sending cluster head. It means that the statistic nodes are too sparse there.
That is to say, the sleeping nodes in the cluster have to be taken up in order
to relay the message
The construction of the information deliver is shown in Figure 5
The static nodes in this system adopt tmote Sky [6] made by Moteiv
Corporation. They are ultra-low power wireless module for use in sensor
networks, monitoring applications, and rapid application prototyping. In an
open field, the communication distance of tmote nodes can reach 100 meters.

Figure 5. The construction of information deliver in a cluster.

But in this system, the communication distance of tmote nodes is
restrained to 10 meters in or- der to saving energy. In a cluster, there is one
tmote node working as the cluster and the other nodes working as sleeping
nodes. The event occurs, such as the temperature is over the threshold, the
cluster head sends the message to the next cluster head. If the next cluster
head receives the message, it sends back a message to the former cluster.
 A Sensing and Robot Navigation of Hybrid Sensor ... 217

Consequently, message can be relayed on. If the sending cluster head does
not receive the backing message of the next cluster head, the sending cluster
head will send a signal to the sleeping nodes in its cluster, in order to wake
up these sleeping nodes. That is to say, the low power costs’ nodes (sleeping
nodes) will transfer from the low power costs’ state (sleeping) to working
state (waken up). Once the sleeping nodes are waken up, they can afford the
message delay events.
After the works have been accomplished, we simulate the sensing and
navigation of hybrid sensor network by using. Figure 6 and Figure 7 is result
of simulating experiments.
It can be seen that the energy saving of mobile work is better than the
static network. While, the energy saving of hybrid network is much better
than the mono-static and mono-mobile network.
The practical experiments are performed with the setup shown in Figure
8. A sequence process of navigation is shown from (a) to (d). An emergent
event occurs by a node (the temperature is over the threshold), the robot
(Pioneer Ⅱ-AT) should find the node and arrive the place in time. Part (a)
shows the initial experiments situation. There are 100 tmote nodes (static
nodes) and 8 small robots with the tmote nodes on them (mobile nodes) in
a 25m2 (5m×5m) filed. The lawn simulates the dangerous fields, and the
scope can be detected by moisture sensor on tmote nodes.

Figure 6. The result of simulating and the trajectory of robot.

218 Mechatronics and Robotics

Figure 7. The energy conservation of different network.

Figure 8. The practical experiments of the navigation.

Therefore, the range of dangerous fields can be defined by them.
Consequently, part (b) shows the dangerous fields are defined. Part (c) shows
the process of navigation, and part (d) shows the robot arrive the emergency
place (red circle) finally.
 A Sensing and Robot Navigation of Hybrid Sensor ... 219

Despite the difficulties with the dangerous fields’ definition, the navigation
is successfully performed many times. Although arriving the final location
of the robot is not so quickly (about 3 minutes), the robot avoids the danger
fields and being guided by this hybrid sensor network system indeed. We
plan to conduct further re- search to get higher speed of this navigation
system.

CONCLUSIONS
This implementation demonstrated robot guidance by a hybrid sensor
network. The energy saving in sensor net- work is an important problem.
We take it as the primary problem in this system. The skeleton of adversary
or dangerous fields can be outlined by using the mobile sensor nodes and the
robot navigation comes into reali- ties on the ad-hoc method. In spite of the
less optimal routine, it is an effective scheme when there is no map of the
whole detecting fields in advance.
However, the precision of the navigation system greatly depends on the
robot hardware. Navigation can be problematic if environmental factors are
not so ideal, such as electrical cables, higher power electrical equipment and
steel structures exist. Although it is possible to compensate for these effects
in a known environment, a search situation may not permit elaborate offline
calibration. In addition, the response speed of whole system should be raised
higher. We are going to do further re- search on them.

ACKNOWLEDGMENT
We would like to give special thanks to Dr. Jindong Tan, the professor of
Michigan Technological University in
U.S.A. for his providing of experimental equipment and environment.
The same thanks go to Qi An, who helped
me with studies of the project and sensor network,
TinyOS.
220 Mechatronics and Robotics

REFERENCES
1. D. Moore, J. Leonard, D. Rus and S. Teller, “Robust Distributed
Networklocalization with Noisy Range Meas- urements,” Proceedings
of the 2nd International Confer- ence on Embedded networked sensor
systems, Baltimore, 2004, pp. 50-61.
2. R. Szewczyk, A. Mainwaring, J. Polastre and D. Culler, “An Analysis
of a Large Scale Habitat Monitoring Ap- plication,” Proceedings of
the 2nd International Confer- ence on Embedded networked sensor
systems, Baltimore, 2004, pp. 214-226.
3. J. Polastre, J. Hill and D. Culler, “Versatile Low Power Media Access
for Wireless Sensor Networks,” Proceed- ings of the 2nd International
Conference on Embedded networked sensor systems, Baltimore, 2004,
pp. 95-107.
4. M. Batalin, M. Rahimi, Y. Yu, D. Liu, A. Kansal, G. Sukhatme, W.
Kaiser, M. Hansen, G. Pottie, M. Srivastava and D. Estrin, “Call and
Response: Experiments in Sampling the Environment,” Proceedings of
ACM Sensor Systems, Los Angeles, 2004.
5. M. Rahimi, R. Pon, W. J. Kaiser, G. S. Sukhatme, D. Estrin and
M. Srivastava, “Adaptive Sampling for Environmental Robotics,”
Proceedings of IEEE International Conference on Robotics and
Automation, Los Angeles, 2004.
6. http://www.moteiv.com
7. http://www.crossbow.com
8. R. Willett, A. Martin and R. Nowak, “Backcasting: Adaptive Sampling
for Sensor Networks,” Proceedings of IEEE International Conference
on Information process- ing in sensor networks, New York, 2004, pp.
124-133.
9. R. Castro, R. Willett and R. Nowak, “Faster Rates in Re- gression via
Active Learning,” Proceedings of Neural In- formation Processing
Systems, 2005. http://homepages.cae. wisc.edu/ rcastro/ ECE-05-3.pdf
10. M. Horstein, “Sequential Decoding Using Noiseless Feed- back,” IEEE
Transactions Information Theory, Vol. 9, No. 3, 1963, pp. 136-143.
11. J. Borenstein and H. R. Everett, “Navigating Mobile Ro- bots: Sensors
and Techniques,” John Wiley & Sons, 1992.
12. M. Burnashev and K. S. Zigangirov, “An Interval Esti- mation Problem
for Controlled Observations,” Problems in Information Transmission,
Vol. 10, 1974, pp. 223-231.
 A Sensing and Robot Navigation of Hybrid Sensor ... 221

13. O. Khatib, “Real-Time Obstacle Avoidance for Manipu- lators and

Mobile Robots,” The International Journal of Robotics Research, Vol.
5, No. 1, 1986, pp. 90-98.
14. J. Cortes, S. Martinez, T. Karatas and F. Bullo, “Cover- age Control
for Mobile Sensing Networks,” IEEE Trans- actions on Robotics and
Automation, New Orleans, April 2004, pp. 243-255.
15. A. Korostelev, “On Minimax Rates of Convergence in Image Models
under Sequential Design,” Statistics and Probability Letters, Vol. 43,
1999, pp. 369-375.
16. G. Golubev and B. Levit, “Sequential Recovery of Ana- lytic Periodic
Edged in the Binary Image Models,” Mathe- matical Methods of
Statistics, Vol. 12, 2003, pp. 95-115.
17. R. Castro, R. Willett and R. Nowak, “Faster Rates in Re- gression
via Active Earning,” Proceedings of Neural In- formation Processing
Systems, Vancouver, 2005. http:// homepages.cae.wisc.edu/rcastro/
ECE-05-3. Pdf
 CHAPTER

12
A Study on Vehicle Detection and
Tracking Using Wireless Sensor
Networks

G. Padmavathi1, D. Shanmugapriya2, M. Kalaivani1

Department of Computer Science, Avinashilingam University for Women,
1

Coimbatore, Tamil Nadu, India 2Department of Information Technology,

Avinashilingam University for Women, Coimbatore, Tamil Nadu, India

ABSTRACT
Wireless Sensor network (WSN) is an emerging technology and has great
potential to be employed in critical situations. The development of wireless
sensor networks was originally motivated by military applications like
battlefield surveillance. However, Wireless Sensor Networks are also used
in many areas such as Industrial, Civilian, Health, Habitat Monitoring,

Citation:- G. Padmavathi, D. Shanmugapriya and M. Kalaivani, “A Study on Vehicle

Detection and Tracking Using Wireless Sensor Networks,” Wireless Sensor Network,
Vol. 2 No. 2, 2010, pp. 173-185. doi: 10.4236/wsn.2010.22023.
Copyright: © 2010 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
224 Mechatronics and Robotics

Environmental, Military, Home and Office application areas. Detection and

tracking of targets (eg. animal, vehicle) as it moves through a sensor network
has become an increasingly important application for sensor networks. The
key advantage of WSN is that the network can be deployed on the fly and
can operate unattended, without the need for any pre-existing infrastructure
and with little maintenance. The system will estimate and track the target
based on the spatial differences of the target signal strength detected by the
sensors at different locations. Magnetic and acoustic sensors and the signals
captured by these sensors are of present interest in the study. The system is
made up of three components for detecting and tracking the moving objects.
The first component consists of inexpensive off-the shelf wireless sensor
devices, such as MicaZ motes, capable of measuring acoustic and magnetic
signals generated by vehicles. The second component is responsible for the
data aggregation. The third component of the system is responsible for data
fusion algorithms. This paper inspects the sensors available in the market
and its strengths and weakness and also some of the vehicle detection and
tracking algorithms and their classification. This work focuses the overview
of each algorithm for detection and tracking and compares them based on
evaluation parameters.
Keywords: Wireless Sensor Networks, Acoustic and Magnetic Sensors,
Acoustic and Magnetic Signals, Detection and Tracking Algorithms

INTRODUCTION
The Wireless Sensor Networks comprise of relatively inexpensive sensor
nodes capable of collecting, processing, storing and transferring information
from one node to another. These nodes are able to autonomously form a
network through which sensor readings can be propagated. Since the sensor
nodes have some intelligence, data can be processed as it flows through
the network. Sensing devices will be able to monitor a wide variety of
ambient conditions: temperature, pressure, humidity, soil makeup, vehicular
movement, noise levels, lighting conditions, the presence or absence of
certain kinds of objects, mechanical stress levels on attached objects and
so on. These devices will also be equipped with significant processing,
memory and wireless communication capabilities. Emerging low-level
and low-power wireless communication protocols can be used to networks
the sensors. This capability will add a new dimension to the capabilities of
sensors. Sensors will be able to coordinate among themselves on a higher-
level sensing task. The sensors can be deployed in any facility or area,
 A Study on Vehicle Detection and Tracking Using Wireless ... 225

which has to be sensed in three main types. It can either be 1) triangular

sensor deployment, 2) square sensor deployment and 3) irregular sensor
deployment [1]. These deployments are depicted in Figure 1.
Networking inexpensive sensors can revolutionize information gathering
in a variety of situations.
A sensor node usually consists of four sub-systems:
• • A computing subsystem: In a sensor node, the microprocessor
(microcontroller unit, MCU) is responsible for functions such as
control of sensors and execution of communication protocols.

Figure 1. (a) Triangular; (b) Square; (c) Irregular deployments.

• A communication subsystem: This comprises of short range radios
used to communicate with neighbouring nodes and the outside
world. These devices operate under the Transmit, Receive, Idle
and Sleep modes having various levels of energy consumption.
• A sensing subsystem: Low power components can help to
significantly reduce power consumption. Since this subsystem
(sensors and actuators) is responsible for the sharing of
information between the sensor network and the outside world.
• A power supply subsystem: It consists of a battery which supplies
power to the node.

Characteristics of WSN
Some of the unique characteristics of a WSN include:
• Limited power they can harvest or store.
• Ability to withstand harsh environmental conditions
• Ability to cope with node failures
• Mobility of nodes
• Dynamic network topology
226 Mechatronics and Robotics

• Communication failures
• Heterogeneity of nodes large scale of deployment
• Unattended operation

Applications
The various sensor network applications include,
1) Military Applications
• Monitoring Friendly Forces, Equipments and Ammunition
• Battlefield Surveillance
• Targeting
• Battle Damage Assessment
• Nuclear, Biological and Chemical Attack Detection
2) Environmental Applications
• Forest Fire Detection
• Flood Detection
• Monitoring through Internet
• Monitoring Biodiversity
3) Habitat Monitoring Applications
• Habitat Monitoring
• Great Duck Island System
4) Health Applications
• Tele-monitoring of Physical Data
• Tracking and Monitoring Doctors and Patients in a Hospital
• Drug Administration in Hospitals
5) Home and office Applications
• Smart Homes
• Managing Inventory Control
• Home automation
• Environmental control in office buildings
6) Other Applications
• Monitoring Nuclear Reactor.
• Target Tracking
• Suspicious Individual detection: Interactive museums
 A Study on Vehicle Detection and Tracking Using Wireless ... 227

• Fire Fighters Problem (First Responders Problem) The major

contribution of this paper includes classification of sensors
available and vehicle detection and tracking algorithms using
Wireless Sensor Networks.
The paper is organized as follows: Section 2 gives the classification of
sensors used for vehicle detection and mainly concentrates on the acoustic
and magnetic sensors. Section 3 gives the gist about the acoustic and
magnetic signals acquired from the sensors. Section 4 discusses the various
algorithms used in vehicle detection and tracking, analysis about these
algorithms is given in Section 5 and summary of these algorithms given in
Section 6 followed by the conclusion.

CLASSIFICATION OF SENSORS USED IN VEHICLE

DETECTION
There are a wide variety of sensors available in market today [2]. Table 1
shows the variety of current sensor technologies and compares the strengths
and weaknesses with respect to installation, parameters measured, and
performance in bad weather, variable lighting, and changeable traffic flow.
Many over-roadway sensors are compact and mounted above or the side
of the roadway, making installation and maintenance relatively easy.
Some sensor installation and maintenance applications may require the
closing of the roadway to normal traffic to ensure the safety of the installer
and motorist.
All the sensors listed here operate under day and night conditions.
Sensors are broadly classified as Intrusive and Non-Intrusive sensors.
228 Mechatronics and Robotics

Table 1. Strengths and weaknesses of commercially available sensor technolo-

gies. G. PADMAVATHI ET AL. 175

Table 1. Strengths and weaknesses of commercially available sensor technologies.

Technology Strengths Weaknesses
Inductive loop Flexible design to satisfy large variety of applications. Installation requires pavement cut.
Mature, well understood technology. Improper installation decreases pavement life.
Large experience base. Installation and maintenance require lane closure.
Provides basic traffic parameters (e.g., volume, pres- Wire loops subject to stresses of traffic and temperature.
ence, occupancy, speed, headway, and gap). Multiple loops usually required to monitor a location.
Insensitive to inclement weather such as rain, fog, and Detection accuracy may decrease when design requires
snow. detection of a large variety of vehicle classes.
Provides best accuracy for count data as compared
with other commonly used techniques.
Common standard for obtaining accurate occupancy
measurements.
High frequency excitation models provide classifica-
tion data.

Magnetometer Less susceptible than loops to stresses of traffic. Installation requires pavement cut.
(two-axis fluxgate Insensitive to inclement weather such as snow, rain, Improper installation decreases pavement life.
magnetometer) and fog. Installation and maintenance require lane closure.
Some models transmit data over wireless radio fre- Models with small detection zones require multiple units for
quency (RF) link. full lane detection.

Magnetic (induction Can be used where loops are not feasible (e.g., bridge Installation requires pavement cut or boring under roadway.
or search coil mag- decks). Cannot detect stopped vehicles unless special sensor layouts
netometer) Some models are installed under roadway without and signal processing software are used.
need for pavement cuts. However, boring under
roadway is required.
Insensitive to inclement weather such as snow, rain,
and fog.
Less susceptible than loops to stresses of traffic.

Microwave radar Typically insensitive to inclement weather at the Continuous wave (CW) Doppler sensors cannot detect
relatively short ranges encountered in traffic man- stopped vehicles
agement applications.
Direct measurement of speed.
Multiple lane operation available.

Active infrared (laser Transmits multiple beams for accurate measurement Operation may be affected by fog when visibility is less than
radar) of vehicle position, speed, and class. 20 feet (ft) (6 m) or blowing snow is present.
Multiple lane operation available. Installation and maintenance, including periodic lens clean-
ing, require lane closure

Passive infrared Multizone passive sensors measure speed. Passive sensor may have reduced vehicle sensitivity in
heavy rain, snow and dense fog.
Some models not recommended for presence detection.

Ultrasonic Multiple lane operation available Environmental conditions such as temperature change and
Capable of over height vehicle detection. extreme air turbulence can affect performance. Temperature
Large Japanese experience base. compensation is built into some models.
Large pulse repetition periods may degrade occupancy
measurement on freeways with vehicles travelling at moder-
ate to high speeds.
Acoustic Passive detection. Cold temperatures may affect vehicle count accuracy.
Insensitive to precipitation. Specific models are not recommended with slow-moving
Multiple lane operation available in some models. vehicles in stop-and-go traffic.

Video image proces- Monitors multiple lanes and multiple detection Installation and maintenance, including periodic lens clean-
sor zones/lane. ing, require lane closure when camera is mounted over
Easy to add and modify detection zones. roadway (lane closure may not be required when camera is
Rich array of data available. mounted at side of roadway).
Provides wide-area detection when information gath- Performance affected by inclement weather such as fog,
ered at one camera location can be linked to another. rain, and snow; vehicle shadows; vehicle projection into
adjacent lanes; day-to-night transition; vehicle/road con-
trast; and water, salt grime, icicles, and cobwebs on camera
lens.
Reliable night-time signal actuation requires street lighting.
Requires 30- to 50-ft (9- to 15-m) camera mounting height
(in a side-mounting configuration) for optimum presence
detection and speed measurement.
Some models are susceptible to camera motion caused by
strong winds or vibration of camera mounting structure.
Generally cost effective when many detection zones within
the camera field of view or specialized data are required.

Copyright © 2010 SciRes. WSN

Intrusive Sensors
Intrusive sensors are those that need to be installed under the pavement,
in saw-cuts or holes on the roads. Popular intrusive sensors include
inductive loops, magnetometers, micro loop probes, pneumatic road
tubes, piezoelectric cables and other weigh-in-motion sensors. The main
advantage of these sensors is their high accuracy for vehicle detection while
 A Study on Vehicle Detection and Tracking Using Wireless ... 229

the drawbacks include the disruption of traffic for installation and repair,
resulting in high installation and maintenance cost.

Non-Intrusive Sensors
To overcome the disadvantage of intrusive sensors, non-intrusive sensors
are developed eg, above ground vehicle detection sensors. Above ground
sensors can be mounted above the lane of traffic or on the side of a roadway
where they can view multiple lanes of the traffic at angles perpendicular to or
at a slanting angle to the flow direction. Technologies used in aboveground
sensors include video image processing (VIP), microwave radar, laser radar,
passive infrared, ultrasonic, passive acoustic array, and combinations of
these sensor technologies. However, these non-intrusive sensors tend to be
large size and power hunger.

Smart-Dust Sensor Node-Hardware Platform

Smart-Dust is one of the potential sensor nodes which can be used in the
future vehicle detection system. In Smart-Dust sensor node, essential
components for vehicle detection (processor, memory, sensor and radio)
could be integrated together as small as a quarter through MEMS technology.
Together with its low power design [3], the Smart-Dust sensor node is
suitable for implementing the vehicle detection sensor networks. Figure 2
shows the different generations of Smart Dust sensor nodes (Motes). The
left picture is a 1st generation smart dust sensor node “Rene Mote”. From
left to right, the right picture shows the “MICA Mote” (2nd generation),
“MICA2 Mote” (3rd generation) and “MICA2-Dot Mote (3rd generation)”.
Smart-Dust sensor node is designed by EECS department in UC Berkeley
and Intel [4] using modular component approach and it consists of two major
components: mother board and sensor board. Thus, different sensor boards
could be attached to the same mother board for different applications. Thus,
Smart

Figure 2. Dust family.

230 Mechatronics and Robotics

Dust sensor node could potentially be used in a wide range of applications

such as vehicle detection, enemy monitoring in the battlefield, temperature
measurement in a building, environmental monitoring etc.
The basic components of MICA mote shown in Figure 3 belongs to
the Smart-Dust family. The components are listed in Table 2. The mother
board consists of an Atmel 90LS8535 processor, 512KB SRAM, 8KB Flash
RAM and a RF transceiver for wireless communication. The Sensor board
consists of a 10-bit analog to digital converter, a Magnetometer (Honeywell
HMC1002), a temperature sensor, a photo camera and an accelerometer
sensor.
For vehicle detection system, the sensors used are the magnetometer and
acoustic sensors. Next, the basic operating principles of magnetometer and
acoustic sensors are will be reviewed.

Acoustic Sensors
The acoustic sensor in the Smart Dust sensor node is a condenser type
microphone. The schematic for a typical condenser acoustic sensor is shown
in Figure 4. It includes a stretched metal diaphragm that forms one plate of a
capacitor. A metal disk placed close to the diaphragm acts as a backplate. A
stable DC voltage is applied to the plates through a high resistance to keep
electrical charges on the plates. When a sound field excites the diaphragm,
the capacitances between the two plates vary according to the variation in
the sound pressure. The change in the capacitance generates an AC output
proportional to the sound pressure, which shows the ultra low-frequency
pressure variation. A high-frequency voltage (carrier) is applied across the
plates and the acoustic sensor output signal is the modulated carrier. The
photo in the right of Figure 4 shows the Panasonic

Figure 3. MICA mote.

 A Study on Vehicle Detection and Tracking Using Wireless ... 231

Table 2. Components of smart dust mote.

Figure 4. Condenser microphone. AP-the acoustic pressure; C-the variable ca-

pacitance; 1-the metal diaphragm; 2-the metal disk; 3-the insulator; 4-the case.

Figure 5. Waveforms of acoustic signals emitted from truck.

WM-62A condenser microphones used in Smart Dust Motes. Figure 5
shows a typical vehicle acoustic signal waveforms.
232 Mechatronics and Robotics

Magnetic Sensors
Magnetic sensors differ from most other detectors in that they do not
directly measure the physical property of interest. Devices that monitor
properties such as temperature, pressure, strain, or flow provides an output
that directly reports the desired parameter (Figure 6). Magnetic sensors, on
the other hand, detect changes, or disturbances, in magnetic fields that have
been created or modified, and from them derive information on properties
such as direction, presence, rotation, angle, or electrical currents. The output
signal of these sensors requires some signal processing for translation into
the desired parameter. Although magnetic detectors are somewhat more
difficult to use, they do provide accurate and reliable data without physical
contact.
The Honeywell HMC1002 magnetometer on the MICA sensor board is
a magnetoresistive sensor. Honeywell magnetic sensors and magnetometers
offer complete magnetic field sensing solutions that are highly accurate, and
allow for easy integration for virtually any application. Figure 7 shows the
honeywell magnetic sensors.
The Anisotropic Magneto Resistive (AMR) sensor is one type that has a
wide Earth’s field sensing range and can sense both the strength and direction
of the Earth field. The AMR sensor is made of a nickel-iron (Permalloy)
thin film deposited on a silicon wafer and patterned as a resistive strip. The
strip resistance changes about 2%-3% when a magnetic field is applied.
Typically, four of these resistive strips are connected in a wheatstone bridge
configuration so that both magnitude and direction of a field along a single
axis can be measured. The key benefit of AMR sensors is that they can be bulk
manufactures on silicon wafers mounted in commercial integrated circuit
packages. Figure 8 shows a typical vehicle magnetic signal waveforms.

Figure 6. Magnetic sensor parameters.

 A Study on Vehicle Detection and Tracking Using Wireless ... 233

Figure 7. Magnetic sensors.

Figure 8. Waveforms of magnetic signals.

Magnetic Field
A magnetic field is a vector field that surrounds magnets and electric currents,
and is detected by the force it exerts on moving electric charges and on
magnetic materials. When placed in a magnetic field, magnetic dipoles tend
to align their axes parallel to the magnetic field. Magnetic fields also have
their own energy with an energy density proportional to the square of the
field intensity.

Magnetic Data
Magnetic data is the term used for data that is acquired from magnetic (as
opposed to optical) motion capture systems. A central magnet is used to
create a field in which sensors can determine their position and orientation
as they move about in the field. Raw magnetic data has no hierarchy
234 Mechatronics and Robotics

information; the sensors do not know where they are relative to the other
sensors, and know their position in the magnetic field. The file formats for
magnetic motion capture data reflect this. In this respect, they are somewhat
similar to the BVA file format.
The problem associated with magnetometer vehicle detection is similar
to the acoustic sensors but the magnetic signals are much cleaner than
acoustic signals. The magnetometers available today can sense magnetic
fields within the earth’s field-below 1 gauss.
They can be used for detecting the vehicles, which are ferrous objects
that disturb the earth’s field. The earth’s field provides a uniform magnetic
field over wide area in the scale of kilometers and even a ferrous object can
create a local disturbance in this field.
This local field disturbance can be sensed by the magnetometers for
vehicle detection. Figure 9 shows the disturbance of earth’s magnetic field
by a car. After presenting the basic principles of sensors, the characteristics
of the measured acoustic and magnetic signals will be studied and algorithms
are proposed for reliable low cost vehicle detection.

SIGNALS FOR VEHICLE DETECTION AND TRACK-

ING
The signals taken into consideration for study in this proposal are acoustic
and magnetic signals.

Figure 9. Earth’s magnetic field disturbed by a car.

Acoustic Signals
The acoustic signature is made up of a number of individual elements.
These include:
 A Study on Vehicle Detection and Tracking Using Wireless ... 235

• Machinery noise: noise generated by a ships engines, propeller

shafts, fuel pumps, air conditioning systems, etc.
• Cavitation noise: noise generated by the creation of gas bubbles
by the turning of a ship’s propellers.
• Hydrodynamic noise: noise generated by the movement of
water displaced by the hull of a moving vessel.
These emissions depend on a hull’s dimensions, the installed machinery
and ship’s displacement. Therefore, different ship classes will have different
combinations of acoustic signals that together form a unique signature.
Sonar operating in passive mode can detect acoustic signals radiated by
invisible submarines and these signals are used to target attacks.

Magnetic Signals
Magnetic Signals are changes that happen in the magnetic field caused by
the movement of ferrous objects.

ALGORITHMS FOR VEHICLE DETECTION AND

TRACKING
There are a wide variety of algorithms generally available for detection and Tracking of
target. Figure 10 shows the classification of detection and tracking algorithms.

Target Detection Background

The basic idea behind a target detection algorithm is that sensor nodes are
deployed within an area and the nodes determine whether or not a target has
entered or is presently within the area. Nodes usually determine if a target is
present by detecting a change in some sort of signal, whether it is acoustic,
light, temperature or some other type [5]. Since the strength of this signal
will scatter throughout the sensor network the nodes need to collaboratively
work together in order to determine if a target has been detected or not.
Several of these collaborative detection algorithms are discussed in the
following section:

Basic Fusion
This algorithm is presented first since it has an extremely simple design and
implementation and serves as an ideal algorithm to use as the foundation
for explaining other more complex detection algorithms. This algorithm
collects the captured data from each of the nodes in a network. Then, to
236 Mechatronics and Robotics

remove any invalid data provided by faulty nodes, the largest and smallest
data values are dropped. The remaining data is then averaged together. If this
average exceeds a predetermined threshold, then the system is notified that a
target has been detected. This algorithm has two versions in which the data
provided by each node differs. The data is either the original data collected
by the node or it is a detection decision already made by the individual
nodes using their original data [5].

Hybrid
This algorithm is similar to the basic fusion algorithm. In that a detection
decision is based on data collected by numerous nodes throughout the
network. However, this algorithm differs in that either a yes/no target present
bit decision is sent or the collected data is sent. In this algorithm, each node
is given two threshold values for a previously determined detection signal
type, such as sound. One of these threshold values determines that if the
data collected by the node exceeds this value then a target has been detected.
Correspondingly, the other threshold is a minimum in that if the data is below,
then no target is present. If either of these two thresholds is met then a bit
message indicating the presence or lack of a target is transmitted. However,
if neither of these thresholds is exceeded then the raw data collected by the
node is passed on to the base station where it is used with the data of other
nodes to determine if a target is present or not [18].

Predicate Clustering
In this algorithm nodes are divided into groups called clusters in which each
group has a cluster head that all sub nodes of the cluster report their data to.
Typically, one cluster is formed around a single target with multiple clusters
being in the network if multiple targets are present. Cluster heads are chosen
by all nodes within the area of a detected target “electing” to become the
head by broadcasting their intent to become a cluster head. The node with
the peak sensor reading is chosen as the cluster head [6]. Once cluster heads
are chosen, the remaining nodes choose group to join by running a decision
predicate on their captured sensor data and the data of their neighbours. One
particularly interesting feature of this algorithm is that a cluster head can
be responsible for multiple targets. When it is detected, multiple targets are
going to enter the same cluster, both clusters provide the target information
to the “merged” cluster head that is now responsible for tracking both targets.
 A Study on Vehicle Detection and Tracking Using Wireless ... 237

Grids
In this approach the sensor network is broken up into “virtual grids” where
each grid contains points. For each grid point, it is determined which nodes’
sensing area covers the grid point and then these nodes create a schedule so
that they all collaboratively cover this grid point at alternating times. So at
any point of time‘t’ at least one node’s sensing area covers grid point ‘x’.
Since each node covers multiple grid points, each node will contain several
sensing schedules to keep in sync so that at any point of time all grid points
are covered by at least one node. In order to do this, all nodes first determine
the schedules for each grid point, then schedule all their individual point
schedules together so that each node is not constantly waking and sleeping
instead sleeps a little and then wakes for a while. Once each node determines
its own individual schedule, it must be correlated with those of its neighbours
so that as few as possible nodes are awake at any point of time but all grid
points are covered at all times. Once the schedules are completed for each
node, the nodes start to follow their schedule, going through states of being
asleep and awake in order to detect a new target [7].

Wave
The Wave protocol has the same goal in mind as that of the Grid algorithm,
detecting targets while conserving energy by not having all the nodes on
all the time. The basic idea of the wave algorithm is to wake up a series
of nodes at certain times so that only nodes that are awake at one point of
time are in one concentrated area. Then, over time, this section moves over
the network by putting nodes to sleep and waking new ones. This gives the
appearance of an awake section of nodes flowing over the entire network
like a wave [8].
This algorithm proposes three types of waves:
• Two straight lines of nodes are awake at the same time, with the
lines being parallel to each other. The two lines start on opposite
sides of the network from each other and gradually work towards
each other to meet in the centre of the network.
• A wide line of nodes are awake at the same time and this line
moves across the network from one edge to the opposite edge.
• Waves of nodes are awake and the wave traverses from one edge
of the network to the other. This approach differs from the wide
line in that the boundary shape is curvy or wave like instead of
238 Mechatronics and Robotics

being a straight line boundary. The wave is created by an awake

sensor waking all nodes within its communication range on the
side of it that the wave needs transition too. This approach also
differs from the previous two in that it uses radio communication
to wake up the next nodes to start sensing instead of time syncing
the movement of wake/sleep nodes across the network.
Like the other algorithms, when a target is detected, the node(s) pass the
information on for use.

Patrolling
Patrolling algorithms are similar to the wave algorithms in that only a series
of nodes are awake at one point of time. This algorithm just chooses which
nodes to wake differently. This can be done on an on-demand basis or a
coverage-orientated basis. The difference between the two is that on-demand
is some predetermined path that is currently interesting whereas coverage-
orientated just repeatedly watches an area [9].
The algorithm for both approaches is the same. First, a patrol is defined
as a set of information containing the patrol speed, duration, iteration period
and path. Once this information is collected, a Patrol Host node is selected,
which is usually the first node on the patrol path. The duty of the Patrol
Host is to periodically transmit the patrol data along with the current patrol
time. The idea is that over the time the patrol virtually traverses down the
specified path. As the patrol advances, nodes that are closest to the patrol
path wake up and collect information. Nodes determine if they are near the
patrol path by using the patrol information that the Patrol Hosts broadcast.
Additionally, as the patrol moves, new Patrol Hosts must be elected. As the
patrol virtually moves out of the current patrol host’s range, a new patrol
host is selected on the path which then takes over the patrol host duties and
the previous patrol host can now sleep. This process of waking nodes and
handing off patrol host duties continues until the end of the path is reached.
Once the path end is reached, the patrol is reversed in order to traverse the
path back to the original starting point. This process continues according to
the duration and iteration parameters are set in the original patrol parameters
[9].

Mobile Nodes
Until this point, all the target detection algorithms discussed so far have
assumed that the sensor nodes remain stationary. This section looks at a
 A Study on Vehicle Detection and Tracking Using Wireless ... 239

few target detection algorithms in which the nodes are mobile within the
network. Even though these algorithms allow nodes to be mobile, their goal
remains the same as the previously mentioned target detection algorithms.
These algorithms are still trying to detect targets in a timely and efficient
manner while trying to cover as much of the network as possible using as
little energy as possible. The theory behind allowing nodes to be mobile is
that they may now detect a target that would have otherwise gone undetected
in a stationary node network [10].

Past Detections
Like other algorithms, the network in this algorithm is divided into cell
areas. Each cell is unique in that it maintains the current state and past
history of targets that have been in the cell. This includes the location in
which the target was first detected within in the cell. Using this data, the
system calculates the areas of the cell that are most likely to contain a target
based on the locations of previously detected targets [11].
Once these locations are determined, each node is assigned an area
to watch and a priority. These priorities are used in the “coordination
mechanism” part of the algorithm. The coordination part of the algorithm is
responsible for altering the path nodes so that their observation paths overlap
as little as possible. For example, the node that watches the area where
targets are the most likely to occur is given the highest priority of the system
and its path is not altered. A second node, whose priority is slightly less than
the first node, whose path also slightly intersects that of the first node has its
path altered so that the two paths will not cross. This path altering continues
throughout the system from highest priority down to lowest priority until all
node paths are altered to prevent too much overlapping of observation areas
[11].

Collaborative Coverage
This algorithm is little different in that the paths the nodes traverse are not
coordinated in any way. Each node is allowed to create and follow its own
arbitrary path. While the node is travelling on this path at different locations
and times, it collects data and saves. The sensed data are saved with time
and location details [12]. While the nodes move across their paths they share
their collected data with those nodes they come into radio communication
with. Data is usually shared on a request basis. Each node is provided a
coverage threshold and confidence threshold. Actual target detection in this
240 Mechatronics and Robotics

system is determined by examining the data taken at several points that are
close to each other. If the combined data collected by this group of close
points exceeds some predetermined threshold then it is determined that a
target is detected at this area. Note that the value of this detection threshold
will determine how sensitive/ insensitive the system is at detecting targets.

Bayesian Estimation Track Before Detect (TBD) Estimator

Integrated tracking and detection, based on unthresholded measurements,
also referred to as Track Before Detect (TBD) is a hard nonlinear and non-
Gaussian dynamical estimation and detection problem. However, it is a
technique that enables the user to track and detect targets that would be
extremely hard to track and detect, if possible at all with ‘‘classical’’ methods.
TBD enables one to be better able to detect and track weak, stealthy or dim
targets in noise and clutter. The particle filters have shown to be very useful
in the implementation of TBD algorithms.In TBD, the detection problem is
done using the track output over multiple scans. The detection decision will
be made at the end of the processing chain, i.e., when all information has
been used and integrated over time [15].
Although it is called track before detect, the tracking and detection
processes occur simultaneously. In this way, the energy of a (weak) target
is integrated and correlated over time and position. The concept will lead to
a better performance when detecting and tracking weak targets. TBD also
implicitly solves the problem with data association.

Target Tracking Background

Target tracking algorithms usually focus on the aspect of the sensor nodes’
interaction with a target after the target has already been detected within the
area the sensor nodes cover. Once the object has been detected, the nodes
collect information and then use one of many different types of algorithms
to calculate the current location of the object relative to the sensor nodes’
locations. From here, it is the goal of the sensor network to track the object
as it moves through the network. This may or may not involve predicting the
next location of the object when it moves. In order to forewarn those nodes
it will be heading towards to prepare to capture data [14]. Several of these
tracking algorithms are discussed in the following section:

Simple Triangulation
This algorithm is an extremely simple design and implementation and
 A Study on Vehicle Detection and Tracking Using Wireless ... 241

serves as an ideal algorithm to use as the foundation for explaining other

more complex tracking algorithms. The whole goal of this algorithm [14]
is to provide a simple algorithm that uses simple computation in order to
calculate an object’s current location and predict where it is headed and
notify those nodes near the predicted next location of the object.
This algorithm first assumes that all nodes in the network are localized
to a common reference point and can detect and estimate the distance to a
target using signal strength [14]. When a node detects an object within its
range it broadcasts a Target Detected message.
This message contains the location of the sensor node and the distance
to the target. All nodes that hear this message store its data in their local
memory. When a node that has detected the target hears two other
Target Detected messages from two other nodes it performs triangulation
on the three coordinates to calculate the location of the target. (Note that this
means that more than one node may perform this calculation for the same
target at the same time.)
This node then continues to project the trajectory of the target. When
the estimated target trajectory has been calculated, all nodes that are within
some distanced’ perpendicular to the target’s trajectory are sent a Warning
message to alert them that the target is headed towards them. These newly
awoken nodes then track the object as it enters their area and repeat the
Target Detected and Warning message sending process.

Clusters
The cluster target tracking algorithm has been widely discussed through
many research papers. In this section, the basic idea of target tracking using
clusters is discussed, followed by an overview of the variances in each of the
different cluster algorithm research papers.

Basic Cluster Algorithm

The basic algorithm for tracking an object using clusters is as follows [20]:
• Some (or all) of the nodes in a cluster detect the object and report
their data to a cluster head. Note that each cluster has only one
cluster head.
• The cluster head node uses all the target detection information
from the sensor nodes to estimate the target’s location.
• The cluster head uses the calculated target location and past
242 Mechatronics and Robotics

locations of the target to predict the next location of the target.

• Those sensors around the predicted location are woken up to form
a new cluster (if not already in one) to detect the target.
• When the target is detected in this new cluster, the previous
cluster’s nodes are all put into a sleep state. This new cluster then
continues the cluster tracking algorithm.

Hierarchical Super nodes

This algorithm deviates from the basic algorithm in that the cluster heads
(called super nodes) have a higher Communication range and more
computational power [21]. These “super nodes” are distributed throughout
the network and the otherwise normal nodes are assigned to super nodes.
Clusters are not dynamically generated in this algorithm. Interestingly
enough, super nodes do share target location information among each other,
whereas regular sensor nodes do not.

Dynamic Clustering
Like the super node algorithm, this algorithm also assumes that cluster head
nodes have more power than normal sensor nodes. However, sensor nodes
are not assigned to clusters in this algorithm. Instead, they are invited to join
a cluster by the cluster head. The cluster head does this by broadcasting a
join message that includes the time and signature of the target the cluster
head detected [19]. Those sensor nodes that have stored data that matches
the data in the broadcast message respond to the broadcast by sending their
captured data to the cluster head node. Interestingly, the cluster head only
waits for a certain number of replies and when the required numbers of
replies are received, the cluster head calculates the target’s location. Unlike
the super node algorithm, this algorithm has one active cluster head at a
time. In other words, cluster heads do not work together [22].

Cell Collaboration
Like predicate clustering all nodes in this algorithm are of the same type.
Nodes are formed into clusters (called cells) in which all nodes within the
cells collaboratively decide when a target has entered their cell and if they
should track it [22]. It is interesting to note that cells can have different
sizes; the size of cells is determined by the observed velocity of the target.
So cell size will increase for faster moving targets.
 A Study on Vehicle Detection and Tracking Using Wireless ... 243

Probabilistic Localization
This algorithm deviates even further from the previously discussed
algorithms. Cluster heads are special high powered nodes that know the
location of every node within their cluster [23]. This algorithm takes
advantage of the fact that the cluster head knows these locations. When
sensor nodes detect a target, they send a very small notification message to
their cluster head and store the target location data, time and other relevant
data in their local memory. Upon receiving several notifications, the cluster
head performs a probabilistic localization algorithm to determine which
sensor nodes to query saved data from. In other words, the cluster head
runs an algorithm that helps to predict which sensor nodes are closest to the
target. Therefore, the cluster head has the best data to calculate an accurate
estimation of the target location. When the cluster head determines which
nodes to query, they are asked for the data they saved in their local memory
and the cluster head uses this information to calculate the location of the
target.

Distributes Predictive Tracking

This algorithm is very similar to the predicate localization algorithm in that
cluster head knows the ID, location and energy level of each of the nodes
within their cluster. However, not all nodes belong to a cluster. Nodes on
the border of the network are not in a cluster and those are on sense at all
times. Similarly, nodes next to the border are not in a cluster. Moreover the,
cluster heads only choose 3 nodes from target tracking data. Cluster heads
determine which 3 nodes to query based on their location in relation to the
predicted target location calculated by a previous cluster [24]. When the
cluster head has chosen which nodes to query, those nodes are woken up to
prepare to detect the approachable target. This algorithm differs even further
in that if a cluster head cannot find 3 suitable nodes within its own cluster, it
can seek the help of neighbouring clusters and ask for one or more of their
nodes to be turned on and report information.

Rooted Tree
Related to the clusters tracking algorithm is the idea of the rooted tree
tracking algorithm. This algorithm differs somewhat in that instead of
having multiple cluster heads, there is only one head node and it is referred
to as the root. The root node is the node that is closest to the target.
244 Mechatronics and Robotics

When a new target location is predicted, if there is another node that is

closer to that position than the current root then new node becomes the root.
Other nodes around the root work together to form a tree in which their sensed
data is collected and passed up through the tree (children to parents) until all
data reaches the root [25]. When the root receives all the data it calculates the
target position and predicts the new target position as mentioned in the basic
target tracking algorithm. The tree itself is reconfigured on every root change.
When a root change occurs, a “reconfigure” message is broadcast containing
the location of the new root. When a node detects the “reconfigure” message, it
detaches itself from the old tree and attaches itself to the new tree by recognizing
its neighbour node that is closest to the new root node as its new parent. New
data is then collected and passed up the tree for the new root to use.

Particle Filtering
The basic idea behind a particle filtering tracking algorithm is that numerous
object state descriptions are saved that contain data necessary to calculate
the target position at a certain time. These state descriptions are referred to
as particles and each particle has its own weight. The weight of a particle
determines how much the data it contains will contribute to the location
estimation of an object. When new particles are created, the weights of
the pre-existing particles are adjusted and then all the particles are used to
calculate a new target location [26]. Eventually, particles with weights that
are below a certain threshold are eliminated as duplicate particles.
Some of the particles filtering algorithms differ a little. Certain algorithms
have all the particles stored in a central node and this node does all the target
location processing [28]. Other algorithms distribute the stored particles
across the network nodes.

ANALYSIS OF ALGORITHMS
This section analyses each target tracking algorithm mentioned with a focus
on how it could be improved by combining it with one or more of the ideas
presented in the target detection algorithms.

Simple Triangulation
The main problem with this algorithm lies in the fact that when a node
has detected a target and hears two other nodes broadcast the same
target detection, the nodes starts to perform the triangulation localization
 A Study on Vehicle Detection and Tracking Using Wireless ... 245

calculation. This means that at the same point of time there will be multiple
nodes calculating the same target location. This is a waste of computation and
energy resources. This algorithm needs to be improved so that when a node
starts to calculate the location of the target, it first broadcasts its intention to
calculate the location to all the other nodes, similar to the data sharing that
is done in the collaborative coverage mobile detection algorithm. Then all
the other nodes that would have performed this same calculation will not
perform it once they hear the broadcasted intention of the first node.

Basic Cluster
The majority of the cluster algorithms had all the nodes on during their
target detection phase. It would be better if they instead adopt a technique
similar to that used in the wave detection algorithm to detect targets. Since
each area of the network is already divided into clusters, each node within
the cluster could take turns monitoring the area for the arrival of a target
using one of the wave detection methods mentioned. This would increase
the lifetime of the network.However, using one of the mobile detection
algorithms is not recommended since the nodes have to register with the
cluster head. Depending on the mobile detection algorithm used, the nodes
would constantly be moving in and out of clusters. This would cause an
increase in the amount of message communications with these nodes in
order keep track of the nodes within the cluster. Mobile detection algorithm
of this type can be applied to the situation only if the mobility of the nodes
is limited and the node always stay within the same cluster.Additionally, the
cluster algorithms could benefit from the basic fusion detection algorithm.
Once a target has been detected and all the information is passed to the
cluster head the cluster head could use the technique used in the basic fusion
algorithm to eliminate any data provide by faulty nodes. This would help
make the network less vulnerable to faulty nodes.

Dynamic Clustering
The biggest problem with this algorithm is that it will not work with mobile
nodes. The network calculates the sensor node’s locations at start-up. Since
these locations need to be known by the cluster head and are calculated
only once, this prevents any nodes from moving or even being added to
the network. This problem could be easily fixed by periodically refreshing
the locations of the sensor nodes, although this will decrease the network’s
lifetime. So it is better to use a detection algorithm and constantly refreshing
the nodes’ location isn’t an optimal solution is better in this case.
246 Mechatronics and Robotics

Cell Collaboration
This clustering algorithm is already perfectly setup to use the grid detection
algorithm for detection of targets. Since the nodes in this algorithm are
already divided into cells, the basic structure to use the grid algorithm is
already in place. All that needs to be done is to add the grid algorithm for
the detection part. Once, this is done the algorithm would be more efficient
since the nodes would be alternating sleep schedules with each other in order
to detect the target instead of all being awake at the same time.

Probabilistic Localization
This algorithm already sounds like that it has incorporated parts of the hybrid
detection algorithm in it. When a target is detected, this algorithm sends a
small message to the cluster head. Similarly, when a target is detected in the
hybrid algorithm a small yes/no bit message is sent to the decision maker in
the network, or cluster head.
Then the cluster head uses this data to determine which nodes to query
for further data. In determining which nodes to query, it would be a good
idea to use an averaging technique similar to that used in the basic fusion
detection algorithm in order to eliminate the impact of any data provided by
a faulty node. The similarities to the hybrid algorithm continue. When the
cluster head determines which nodes to query, the nodes respond back to the
cluster head with more detailed information.
However, this algorithm should not be combined with a mobile node
detection algorithm because of the way it decides which nodes to query.
In this algorithm, the cluster heads create probability tables in order to
determine which sensors to query [23]. The use of a table is acceptable until
nodes move outside or are added to the system. When this happens, the
probability table must be updated. Therefore, the use of a mobile detection
scheme would cause a lot of additional computational overhead since these
tables are to be constantly refreshed.

Distributed Predictive Tracking

This algorithm is unique in that it is already doing its own target detecting.
Instead of having the border nodes being on at all times to detect nodes
it would be beneficial to have these nodes adopt a wave or patrol like
technique. This way, they would still detect targets but would be consuming
a considerably smaller amount of energy.
 A Study on Vehicle Detection and Tracking Using Wireless ... 247

Rooted Tree
One of the basic problems with the rooted tree technique is that it doesn’t
provide any fault tolerance facts. In order to make this algorithm less
vulnerable to problems caused by faulty nodes, it should use an averaging
technique similar to that used in the basic fusion detection algorithm.
Additionally, this algorithm encounters problems when the velocity of
the target is very high. As the velocity of the target increases so does the
number of times in which the tree must be reconfigured [25]. This will cause
problems when a threshold velocity of the target is reached in which the
tree can no longer be generated quickly enough to keep up with the moving
target. This problem could be eliminated by altering this algorithm to work
more like the patrolling algorithm. Instead of constantly rebuilding the tree,
the tree is treated as a path that can have steps (or nodes) along the path
added or deleted from it. This way when the target moves, just add a new
node to the path and remove any one that is too far away to contribute any
longer. The basic nodes needed to collect data are still members of the path
and new ones are added. It saves an enormous amount of time and message
communication needed to periodically rebuild the tree structure each time.

Particle Filtering
In order to track targets, this algorithm saves an enormous amount of data
in the form of particles. Over time, these particles take up a large amount
of memory space. The amount of memory actually used could be decreased
if this algorithm is to adopt the idea behind the TBD algorithm. Bayesian
Estimation Track Before Detect (TBD) based on particle filtering gives high
performance and hybrid detection algorithm stops recording data once the
data it currently contains exceeds a threshold to know whether a target is
present or not. The particle filtering algorithm could use the same principle.
Once enough particle data is collected to exceed some threshold, the node
would stop collecting data. This would not only decrease the amount of
memory used but also decrease the amount of work to be done to save energy.

Evaluation of Target Detection and Tracking Algorithms

The best algorithm is established based on the evaluation criteria. The
parameters used for evaluation are time, energy, propagation delay,
vulnerability, memory used, stability and scalability. Particle Filtering based
on the Bayesian TBD estimator algorithm that sounds the best for vehicle
detection and tracking in WSNs. Table 4 shows the evaluation summary.
248 Mechatronics and Robotics

SUMMARY OF ALGORITHMS
It is observed that, one of the best target detection and tracking algorithms
is the combination of the Bayesian Estimation Track before Detect (TBD)
Estimator and particle filtering algorithm. The fact that the TBD covers the
network area in a short amount of time and enables nodes to sleep at times
saves a huge amount of energy. This enables a target to be detected, without
too much delay, in an energy efficient manner.
The particle filtering algorithm is also on the right track since the pre-
existing particles are adjusted when the new particles are created and hence
particles with weights that are below a certain threshold are eliminated as
duplicate particles, which also conserves energy. Additionally, the use of
several central localization calculating points instead of one helps to reduce
the vulnerability of the system to an attacker. It eliminates the one point
of failure problem. It also shares the burden of energy usage that one node
would face across several nodes.
Particle filtering approach allows all the nodes to share the energy burden.
Hence particle filtering algorithm combined with Bayesian Estimation Track
before Detect (TBD) Estimator algorithm can be used for target detection
and tracking.

CONCLUSIONS AND FUTURE DIRECTIONS

In this paper an attempt is made to gather the information about the
unauthorized vehicle detection and tracking in the battlefield surveillance
and also to survey the sensors that are widely available for vehicle
detection. The effort is also set to survey and evaluate the detection and
tracking algorithms intended for target (eg vehicle or animal) detection and
tracking using Wireless Sensor Networks. An overview of each algorithm
type was presented. The combinations of target detection algorithms with
target tracking algorithms are also discussed. The various combinations of
algorithms have been classified and each category is evaluated according
to the identified criteria. Particle Filtering based on the Bayesian TBD
estimator algorithm is an interesting one for target detection and tracking
using WSNs. It has good resource efficiency, propagation delay is minimum.
It also increases scalability and particle based approach increases stability.
Real time experiments are required in order to conclude whether a Particle
Filtering based on Bayesian TBD estimator algorithm will work in a practical
scenario. This paper aims to give a brief overview of sensors and algorithms
used for Vehicle detection and tracking which directs the future researchers
 A Study on Vehicle Detection and Tracking Using Wireless ... 249

to show new directions used in battlefield surveillance or in any place

where human monitoring is not possible.

ACKNOWLEDGMENT
The authors would like to thank the Armament Research Board (ARMREB-
DRDO) for supporting this Research project by funding.
250 Mechatronics and Robotics

REFERENCES
1. Y. C. Tseng, S. P. Kuo, H. W. Lee, and C. F. Huang, “Location tracking
in a wireless sensor network by mobile agents and its data fusion
strategies,” Lecture notes in computer science, Springer Link, pp. 2–3,
2003.
2. E. Y. Luz and A. Mimbela, “Summary of vehicle detection and
surveillance technologies used in intelligent transportation systems,”
The Vehicle Detector Clearinghouse, Southwest Technology
Development Institute (SWTDI) at New Mexico State University
(NMSU), Fall 2007. http://www.nmsu.edu/traffic/.
3. S. Coleri, M. Ergen, and T. J. Koo, “Lifetime analysis of a sensor
network with hybrid automata modelling,” Processings of ACM
International Workshop on Wireless Sensor Networks and Applications
(Atlanta, GA), pp. 98–104, 2002.
4. D. Jiagen, S. Y. Cheung, C. W. Tan, and V. Pravin, “Signal processing of
sensor node data for vehicle detection,” IEEE Intelligent Transportation
Systems, pp. 70–75, 2004.
5. C. Thomas, K. S. Kewel, and R. Parameswaran, “Fault tolerance in
collaborative sensor networks for target detection,” IEEE Transactions
on Computers, pp. 320–333, 2004. http://citeseer.ist.psu.edu/
clouqueur03fault.html.
6. F. Qing, Z. Feng, and L. Guibas, “Lightweight sensing and
communication protocols for target enumeration and aggregation,”
Proceedings of the 4th ACM international symposium on Mobile ad
hoc networking & computing, pp. 165–176, 2003.
7. Y. Ting, H. Tian, and A. S. John, “Differentiated surveillance for
sensor networks,” Proceedings of the 1st international conference on
Embedded networked sensor systems, pp. 51–62, 2003. http://www.
cs.virginia.edu/papers/ diff_surveillance_sn_p51-yan.pdf.
8. S. S. Ren, Q. Li, H. N. Wang, and X. D. Zhang, “Design and analysis of
wave sensing scheduling protocols for object-tracking applications,”
Lecture notes in computer science, SpringerLink, pp. 228–243, 2005.
http://www.cs. wm. edu/~liqun/paper/dcoss05.pdf.
9. C. Gui and P. Mohapatra, “Virtual patrol: a new power conservation
design for surveillance using sensor networks,” Information Processing
in Sensor Networks, pp. 246–253, 2005.
10. B. Y. Liu, P. Brass, O. Dousse, P. Nain, and D. Towsley, “Mobility
 A Study on Vehicle Detection and Tracking Using Wireless ... 251

improves coverage of sensor networks. international symposium on

mobile ad hoc networking & computing,” pp. 300–108, 2005.
11. M. K. Krishna, H. Hexmoor, and S. Sogani, “A t-step ahead constrained
optimal target detection algorithm for a multi sensor surveillance
system,” IEEE Intelligent Robots and Systems, pp. 357–362 2005.
http://arxiv.org/ ftp/cs/papers/0505/0505045.pdf.
12. K. C. Wang and P. Ramanathan, “Collaborative sensing using sensor
of uncoordinated mobility,” Lecture notes in computer science,
SpringerLink, pp. 293–306, 2005. http: //www.ece.wisc.edu/~wander/
papers/dcoss05_wang.pdf
13. E. B. Ermis and V. Saligrama, “Adaptive statistical sampling methods
for decentralized estimation and detection of localized phenomena,”
ACM’05. http://iss.bu.edu/ srv/Publications/Manuscript-final.pdf.
14. R. Gupta and S. R. Das, “Tracking moving targets in a smart sensor
network,” Vehicular Technology Conference, pp. 3035–3039, 2003.
http://www.cs.sunysb.edu/ ~samir/Pubs/VTC03-long.pdf.
15. S. J. Davey, M. G. Rutten, and B. Cheung, “A com parison of detection
performance for several track-bef ore-detect algorithms,” EURASIP
Journal on Advance s in Signal Processing, 2008, http://www.hindawi.
com /journals/asp/2008/428036.html.
16. B. Tatiana, H. Wen, K. Salil, R. Branko, G. Neil, B. Travis, R. Mark,
and J. Sanjay, “Wireless sensor networks for battlefield surveillance,”
Land Warfare Conference, 2006.
17. L. A. Klein, M. K. Mills, and D. R. P. Gibson, “Traffic detector
handbook,” Operations and Intelligent Transportation Systems
Research, 2006. http://www.tfhrc.gov/ its/pubs/06108/index.htm.
18. L. Yu, L. Yuan, G. Qu, and A. Ephremides, “Energydriven detection
scheme with guaranteed accuracy” Information Processing in Sensor
Networks, pp. 284–291, 2006.
19. W. P. Chen, J. C. Hou, and L. Sha, “Dynamic clustering for acoustic
target tracking in wireless sensor networks,” 11th IEEE International
Conference on Network Protocols (ICNP’03), November 2003.
20. D. Li, K. D. Wong, Y. H. Hu, and A. M. Sayeed, “Detection,
classification and tracking of targets in distributed sensor networks”
IEEE Signal Processing Magazine, pp. 1163–117, 2002.
21. S. Oh and S. Sastry, “A hierarchical multiple-target tracking algorithm
for sensor networks,” IEEE Robotics and Automation,” pp. 2197–2202,
252 Mechatronics and Robotics

2005. http://www.eecs. berkeley.edu/~sho/papers/icra05_hmttsn.pdf.

22. R. B. Richard, R. Parameswaran, and M. S. Akbar, “Distributed target
classification and tracking in sensor networks,” In Proceedings of the
IEEE, Vol. 91, No. 8, pp. 1163–1171, 2003. http://www-net.cs.umass.
edu/cs791_ sensornets/papers/brooks.pdf.
23. Y. Zou, and K. Chakrabarty, “Target localization based on energy
considerations in distributed sensor networks,” 1st IEEE Workshop
Sensor Network Protocols and Applications (SNPA’03), pp. 51–58,
2003.
24. H. Yang and B. Sikdar. “A protocol for tracking mobile targets using
sensor networks, sensor network protocols and applications,” pp. 71–
81, 2003. http://networks.ecse. rpi.edu/~bsikdar/papers/hua_icc03.pdf.
25. W. S. Zhang and G. H. Cao, “Optimizing tree reconfiguration for
mobile target tracking in sensor networks,” Infocom, pp. 2434–2445,
2004.
26. G. Ing, “Distributed particle filters for object tracking in sensor
networks,” Proceedings of the 3rd international symposium on
Information processing in sensor networks pp. 99–107, 2004.
 CHAPTER

13
Deployment of Pre-Industrial
Autonomous Microbe Sensor in Saudi
Arabia’s Injection Seawater System

Mohammed A. Al-Moniee1, Xiangyang Zhu1, Rikke Markfoged2,

Aabdullah H. Al-Wadei1, Poul L. Pedersen2, Anders K. Tuxen2,
Fuad I. Al-Nuwaiser1, Lone Tang2, Tinna Staghøj Roesen2, Thomas
Lundgaard2
1
Saudi Arabian Oil Company (Saudi Aramco), Dhahran, Saudi Arabia
2
Danish Technological Institute (DTI), Aarhus, Denmark

ABSTRACT
Microbial growth in water injection systems can lead to many problems,
including biofouling, water quality deterioration, injectivity loss, microbial
corrosion, and reservoir formation damage. Monitoring of microbial

Citation:- A. Al-Moniee, M. , Zhu, X. , Markfoged, R., Al-Wadei, A. , Pedersen, P.,

Tuxen, A., Al-Nuwaiser, F., Tang, L., Staghøj Roesen, T. and Lundgaard, T. (2018),
“Deployment of Pre-Industrial Autonomous Microbe Sensor in Saudi Arabia’s Injection
Seawater System”. Journal of Sensor Technology, 8, 1-17. doi: 10.4236/jst.2018.81001.
Copyright: © 2018 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
254 Mechatronics and Robotics

activities is required in any mitigation strategy, enabling operators to

apply and adjust countermeasures properly and in due time. In this study,
the pre-industrial autonomous microbe sensor (AMS) was constructed
with technical improvements from the prototype for increased sensitivity,
durability, robustness, and maintainability. The pre-industrial AMS was
lab validated, field proven, and deployed at critical locations of seawater
injection network for automated detection of microorganisms under the
Saudi Arabia’s harsh environment. An excellent correlation between AMS
measurement data (fluorescence count) and actual count of microbial cell
number under microscope was established (coefficient of determination,
R2 > 0.99) for converting AMS fluorescence count to cell numbers (cell
mL−1) in the injection seawater. The pre-industrial AMS only required
monthly maintenance with solutions refill, and was able to cope with hot
summer months even without protection in an air-conditioned shelter. The
study team recommended wider deployment of the online AMS for real-time
monitoring of bacteria numbers in the various strategic locations in Saudi
Aramco’s complex seawater injection network, as an integral component of
pipeline corrosion and leak mitigation program.
Keywords: Automated Monitoring, Autonomous Microbe Sensor,
Microbial Sensor, DNA Staining, Injection Seawater

INTRODUCTION
Saudi Arabian Oil Company (Saudi Aramco) has the largest seawater injection
system in the world. The growth of microorganisms and the formation
of biofilm on the pipeline’s inner surfaces and process equipment lead to
an array of challenges, including biofouling, water quality deterioration,
infectivity loss, reservoir plugging, and microbiologically influenced
corrosion (MIC) [1] [2] . A critical cornerstone in an effective mitigation
strategy is the routine surveillance of the system microbiology, enabling
operators to apply and adjust microbiological countermeasures accordingly.
The microbial number and activity are traditionally monitored with
conventional growth methods that require manual sampling and handling.
Monitoring the microbial activity with conventional methods in this large
system, especially at the remote locations, presents a long-standing challenge
to Saudi Aramco. In an attempt to ensure the injection water quality and
maintain the integrity of injection pipelines network, Saudi Aramco and the
Danish Technological Institute (DTI) embarked on a three-phase project
 Deployment of Pre-Industrial Autonomous Microbe Sensor in ... 255

with an overall objective of developing and deploying an online sensing

technology for automatic and real-time monitoring of microbial numbers in
Saudi Aramco’s large injection seawater system.
In the feasibility phase study [3] [4], five single-analyte methods were
evaluated in the laboratory setup for the suitability of automation for detection
of microbial activity in the Saudi Aramco injection seawater. Staining
of cells with DNA binding fluorescent dyes (SYBR Green), followed by
quantification of fluorescence signals, was identified as a reliable single-
analyte method for automated, online determination of microbial cell
numbers in the injection seawater system.
In the subsequent phase, an autonomous microbe sensor (AMS)
prototype was constructed, tested, and optimized in the laboratory, followed
by a validation in the field for automated detection of microorganisms in the
harsh Saudi Arabia desert environment [5]. In the field validation, the AMS
prototype was able to monitor and follow the general microbial status in the
system, including detection of periods with increased microbial growth or
decreased microbial numbers following biocide injection. The infield AMS
prototype detection limit was 105 cells/ml.
The objective of the current phase study was to implement the technical
improvements identified in the prototype field validation [5], construct two
pre-industrial AMS devices, validate and deploy the pre-industrial AMS in
Saudi Aramco’s seawater injection network. The technical improvements
aimed to increase the sensitivity, durability, robustness, and maintainability,
while reducing the uncertainties related to AMS operation and function.

TECHNICAL IMPROVEMENTS OF PRE-INDUSTRI-

AL AMS
The pre-industrial AMS is an instrument that provides an online and real-
time detection of total microbial numbers in seawater pipelines. It is provided
with a flow cell that measures the fluorescence intensity of microorganisms
in the seawater sample after mixing with DNA staining dye SYBR Green
[6] [7] [8] . The sensor contains all necessary ancillary systems to sample
seawater and fluorescence dye, inject them in precise volumes, mix them and
dispose them after measurement, to a waste container. The microbial number
is derived from the fluorescence intensity through a calibration procedure
that can be checked periodically. The calibration procedures involve the
fluorescence intensity measurement with AMS, the temperature during
256 Mechatronics and Robotics

measurement and the direct microscopy counting of known concentrations

of bacteria dilutions after staining with 4, 6-diamindino-2-phenylindole
(DAPI) [9] [10] [11] .
The areas for technical improvement and optimization identified in the
field testing of the AMS prototype [5] have been incorporated in the design and
construction of the pre-industrial AMS. Some of the improvement areas were:
1) Optimization of cleaning procedures: The internal AMS cleaning
procedures were optimized to minimize the risk of bio fouling
and iron precipitation. Alkaline (NaOH) and acid (oxalic acid)
cleaning procedures were incorporated into the measurement
cycle to prevent the biofouling and formation of iron precipitations
in liquid handling components.
2) Incorporation of an internal DNA control: The internal DNA
control was used to validate the AMS functionality and
compensate the temperature effect on fluorescence counts.
3) Optimization of fluorescence detection procedure and software: The
sampling procedure and software were optimized in the pre-industrial AMS
to minimize the scattering of data points caused by uneven cell distribution.
4) Sensitivity improvement: The detection limit of pre-industrial
sensor was improved by optimizing hardware components
(spectrophotometer, optical fiber, and flow cell) and microbial
staining and measurement procedures.
5) Incorporation of temperature correction factor: The temperature
inside the AMS was measured simultaneously with the
fluorescence measurements. The inside temperature and the
fluorescence count of internal DNA control were used to develop
a temperature compensation factor to compensate the temperature
effect on the fluorescence dye.
6) Redesigned user interface: The new software interface, layout
and data output allowed data export and automatic plotting and
display in remote computer.

DESIGN AND CONSTRUCTION OF PRE-INDUSTRI-

AL AMS
The design of pre-industrial AMS has incorporated the technical improvement
areas identified from AMS prototype. Design changes from the prototype to
the pre-industrial AMS are summarized in Table 1.
 Deployment of Pre-Industrial Autonomous Microbe Sensor in ... 257

The pre-industrial AMS consists of three separate modules: hardware

module, consumables module, and power module:
1) Hardware module: includes a custom-made flow cell for achieving
flow during quantification, a heating chamber for improved
staining conditions and two dedicated pumps.

Table 1. Design changes from the prototype to the pre-industrial AMS.

Part Prototype AMS Pre-industrial AMS

Unit layout Single box Three modules: hardware, consum-
ables, and battery
Internal DNA N/A Synthetic DNA in
control ethyleneiaminetetraacetic acid
(EDTA) and ascorbate buffer
Sample stain- 40 min at ambient tempera- 13 min at 60˚C
ing ture
Flow cell Commercially available Custom-built flow cell with mini-
flow cell mal volume and large numerical
aperture
Measurement Sample is static when Sample flows slowly when illumi-
illuminated nated to avoid bleaching of dye
Dye dispensing Fixed volume Dedicated syringe pump for dye
Lower limit of ~10 5
Low limit of detection, LLOD
field detection − AMS 1: 1.9 × 104; AMS 2: 1.2
× 104
Low limit of quantification, LLOQ
− AMS 1: 5.0 × 104; AMS 2: 3.2
× 104
Cleaning NaOH NaOH and oxalic acid
Power AC mains Battery, charged either by AC
mains or solar panels
Remote com- Team Viewer through Saudi Permanent RS-485 connection.
munication Aramco intranet when AMS Data retrieval and AMS control
is on through dedicated software on re-
mote computer
Pump A is for dispensing water sample and cleaning solutions; while
Pump B is dedicated for SYBR Green dye distribution.
2) Consumables module: includes four containers for water, oxalic
acid, sodium hydroxide and waste.
258 Mechatronics and Robotics

3) Power module: A 24 V DC battery pack was chosen as the power

supply. The batteries are connected to three solar panels or 230 V
AC mains power.
The modular design of AMS improves the equipment functionality and
durability in the challenging Saudi Arabia dessert environment and operation
conditions of seawater injection system [4]. Each of these three modules are
mounted in a stainless-steel cabinet and may be stacked with the battery module
at the bottom, the consumables module in the middle and the hardware module
at the top (Figure 1). The AMS is designed to run on batteries that are charged
by solar panels or 230 V mains power. The AMS device installed in the field is
connected to the remote computer through a MODBUS protocol using RS-485
cable. The Lab View application [12] on the remote computer controls the sensor
operation, equipment shut down following measurements, and data retrieval.

PERFORMANCE ASSESSMENT

Lab validation
After construction of two pre-industrial AMS units, the AMS was thoroughly
tested and validated in the laboratories at DTI premises. The repeatability
and functionality of the units were tested with respect to procedures as well
as software. The major improvements in lab validation and their effect on
AMS performance are summarized in Table 2.
After implementation of above improvements on AMS units, a correlation
of fluorescence signals from AMS and direct counting of microbial cell
number under microscope after DAPI staining was established (Table 3).

Figure 1. Hardware module, consumables module, and power module of pre-

industrial AMS.
 Deployment of Pre-Industrial Autonomous Microbe Sensor in ... 259

Table 2. Major improvements incorporated during the lab validation.

Improvement Effect
Decrease of Pump A rate to 3 Eliminate air from surroundings into sy-
mL・min−1 ringe A
New loading procedure for internal Improved mixing of internal control (IC)
DNA control DNA and Milli-Q water
Increased total volume of diluted in- Improved loading of IC DNA
ternal DNA control to sample volume
New dye loading procedure. Dead Accurate dye dosing
space in Pump B valve used as dye
dosing volume
Selection of incubation conditions of Improved staining of cells with SYBR
13 min at 60˚C Green I
No air is moved through flow cell Eliminate introduction of air bubbles in
flow cell
Flow direction of flow cell changed Remove potential air bubbles in flow cell
New flow rate through flow cell Optimal flow rate for maximum reading
Increased rate of dispensing and Minimize total AMS measurement cycle
aspirating with Pump A for all non- time to 75 min
critical steps
Incorporated data export function For automatic plot of raw and calibrated
readings on graph in Excel

The correlation is performed using a series of dilutions of sulfate-

reducing bacteria (SRB) enriched from Saudi Aramco injection seawater
and diluted in phosphate buffer (pH 8, 5.5% NaCl).
The correlation is required to convert the data obtained by the AMS, in
terms of fluorescence counts, into cell count per volume (cells mL−1).
The calibrations show that the AMS units are linear in a wide range of
at least two orders of magnitude, with a laboratory detection limit of around
103.
From the performed calibrations, the AMS units were accepted for field
installation, calibration, and validation.
At the end of lab validation, the optimal AMS measurement settings
were determined and implemented in both AMS units (Table 4).
260 Mechatronics and Robotics

Table 3. Laboratory calibration data and equation for AMS1 and AMS2.

AMS1 AMS2
Fluorescence Cells mL −1
Fluorescence Cells mL−1
count (AMS) count (AMS)
Sterile filtered sys- 154 ± 7, n = 3 98 ± 3, n = 3
tem water (blank)
LLOD 167 6.0 × 103 104 8.8 × 102
LLOQ 188 1.6 × 104 114 2.3 × 103
Regression equa- 473X − 72,788 146X − 14,265
tion
Regression, R2 0.9902 0.9490
Linear range 6.4 × 10 - 5.0 × 10
3 6
1.1 × 104 - 2.4 × 106

LLOD: Lower Limit of Detection; LLOQ: Lower Limit of Quantification.

Table 4. Final settings for one full AMS measurement cycle.

Improvement Effect
Volume of sample 4 ml
Volume of IC DNA 4 ml (0.1 ml concentrated IC discarded at first,
0.2 ml IC in 3.8 ml sterile H2O)
SYBR Green I Diluted 1:20 in Dimethyl sulfoxide (DMSO).
Volume equals to dead space in Pump B (20 μL)
Incubation of sample 13 min at 60˚C
Incubation of IC Less than one minute at ambient temperature
Flow rate and time 3.5 μL・s−1 of 180 sec and 10 sec
Measurement wavelength inter- 520 - 530 nm
val
Fluorescence count Average of readings every one sec for 180 and 10
sec
Background reading (auto-fluo- To be subtracted from sample reading. Fluorescence
rescence) reading of sample without dye, for correction of any
auto-fluorescence of the sample water. Background
reading is not subtracted from internal control readings.

Cleaning Sterile H2O and air between sample and IC mea-

surement. Acid, alkaline, sterile water and air at
the end of a cycle. No air through the flow cell.

Time consumption 75 min for one full cycle, including washing/

cleaning
 Deployment of Pre-Industrial Autonomous Microbe Sensor in ... 261

Field installation and Calibration

After laboratory validation at DTI premises, both AMS units were shipped to
Saudi Aramco, and tested for functionality and stability after transportation.
After functionality check, AMS1 was connected to the side-stream of a 60”
shipping line at an injection facility and AMS2 was installed at an injection
plant further downstream of AMS1, connecting to a side-stream of 40”
seawater pipeline. The seawater flow continuously passes AMS through a
bypass connection to the disposal when AMS is not doing the measurement.
During the measurement cycle, a fixed volume of seawater sample is taken
into the AMS for staining and measurement of microbial cell number. During
the field testing and validation, the sensors were programmed to take two
measurements a day, with a monthly maintenance for solutions refill. The
data were retrieved by the operators in the Control Room.
The fluorescence intensity from a stained sample depends on the amount
of DNA present in the sample, which in turn, depends on the cell size and
the metabolic status of the cell, as an actively dividing cell contains more
DNA than a dying cell [13] [14] [15] . Therefore, the fluorescence intensity
from the AMS was re-calibrated in Saudi Aramco with active aerobic
heterotrophic bacteria enriched from injection system seawater. A 3-point
field calibration was performed on AMS1 and AMS2 using the bacteria
dilution series prepared with sterile filtered system water.
The calibration correlates the cell number (cell mL−1) of the system water
determined by DAPI staining and counting to AMS fluorescence counts
from dilution series of the sample measured in the field. A typical DAPI
staining for counting microbial cell number under microscope is shown
in Figure 2. The results of 3-point field calibration between cell number and
fluorescence count is shown in Figure 3.

Figure 2. DAPI staining and counting under microscope.

262 Mechatronics and Robotics

Figure 3. Relation between cell number and fluorescence count of AMS1 (blue)
and AMS2 (red) using enrichment culture from the system, diluted in sterile
filtered system water. All AMS readings are performed with an AMS interior
temperature of 30˚C.
A linear regression between AMS fluorescence measurement and DAPI
count for converting AMS fluorescence count to cell numbers in the sample
is summarized in Table 5. Compared to the calibration performed in the
laboratory, these data show that the fluorescence count relative to the cell
number is lower in the field, and therefore the regression equation in the lab
at 25˚C is not recommended for field trial. The field regression equations for
each AMS unit were loaded in the specific AMS units, and used for the data
conversion from fluorescence count to cell number.

Temperature effect on Fluorescence Intensity

During the field validation of the AMS units, a tendency of higher
fluorescence counts with lower AMS inside temperatures has been observed.
A full correlation of all internal DNA control readings shows an exponential
decrease in fluorescence with increasing temperature, which fits well with
an exponential function with both a vertical and horizontal offset [Figure
4 and Equation (1)].

(1)
where the fitted parameters are A = 2.4 × 104 s−1, B = −0.10˚C−1, x0 =
4.16˚C, C = 2.49 × 103 s−1, and x = AMS inside temperature (˚C).
 Deployment of Pre-Industrial Autonomous Microbe Sensor in ... 263

The function f(x) between temperature and fluorescence intensity was used
to temperature compensate internal control (IC) readings. The compensation
is done by multiplying the IC reading with the ratio f(30)/f(T), where 30 is
the reference temperature, and T is the AMS inside temperature of the given
data point.

Table 5. Field calibration data and equation for AMS1 and AMS2.

AMS1 AMS2
Fluorescence Cells mL −1
Fluorescence Cells mL−1
count (AMS) count (AMS)
Sterile filtered sys- 114 ± 12, n = 3 123 ± 7, n = 3
tem water (blank)
LLOD 136 1.9 × 104 137 1.2 × 104
LLOQ 173 5.0 × 104 160 3.2 × 104
Regression equa- 843X − 96,020 878X − 107,935
tion
Regression, R2 0.9939 0.9940

Figure 4. Internal control DNA fluorescence readings versus temperature inside

AMS during field readings. October-December data have been fitted with an
exponential function with both a vertical and horizontal offset.
The temperature effect on fluorescence intensity is likely related to
the molecular structure of SYBR Green I, as the supplier of SYBR Green
(Invitrogen) explained that “the temperature effect is likely caused by
less rigid bonds in the fluorescent molecule as the atoms become excited
264 Mechatronics and Robotics

and move around.” Therefore, the observed temperature response is a

direct consequence of the reduced SYBR Green fluorescence at elevated
temperatures. In this case, the fluorescence intensity of stained IC DNA or
stained cells are equally affected by temperature.
An IC measurement is performed for each AMS run with equal DNA
content for each run. As the IC and system water samples are affected
equally by temperature, the IC can, besides validating the performance of
the AMS unit, also be used for an internal temperature compensation. Figure
5 and Figure 6 showed the raw fluorescence readings and readings after
temperature compensation from IC DNA and system seawater, respectively.
Other possible causes for the temperature dependency (melting of
oligonucleotides DNA and heating of spectrometer) have been considered
unlikely. Internal control DNA (a synthetic 55-base pair oligonucleotides
with hairpin structure) was tested for the stability at a wide range of
temperature. The synthetic DNA in EDTA and ascorbate buffer was found
to be stable from 20˚C to 80 ˚C for at least 28 days when stored in a sealed
container.

Figure 5. Internal control DNA fluorescence readings (October-December)-

raw (top) and temperature compensated (bottom).

Figure 6. System seawater fluorescence readings (October-December)-raw

(top) and temperature compensated (bottom).
 Deployment of Pre-Industrial Autonomous Microbe Sensor in ... 265

Melting of the double stranded (ds) Oligo DNA (internal control) would
result in a lower binding between the SYBR molecule and the DNA, as
SYBR Green I only binds to ds DNA [6] [16] . The oligo DNA as well as the
internal control buffer have been designed to withstand high temperatures,
and based on calculations from the Oligo DNA supplier (Integrated DNA
Technologies), the melting temperature (Tm, 50%) is 73˚C for the conditions
provided from the internal control buffer solution (Figure 7).
The spectrometer must be cooled for proper readings, and cooling has been
included in the design of the AMS. To test whether the AMS spectrometer is
affected by AMS inside temperature, fluorescein was injected manually into
the flow cell at high and low temperatures (40˚C and 25˚C, respectively).
Fluorescein is a dye that emits fluorescent light with excitation and emission
spectra comparable to those of SYBR Green when bound to DNA [17]. The
effect of AMS inside temperature on spectrometer readings of fluorescein at
two concentrations was not detected (data not shown).

Field measurements
The AMS1 measurement data, from October 2016 to February 2017, on the
fluorescence count of the IC DNA and microbial cell number in the seawater
are plotted in Figure 8 and Figure 9, respectively. Figure 8 shows that the
IC readings are relatively stable in this 4-month period, after temperature
compensation of fluorescence intensity, verifying the stability of both IC
DNA and SYBR Green dye, and the functionality of AMS in the field. Figure
9 shows that the majority of the measurements of seawater cell number are
above the detection limit. When system water fluorescence count has been
temperature compensated, the cell numbers from winter are in the range
comparable to fall cell numbers.
Temperature effect on fluorescence intensity is determined to be
significant when SYBR Green I dye is used to stain IC DNA and cell DNA in
seawater samples. As a result, temperature compensation for all fluorescence
counts of both IC DNA and microbial cells is required. The temperature
compensation factor was incorporated in the software for converting the raw
data to temperature compensated data for data exporting and plotting.

LESSONS LEARNED
During the field testing and validation, a number of technical challenges
have been either encountered or identified, and the solutions have been
implemented to address these challenges (Table 6). Among the challenges,
266 Mechatronics and Robotics

the two main threats are the pressure regulator clogging and leakage from
the fluid handling system, which are discussed in detail in the following
text, to illustrate how the technical challenges have been resolved.
The pre-industrial AMS is equipped with a restrictor to reduce the side-
stream pressure from 10 bars to less than 1 bar before water enters the AMS
system. The restrictor is equipped with a pressure gauge to monitor inlet
pressure. Since the seawater pipeline pressure is much higher, it is important
that another pressure regulator is installed on the side-stream to reduce the
pressure to about 10 bars for prevention of damage to AMS hardware.

Figure 7. Hybridization profile of the Oligo DNA used as internal control. The
total DNA is intact up to 70˚C, and the Tm 50% is 73˚C.

Figure 8. AMS1 autonomous sampling and measurement for fluorescence sig-

nals of IC DNA from October 2016 to February 2017-raw (top) and temperature
compensated (bottom).
 Deployment of Pre-Industrial Autonomous Microbe Sensor in ... 267

Figure 9. AMS1 autonomous sampling and measurement for microbial cell

number from October 2016 to February 2017-raw (top) and temperature com-
pensated (bottom).
Due to the harsh conditions in the seawater environment, the needle type
pressure regulator tends to be easily clogged. To avoid clogging, the ball
valve type or spring type pressure regulator is required for both side-stream
and AMS. After the right type of pressure regulator is selected and installed,
the regulator clogging and potential AMS damage due to high pressure are
eliminated.
AMS liquid handling system consists of two syringe pumps, two multi-
channel valves, and many tubings to transport various liquids around the
AMS system for measurement, cleaning, and disposal. Loose fittings, wear
and tear, and corrosion have been noticed during long-term field trials.
Replacement of original Delrin fittings with PEEK fittings has solved the
corrosion problem caused by oxalic acid and reduced the leaking occurrence.
Occasional tightening and close monitoring of the fittings are required when
there is large ambient temperature fluctuation.
268 Mechatronics and Robotics

Table 6. Lessons learned from field test of the AMS.

Element Issue Lessons Learned Action Required

name
Reagents and solutions
SYBR dye Experienced inac- Solvent DMSO in Consider insulating SYBR
curate volume of dye the dye will freeze vial or enclosing AMS
collected by syringe. below 12˚C. modules within an air condi-
tioned shelter.
Reaction and measuring system

Flow cell Biocide entering Need to coordinate Programmed AMS measure-

flow cell. with corrosion en- ment time around the esti-
gineer for biocide mated biocide arrival time.
injection time.
Liquid handling system
Pressure Experienced clog- Select right type of Replaced original needle
regulator ging of pressure regulator. type pressure regulator with
regulator inside a ball valve to reduce the
AMS. pressure. Increased sampling
tubing size
from 1/8” to 6 mm.

Pressure Experienced damage Avoid overpressure Installed a pressure regulator

regulator of pressure regulator of side-stream. on side-stream to reduce the
inside AMS. pressure to below 10 bars before
connecting side-stream to AMS.
Installed a restrictor with a
pressure gauge to monitor inlet
pressure.
Corrosion Experienced corro- Delrin fittings is Replaced all Delrin fittings
sion of fittings. prone to oxalic with PEEK fittings.
acid corrosion.
Leakage Experienced leakage Loosening of fit- Equipped all fittings with
and precipitates tings due to large lock nut. Tighten fittings, if
around fittings. Temperature fluc- necessary.
tuation.
Electronics and power system

Solar panel Experienced lots of Required monthly Consider installing solar

dust collection on cleaning. panels at proper height for
solar panels. easy access during monthly
maintenance.
Power Power socket is not Risk of water Replaced power socket with
socket water proof. splash entering built-in power cable.
power socket.
 Deployment of Pre-Industrial Autonomous Microbe Sensor in ... 269

For AMS field deployment, an air conditioned (AC) shelter is recommended

to maintain a relatively stable temperature in AMS modules, which will have
many beneficial effects: eliminate fitting loosening due to large temperature
change, prevent SYBR Green dye freezing, prevent temperature-regulated
AMS shutdown, and extend the life span of the AMS device.
In general, pre-industrial AMS sampling, conditioning, measurement
sequence and method, solution and reagent refilling were found suitable
for field application. When in operation, the AMS was subjected to routine
maintenance from the operation manual (Table 7).

Table 7. Pre-industrial AMS routine maintenance schedule.

Maintenance items Schedule

Check instrument inlet pressure Monthly
Check instrument sample flow rate Monthly
Inspect internal sample leak and blockage Monthly
Clean solar panels Monthly
Refill solutions and reagents Monthly
Inspect tubings and fittings, and replace if neces- Half-year ser-
sary vice
Validate performance and recalibrate if neces- Half-year ser-
sary vice

CONCLUSION AND RECOMMENDATIONS

The online AMS, based on fluorescent DNA staining technology, is
developed in three phases for real-time detection of microbial cell numbers in
large seawater injection systems under Saudi Arabian conditions. Numerous
technical improvements have been implemented in each developmental
phase for increased sensitivity, durability, robustness, and maintainability,
while reducing the uncertainties related to AMS operation and function.
Two pre-industrial AMS units have been constructed and deployed at
critical field locations for automated monitoring of microbial numbers in
injection seawater system (salinity ~5.5%). The two units are currently
operational, including the hot summer months in Saudi Arabia without AC
shelters, requiring only minimal monthly intervention from field operators
and engineers. The wider deployment at more strategic locations in large
Saudi Aramco sea water injection network is planned. In addition, a renewed
270 Mechatronics and Robotics

market study is on-going to assess the market opportunities, technology

landscape, and commercialization strategy.
For long-term reliability and maintainability, it is recommended to
house the AMS hardware and solution modules inside an AC shelter. AC
shelter is expected to provide two major benefits. One is to prevent solvent
DMSO of SYBR Green dye from freezing when temperature in Saudi
Arabia winter drops to below 12˚C. DMSO freezing affects the delivery of
SYBR Green dye for bacteria staining, and hence the fluorescence signals.
For hardware protection, AMS is programmed to shut down automatically
when temperature is above 40˚C, resulting in many missing measurements
during hot Saudi Arabia summer. An AC shelter increases the flexibility of
programming measurement time and frequency throughout the year. It is
also recommended to use dual power supply (mains power and solar panel)
for battery charge, whenever mains power is available at the installation
site. This will ensure the sufficient battery charge to power AMS operation
in case of reduced solar panel efficiency caused by adverse weather and
environmental conditions (e.g., rain, clouds, dust, and sand storms). The
cleaning of solar panels should be maintained even when the mains power
is connected.
Calibration is sensitive to the specific water system in question, therefore
specific calibration shall be performed during commissioning of each water
system, to ensure the accurate conversion of fluorescence count to cell
number in the sample. In addition, the performance of the AMS shall be
validated after major maintenance and service, and re-calibrated if necessary.

ACKNOWLEDGMENTS
The authors would like to acknowledge the Saudi Arabian Oil Company
(Saudi Aramco) and the Danish Technological Institute for granting
permission to publish this paper.
 Deployment of Pre-Industrial Autonomous Microbe Sensor in ... 271

REFERENCES
1. Ren, H., Wang, W., Liu, Y., Liu, S., Lou, L., Cheng, D., He, X., Zhou,
X., Qiu, S., Fu, L., Liu, J. and Hu, B. (2015) Pyrosequencing Analysis
of Bacterial Communities in Biofilms from Different Pipe Materials
in a City Drinking Water Distribution System of East China. Applied
Microbiology and Biotechnology, 99, 10713-10724. https://doi.
org/10.1007/s00253-015-6885-6
2. Moniee, M.A., Juhler, S., Sorensen, K., Abeedi, F.N., Lundgaard, T.
and Sanders, P.F. (2014) A Review of Saudi Aramco’s Water Flooding
System and Methods for Monitoring Microbial Activity. Proceedings
15th Middle East Corrosion Conference, Bahrain, 2-5 February 2014,
1-19.
3. Moniee, M.A., Juhler, S., Sorensen, K., Zhu, X.Y., Lundgaard, T.,
Abeedi, F.N. and Sanders, P.F. (2016) Laboratory-Scale Evaluation
of Single Analyte Bacterial Monitoring Strategies in Water Injection
Systems. Journal of Sensor Technology, 6, 11-26. https://doi.
org/10.4236/jst.2016.62002
4. Moniee, M.A., Zhu, X.Y., Tang, L., Juhler, S., Nuwaiser, F.I., Sanders,
P.F. and Abeedi, F.N. (2016) Optimization of DNA Staining Technology
for Development of Autonomous Microbe Sensor for Injection
Seawater Systems. Journal of Sensor Technology, 6, 27-45. https://doi.
org/10.4236/jst.2016.63003
5. Moniee, M.A., Zhu, X.Y., Tang, L., Nuwaiser, F.I., Voigt, N.V., Sanders,
P.F., Abeedi, F.N. and Habboubi, H.H. (2016) Validation of Autonomous
Microbe Sensor Prototype for Monitoring of Microorganisms in
Injection Seawater Systems. Journal of Sensor Technology, 6, 81-100.
https://doi.org/10.4236/jst.2016.64007
6. Dragan, A.I., Pavlovic, R., McGivney, J.B., Casas-Finet, J.R., Bishop,
E.S., Strouse, R.J., Schenerman, M.A. and Geddes, C.D. (2012) SYBR
Green I: Fluorescence Properties and Interaction with DNA. Journal
of Fluorescence, 22, 1189-1199. https://doi.org/10.1007/s10895-012-
1059-8
7. Noble, R.T. and Fuhrman, J.A. (1998) Use of SYBR Green I for Rapid
Epifluorescence Counts of Marine Viruses and Bacteria. Aquatic
Microbial Ecology, 14, 113-118. https://doi.org/10.3354/ame014113
8. Shibata, A., Goto, Y., Saito, H., Kikuchi, T., Toda, T. and Taguchi,
S. (2006) Comparison of SYBR Green I and SYBR Gold Stains for
272 Mechatronics and Robotics

Enumerating Bacteria and Viruses by Epifluorescence Microscopy.

Aquatic Microbial Ecology, 43, 223-231. https://doi.org/10.3354/
ame043223
9. Maruyama, A. and Sunamura, M. (2000) Simultaneous Direct Counting
of Total and Specific Microbial Cells in Seawater, Using a Deep-Sea
Microbe as Target. Applied Environmental Microbiology, 66, 2211-
2215. https://doi.org/10.1128/AEM.66.5.2211-2215.2000
10. Kapuscinski, J. (1995) DAPI: A DNA-Specific Fluorescent Probe.
Biotechnic & Histochemistry, 70, 220-233. https://doi.org/10.3109/1
0520299509108199
11. Tarnowski, B.I., Spinale, F.G. and Nicholson, J.H. (1991) DAPI as a
Useful Stain for Nuclear Quantitation. Biotechnic & Histochemistry,
66, 296-302. https://doi.org/10.3109/10520299109109990
12. LabVIEW 2015, Version 15.0f2 (64-bit), National Instruments
Corporation, Austin, Texas, US. http://www.ni.com/labview/
13. Gasol, J.M., Zweifel, U.L., Peters, F., Fuhrman, J.A. and Hagstro, A.
(1999) Significance of Size and Nucleic Acid Content Heterogeneity as
Measured by Flow Cytometry in Natural Planktonic Bacteria. Applied
Environmental Microbiology, 65, 4475-4483.
14. Sgorbati, S., Barbesti, S., Citterio, S., Bestetti, G. and de Vecchi, R.
(1996) Characterization of Number, DNA Content, Viability and Cell
Size of Bacteria from Natural Environments using DAPI PI Dual
Staining and Flow Cytometry, Minerva Biotechnology, 8, 9-15.
15. Gasol, J.M., del Giorgio, P.A., Massana, R. and Duarte, C.M. (1995)
Active versus Inactive Bacteria: Size-Dependence in a Coastal Marine
Plankton Community, Marine Ecology Progress Series, 128, 91-97.
https://doi.org/10.3354/meps128091
16. Marie, D., Partensky, F., Jacquet, S. and Vaulot, D. (1997) Enumeration
and Cell Cycle Analysis of Natural Populations of Marine Picoplankton
by Flow Cytometry Using the Nucleic Acid Stain SYBR Green I.
Applied Environmental Microbiology, 63, 186-193.
17. Thermo Scientific, FITC and TRITC, Instructions. https://assets.
thermofisher.com/TFS-Assets/LSG/manuals/MAN0011224_FITC_
TRITC_UG.pdf
 CHAPTER

14
A Survey of Modeling and Control
of Piezoelectric Actuators

Jingyang Peng, Xiongbiao Chen

Department of Mechanical Engineering, University of Saskatchewan,
Saskatoon, Canada

ABSTRACT
Piezoelectric actuators (PEAs) have been widely used in microand
nanopositioning applications due to their fine resolution, fast responses,
and large actuating forces. However, the existence of nonlinearities such
as hysteresis makes modeling and control of PEAs challenging. This paper
reviews the recent achievements in modeling and control of piezoelectric
actuators. Specifically, various methods for modeling linear and nonlinear

Citation:- J. Peng and X. Chen, “A Survey of Modeling and Control of Piezoelec-

tric Actuators,” Modern Mechanical Engineering, Vol. 3 No. 1, 2013, pp. 1-20. doi:
10.4236/mme.2013.31001.
Copyright: © 2013 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
274 Mechatronics and Robotics

behaviors of PEAs, including vibration dynamics, hysteresis, and creep, are

examined; and the issues involved are identified. In the control of PEAs as
applied to positioning, a review of various control schemes of both model-
based and non-model-based is presented along with their limitations. The
challenges associated with the control problem are also discussed. This
paper is concluded with the emerging issues identified in modeling and
control of PEAs for future research.
Keywords: Piezoelectric Actuators; Modeling; Control

INTRODUCTION
Piezoelectric actuators (PEAs) and PEA-driven positioning systems
have been widely used in the fields of microand nanopositioning such as
atomic force microscopes [1-3], adaptive optics [4], computer components
[5], machine tools [6], aviation [7], internal combustion engines [8],
micromanipulators [9], and synchrotron-based imaging systems [10] due
to their high displacement resolution (sub nanometer) and large actuating
force (typically a few hundreds of N). PEA-driven positioning systems have
also been developed with various configurations, for example, 1-degree-
of-freedom (1-DOF) positioning systems with flexure hinge mechanisms
[11,12], stick-slip actuators [13], multiple PEAs-driven inchworms [14],
or walking actuators [15]; multi-DOF positioning systems with series
mechanism [16], parallel mechanism [17], or stick-and-clamping actuators
[18].
In all positioning control applications, the hysteresis and creep effects of
PEAs have shown to be able to significantly degrade the system performance
and even system stability [19]. For improvement, models are desirable to
represent these effects and on this basis, tracking controllers can be designed
and implemented to achieve desired positioning requirements.
This paper reviews the modeling and control methods of PEAs in
microand nanopositioning applications and some remaining issues to be
solved, and is arranged as follows.
Firstly, the working principles and behaviors of PEAs are briefly
introduced in Section 2. In Section 3 various models and modeling techniques
for PEAs are examined, along with their limitation; and in Section 4 various
control schemes for PEAs are reviewed with associated challenges. Section
5 concludes with emerging issues identified in modeling and control of
PEAs for future research.
 A Survey of Modeling and Control of Piezoelectric ... 275

PIEZOELECTRIC ACTUATORS AND THEIR BEHAV-

IORS

Working Principles of Piezoelectric Actuators

PEAs utilize the converse piezoelectric effect of piezoelectric materials to
generate displacement and force, i.e. a piece of piezoelectric material will
be mechanically strained if subject to an electric field (by placing it into
the electric field or applying voltages to its surfaces) [20]. This property
is resulted from the microscopic structure of piezoelectric materials as
explained in the following. Most piezoelectric materials used in PEAs,
e.g. lead zirconate titanate, undergo a structural phase transition as its
temperature drops through the so-called Curie temperature, during which
their structurally and electrically symmetric cubic unit cells deform into
structurally and electrically asymmetric tetragonal unit cells, resulting in
spontaneous strain and polarization [21]. Groups of adjoining unit cells
with uniformly oriented spontaneous polarization are called Weiss [20] or
ferroelectric [21] domains. The directions of spontaneous polarization of
different domains thus developed are random, so in this state the piezoelectric
materials exhibit no overall piezoelectric behavior.
To fabricate PEAs from such piezoelectric materials, they are further
exposed to a strong electric field (106 to 107 V/m) at a temperature just
below the Curie temperature, forcing the directions of spontaneous
polarization of the domains to align with the electric field. Such alignment
can be approximately preserved after the removal of the electric field. This
process is referred to as poling [21,20]. After poling, due to the approximate
alignment of the spontaneous polarization of the domains, the deformations
of the domains in the direction of their respective spontaneous polarization
resulting form the application of an voltage to the piece of piezoelectric
material (which generates an electric field that either enhance or suppress the
spontaneous polarization of the domains, causing their dimension to change
in the direction of the spontaneous polarization) can accumulate and causes
an overall deformation or displacement. Thus the piece of piezoelectric
material now possesses overall piezoelectric property including the converse
piezoelectric effect and can be used as a PEA [20].

Behaviors of Piezoelectric Actuators

In microand nanopositioning applications, typical behaviors of PEAs
concerned include hysteresis, creep, and vibration dynamics.
276 Mechatronics and Robotics

Hysteresis is the nonlinear dependence of a system not only on its current

input but also on its past input. In PEAs, hysteresis exists in both the electric
field (voltage)-polarization relationship and the electric field (voltage)-strain
(deformation or displacement) relationship, with the latter being mostly
concerned in microand nanopositioning, and it is caused by the nonlinearities
in the converse piezoelectric effect of the unit cells and the switching and
movement of domain walls [21]. According to [22], the relationship between
the actuating force exerted on the PEA and the resulting displacement of
the PEA is linear, as such the electric field (voltage)-strain (displacement)
hysteresis can also be treated as the result of the voltage-internal actuating
force hysteresis.
The strain (deformation or displacement)-electric field (voltage)
hysteresis, which is usually also the input-output hysteresis relationship, of
a typical PEA is shown in Figure 1(a). It can be seen in Figure 1(a) that the
hysteresis trajectory of a PEA can be treated as being composed of three
types of components: 1) the major loop which is the hysteresis loop that
spans the whole input (voltage) range, 2) the minor loops which are the
hysteresis loops that only span portions of the input range, and 3) the initial
ascending curve which is the input/output trajectory traversed whenever the
magnitude of the input applied to the PEA exceeds the maximum magnitude
in the input history.

Figure 1. (a) Hysteresis of a PEA and (b) creep of a PEA subject to a 30 V step
input.
 A Survey of Modeling and Control of Piezoelectric ... 277

Hysteresis in PEAs occurs in both relatively static operating conditions

(i.e. with constant or slow varying input) and dynamic operating conditions
(i.e. with fast varying input). If the influence of the rate of change of the
input on the hysteresis can be ignored, then the hysteresis is referred to
as rate-independent, otherwise ratedependent. The latter is usually treated
as a combined effect of the rate-independent hysteresis and the vibration
dynamics.
As hysteresis being the major nonlinearity of PEAs and possessing
detrimental effects on the positioning accuracy and stability margins of
feed-back control systems [23], compensation of hysteresis has always been
a major concern in modeling and control of PEAs.
Creep is the slow variation in the PEA displacement that occurs without
any accompanying change in the input voltage [24] as shown in Figure
1(b). It is caused by the same piezoelectric material properties as PEA
hysteresis [20]. Being a slow and a small effect (on the order of 1% of the
last displacement per time decade [20]), creep is sometimes neglected in
closed loop and high frequency operations, e.g. in [25,26]. However, for
slow and open-loop operations of PEAs, creep must be considered to avoid
large positioning error [27].
Finally, PEAs also exhibit linear vibration dynamics, which is the
dynamic relationship between the total force exerted on a PEA and the
displacement of the PEA, and it resembles a distributed parameter system
as the mass of the PEA is not concentrated at certain points [22]. However,
if the positioning mechanism attached to the PEA, which is usually much
more massive than the PEA, is also taken into account, the resulting overall
vibration dynamics will approach that of a lumped parameter system [28].

MODELING OF PIEZOELECTRIC ACTUATORS

A large number of PEA models have been developed to mathematically
represent the behaviors of PEAs mentioned in Section 2.2, and they can be
generally classified into macroscopic models, which models a PEA as a whole,
microscopic models, which models a PEA as a combination of a series of
ferroelectric domains or discretized cells [21,29-39], and hybrid models, which
combines the ideas behind these two categories has also been reported in [40].
Due to the requirement of using finite element methods, which is computationally
expensive, to obtain the overall PEA behaviors, microscopic models and hybrid
models are not suitable for the use in microand Nano positioning applications.
As such, the following discussion only concerns macroscopic models.
278 Mechatronics and Robotics

The linear electromechanical model reported in [41] is an early example

of macroscopic PEA models. But it can represent neither the nonlinear
behaviors (hysteresis and creep) nor the linear vibration dynamics in PEAs
due to its linear and static nature. To solve this, various submodels have
been developed with each representing one or two of the linear/nonlinear
behaviors of PEAs mentioned in Section 2.2, and then connected in
appropriate manners to form (or used alone as) macroscopic PEA models.
These sub-models, the methods of constructing macroscopic PEA models
based on such sub-models, and the methods of inverting such sub-models
for the use in model-based open-loop feedforward control of PEAs are
reviewed as follows.

Hysteresis Sub-Models

Models for the Major Hysteresis Loop

At the early stage, hysteresis models were developed only to represent
the major loop in the hysteresis trajectories of PEAs (refer to Figure 1).
For example, in [42] polynomials were used to represent the major loop.
The JilesAtherton hysteresis model, which was initially developed for
representing ferromagnetic hysteresis [43] and afterwards adopted for the
ferroelectric hysteresis in piezoelectric materials [44], is also limited to
represent the major loop [45]. The inability of such models to represent
minor loops in the hysteresis curves of PEAs limits their applications.

Rate-Independent Hysteresis Models

To model both the major loops and the minor loops in the hysteresis of PEAs
without concerning the influence of the rate of chance of the input, in other
words, to model the rate-independent hysteresis of PEAs, both existing and
newly developed rate-independent hysteresis models have been adopted.
Among which the Preisach hysteresis model [46] and its modifications [47-
58], the PrandtlIshlinskii (PI) hysteresis model [25,59-63], and the Maxwell
resistive capacitor (MRC) model [64-67] are among the most widely used.
These rate-independent hysteresis models are reviewed as follows.
The Preisach hysteresis model represents the hysteresis by the combined
effect of an infinite number of Preisach hysteresis operators , as
shown in Figure 2. Two parameters are used to characterize a Preisach
hysteresis operator: the up switching value and the down switching value
 A Survey of Modeling and Control of Piezoelectric ... 279

, with . Each operator has two saturation values: 0 and 1, and its
contribution to the model output is adjusted by , referred to as the
Preisach weighting function.

(1)

Figure 2. (a) Preisach hysteresis operator and (b) working principle of the Prei-
sach hysteresis model: the sum of two hysteresis operators of different switch-
ing values as shown in (1) can represent a simple but unsmooth hysteresis loop
as shown in (2); by using the sum of an infinite number of hysteresis operators
of different switching values, smooth hysteresis loops including both major and
minor loops can then be represented as shown in (3).
As such, the resulting Preisach hysteresis model is expressed as where
is the input and is the output of the hysteretic system [47]. To
reduce the computational effort due to the involvement of the double
integration, alternative expressions of the Preisach hysteresis model that
only involves arithmetic operations have also been derived from Equation
(1) and the properties of the Preisach hysteresis operator [68,69].
280 Mechatronics and Robotics

In PI hysteresis models, hysteresis is represented by the combined effect of a

finite number of play or backlash operators as shown in Figure 3. Following
this idea, the expression of a PI hysteresis model involving n play operators
is given as

(2)

where is the model output; i denotes the index of the play operator;
and are the weight parameter and the threshold parameter of the i-th
play operator, respectively.
Similar to the PI hysteresis model, the MRC hysteresis model represents
the hysteresis by the combined effect of a finite number of elasto-slide
elements or operators as shown in Figure 4 [64]. Each elasto-slide element
or operator (Figure 4(a)(1)) is composed of 1) a mass sliding on a surface
with a Coulomb friction , where is the friction coefficient and N is
the normal force between the mass and the surface, and 2) a spring of
stiffness k with one end connected to the mass whilst the displacement of
the other end (the free end) can be freely assigned and use as input to the
elasto-slide element.

Figure 3. (a) Play operator and (b) working principle of the PI hysteresis mod-
el: the sum of two play operators of different parameters as shown in (1) form a
simple PI hysteresis model as shown in (2).
 A Survey of Modeling and Control of Piezoelectric ... 281

Figure 4. (a) An elasto-slide element and its input-output relationship; (b) The
physical interpretation of the MRC hysteresis model [63].
As such, hysteresis relationship exists between the (input) displacement
of the free end of the spring and the resulting force in the spring , as
can be seen in Figure 4(a)(2).
Connecting n elasto-slide elements as in Figure 4(b), the resulting MRC
hysteresis model or the relationship between the input displacement of the
free ends of the springs and the total force experienced at the free ends
of the springs can be expressed as

(3)
where is the force in the spring of the i-th elasto-slide element
induced by .

Rate-Dependent Preisach and PI Hysteresis Models

The rate-independent hysteresis models discussed in Section 3.1.2 are only
capable of representing the hysteretic behavior of a PEA in narrow (a few
Hz) frequencies bands. In wide band applications where both hysteresis
and vibration dynamics are significant, the rate-independent hysteresis
models need either to be combined with vibration dynamics models or to be
modified into rate-dependent hysteresis models to represent the behaviors
of the PEA. In the literature, both the Preisach hysteresis model and the PI
hysteresis model have been modified into rate-dependent hysteresis models.
282 Mechatronics and Robotics

In the rate-dependent Preisach hysteresis models, the Preisach weighting

function are adjusted to account for the input-rate dependency by using
rate-dependent multiplicative modifying factors [68,70] or neural network
approaches [69]. In [68], the rate-dependent multiplicative modifying
factor for adjusting is a function of 1) the average voltage input rate
between two input extrema and 2) the variation of the voltage between two
input extrema. Even though this model can effectively represent the PEA
hysteresis up to 800 Hz, it requires the a priori knowledge of the input voltage
waveform (at least to the next extremum), and thus being not applicable
to conventional situations in which the future input voltage is unknown.
To avoid this problem, in [70] a multiplicative modifying factor which
depends only on the input-rate or the first derivative of the input voltage is
used instead. However, the small differences between the input-rate at the
input extrema at different input frequencies limits the model accuracy at
higher input frequencies (the model is effective typically below 10 Hz). To
achieve better accuracy (error of the model prediction being less than 5%
of the maximum displacement of the PEA) in operations involving higher
frequencies (~30 Hz), in [69] a neural network with rate-dependent output
is used to account for the input-rate dependence of the Preisach weighting
function. But the applicable frequency range remains low compared to the
resonant frequencies of conventional PEAs (typically over 1 KHz).
Similarly, the rate-dependent PI hysteresis models, which are modified
from the PI models by modeling the parameters as functions of the input rate,
are also limited to represent low frequency (<30 Hz) PEA operations or the
error becomes excessively high (over 10% of the peak-to-peak sinusoidal
displacement of the PEA) [25,26].
So, in conventional applications where the a priori knowledge of the
input voltage is not available, the rate dependent Preisach and the rate-
dependent PI hysteresis models are both limited to represent PEA operations
over frequency bands only slightly wider than the applicable frequency
bands of the rate-independent models.

Inverse Preisach and PI Hysteresis Models

Inverse hysteresis models are essential to the model inversion-based feed
forward hysteresis compensation technique for PEAs. The inverses of the
two most widely used hysteresis models, i.e. the Preisach hysteresis model
and the PI hysteresis model, have been established by employing different
methods in the literature.
 A Survey of Modeling and Control of Piezoelectric ... 283

The Preisach hysteresis model cannot be inversed analytically; hence

several methods of approximately inverting the Preisach hysteresis model
have been developed instead [71]. Such approximate inverting methods can
be classified into two categories.
The first category involves directly identifying the Preisach hysteresis
model inversely (using the measured plant output as the input to the
Preisach hysteresis model while using the plant input as the output to the
Preisach hysteresis model) [1,72,73] or directly identifying the inverses
of some intermediate functions that constitute the alternative expressions
of the Preisach hysteresis model and then expressing the inverse Preisach
hysteresis model based on them [56,71]. The advantage of this category of
methods is that the computational effort of the inverse Preisach hysteresis
model is the same as the Preisach hysteresis model, not more. However, in
applications that also require the Preisach hysteresis model in additional
to the inverse one, a different parameter identification procedure has to be
performed (assume without using iterative methods) since the former cannot
be analytically derived from the latter, and the two resulting models are
not necessarily accurate inversions of each other due to the different model
errors introduced in their respective parameter identification procedures.
The second category involves iterative-algorithms-based methods [71],
which find the input to an identified Preisach hysteresis model iteratively
until the output of the hysteresis model converges to the desired value. Such
methods are developed based on the contraction mapping principle, which
are firstly proposed in [74,75], and they have been successfully applied to
inverse the Preisach hysteresis modeled hysteresis in a magnetostrictive
actuator [71] and a PEA [76]. The advantage of such methods is that both the
Preisach hysteresis model and the inverse one can be obtained via a single
parameter identification procedure and the inversion can be highly accurate.
However, the computational effort involved in the iterative-algorithms-
based methods is generally much higher than that of calculating the Preisach
hysteresis model output since the Preisach hysteresis model output needs to
the calculated in each iteration.
The PI hysteresis model, in the contrary, can be inversed analytically, so
good accuracy and low computational effort can be achieved simultaneously,
making it much easier to implement than the inverse Preisach hysteresis
models. The inversion algorithms for PI hysteresis models can be found,
for example, in [63]. In [25,26] the inversion of the rate-dependent PI
hysteresis model was also presented. In either case of rate-independent or
284 Mechatronics and Robotics

rate-dependent, the PI hysteresis model inversion algorithms fail if the input

frequency is so high that the rounding of the hysteresis loops due to phase lag
introduced by other dynamics causes the hysteresis loading and unloading
curves to be no longer monotonic. This problem was later solved in [77] by
additional mappings to and back from a domain where the inversion of the
rate-dependent PI hysteresis model always exists.

Other Hysteresis Models

Other hysteresis models, such as a first-order nonlinear differential equation
model [78] and the Bouc-Wen hysteresis model [79-81] have also been used to
model hysteresis in PEAs with promising results [22,82-84]. In [85], a non-linear
auto-regressive moving average model with exogenous inputs (NARMAX) was
employed to represent the hysteresis of a PEA. In [73] a hysteresis model was
developed for PEAs by using a polynomial based linear mapping strategy. In
[86], a hysteresis model based on a new hysteresis operator, which in turn is
based on two hyperbola functions, are developed for PEAs along with its inverse
model. Such hysteresis models have not seen wide application mainly due to
their high complexity but limited improvements on accuracy as compared to
other widely used hysteresis models discussed before.

Creep Sub-Models

Frequently Used Creep Models

The creep response in a PEA has a logarithmic shape over time, so it is
sometimes referred to as the Log(t)- type creep, and the nonlinear creep
model given by Equation (4) has been used to represent such creep behavior
[87,88].

(4)
where is the creep model output, is the creep output at after the
static input voltage is applied, and is a coefficient depending on the input
voltage. The dependence of on the input voltage and the unsoundness of
such a nonlinear creep model as and pose difficulties in its
implementation [19].
To avoid the problem of unboundness encountered in the nonlinear
creep model Equation (4) and to avoid the nonlinearity, a linear creep model
was proposed as.
 A Survey of Modeling and Control of Piezoelectric ... 285

(5)
where the creep transfer function is represented by the combined
effects of spring damper systems [1,88]. Besides, an operator-
based linear creep model was also reported in [60].
However, these linear creep models have linear and memory free
equilibrium values, whist the actual equilibrium value of the creep effect of
PEA exhibits hysteresis nonlinearity [89].
To further solve the problems encountered in both the nonlinear and
linear creep model above, the PI hysteresis model design approach was
adopted in [63] to develop an operator-based nonlinear creep model, whose
output is a weighted sum of the values of a series of elementary creep
operators.

Inverse Creep Models

Similar to the case of inverse hysteresis models, inverse creep models are
essential in feed forward creep compensation.
While the analytic inversion of the nonlinear creep model was not
mentioned in the literature, a method of feeding a voltage with an inverse
Log(t)-type creep itself to a PEA to compensate for the creep in the output
displacement was found to be effective [87].
On the other hand, the inversion of the linear creep model (Equation (5),
not operator-based) is usually carried out together with the inversion of the
vibration dynamics model since they can be combined into a single linear
dynamics model.
An optimal inversion approach, which minimizes a quadratic cost
function representing the weighted sum of the input voltage energy and the
output displacement error energy [90], can be used to inverse the combined
vibration dynamics and creep model in [1].
Linear creep models can also be combined with hysteresis models and
then inverted, e.g. via iterative methods [88].
The inversion of the operator-based linear and nonlinear creep models
together with other dynamics are reported in [60,63], respectively.
286 Mechatronics and Robotics

Vibration Dynamics Sub-Models

Frequently Used Vibration Dynamics Sub-Models

According to Section 2.2, for the consideration of accuracy, the vibration
dynamics of a PEA should be modeled as a distributed linear system for a
stand-alone PEA [22]. But for the convenience of practical uses, the vibration
dynamics of a PEA and the attached positioning mechanism is often modeled
as a whole as a single second-order system, as in [28,64,91,92], whilst higher
order linear dynamical systems are only used in applications requiring higher
level of accuracy [1]. Such lumped linear vibration dynamics models of
PEAs are usually identified through fitting the measured frequency response
of the PEA to that of a specific model structure by using a dynamic signal
analyzer [1,93].
On the other hand, the axiomatic design method has also been used to
find a higher order linear system representation of the vibration dynamics of
a PEA with an attached positioning mechanism [94,95].

Inverse Vibration Dynamics Sub-Models

Inverse vibration dynamics are usually used in open-loop feed forward
control of PEAs, especially in the case of controlling piezoelectric tube
actuators as in [1]. Methods for inverting the vibration dynamics have been
reviewed in Section 3.2.2.

Model Structures
With different sub-models representing hysteresis, creep, and vibration
dynamics of PEAs defined, macroscopic PEA models are constructed by
connecting such sub models according to certain model structures. There are
two categories of model structures for macroscopic PEA models.
In the first category, the behavior of a PEA, when subject to a voltage
input and an external loading force, is decoupled into several effects such as
rate-independent hysteresis, vibration dynamics, and creep based on physical
laws. Such effects are then represented by individual sub-models and such
sub-models are connected in various manners to form a comprehensive
model of the PEA.
The most well-known model structure of PEAs belonging to this
category is the electromechanical model proposed in [64], which is further
refined in [22]. Figure 5 shows the schematic representation of this model.
 A Survey of Modeling and Control of Piezoelectric ... 287

There are 3 effects in Figure 5, namely 1) the hysteresis, H, between the

charge in the PEA, q, and the resulting voltage, ; 2) the electromechanical
transducer, , converts the mechanical deformation of the PEA, y, into
charge, (inverse piezoelectric effect), and converts the voltage, , into an
actuating force, (piezoelectric effect); and 3) the vibration dynamics, M,
relating the deformation of the PEA, y, to the internal actuating force, ,
and the external load, . Other symbols in Figure 5 include , which is
the voltage across the PEA, L, which is the original length of the PEA, and
C, which is the capacitance of the PEA. The model can then be expressed as

(6)

(7)

(8)

(9)

(10)

(11)
The hysteresis effect, H, and the vibration dynamics, M, can be modeled
by different kinds of sub-models. Specifically, His usually represented by
rate-independent hysteresis models, e.g. the MRC model was used in [64]
whist the first-order nonlinear differential equation model, which was firstly
proposed in [78], was employed in [22].

Figure 5. Schematic representation of a electromechanical model of a PEA

[22].
288 Mechatronics and Robotics

The vibration dynamics, M, can be modeled by any models discussed in

Section 2.1.3. It is noted that the use of this electromechanical model usually
requires the measurements of the charge, q, in the PEA for model parameter
estimation, which is difficult in practice. So this model is usually employed
in applications that the relationship between and , which is difficult to
identify, can be approximated by a simple function (e.g. an affine function)
with an error term as model uncertainties, such as in [91,92]. Also, creep is
usually not considered in this electromechanical model [22,78] expect being
treated as model uncertainties [91,92].
Another model structure proposed in [1] models the relationship
between and by a rate-independent hysteresis model directly, based
on a conclusion drawn from the analysis in [64] that the rate-independent
hysteresis of a PEA only exists between the input voltage and the resulting
internal actuating force. This hysteresis sub-model is then cascaded to a
vibration dynamic submodel and a creep sub-model to form the complete
model of a PEA, as shown in Figure 6. Without the involvement of charge
measurement for parameter estimation, this model structure is much easier
to apply than the electromechanical model. Similar to the case of the
electromechanical model, this model structure is highly flexible that each
sub-model can be implemented in various ways, e.g. in [1], the Preisach
hysteresis model was used for H and a fourth-order and a third-order linear
dynamic systems are used to represent M and the creep effect, , whist in
[63], a modified PI hysteresis model is used for H and an operator based
model of complex Log(t)- type creep for . Also, due to the cascading
connection between the sub-models, this model structure is very suitable for
implementing control schemes with inverse nonlinearity and/or vibration
dynamics compensation as in [1,63,93,96,97]. A similar model structure
constructed using a bond-graph representation was also presented in [67].
In the second category, the different behaviors of a PEA are not decoupled.
In such cases, rate-dependent hysteresis models such as the rate-dependent
Preisach hysteresis model, the rate-dependent PI hysteresis model (Section
3.1.3), and the rate-dependent neural-networkbased hysteresis model [98]
are used directly as models of the PEA. However, as mentioned in Section
3.1.3, such models are only accurate over small frequency ranges. And, as
can be seen above, creep is usually neglected. These disadvantages seriously
limit their usage.
 A Survey of Modeling and Control of Piezoelectric ... 289

Figure 6. Model structure of a PEA as in [1].

CONTROL OF PEAS
Over the years, a number of control schemes have been reported for positioning
control of PEAs. Due to the detrimental effects of the nonlinearities
(especially hysteresis) of PEAs on positioning control performances in
terms of positioning error and stability [19], the compensation or rejection
of such nonlinear effects has been a major consideration in all of these
control schemes. In the following, the typical control schemes for PEAs are
reviewed.

Open-Loop Control Schemes

Open-loop control schemes are usually employed in applications in which
position feedback are difficult to implement due to mechanical constraints,
e.g. atomic force microscopes [1-3]. In such control schemes, inverse model
of the PEA to be controlled is found and then cascaded to the PEA. The
methods for finding such inverse models have been reviewed in Sections
3.1.4, 3.2.2, and 3.3.2. The inverse model generates an input voltage to
the PEA according to the desired displacement , such that the PEA subject
to produces an output displacement of PEA y that follows . Figure
7 shows a typical open-loop control scheme for PEAs.
In some of the open-loop control schemes, all three effects in the PEA
shown in Figure 6 are compensated for by the inverse model of PEA to
achieve best tracking accuracy (as the case in Figure 7). For example, in
[1,67], inverse models of PEA representing hysteresis, vibration dynamics,
and creep are used to compensate for these effects; whist in [93] a charge
control technique was firstly employed to remove the hysteresis effect and
then an inverse model was used to compensate for the linear dynamics.
290 Mechatronics and Robotics

Figure 7. Open-loop control scheme for a PEA [1].

Other open-loop control schemes only focus on compensating for some
of the three effects shown in Figure 6. For example, in [72,73,86,99],
only hysteresis was compensated for by using the inverses of the Preisach
hysteresis model, a hysteresis model developed by using a polynomial-based
linear mapping strategy, a hysteresis model developed based on hyperbola-
shaped hysteresis operators, and a hysteresis model developed in [100],
respectively, since hysteresis is the most significant nonlinearity in PEAs.
In [90,101], creep is also compensated for in additional to hysteresis by
using operator-based inverse creep models and inverse PI hysteresis models
to achieve better tracking performance. However, due to the negligence
of the vibration dynamics in these schemes, their performances are only
guaranteed in fixed frequency or narrow band operations unless inverse
ratedependent hysteresis models are employed, e.g. the use of inverse rate-
dependent PI hysteresis models in [25,26,77]. Nonetheless, the applicable
frequency ranges (with acceptable tracking error, e.g. <5% of the maximum
displacement of the PEA) of the schemes using inverse ratedependent
hysteresis models are still relatively small (less than 50 Hz) compared with
those also inverting the vibration dynamics (up to a few hundreds of Hz)
[1,25,77].
In no load condition the open-loop control schemes have been shown to
be highly effective in their applicable frequency ranges [1].
However, one major disadvantage of the open-loop control schemes
is that their positioning performances are highly sensitive to unknown
effects such as model errors, external loads, and changes in the dynamics
of the PEA. To partially alleviate the problem, reference [65] proposed to
use a disturbance-observer-based scheme in addition to the inverse MRC
hysteresis model to compensate for the hysteresis and unknown effects
simultaneously. Specifically, the estimated disturbance input to the PEA,
 A Survey of Modeling and Control of Piezoelectric ... 291

which has the same effect as the aforementioned unknown effects on the
PEA output, is subtracted from . However, the effectiveness of this
scheme was not examined through experiments. In [102], the inverse
hysteresis model is updated adaptively to account for the unknown effects.
However, the influence of the unknown effects on the PEA output, though
largely suppressed, can still be clearly observed in the experiment results.
As such, the best solution to this problem remains introducing feedback into
the control schemes, which is to be discussed in the following sections.

Feedback
Feedback control schemes as shown in Figure 8 lead to strong suppression
of the unknown effects including model errors, external loads, and changes
in the dynamics of the PEA on the position control performances, hence they
are widely used.

Figure 8. A feedback control scheme for PEAs [19].

In static or low frequency operations, classical control techniques such as
PID control or using multiple integrators for tracking desired displacement
profiles are widely used because of their simplicity and capability of
eliminating steady state errors in such applications [20,103,104]. Various PID
tuning techniques has been reported in PEA positioning control applications,
e.g. by trial and error [20], by grey relational analysis [105], using an
optimal linear quadratic regulation method [106], by a semiautomatic tuning
technique [107], and by an automatic tuning technique [108]. However, in
broadband operations with large system uncertainties including modeling
errors, nonlinearities, external loads, etc., advanced control techniques are
required because PID control is limited in bandwidth while dealing with
uncertainties [109].
Among all of such advanced control techniques, sliding mode control
has drawn a lot of attention in the recent years. This is due to the fact that
sliding mode controllers can completely reject the effects of the so-call
matched uncertainties or uncertainties in the input channel of the plant,
resulting in strong robustness [110].
292 Mechatronics and Robotics

Specifically, in its basic form, the control signal generated by a sliding mode
controller can be divided into two parts. The first part is discontinuous,
referred to as the switching control, for compensating for the matched
uncertainties while driving the plant trajectory to a prescribed geometric
entity in the phase space (a space with each dimension corresponding to
one state variable in the plant model), referred to as the sliding surface.
The second part is continuous, referred to as the equivalent control, for
keeping the plant trajectory on the sliding surface after the latter is reached.
Since the sliding surface and hence the closed loop system dynamics after
reaching the sliding surface can be freely designed, the closedloop system
becomes insensitive to the matched uncertainties after the sliding surface is
reached [110]. The major problem of implementing the sliding mode control
techniques on PEAs is chattering and the need of the knowledge about the
bounds of the matched uncertainties for designing the discontinuous control
signal [110], which is to be discussed in more detail in Section 5.2.
Based on the model structure shown in Figure 6, hysteresis and creep
in a PEA can be treated as (matched) uncertainties in the input channels
of the vibration dynamics sub-model of the PEA (given that a linear creep
model is used). As such, a sliding mode controller can be designed using
the vibration dynamics sub-model as a nominal model of the PEA, and
the effects of hysteresis and creep on the PEA output displacement can be
strongly suppressed without knowing the exact expression of the hysteresis
sub-model and the creep sub-model, thus largely reducing the efforts
involved in hysteresis and creep modeling and compensation [95,111-114].
Various modifications of the basic sliding mode control technique have also
been reported for controlling PEAs for improved performance. For instance,
adaptive methods are combined with sliding mode control to remove the
requirement of model parameter estimation or lead to better compensation
for the nonlinearities as in [115-117]. And in [118] the nonlinearities of the
positioning mechanism driven by a PEA in additional to those of the PEA
itself were considered in sliding mode controller design.
Robust control techniques constitute another branch of widely used
advanced control techniques for positioning control of PEAs. They try to
find the control law via optimizing an objective function that incorporates
the robustness objective, for example, minimizing the H∞ norm [88,119-
122] or the H2 norm [76] of the transfer functions relating the disturbances
to the plant output, thus minimizing the effects of the disturbances on the
plant output and enhancing robustness. Robust control techniques can be
combined with sliding mode control to suppress unmatched uncertainties
 A Survey of Modeling and Control of Piezoelectric ... 293

(which the sliding mode control techniques cannot compensate for), while
sliding mode control are considered to be superior in dealing with matched
uncertainties because robust control techniques can only minimize the
effects of the matched uncertainties on the plant output whilst sliding mode
control techniques can completely reject such effects in theory [110].
Other advanced control techniques such as state feedback [123], optimal
control [109,124], adaptive control [125-128], and neuron network methods
[129-131] have also been applied to control PEAs.
Inverse sub-models of the nonlinearities of PEA are sometimes inserted
between the feedback controller and the controller in Figure 8 to linearize
the PEA thus facilitating the use of linear feedback controllers. For example,
an inverse Preisach hysteresis model is used to cancel out the hysteresis
and then a linear H2 controller is designed based on the remaining linear
dynamics of a PEA in [76], while an inverse PI hysteresis model and a
sliding mode controller are used in [113].
One problem with the feedback control schemes for PEAs is that the
resulting closed-loop systems have low gain margins, which limit the use
of high gain controllers. Such low gain margin is largely due to the quick
phase loss near the first resonant peak in the frequency response of a PEA
[23]. This problem can be alleviated by using a notch filter to reduce the gain
around the first resonant peak [20,132,133].

Feedback with Feed forward

Feed forward is sometimes used to augment feedback controllers for
nonlinearity compensation. A typical control scheme of this type is shown
in Figure 9. An advantage of this scheme is that the low gain margin problem
can be alleviated such that high frequency positioning performances can be
improved over the feedback control schemes [124,134,135].
Recent researches adopting this scheme can be found, e.g. in [136] a
hysteresis compensator based on the inverse Preisach hysteresis model
was used as the feedforward controller to compensate for hysteresis whist
a PID feedback controller was used to account for other effects, in [137] a
inverse nonlinear differential equation hysteresis model was used as the feed
forward controller and a PI controller was used as a feedback controller,
and in [16] a inverse extended Coleman-Hodgdon model was used as the
feed forward controller and a feedback controller designed via loop shaping
techniques was employed.
294 Mechatronics and Robotics

Figure 9. A feedback augmented with feedforward control scheme for PEAs

[19].
Another control scheme involving feedback and inverse model feed
forward is shown in Figure 10 [97,138]. In this scheme, the feedback
controller is a high gain controller used for nonlinear effect suppression
such that the closed loop system has linear dynamics.
Then a feed forward controller, which is the inverse linear dynamics of
the closed loop system, is applied to make the output y follow the desired
output . This control scheme can be seen as being evolved from the open-
loop control scheme shown in Figure 7. The rationale behind this scheme
is that the accurate modeling and inversion of the nonlinear effects such as
hysteresis are complicated whist the inversion of a linear plant is relatively
easy to compute [139]. However the low gain margin problem with the
feedback loop exists in this scheme [19]. Also, similar to the open-loop
control schemes, this scheme is sensitive to the disturbances acting outside
the closed-loop system.

Disturbance Observer Based Schemes

Recently, a technique call disturbance observer is used for the compensation
of hysteresis and other nonlinear/uncertain effects in PEAs. In this scheme,
a PEA is modeled as a linear dynamic system , so when subject to the
same input voltage , the output of the PEA y and that of the corresponding
are different due to hysteresis and other nonlinear/uncertain effects.
The disturbance observer utilizes and y to estimate an additional input
voltage such that if is applied to , the output of is the
same asy. is referred to as an input disturbance to that represents the
hysteresis and other nonlinear/uncertain effects.
 A Survey of Modeling and Control of Piezoelectric ... 295

Figure 10. A feedback controller is used for nonlinear suppression whist a feed
forward controller is the inverse model of the closed loop system for canceling
out the dynamics of the closed loop system, which can be treated as linear [19].
As such, to compensate for hysteresis and other nonlinear/uncertain
effects in the PEA, one only needs to subtract from before being
applied to the PEA. A control scheme with such a disturbance observer is
shown in Figure 11. This scheme has been shown to be effective in both step
[140] and high frequency tracking (200 Hz sinusoidal) [141] operations.

EMERGING ISSUES IN PEA MODELING AND CON-

TROL
Based on the reviews in the previous sections, some emerging issues
regarding the modeling and control of PEAs and PEA-driven positioning
stages are discussed in this section.

Modeling Issues

Limitations of Some Existing Hysteresis Models of PEAs

While the modeling of creep and vibration dynamics of PEAs have been
relatively well address, the existing models of hysteresis, especially the rate-
independent hysteresis models reviewed in Section 3.1.2, still possess issues,
as follows. Though being highly effective in representing both the major
hysteresis loops and the minor hysteresis loops, it is noted in the literature
[40,53-56,58] that the Preisach hysteresis model cannot at the same time
represent the initial ascending curve when the PEA is subject to nonnegative
voltage inputs. This is due to the fact that the Preisach hysteresis operator
has only two saturation values and without a separate initial section that
connects the origin to the hysteresis loop (Figure 2(a)). The PI hysteresis
models and the MRC hysteresis model do not suffer from this problem due
to the existence of initial sections in their respective hysteresis operator
296 Mechatronics and Robotics

besides the hysteresis loop (Figures 3(a) and 4(a)(2)). However, to limit the
computational effort, the number of hysteresis operators used in PI or MRC
hysteresis models in practice is usually very small (typically around 10)
compared to the Preisach hysteresis model (infinity).
As a result, the PI and MRC models tend to be less accurate than the
Preisach hysteresis model in representing the hysteresis loops, especially
in the cases of the small minor loops. So, given a tractable amount of
computational effort, there is a contradiction between the capability of
representing all hysteresis phenomena, including one-sided hysteresis (the
PI and MRC hysteresis models), and the accuracy (Preisach hysteresis
models) in the existing rate-independent hysteresis models. This remains an
issue to be solved in future research.

Applications of Different Actuator Structures

The applications of PEA-driven positioning systems with structures different
from the conventional ones (those can be treated as a mass fixed to one
end of the PEA) also pose modeling problems. A typical example is the
PEAdriven stick-slip actuator shown in Figure 12, in which an end-effector
is supported and guided by a movable platform that is driven by a PEA, and
it works as follows.
During the course of slow expansion of the PEA, the end-effector moves
along with the platform. If the PEA suddenly contracts, the end-effector
slides on the platform because the force due to inertia becomes larger than
the friction between the end-effector and the movable platform. As a result,
the end-effector moves a step, , with respect to its original position. Such
steps can be accumulated to achieve a theoretically unlimited displacement
(actually limited by the size of the moveable platform).
The modeling of the dynamics of the endeffector displacement has not
been well addressed in the literature. Chang and Li [142] developed a model
for the PEA-driven stick-slip actuator without considering the dynamics
of the PEA while the friction involved was modeled as Coulomb friction,
which is over simplified.
By taking into account presliding friction, models were developed and
reported in more recent studies on stickslip actuators, e.g. [143]. However,
issues including the influence of PEA nonlinearities and the end-effector
mass on the performance of the PEA-driven stick-slip actuator remain to be
addressed.
 A Survey of Modeling and Control of Piezoelectric ... 297

Figure 11. Augmenting the feedback control loop with a disturbance observer.

Figure 12. Actuation sequence of a PEA-driven stick-slip actuator; (a) Start; (b)
Slow expansion and (c) Fast contraction.

Control Issues

Controller Design
As can be seen from Section 4.2, sliding mode control has been one of the most
promising techniques for PEA control applications due to their capability to
largely suppress hysteresis and creep in practice. However, there are two
remaining problems. One problem of implementing the sliding mode control
techniques on PEAs is chattering. This is a kind of high frequency vibration
in the plant output displacement induced by the discontinuous control signal
generated by a sliding mode controller, and it can induce wear, noises, and
even resonance and instability in the plant. Chattering is especially severe
in plants with fast responses, such as PEAs, because they cannot filter out
the high frequency component in the switching control signal. Another
problem is that the bounds of the matched uncertainties need to be known
for designing the discontinuous control signal, but such bounds are usually
difficult to determine in practice while conservatively setting the bounds to
large values only leads to more severe chattering [110]. Existing research
efforts have been focused on solving the chattering problem. For example,
298 Mechatronics and Robotics

a method referred to as the sliding mode with boundary layer control

was developed by approximating the discontinuous control signal with a
continuous one, but with the cost of introducing nonzero steady state error,
and chattering is still significant if the bounds of the matched uncertainties
are large [110]. Higher order sliding surfaces are also used for chattering
suppression, such as the integral sliding mode technique in [92,144] and the
PID sliding surface in [115]. However, the second problem remains to be
solved.

State Estimation
Many advanced control techniques reviewed in Section 4, such as sliding
mode control [114], require state feedback. Since usually not all of the states
of a PEA model are measurable in application, state observers must be used.
However, the problem of estimating the states of PEA models has not been
well addressed in the literature, as discussed below.
Currently, many kinds of state observers are applicable to PEA control
and they can be classified into two categories: non-model-based and model-
based. Non-modelbased filters/differentiators, such as the low-pass filter
plus ideal differentiator, the α-β filter [145], the highgain differentiator [146],
the integral-chain differentiator [146], and the sliding mode differentiator
based on the super-twisting algorithm [147] (given that the states needed are
the derivatives of the measured plant output), usually generate large phase
lags if the desired level of noise suppression is enforced (for the first four
aforementioned methods) or excessive chattering in noisy systems (for the
last method). Compared to non-modelbased filters/differentiators, model-
based observers, e.g., the extended Kalman filter [148], the unscented
Kalman filter [149,150], and the high-gain observer [114], can generate
more accurate estimations if the imperfection of the model can be ignored.
However, with the presence of uncertainties such as the effects of hysteresis,
creep, and external load on a PEA, however, the performances of the model-
based observers degrade. For improvement, in the cases that the system
uncertainties can be treated as a lumped unknown input to the system model,
a kind of model-based observers should be used to estimate the system
states even with the presence of the unknown input, which is referred to as
unknown input observers (UIOs) in the literature. Many UIOs have been
reported in the past three decades, including the full-order UIO [151], the
reduced-order UIO [152], the UIO designed based on projection operator
[153], and the sliding mode-based observers (SMOs) [110,153-155].
 A Survey of Modeling and Control of Piezoelectric ... 299

Applications of these UIOs require that the observer matching condition be

satisfied [156,157], which states that the rank of the product of the output
matrix and the unknown input matrix in the state space model of the system
must be equal to that of the unknown input matrix [110]. However, existing
PEA models, such as the one in [92], do not meet this condition. Attempts
to relax the observer matching condition have been reported [156,157], but
the resultant UIOs were very complicated. As such, a UIO with a simpler
structure and the capability of relaxing the observer matching condition still
needs to be developed in future research for the use in PEA tracking control.

Modeling and Control of Multi-Degree-of-Freedom (Multi-

DOF) PEA-Driven Positioning Systems
All of the above discussions have been focused on single-DOF PEA-
driven positioning systems. However, there are modeling and control issues
uniquely related to multi-DOF PEA-driven positioning systems and they
are briefly discussed in this section.The multi-DOF PEA-driven positioning
systems reported in the literature can be divided in to two categories: series
mechanisms and parallel mechanisms. Among which, multi-DOF PEA-driven
series mechanisms [16] and the xyz-type parallel mechanisms (translations
along x-, yand z-axis) such as those in [16,158-160] can usually be treated
as a series of individual 1-DOF PEAdriven mechanisms without significant
coupling between the moving axes, as such their modeling and control can
be well address with the techniques applied to the single-DOF systems.The
remaining multi-DOF PEA-driven parallel mechanisms whose moving axes
are significantly coupled can be further divided into fully-actuated systems,
i.e. the number of actuation is the same as the number of DOF [17,114,161-
167], and over-actuated systems, i.e. the number of actuation is the more than
the number of DOF [20,112].For the fully-actuated systems, their models
are usually established based on the geometric structure of the mechanism
and mechanics. Since most of such mechanisms use flexure hinges instead
of conventional hinges to avoid friction and backlash [17,114,158,159-166],
the modeling of the flexure hinges becomes an important part in modeling
such mechanisms [114,160,166]. Two kinds of workflows can be adopted
for modeling the multi-DOF PEA-driven positioning systems. For the first
kind of workflows, a kinematic model is firstly constructed based on the
geometric structure of the mechanism, and then either a phenomenological
[114] or physical [160] dynamic model can be established based on the
kinematic model. The second kind of workflows involves direct identification
of the dynamic model of the mechanism without establishing the kinematic
300 Mechatronics and Robotics

model [16].Many control schemes that have been applied to control single-
DOF PEA-driven positioning systems can also be extended to control
fully-actuated PEA-driven positioning systems, e.g. PID control [112,168],
H2 control [168], and sliding mode control [112] since such methods can be
easily extended to fully-actuated multi-in putmulti-output cases.
Compared to the fully-actuated systems, the over actuated systems
have certain advantages in the performance point of view, e.g. singularity
elimination, dexterity improvement, and better load carrying ability [169].
However, the over-actuated structure poses difficulties in both modeling and
control. Specifically, the PEAs in the mechanism need to cooperate well or
excessive internal forces will occur and may damage the mechanism, whilst the
involvement of PEA nonlinearities makes such cooperation difficult to model
and control. As a result, very few researches on the modeling and control of
over-actuated PEA-driven parallel mechanism have been reported compared
to its popularity in macro parallel mechanisms. The only example is [112], in
which the modeling and control of a xθyθz 3-DOF positioning system driven
by 4 PEAs are concerned. Specifically, in [112] each PEA in the mechanism is
modeled and controlled individually and uses only the output displacements
of the corresponding PEA as feedback, while the inverse kinematic model
of the mechanism is used to generate the desired displacement for each PEA
according to the desired position of the end-effector. The cross-axis couplings
are considered as disturbances. As such, the control problem is transformed
into several 1-DOF control problems, which are much simpler. However, in
this scheme the actual position of the end-effector is not closed-loop controlled,
and the requirement of being able to measure the displacement of each PEA
in the mechanism limits the generality of this scheme. As such, more general
schemes for modeling and controlling over-actuated multi DOF PEA-driven
positioning systems that can relax the aforementioned requirement are still to
be developed in the future.

CONCLUSIONS
Jingyang Peng, Xiongbiao Chen (chattering) and the corresponding state
estimators (the incompatibility between the observer matching condition
and the existing PEA models); finally, 6) the modeling and control of
some PEA-driven positioning systems with unconventional structures, e.g.
PEA-driven stick-slip actuators and over-actuated multiDOF PEA-driven
positioning systems, are largely untouched in the literature, thus remain to
be addressed in future research.
 A Survey of Modeling and Control of Piezoelectric ... 301

REFERENCES
1. D. Croft, G. Shed and S. Devasia, “Creep, Hysteresis, and Vibration
Compensation for Piezoactuators: Atomic Force Microscopy
Application,” Journal of Dynamic Systems, Measurement, and Control,
Vol. 123, No. 1, 2001, pp. 35-43. doi:10.1115/1.1341197   
2. Q. Zou, K. K. Leang, E. Sadoun, M. J. Reed and S. Devasia, “Control
Issues in High-Speed AFM for Biological Applications: Collagen
Imaging Example,” Asian Journal of Control, Vol. 6, No. 2, 2004, pp.
164-178. doi:10.1111/j.1934-6093.2004.tb00195.x
3. R. Kassies, K. O. Van der Werf, A. Lenferink, C. N. Hunter, J. D.
Olsen, V. Subramaniam and C. Otto, “Combined AFM and Confocal
Fluorescence Microscope for Applications in Bio-Nanotechnology,”
Journal of Microscopy, Vol. 217, No. 1, 2005, pp. 109-116. doi:10.1111/
j.0022-2720.2005.01428.x
4. H. Song, G. Vdovin, R. Fraanje, G. Schitter and M. Verhaegen,
“Extracting Hysteresis from Nonlinear Measurement of Wavefront-
Sensorless Adaptive Optics System,” Optics Letters, Vol. 34, No. 1,
2009, pp. 61-63. doi:10.1364/OL.34.000061   
5. W. Yang, S.-Y. Lee and B.-J. You, “A Piezoelectric Actuator with
a Motion-Decoupling Amplifier for Optical Disk Drives,” Smart
Materials and Structures, Vol. 19, No. 6, 2010, pp. 065027.1-
065027.10. doi:10.1088/0964-1726/19/6/065027   
6. G. Stöppler and S. Douglas, “Adaptronic Gantry Machine Tool with
Piezoelectric Actuator for Active Error Compensation of Structural
Oscillations at the Tool Centre Point,” Mechatronics, Vol. 18, No. 8,
2008, pp. 426-433. doi:10.1016/j.mechatronics.2008.03.002   
7. S. R. Viswamurthy, A. K. Rao and R. Ganguli, “Dynamic Hysteresis
of Piezoceramic Stack Actuators Used in Helicopter Vibration Control:
Experiments and Simulations,” Smart Materials and Structures, Vol.
16, No. 4, 2007, pp. 1109-1119. doi:10.1088/0964-1726/16/4/020   
8. M. S. Senousy, F. X. Li, D. Mumford, M. Gadala and R. K. N. D.
Rajapakse, “Thermo-Electro-Mechanical Performance of Piezoelectric
Stack Actuators for Fuel Injector Applications,” Journal of Intelligent
Material Systems and Structures, Vol. 20, No. 4, 2009, pp. 387-
399. doi:10.1177/1045389X08095030   
9. J.-J. Wei, Z.-C. Qiu, J.-D. Han and Y.-C. Wang, “Experimental
Comparison Research on Active Vibration Control for Flexible
302 Mechatronics and Robotics

Piezoelectric Manipulator Using Fuzzy Controller,” Journal of

Intelligent and Robotic Systems, Vol. 59, No. 1, 2010, pp. 31-
56. doi:10.1007/s10846-009-9390-2   
10. D. Shu, J. Maser, M. Holt, R. Winarski, C. Preissner, A. Smolyanitskiy,
B. Lai, S. Vogt and G. B. Stephenson, “Optomechanical Design of a
Hard X-Ray Nanoprobe Instrument with Nanometer-Scale Active
Vibration Control,” AIP Conference Proceedings, Vol. 879, 2007, pp.
1321-1324. doi.org/10.1063/1.2436307   
11. H. Zhou and B. Henson, “Analysis of a Diamond-Shaped Mechanical
Amplifier for a Piezo Actuator,” The International Journal of Advanced
Manufacturing Technology, Vol. 32, No. 1-2, 2007, pp. 1-7. doi:10.1007/
s00170-005-0303-7
12. M. Muraokaa and S. Sanada, “Displacement Amplifier for Piezoelectric
Actuator Based on Honeycomb Link Mechanism,” Sensors and
Actuators A: Physical, Vol. 157, No. 1, 2010, pp. 84-90. doi:10.1016/j.
sna.2009.10.024
13. Y. Zhang, G. Liu and J. Hesselbach, “On Development of a Rotary-
Linear Actuator Using Piezoelectric Translators,” IEEE/ASME
Transactions on Mechatronics, Vol. 11, No. 5, 2006, pp. 647-
650. doi:10.1109/TMECH.2006.882998   
14. J. Li, R. Sedaghati, J. Dargahi and D. Waechter, “Design and
Development of a New Piezoelectric Linear Inchworm Actuator,”
Mechatronics, Vol. 15, No. 6, 2005, pp. 651-681. doi:10.1016/j.
mechatronics.2005.02.002   
15. R. Merry, R. Van de Molengraft and M. Steinbuch, “Modeling of a
Walking Piezo Actuator,” Sensors and Actuators A: Physical, Vol. 162,
No. 1, 2010, pp. 51-60. doi:10.1016/j.sna.2010.05.033   
16. R. Merry, M. Uyanik, R. Van de Molengraft, R. Koops, M. Van
Veghel and M. Steinbuch, “Identification, Control and Hysteresis
Compensation of a 3 DOF Metrological AFM,” Asian Journal of
Control, Vol. 11, No. 2, 2009, pp. 130-143. doi:10.1002/asjc.89   
17. T.-F. Lu, D. C. Handley, Y. K. Yong and C. Eales, “A Three-DOF
Compliant Micromotion Stage with Flexure Hinges,” Industrial
Robot: An International Journal, Vol. 31, No. 4, 2004, pp. 355-
361. doi:10.1108/01439910410541873   
18. S. W. Lee, K.-G. Ahn and J. Ni, “Development of a Piezoelectric Multi-
Axis Stage Based on Stick-and-Clamping Actuation Technology,”
 A Survey of Modeling and Control of Piezoelectric ... 303

Smart Materials and Structures, Vol. 16, No. 6, 2007, pp. 2354-
2367. doi:10.1088/0964-1726/16/6/040   
19. S. Devasia, E. Eleftheriou and S. O. R. Moheimani, “A Survey of
Control Issues in Nanopositioning,” IEEE Transactions on Control
Systems Technology, Vol. 15, No. 5, 2007. pp. 802-823. doi:10.1109/
TCST.2007.903345   
20. P. Instrumente, “The World of Nanopositioning and Micropositioning
2005/2006,” Physik Instrumente, Karlsruhe, 2005.   
21. I. Mayergoyz and G.Bertotti (Eds.), “The Science of Hysteresis,” Vol.
3, Elsevier, St. Louis, 2005.   
22. H. J. M. T. S. Adriaens, W. L. De Koning and R. Banning, “Modeling
Piezoelectric Actuators,” IEEE/ASME Transactions on Mechatronics,
Vol. 5, No. 4, 2000, pp. 331- 341. doi:10.1109/3516.891044   
23. J. A. Main and E. Garcia, “Piezoelectric Stack Actuators and Control
System Design: Strategies and Pitfalls,” Journal of Guidance, Control,
and Dynamics, Vol. 20, No. 3, 1997, pp. 479-485. doi:10.2514/2.4066   
24. T. Fett and G. Thun, “Determination of Room-Temperature Tensile
Creep of PZT,” Journal of Materials Science Letters, Vol. 17, No. 22,
1998, pp. 1929-1931. doi:10.1023/A:1006608509876   
25. W. T. Ang, F. A. Garmón, P. Khosla and C. N. Riviere, “Modeling
Rate-Dependent Hysteresis in Piezoelectric Actuator,” 2003 IEEE/
RSJ International Conference on Intelligent Robots and Systems, Las
Vegas, 27 October-1 November 2003, pp. 1975-1980. doi:10.1109/
IROS.2003.1248937
26. W. T. Ang, P. K. Khosla and C. N. Riviere, “Feedforward Controller
with Inverse Rate-Dependent Model for Piezoelectric Actuators
in Trajectory-Tracking Applications,” IEEE/ASME Transactions
on Mechatronics, Vol. 12, No. 2, 2007, pp. 134-142. doi:10.1109/
TMECH.2007.892824
27. R. S. Robinson, “Interactive Computer Correction of Piezoelectric
Creep in Scanning Tunneling Microscopy Images,” Journal of
Computer-Assisted Microscopy, Vol. 2, No. 1, 1996, pp. 53-58.   
28. X. B. Chen, Q. Zhang, D. Kang and W. Zhang, “On the Dynamics of
Piezoactuated Positioning Systems,” Review of Scientific Instruments,
Vol. 79, No. 11, 2008, pp. 116101.1-116101.3. doi:10.1063/1.2982238   
29. W. Chen and C. S. Lynch, “A Micro-Electro-Mechanical Model for
Polarization Switching of Ferroelectric Materials,” Acta Materialia, Vol.
304 Mechatronics and Robotics

46, No. 15, 1998, pp. 5305- 5311. doi:10.1016/S1359-6454(98)00207-

9
30. W. Seemann, A. Arockiarajan and B. Delibas, “Modeling and
Simulation of Piezoceramic Materials Using Micromechanical
Approach,” European Congress on Computaional Methods in Applied
Sciences and Engineering 2004 (ECCOMAS 2004), Jyväskylä, 24-28
July 2004.
31. A. Arockiarajan, B. Delibas, A. Menzel and W. Seemann, “Studies
on Nonlinear Electromechanical Behavior of Piezoelectric Materials
Using Finite Element Modeling,” IWPMA 2005, 2nd International
Workshop on Piezoelectric Materials and Applications in Actuators,
Heinz Nixdorf Institute, Paderborn, 22-25 May 2005.
32. B. Delibas, A. Arockiarajan and W. Seemann, “Intergranular Effects
on Domain Switchings in Polycrystalline Piezoceramics,” IWPMA
2005, 2nd International Workshop on Piezoelectric Materials and
Applications in Actuators, Heinz Nixdorf Institute, Paderborn, 22-25
May 2005.
33. M. Kamlah and Q. Jiang, “A Constitutive Model for Ferroelectric PZT
Ceramics under Uniaxial Loading,” Smart Materials and Structures,
Vol. 8, No. 4, 1999, pp. 441- 459. doi:10.1088/0964-1726/8/4/302
34. M. Kamlah and C. Tsakmakis, “Phenomenological Modeling of
the Non-Linear Electro-Mechanical Coupling in Ferroelectrics,”
International Journal of Solids and Structures, Vol. 36, No. 5, 1999, pp.
669-695. doi:10.1016/S0020-7683(98)00040-7
35. M. Kamlah and U. Böhle, “Finite Element Analysis of Piezoceramic
Components Taking into Account Ferroelectric Hysteresis Behavior,”
International Journal of Solids and Structures, Vol. 38, No. 4, 2001, pp.
605-633. doi:10.1016/S0020-7683(00)00055-X
36. C. M. Landis, J. Wang and J. Sheng, “Microelectromechanical
Determination of the Possible Remanent Strain and Polarization
States in Polycrystalline Ferroelectrics and the Implications for
Phenomenological Constitutive Theories,” Journal of Intelligent
Material Systems and Structures, Vol. 15, No. 7, 2004, pp. 513-
525. doi:10.1177/1045389X04041653
37. J. Schröder and H. Romanowski, “A Simple Coordinate Invariant
Thermodynamic Consistent Model for Nonlinear Electro-Mechanical
Coupled Ferroelectrica,” European Congress on Computaional
Methods in Applied Sciences and Engineering 2004 (ECCOMAS
 A Survey of Modeling and Control of Piezoelectric ... 305

2004), Jyvä- skylä, 24-28 July 2004.

38. J. Schröder and H. Romanowski, “A Thermodynamically Consistent
Mesoscopic Model for Transversely Isotropic Ferroelectric Ceramics
in a Coordinate-Invariant Setting,” Achive of Applied Mechanics, Vol.
74, No. 11-12, 2005, pp. 863-877. doi:10.1007/s00419-005-0412-7
39. S. Klinkel, “A Thermodynamic Consistent 1D Model for Ferroelastic and
Ferroelectric Hysteresis Effects in Piezoceramics,” Communications
in Numerical Methods in Engineering, Vol. 22, No. 7, 2006, pp. 727-
739. doi:10.1002/cnm.845
40. T. Hegewald, B. Kaltenbacher, M. Kaltenbacher and R. Lerch, “Efficient
Modeling of Ferroelectric Behavior for the Analysis of Piezoceramic
Actuators,” Journal of Intelligent Material Systems and Structures, Vol.
19, No. 10, 2008, pp. 1117-1129. doi:10.1177/1045389X07083608   
41. IEEE, “ANSI/ IEEE Std. 176-1987: IEEE Standard on Piezoelectricity,”
The Institute of Electrical and Electronics Engineers, New York, 1988.   
42. P. G. Harper, “Kinematic Theory of Piezoelectric Hysteresis,”
Journal of Applied Physics, Vol. 52, No. 11, 1981, pp. 6851-
6855. doi:10.1063/1.328677   
43. D. C. Jiles and D. L. Atherton, “Theory of the Magnetisation Process in
Ferromagnets and Its Application to the Magnetomechanical Effect,”
Journal of Physics D: Applied Physics, Vol. 17, No. 6, 1984, pp. 1265-
1281. doi:10.1088/0022-3727/17/6/023   
44. R. C. Smith and C. L. Hom, “Domain Wall Model for Ferroelectric
Hysteresis,” Technical Report CRSC-TR99- 07, Center for Research
in Scientific Computation, New York, 1999.   
45. T. Hegewald, M. Kaltenbacher and R. Lerch, “Characterization and
Modeling of Piezoelectric Stack Actuators,” IWPMA 2005, 2nd
International Workshop on Piezoelectric Materials and Applications in
Actuators, Heinz Nixdorf Institute, Paderborn, 22-25 May 2005, pp.
99- 106.   
46. E. Preisach, “On the Magnetic Aftereffect,” Zeitschrift für Physik A
Hadrons and Nuclei, Vol. 94, No. 5-6, 1935, pp. 277-302.   
47. I. D. Mayergoyz, “Mathematical Models of Hysteresis,” Physical
Review Letters, Vol. 56, No. 15, 1986, pp. 1518- 1521. doi:10.1103/
PhysRevLett.56.1518   
48. I. D. Mayergoyz and G. Friedman, “Generalized Preisach model of
Hysteresis,” IEEE Transactions on Magnetics, Vol. 24, No. 1, 1988, pp.
306 Mechatronics and Robotics

212-217. doi:10.1109/20.43892
49. P. Hejda and T. Zelinka, “Generalized Preisach Model of Hysteresis—
Theory and Experiment,” Czechoslovak Journal of Physics, Vol. 40,
No. 1, 1990, pp. 57-68. doi:10.1007/BF01598355
50. M.-J. Jang, C.-L. Chen and J.-R. Lee, “Modeling and Control of a
Piezoelectric Actuator Driven System with Asymmetric Hysteresis,”
Journal of the Franklin Institute, Vol. 346, No. 1, 2009, pp. 17-
32. doi:10.1016/j.jfranklin.2008.06.005
51. P. N. Sreeram, G. Salvady and N. G. Naganathan, “Hysteresis
Prediction for a Piezoceramic Material System,” The ASME Winter
Annual Meeting, New Orleans, 28 November-3 December 3 1993, pp.
35-42.
52. D. C. Hughes and J. T. Wen, “Preisach Modeling and Compensation
for Smart Material Hysteresis,” Proceedings of SPIE, Vol. 2427, 1995,
pp. 50-64.
53. P. Ge and M. Jouaneh, “Generalized Preisach Model for Hysteresis
Nonlinearity of Piezoceramic Actuators,” Precision Engineering, Vol.
20, No. 2, 1997, pp. 99-111. do:10.1016/S0141-6359(97)00014-7
54. H. Hu and R. B. Mrad, “On the Classical Preisach Model for Hysteresis
in Piezoceramic Actuators,” Mechatronics, Vol. 13, No. 2, 2002, pp.
85-94. doi:10.1016/S0957-4158(01)00043-5
55. G. Robert, D. Damjanovic and N. Setter, “Preisach Modeling of
Piezoelectric Nonlinearity in Ferroelectric Ceramics,” Journal ofApplied
Physics, Vol. 89, No. 9, 2001, pp. 5067-5074. doi:10.1063/1.1359166
56. G. Song, J. Zhao, X. Zhou and J. A. De Abreu-Garcia, “Tracking Control
of a Piezoceramic Actuator with Hysteresis Compensation Using
Inverse Preisach Model,” IEEE/ASME Transactions on Mechatronics,
Vol. 10, No. 2, 2005, pp. 198-209. doi:10.1109/TMECH.2005.844708
57. X. Zhao and Y. Tan, “Neural Network based Identification of Preisach-
Type Hysteresis in Piezoelectric Actuator Using Hysteretic Operator,”
Sensors and Actuators A: Physical, Vol. 126, No. 2, 2006, pp. 306-
311. doi:10.1016/j.sna.2005.10.023
58. X. Zhao and Y. Tan, “Modeling Hysteresis in Piezo Actuator based on
Neural Networks,” Advances in Computation and Intelligence, Lecture
Notes in Computer Science, Vol. 5370, 2008, pp. 290-296.
59. K. Kuhnen and H. Janocha, “Adaptive Inverse Control of Piezoelectric
Actuators with Hysteresis Operators,” Proceedings of European
 A Survey of Modeling and Control of Piezoelectric ... 307

Control Conference (ECC), Karsruhe, 31 August-3 September 1999,

paper F 0291.
60. P. Krejci and K. Kuhnen, “Inverse Control of Systems with Hysteresis
and Creep,” IEE Proceedings—Control Theory and Applications, Vol.
148, No. 3, 2001, pp. 185- 192. doi:10.1049/ip-cta:20010375   
61. K. Kuhnen and H. Janocha, “Complex Hysteresis Modeling of a Broad
Class of Hysteretic Nonlinearities,” Proceedings of the 8th international
conference on New Actuators, Bremen, 10-12 June 2002, pp. 688-691.   
62. K. Kuhnen and F. Previdi, “Modeling, Identification and Compensation
of Complex Hysteretic Nonlinearities: A Modified Prandtl-Ishlinskii
Spproach,” European Journal of Control, Vol. 9, No. 4, 2003, pp. 407-
421. doi:10.3166/ejc.9.407-418   
63. K. Kuhnen, “Modeling, Identification and Compensation of Complex
Hysteretic Nonlinearities and Log(t)-Type Creep Dynamics,” Control
and Intelligent System, Vol. 33, No. 2, 2005, pp. 134-147. doi:10.2316/
Journal.201.2005.2.201-1420  
64. M. Goldfarb and N. Celanovic, “Modeling Piezoelectric Stack
Actuators for Control of Micromanipulation,” IEEE Control Systems
Magazine, Vol. 17, No. 3, 1997, pp. 69- 79. doi:10.1109/37.588158  
65. S.-H. Lee, T. J. Royston and G. Friedman, “Modeling and Compensation
of Hysteresis in Piezoceramic Transducers for Vibration Control,”
Journal of Intelligent Material Systems and Structures, Vol. 11, No. 10,
2000, pp. 781- 790. doi:10.1106/GQLJ-JGEU-MHG1-7JDF   
66. T.-J. Yeh, S.-W. Lu and T.-Y. Wu, “Modeling and Identification
of Hysteresis in Piezoelectric Actuators,” Journal of Dynamic
Systems, Measurement, and Control, Vol. 128, No. 2, 2006, pp. 189-
196. doi:10.1115/1.2192819   
67. T.-J. Yeh, R.-F. Hung and S.-W. Lu, “An Integrated Physical Model
that Characterizes Creep and Hysteresis in Piezoelectric Actuators,”
Simulation Modelling Practice and Theory, Vol. 16, No. 1, 2008, pp.
93-110. doi:10.1016/j.simpat.2007.11.005   
68. R. B. Mrad and H. Hu, “A Model for Voltage to Displacement Dynamics
in Piezoceramic Actuators Subject to Dynamic Voltage Excitations,”
IEEE/ASME Transactions on Mechatronics, Vol. 7, No. 4, 2002, pp.
479-489. doi:10.1109/TMECH.2002.802724   
69. D. Song and C. J. Li, “Modeling of Piezo Actuator’s Nonlinear and
Frequency Dependent Dynamics,” Mechatronics, Vol. 9, No. 4, 1999,
308 Mechatronics and Robotics

pp. 391-410. doi:10.1016/S0957-4158(99)00005-7   

70. Y. Yu, Z. Xiaob, N. G. Naganathan and R. V. Dukkipati, “Dynamic
Preisach Modeling of Hysteresis for the Piezoceramic Actuator
System,” Mechanism and Machine Theory, Vol. 37, No. 1, 2002, pp.
75-89. doi:10.1016/S0094-114X(01)00065-9   
71. X. Tan, R. Venkataraman and P. S. Krishnaprasad, “Control of
Hysteresis: Theory and Experimental Results,” Pentagon Technical
Report A687934, 2001.   
72. H. Hu and R. Ben Mrad, “A Discrete-Time Compensation Algorithm
for Hysteresis in Piezoceramic Actuators,” Mechanical Systems and
Signal Processing, Vol. 18, No. 1, 2004, pp. 169-185. doi:10.1016/
S0888-3270(03)00021-9   
73. S. Bashash and N. Jalili, “A Polynomial-Based Linear Mapping
Strategy for Feedforward Compensation of Hysteresis in Piezoelectric
Actuators,” Journal of Dynamic Systems, Measurement, and Control,
Vol. 130, No. 3, 2008, pp. 031008.1-031008.10. doi:10.1115/1.2907372   
74. R. Venkataraman and P. S. Krishnaprasad, “Novel Algorithm for the
Inversion of the Preisach Operator Smart Structures and Materials,”
Proceedings of SPIE 3984, Smart Structures and Materials 2000:
Mathematics and Control in Smart Structures, 2000, pp. 404-
414. doi:10.1117/12.388785   
75. R. Venkataraman and P. S. Krishnaprasad, “Approximate Inversion of
Hysteresis: Theory and Numerical Results,” Proceedings of the 39th
IEEE Conference on Decision and Control, Sydney, Australia, 12-15
December 2000, pp. 4448-4454. doi:10.1109/CDC.2001.914608   
76. J. Y. Peng and X. B. Chen, “H2-Optimal Digital Control of Piezoelectric
Actuators,” Proceedings of the 8th World Congress on Intelligent
Control and Automation, WCICA- 2010, Jinan, 7-9 July 2010, pp.
3684-3690. doi:10.1109/WCICA.2010.5553921   
77. U.-X. Tan, W. T. Latt, C. Y. Shee, C. N. Riviere and W. T. Ang,
“Feedforward Controller of Ill-Conditioned Hysteresis Using
Singularity-Free Prandtl-Ishlinskii Model,” IEEE/ASME Transactions
on Mechatronics, Vol. 14, No. 5, 2009, pp. 598-605. doi:10.1109/
TMECH.2008.2009936   
78. B. D. Coleman and M. L. Hodgdon, “A Constitutive Relation for
Rate-Independent Hysteresis in Ferromagnetically Soft Materials,”
International Journal of Engineering Science, Vol. 24, No. 6, 1986, pp.
 A Survey of Modeling and Control of Piezoelectric ... 309

897-919. doi:10.1016/0020-7225(86)90023-6   
79. R. Bouc, “Forced Vibrations of a Mechanical System with Hysteresis,”
Proceedings of the 4th Conference on Non-linear Oscillations, Prague,
5-9 September 1967, pp. 315.   
80. Y.-K. Wen, “Method of Random Vibration of Hysteretic Systems,”
Journal of the Engineering Mechanics Division, Vol. 102, No. 2, 1976,
pp. 249-263.   
81. M. Ismail, F. Ikhouane and J. Rodellar, “The Hysteresis Bouc-Wen
Model, a Survey,” Archives of Computational Methods in Engineering,
Vol. 16, No. 2, 2009, pp. 161- 188. doi:10.1007/s11831-009-9031-8   
82. O. Gomis-Bellmunt, F. Ikhouane and D. MontesinosMiracle, “Control
of Bouc-Wen Hysteretic Dystems: Application to a Piezoelectric
Actuator,” The 13th Power Electronics and Motion Control Conference,
2008. EPEPEMC 2008, Poznan, 1-3 September 2008, pp. 1670-
1675. doi:10.1109/EPEPEMC.2008.4635507   
83. M. Rakotondrabe, “Bouc-Wen Modeling and Inverse Multiplicative
Structure to Compensate Hysteresis Nonlinearity in Piezoelectric
Actuators,” IEEE Transactions on Automation Science and Engineering,
Vol. 8, No. 2, 2011, pp. 428-431. doi:0.1109/TASE.2010.2081979   
84. D. H. Wang, W. Zhu and Q. Yang, “Linearization of Stack Piezoelectric
Ceramic Actuators Based on BoucWen Model,” Journal of Intelligent
Material Systems and Structures, Vol. 22, No. 5, 2011, pp. 401-
413. doi:10.1177/1045389X10386132   
85. L. Deng and Y. Tan, “Modeling Hysteresis in Piezoelectric Actuators
Using NARMAX Models,” Sensors and Actuators A: Physical, Vol.
149, No. 1, 2009, pp. 106- 112. doi:10.1016/j.sna.2008.09.022   
86. A. Badel, J. Qiu and T. Nakano, “A New Simple Asymmetric Hysteresis
Operator and Its Application to Inverse Control of Piezoelectric
Actuators,” IEEE Transactions on Ultrasonics, Ferroelectrics and
Frequency Control, Vol. 55, No. 5, 2008, pp. 1086-1094. doi:10.1109/
TUFFC.2008.761   
87. H. Jung, J. Y. Shim and D. Gweon, “New Open-Loop Actuating
Method of Piezoelectric Actuators for Removing Hysteresis and
Creep,” Review of Scientific Instruments, Vol. 71, No. 9, 2000, pp.
3436-3440. doi:10.1063/1.1287627   
88. K. Kuhnen and H. Janocha, “Compensation of Creep and Hysteresis
Effects of Piezoelectric Actuators with Inverse Systems,” 6th
310 Mechatronics and Robotics

International Conference on New Actuators, Actuator’98, Bremen, 17-

19 June 1998, pp. 309-312.   
89. H. Janocha and K. Kuhnen, “Real-Time Compensation of Hysteresis
and Creep in Piezoelectric Actuators,” Sensors Actuators A: Physical,
Vol. 79, No. 2, 2000, pp. 83-89. doi:10.1016/S0924-4247(99)00215-0   
90. J. S. Dewey, K. Leang and S. Devasia, “Experimental and Theoretical
Results in Output-Trajectory Redesign for Flexible Structures,” Journal
of Dynamic Systems, Measurement, and Control, Vol. 120, No. 4,
1998, pp. 456-461. doi:10.1115/1.2801486   
91. K. S. Abidi and A. Şabanoviç, “Sliding Mode Control for High-Precision
Motion of a Piezostage,” IEEE Transactions on Industrial Electronics,
Vol. 54, No. 1, 2007, pp. 629-637. doi:10.1109/TIE.2006.885477   
92. H. C. Liaw, B. Shirinzadeh and J. Smith, “Enhanced Sliding Mode
Motion Tracking Control of Piezoelectric Actuators,” Sensors and
Actuators A: Physical, Vol. 138, No. 1, 2007, pp. 194-202. doi:10.1016/j.
sna.2007.04.062   
93. G. M. Clayton, S. Tien, A. J. Fleming, S. O. R. Moheimani and S.
Devasia, “Inverse-Feedforward of ChargeControlled Piezopositioners,”
Mechatronics, Vol. 18, No. 5-6, 2008, pp. 273-281. doi:10.1016/j.
mechatronics.2007.07.006   
94. J. W. Li, X. B. Chen and W. J. Zhang, “A New Approach to Modeling
System Dynamics—In the Case of a Piezoelectric Actuator with a Host
System,” IEEE/ASME Transactions on Mechatronics, Vol. 15, No. 3,
2010, pp. 371- 380. doi:10.1109/TMECH.2009.2026473   
95. J. W. Li, X. B. Chen and W. J. Zhang, “Axiomatic-Design-Theory-
Based Approach to Modeling Linear High Order System Dynamics,”
IEEE/ASME Transactions on Mechatronics, Vol. 16, No. 2, 2011, pp.
341-350. doi:10.1109/TMECH.2010.2043535   
96. K. Kuhnen, H. Janocha, D. Thull and A. Kugi, “A New Drive Concept
for High-Speed Positioning of Piezoelectric Actuators,” Proceedings
of the 10th International Conference on New Actuators, Bremen, 14-16
June 2006, pp. 82-85.   
97. K. K. Leang and S. Devasia, “Feedback-Linearized Inverse Feedforward
for Creep, Hysteresis, and Vibration Compensation in AFM
Piezoactuators,” IEEE Transactions on Control Systems Technology,
Vol. 15, No. 5, 2007, pp. 927-935. doi:10.1109/TCST.2007.902956   
98. X. Dang and Y. Tan, “An Inner Product-Based Dynamic Neural Network
 A Survey of Modeling and Control of Piezoelectric ... 311

Hysteresis Model for Piezoceramic Actuators,” Sensors and Actuators

A: Physical, Vol. 121, No. 2, 2005, pp. 535-542. doi:10.1016/j.
sna.2005.04.003   
99. C. Ru, L. Chen, B. Shao, W. Rong and L. Sun, “A Hysteresis
Compensation Method of Piezoelectric Actuator: Model, Identification
and Control,” Control Engineering Practice, Vol. 17, No. 9, 2009, pp.
1107-1114. doi:10.1016/j.conengprac.2009.04.013   
100. L. Sun, C. Ru, W. Rong, L. Chen and M. Kong, “Tracking Control of
Piezoelectric Actuator Based on a New Mathematical Model,” Journal
of Micromechanics and Microengineering, Vol. 14, No. 11, 2004, pp.
1439-1444. doi:10.1088/0960-1317/14/11/001   
101. C. Ru and L. Sun, “Hysteresis and Creep Compensation for Piezoelectric
Actuator in Open-Loop Operation,” Sensors and Actuators A: Physical,
Vol. 122, No. 1, 2005, pp. 124-130. doi:10.1016/j.sna.2005.03.056   
102. H. H. Najafabadi, S. M. Rezaei, S. S. Ghidary, M. Zareinejad, K.
Razi and R. Seifabadi, “Hysteresis Compensation of Piezoelectric
Actuators under Dynamic Load Condition,” IEEE/RSJ International
Conference on Intelligent Robots and Systems, 2007. IROS 2007, San
Diego, 29 October-2 November 2007, pp. 1166-1171. doi:10.1109/
IROS.2007.4399048   
103. E. Kouno, “A Fast Response Piezoelectric Actuator for Servo
Correction of Systematic Errors in Precision Machining,” CIRP
Annals-Manufacturing Technology, Vol. 33, No. 1, 1982, pp. 369-
372. doi: 10.1016/S0007-8506(07)61444-9   
104. J. F. Cuttino, A. C. Miller and D. E. Schinstock, “Performance
Optimization of a Fast Tool Servo for Single-Point Diamond Turning
Machines,” IEEE/ASME Transactions on Mechatronics, Vol. 4, No. 2,
1999, pp. 169-179. doi:10.1109/3516.769543   
105. J. Lin, H. Chiang and C. C. Lin, “Tuning PID Control Gains for Micro
Piezo-Stage in using Grey Relational Analysis,” 2008 International
Conference on Machine Learning and Cybernetics, Kunming, 12-15
July 2008, pp. 3863-3868. doi: 10.1109/ICMLC.2008.4621078   
106. H.-J. Shieh, Y.-J. Chiu and Y.-T. Chen, “Optimal PID Control System
of a Piezoelectric Microospitioner,” 2008 IEEE/SICE International
Symposium on System Integration, Nagoya, 4 December 2008, pp.
1-5. doi:10.1109/SI.2008.4770417   
107. D. Y. Abramovitch, S. Hoen and R. Workman, “SemiAutomatic
312 Mechatronics and Robotics

Tuning of PID Gains for Atomic Force Microscopes,” Asian Journal of

Control, Vol. 11, No. 2, 2009, pp. 188-195. doi:10.1002/asjc.95   
108. K. K. Tan, T. H. Lee and H. X. Zhou, “Micro-Positioning of Linear-
Piezoelectric Motors Based on a Learning Nonlinear PID Controller,”
IEEE/ASME Transactions on Mechatronics, Vol. 6, No. 4, 2001, pp.
428-436. doi:10.1109/3516.974856   
109. W. S. Oates and R. C. Smith, “Nonlinear Control Design for a
Piezoelectric-Driven Nanopositioning Stage,” Pentagon Technical
Report A591444, 2005.   
110. C. Edwards and S. K. Spurgeon, “Sliding Mode Control: Theory and
Applications,” Taylor & Francis, Abingdon, 1998.   
111. P.-K. Huang, P.-H. Shieh, F.-J. Lin and H.-J. Shieh, “Sliding-
Mode Control for a Two-Dimensional PiezoPositioning Stage,”
Control Theory & Applications, IET, Vol. 1, No. 4, 2007, pp. 1104-
1113. doi:0.1049/iet-cta:20060371   
112. J.-C. Shen, W.-Y. Jywe1, C.-H. Liu, Y.-T. Jian and J. Yang, “Sliding-
Mode Control of a Three-Degrees-ofFreedom Nanopositioner,” Asian
Journal of Control, Vol. 10, No. 3, 2008, pp. 267-276. doi:10.1002/
asjc.33   
113. J.-C. Shen, J.-C. Shen, H.-K. Chiang and Y.-L. Shu, “Precision Tracking
Control of a Piezoelectric-Actuated System,” Precision Engineering,
Vol. 32, No. 2, 2008, pp. 71-78. doi:10.1016/j.precisioneng.2007.04.002   
114. Q. Xu and Y. Li, “Dynamics Modeling and Sliding Mode Control of
an XY Micropositioning Stage,” The 9th International Symposium on
Robot Control (SYROCO’09), Gifu, 9-12 September 2009, pp. 781-
786.   
115. Y. Li and Q. Xu, “Adaptive Sliding Mode Control with Perturbation
Estimation and PID Sliding Surface for Motion Tracking of a
Piezo-Driven Micromanipulator,” IEEE Transactions on Control
Systems Technology, Vol. 18, No. 4, 2010, pp. 798-810. doi:10.1109/
TCST.2009.2028878   
116. H. C. Liaw, B. Shirinzadeh and J. Smith, “Sliding-Mode Enhanced
Adaptive Motion Tracking Control of Piezoelectric Actuation Systems
for Micro/Nano Manipulation,” IEEE Transactions on Control
Systems Technology, Vol. 16, No. 4, 2008, pp. 826-833. doi:10.1109/
TCST.2007.916301   
117. X. Chen and T. Hisayama, “Adaptive Sliding-Mode Position Control
 A Survey of Modeling and Control of Piezoelectric ... 313

for Piezo-Actuated Stage,” IEEE Transactions on Industrial Electronics,

Vol. 55, No. 11, 2008, pp. 3927-3934. doi:10.1109/TIE.2008.926768   
118. H. C. Liaw and B. Shirinzadeh, “Enhanced Sliding-Mode Constrained
Motion Tracking Control of Piezo-Actuated Flexure-Based
Mechanisms,” IEEE/ASME International Conference on Advanced
Intelligent Mechatronics, 2009. AIM 2009, Singapore, 14-17 July
2009 2009, pp. 1879- 1884. doi:10.1109/AIM.2009.5229763   
119. G. Schitter, P. Menold, H. F. Knapp, F. Allgower and A. Stemmer, “High
Performance Feedback for Fast Scanning Atomic Force Microscopes,”
Review of Scientific Instruments, Vol. 72, No. 8, 2001, pp. 3320-
3327. doi:10.1063/1.1387253   
120. M. S. Tsai and J. S. Chen, “Robust Tracking Control of a Piezoactuator
Using a New Approximate Hysteresis Model,” Journal of Dynamic
Systems, Measurement, and Control, Vol. 125, No. 1, 2003, pp. 96-
102. doi:10.1115/1.1540114   
121. G. Schitter, A. Stemmer, and F. Allgöwer, “Robust TwoDegree-of-
Freedom Control of an Atomic Force Microscope,” Asian Journal
of Control, Vol. 6, No. 2, 2004, pp. 156-163. doi: 10.1111/j.1934-
6093.2004.tb00194.x   
122. A. Sebastian and S. M. Salapaka, “Design Methodologies for Robust
Nano-Positioning,” IEEE Transactions on Control Systems Technology,
Vol. 13, No. 6, 2005, pp. 868-876. doi:10.1109/TCST.2005.854336   
123. Y. Okazaki, “A Micro-Positioning Tool Post Using a Piezoelectric
Actuator for Diamond Turning Machines,” Precision Engineering, Vol.
12, No. 3, 1990, pp. 151-156. doi:10.1016/0141-6359(90)90087-F   
124. D. Croft, S. Stilson and S. Devasia, “Optimal Tracking of Piezo-Based
Nanopositioners,” Nanotechnology, Vol. 10, No. 2, 1999, pp. 201-
208. doi:10.1088/0957-4484/10/2/316   
125. C. J. Li, H. S. M. Beigi, S. Li and J. Liang, “Nonlinear Piezo-Actuator
Control by Learning Self Tuning Regulator,” Journal of Dynamic
Systems, Measurement, and Control, Vol. 115, No. 4, 1993, pp. 720-
723. doi:10.1115/1.2899203   
126. H.-J. Shieh, F.-J. Lin, P.-K. Huang and L.-T. Teng, “Adaptive Tracking
Control Solely Using Displacement Feedback for a Piezo-Positioning
Mechanism,” IEE Proceedings——Control Theory and Applications,
Vol. 151, No. 5, 2004, pp. 653-660. doi:10.1049/ip-cta:20040795   
127. X. Tan and J. S. Baras, “Adaptive Identification and Control of Hysteresis
314 Mechatronics and Robotics

in Smart Materials,” IEEE Transactions on Automatic Control, Vol. 50,

No. 6, 2005, pp. 827-839. doi:10.1109/TAC.2005.849215   
128. H. C. Liaw and B. Shirinzadeh, “Enhanced Adaptive Motion Tracking
Control of Piezo-Actuated FlexureBased Four-Bar Mechanisms for
Micro/Nano Manipulation,” Sensors and Actuators A: Physical, Vol.
147, No. 1, 2008, pp. 254-262. doi:10.1016/j.sna.2008.03.020   
129. J.-H. Xu, “Neural Network Control of a Piezo Tool Positioner,” Canadian
Conference on Electrical and Computer Engineering, Vancouver, 14-
17 September 1993, pp. 333-336. doi:0.1109/CCECE.1993.332324   
130. F.-J. Lin, H.-J. Shieh and P.-K. Huang, “Adaptive Wavelet Neural
Network Control with Hysteresis Estimation for Piezo-Positioning
Mechanism,” IEEE Transactions on Neural Networks, Vol. 17, No. 2,
2006, pp. 432-444. doi:10.1109/TNN.2005.863473   
131. F.-J. Lin, H.-J. Shieh, P.-K. Huang and L.-T. Teng, “Adaptive Control
with Hysteresis Estimation and Compensation Using RFNN for Piezo-
Actuator,” IEEE Transactions on Ultrasonics, Ferroelectrics and
Frequency Control, Vol. 53, No. 9, 2006, pp. 1649-1661. doi:10.1109/
TUFFC.2006.1678193   
132. H. Numasato and M. Tomizuka, “Settling Control and Performance
of a Dual-Actuator System for Hard Disk Drives,” IEEE/ASME
Transactions on Mechatronics, Vol. 8, No. 4, 2003, pp. 431-
438. doi:10.1109/TMECH.2003.819999   
133. K. Leang and S. Devasia, “Hysteresis, Creep, and Vibration
Compensation for Piezoactuators: Feedback and Feedforward
Control,” Proceedings of the 2nd IFAC Conference on Mechatronic
Systems, Berkley, 9-11 December 2002, pp. 283-289.   
134. L. Y. Pao, J. A. Butterworth and D. Y. Abramovitch, “Combined
Feedforward/Feedback Control of Atomic Force Microscopes,”
Proceedings of the 2007 American Control Conference, New York
City, 9-13 July 2007, pp. 3509-3515. doi:10.1109/ACC.2007.4282338   
135. D. Croft and S. Devasia, “Vibration Compensation for High Speed
Scanning Tunneling Microscopy,” Review of Scientific Instruments,
Vol. 70, No. 12, 1999, pp. 4600- 4605. doi:10.1063/1.1150119   
136. H. Hu, H. M. S. Georgiou and R. Ben-Mrad, “Enhancement of Tracking
Ability in Piezoceramic Actuators Subject to Dynamic Excitation
Conditions,” IEEE/ASME Transactions on Mechatronics, Vol. 10, No.
2, 2005, pp. 230-239. doi:10.1109/TMECH.2005.844705   
 A Survey of Modeling and Control of Piezoelectric ... 315

137. C.-J. Lin and S.-R. Yang, “Modeling of a Piezo-Actuated Positioning

Stage Based on a Hysteresis Observer,” Asian Journal of Control, Vol.
7, No. 1, 2005, pp. 73-80. doi:10.1111/j.1934-6093.2005.tb00230.x   
138. S. S. Aphale, S. Devasia and S. O. Reza Moheimani, “High-Bandwidth
Control of a Piezoelectric Nanopositioning Stage in the Presence of
Plant Uncertainties,” Nanotechnology, Vol. 19, No. 12, 2008, pp.
125503.1- 125503.9. doi:10.1088/0957-4484/19/12/125503   
139. D. Croft and S. Devasia, “Vibration Compensation for High Speed
Scanning Tunneling Microscopy,” Review of Scientific Instruments,
Vol. 70, No. 12, 1999, pp. 4600- 4605. doi:10.1063/1.1150119   
140. S. Khan, M. Elitas, E. D. Kunt and A. Sabanovic, “Discrete Sliding Mode
Control of Piezo Actuator in NanoScale Range,” IEEE International
Conference on Industrial Technology, 2006. ICIT 2006, Mumbai, 15-
17 December 2006, pp. 1454-1459. doi:10.1109/ICIT.2006.372418   
141. J. Yi, S. Chang and Y. Shen, “Disturbance-ObserverBased Hysteresis
Compensation for Piezoelectric Actuators,” IEEE/ASME Transactions
on Mechatronics, Vol. 14, No. 4, 2009, pp. 4196-4201. doi:10.1109/
TMECH.2009.2023986   
142. S. Chang and S. Li, “A High Resolution Long Travel Friction-Drive
Micropositioner with Programmable Step Size,” Review of Scientific
Instruments, Vol. 70, No. 6, 1999, pp. 2276-2782. doi:10.1063/1.1149794   
143. A. Bergander and J. Breguet, “Performance Improvements for Stick-
Slip Positioners,” Proceedings of 2003 International Symposium on
Micromechatronics and Human Science, Nagoya, 19-22 October 2003,
pp. 59-66. doi:10.1109/MHS.2003.1249910   
144. F. Castanos and L. Fridman, “Analysis and Design of Integral
Sliding Manifolds for Systems with Unmatched Perturbations,” IEEE
Transactions on Automatic Control, Vol. 51, No. 5, 2006, pp. 853-
858. doi:10.1109/TAC.2006.875008   
145. J. H. Painter, D. Kerstetter and S. Jowers, “Reconciling Steady-
State Kalman and Alpha-Beta Filter Design,” IEEE Transactions on
Aerospace and Electronic Systems, Vol. 26, No. 6, 1990, 1990, pp.
986-991. doi:10.1109/7.62250   
146. X. Wang, “High-Order Integral-Chain Differentiator and Application to
Acceleration Feedback,” Submitted to Computer Science for Possible
Publication, 2011.   
147. A. Levant, “Robust Exact Differentiation via Sliding Mode Technique,”
316 Mechatronics and Robotics

Automatica, Vol. 34, No. 3, 1998, pp. 379-384. doi:10.1016/S0005-

1098(97)00209-4   
148. Q. Xu and Y. Li, “Sliding Mode Control of a PiezoDriven
Micropositioning System Using Extended Kalman Filter,” IEEE
International Conference on Automation and Logistics (ICAL), 2010,
Hong Kong and Macau, 16- 20 August 2010, pp. 427-432. doi:10.1109/
ICAL.2010.5585322   
149. J. L. Minase, T.-F. Lu and F. Wornle, “State Estimation of Nonlinear
Piezoelectric Stack Actuator Hysteresis Model,” Proceedings of SPIE,
Vol. 6414, 2006, pp. 641403.1-641403.10.   
150. J. Minase, T.-F. Lu, B. Cazzolato and S. Grainger, “Adaptive
Identification of Hysteresis and Creep in Piezoelectric Stack Actuators,”
The International Journal of Advanced Manufacturing Technology,
Vol. 46, No. 9-12, 2009, pp. 913-921. doi:10.1007/s00170-009-2033-8   
151. F. Yang and R. W. Wilde, “Observers for Linear Systems with Unknown
Inputs,” IEEE Transactions on Automatic Control, Vol. 33, No. 7, 1988,
pp. 667-681. doi:10.1109/9.1278   
152. M. Hou and P. C. Muller, “Design of Observers for Linear Systems
with Unknown Inputs,” IEEE Transactions on Automatic Control, Vol.
37, No. 6, 1992, pp. 871-875. doi:10.1109/9.256351   
153. S. Hui and S. Żak, “Observer Design for Systems with Unknown
Inputs,” International Journal of Applied Mathematics and Computer
Science, Vol. 15, No. 4, 2005, pp. 431-546.   
154. S. H. Zak and S. Hui, S. “Output Feedback Variable Structure
Controllers and State Estimators for Uncertain/Nonlinear Dynamic
Systems,” IEE Proceedings D: Control Theory and Applications, Vol.
140, No. 1, 1993, pp. 41-50. doi:10.1049/ip-d.1993.0006   
155. J.-L. Chang and T.-C. Wu, “Robust Disturbance Attenuation with
Unknown Input Observer and Sliding Mode Controller,” Electrical
Engineering, Vol. 90, No. 7, 2008, pp. 493-502. doi:10.1007/s00202-
008-0099-1   
156. T. Floquet, C. Edwards and S. K. Spurgeon, “On Sliding Mode
Observers for Systems with Unknown Inputs,” International Journal
of Adaptive Control and Signal Processing, Vol. 21, No. 8-9, 2007, pp.
638-656. doi:10.1002/acs.958   
157. K. Kalsi, J. Lian, S. Hui and S. H. Zak, “Sliding-Mode Observers for
Systems with Unknown Inputs: A HighGain Approach,” Automatica,
 A Survey of Modeling and Control of Piezoelectric ... 317

Vol. 46, No. 2, 2010, pp. 347-353. doi: 10.1016/j.automatica.2009.10.040   

158. X. Liu, J. Jeong and J. Kim, “A Three Translational DoFS Parallel
Cube-Manipulator,” Robotica, Vol. 21, No. 6, 2003, pp. 645-
653. doi:10.1017/S0263574703005198   
159. Q. Xu, and Y. Li, “A Novel Design of a 3-PRC Translational Compliant
Parallel Micromanipulator for Nanomanipulation,” Robotica, Vol. 24,
No. 4, 2006, pp. 527-528. doi:10.1017/S0263574705002559   
160. Y. Yue, F. Gao, X. Zhao and Q. Jeffrey Ge, “Relationship among Input-
Force, Payload, Stiffness and Displacement of a 3-DOF Perpendicular
Parallel Micro-Manipulator,” Mechanism and Machine Theory, Vol. 45,
No. 5, 2010, pp. 756-771. doi:10.1016/j.mechmachtheory.2009.12.006   
161. S. Awtar and A. Slocum, “Constraint-Based Design of Parallel
Kinematic XY Flexure Mechanisms,” Journal of Mechanical Design,
Vol. 129, No. 8, 2007, pp. 816-830. doi:10.1115/1.2735342   
162. Q. Yao, J. Dong and P. Ferreira, “A Novel ParallelKinematics
Mechanisms for Integrated, Multi-Axis Nanopositioning: Part 1.
Kinematics and Design for Fabrication,” Precision Engineering, Vol.
32, No. 1, 2008, pp. 7- 19. doi:10.1016/j.precisioneng.2007.03.001   
163. J. Dong, Q. Yao and P. Ferreira, “A Novel ParallelKinematics Mechanism
for Integrated, Multi-Axis Nanopositioning: Part 2: Dynamics, Control
and Performance Analysis,” Precision Engineering, Vol. 32, No. 1,
2008, pp. 20-33. doi:10.1016/j.precisioneng.2007.03.002   
164. J. Dong, S. Salapaka and P. Ferreira, “Robust Control of a Parallel-
Kinematic Nanopositioner,” Journal of Dynamic Systems,
Measurement, and Control, Vol. 130, No. 4, 2008, pp. 041007.1-
041007.15. doi:10.1115/1.2936861   
165. Y. Tian, B. Shirinzadeh and D. Zhang, “A Flexure-Based Five-Bar
Mechanism for Micro/Nano Manipulation,” Sensors and Actuators
A: Physical, Vol. 153, No. 1, 2009, pp. 96-104. doi: 10.1016/j.
sna.2009.04.022   
166. Y. Tian, B. Shirinzadeh and D. Zhang, “Design and Dynamics
of a 3-DOF Flexure-Based Parallel Mechanism for Micro/Nano
Manipulation,” Microelectronic Engineering, Vol. 87, No. 2, 2010, pp.
230-241. doi:10.1016/j.mee.2009.08.001   
167. K. Hu, J. Kim, J. Schmiedeler and C. Menq, “Design,
Implementation, and Control of a Six-Axis Compliant Stage,”
Review of Scientific Instruments, Vol. 79, No. 2, 2008, pp. 025108.2-
318 Mechatronics and Robotics

025105.11. doi:10.1063/1.2841804   
168. T. W. Seo, D. S. Kang and J. Kim, “Synthesis and Comparison of
Fine Actuator Controllers for a 3-DOF Micro Parallel Positioning
Platform,” Proceedings of the 10th International Conference on
Control, Automation, Robotics and Vision, Hanoi, 17-20 December
2008, pp. 1606- 1611. doi:10.1109/ICARCV.2008.4795765   
169. M. Grotjahn, B. Heimann and H. Abdellatif, “Identification of Friction
and Rigid-Body Dynamics of Parallel Kinematic Structures for Model-
Based Control,” Multibody System Dynamics, Vol. 11, No. 3, 2004,
pp. 273-294. doi:10.1023/B:MUBO.0000029426.05860.c2  
 CHAPTER

15
Experimental Investigation of
Photostrictive Materials for MEMS
Application

Mosfequr Rahman1, Mahbub Ahmed2, Masud Nawaz1, Gustavo

Molina1, Abdur Rahman3
1
Georgia Southern University, Statesboro, GA, USA
2
Southern Arkansas University, Magnolia, AR, USA
3
Edinboro University, Edinboro, PA, USA

ABSTRACT
Micro-Electro-Mechanical-Systems (MEMS) is an emerging technology
that integrates micromachined mechanical structures with integrated circuits
(IC). During the last two decades, it has grown vastly as a prominent research
area among scientists and engineers. Usually the main components of MEMS

Citation:- Rahman, M. , Ahmed, M. , Nawaz, M. , Molina, G. and Rahman, A. (2017),

“Experimental Investigation of Photostrictive Materials for MEMS Application”. Open
Access Library Journal, 4, 1-23. doi: 10.4236/oalib.1103856.
Copyright: © 2017 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
320 Mechatronics and Robotics

are sensors, actuators, and structures with all in microscale. A photostrictive

material has high potential to be used in a MEMS application to develop an
advanced optical actuator. A strain is induced when a photostrictive material
is illuminated by high intensity light, which is known as photostriction.
This phenomenon can be generally described as the combined effect of the
photovoltaic effect and the converge-piezoelectric effect. When an incident
light strikes a photostrictive material, a photovoltage is produced and this
photovoltage develops strain in the material. Lanthanum modified lead
zirconate titanate (Pb, La) (Zr, Ti) O3 ceramic doped with WO3, called
PLZT, is one of the photostrictive ceramics which has an advantageous use
as a wireless remote control over traditional actuators. Traditional actuators
require wire connections to transmit the control signal; these wires yield
noise via external electromagnetic fields. Whereas, PLZT actuators can
transmit the control signal without wires which can eliminate possible
noise due to the external electromagnetic field. In this current research, the
photostrictive effect of a thin PLZT film on a silicon wafer is investigated
experimentally. The transverse deflection of the PLZT optical actuator
cantilever beam has been measured for stationary continuous light as well
as for pulses of light using an optical chopper at various light intensities and
focused locations. Results indicate that transverse deflection increases with
the increase of light intensity. Also, the maximum transverse deflection at
the free end of the cantilever beam for stationary light has been found to be
three times larger than that for the pulses of light.
Keywords: MEMS, Photostriction, PLZT, Actuators

INTRODUCTION
Photostriction is a phenomenon in which strain is induced in photostrictive
materials by incident light. In principle, this effect arises from a superposition
of the photovoltaic effect, i.e., the generation of large voltage from the
irradiation of light, and the converse piezoelectric effect, i.e., expansion
or contraction under the voltage applied. A “bulk photovoltaic effect” is
noticeable in some ferroelectric (or non-centrosymmetric) materials, such
as ferroelectric single crystals or polarized ferroelectric ceramics. Along
with this photovoltage, the converse piezoelectric effect induces mechanical
strain [1]. The combinations of these two effects make these ferroelectric
materials candidates for wireless actuator applications, that is, actuators
excited by incident light. Photostrictive ceramics are one promising material
for micro-fabrication technologies, which have photovoltaic, piezoelectric,
 Experimental Investigation of Photostrictive Materials for ... 321

and pyroelectric properties. Photostrictive materials offer the potential for

development of advanced actuators with many advantages over conventional
electro-mechanical actuators, such as non-contact actuation, remote control,
and immunity from electric/magnetic disturbances [2]. Many types of micro-
fabrication technologies, such as device design, lithography, film formation,
and structure and system integration, have been developed [3] . Traditional
actuators, like shape memory alloy and piezoelectric materials, have
drawbacks for hard-wire connections such as control signal contamination
due to external electro-magnetic disturbances. Photostrictive material
appears to have potential for less friction and less weight for use in micro-
electro mechanical systems (MEMS) due to its ability to be used wirelessly
or by remote control. Piezoelectric ceramics have been used commercially
as actuators in various fields; it requires hard wiring to supply voltage for
actuation. On the other hand, photostrictive actuators can directly convert
photonic energy to mechanical motion; they need neither electric lead wires
nor electric circuits. A photostrictive type of actuator uses pulses of light
instead of electric voltages for actuation. One material that possesses a
significant photostrictive behavior under illumination by high-energy light
is lanthanum-modified lead zirconate titanate (Pb, La) (Zr, Ti) O3 ceramic
doped with WO3, called PLZT. In PLZT ceramics, UV and visible light have
been found to excite free electrons within the crystalline structure that have
greater than band-gap voltages; this is characteristic of a photosensitive
material [4]. Also, like a typical ferroelectric material, PLZT can be poled to
a remnant saturation polarization state, and the polarization can be switched
by the application of an electric field. Due to these photosensitive and
ferroelectric responses, PLZT ceramics are often called photo-ferroelectric
(PFE) materials.
Many researchers have investigated the characteristics of photostrictive
materials and its possible applications for MEMS devices [3] [5] [6] [7].
In this current research, a PLZT optical actuator on a silicon wafer as a
cantilever beam is experimentally investigated to understand its response on
continuous illumination of light and pulses of light.

Motivation of the Research

Micro-Electro-Mechanical Systems or MEMS technology can be defined
as miniaturized mechanical and electro-mechanical elements that are
made using micro fabrication techniques. MEMS functional elements are
miniaturized structures, sensors, actuators, and microelectronics; the most
important elements are the micro-sensors and micro-actuators. Over the last
322 Mechatronics and Robotics

several decades, MEMS researchers have invented large numbers of micro-

sensors with many sensing modalities, including temperature, pressure,
inertial forces, chemical species, magnetic fields, optics, etc. Most recently,
MEMS researchers have developed optical switches and mirrors to redirect
or modulate light beams [8]. The present research will experimentally
investigate the feasibility of using photostrictive materials for switching
applications in MEMS devices. Switching applications in MEMS are used
to transfer data and control signals. A traditional transducer needs electrical
wire connections, which may contaminate the control signal by external
electro-magnetic field interference. Photostrictive materials, on the other
hand, do not need an electrical wire to supply voltage for actuation; they
require only external light to actuate and may be used for remote sensing.

Research Hypotheses
Transverse deflection of the PLZT optical actuator depends on the light
intensity. The deflection of the PLZT optical actuator will vary for the
continuous and pulses of lights. Fast response and significant deflection of
the PLZT optical actuator under the illumination of light could be used in
MEMS applications as a wireless control.

Research Objectives
Photostrictive actuators have the superiority of remote sensing capabilities and
freedom from external electric/magnetic disturbances, unlike conventional
actuators. Based on this inspiration from current research, a PLZT thin
film placed on the top of a silicon wafer is used to support a systematic
characterization of photostrictive properties to develop an actuator. The
main goal of the research is to investigate the potential of the PLZT material
as a wireless actuator to convert photonic energy to mechanical energy.
The objectives of the present research follow.
1) Investigate the photostrictive effect of a continuous high intensity
near-ultraviolet (UV) light on a PLZT optical actuator cantilever
beam, which includes:
• measuring transverse deflection of the photostrictive cantilever
beam for UV light focused at three different locations on the beam;
• for each focused location on the cantilever beam, measuring
the transverse deflection along the beam at three different light
intensities: 250 mW/cm2, 200 mW/cm2 and 100 mW/cm2; and
 Experimental Investigation of Photostrictive Materials for ... 323

• observing the time response of the cantilever beam deflection

at the free end of the cantilever beam at three different light
intensities and focused locations.
2) Investigate the photostrictive effect of the pulses of high intensity
UV light on a PLZT optical actuator cantilever beam using an
optical chopper, which includes:
• measuring transverse deflection of the photostrictive actuator at
three different light intensities: 250 mW/cm2, 200 mW/cm2 and
100 mW/cm2; and
• observing the time response of the cantilever beam at the free end
of the beam at three different light intensities.
3) Compare the transverse deflections along the optical actuator
cantilever beam between continuous and pulses of light conditions
at three light intensities.

Background of Photostrictive Actuators

Characteristics of the PLZT Ceramics

PLZT compositions are represented as Pb1−xLax(ZryTiz)1−x/4O3 or PLZT(X/
Y/Z), where X = 100x, Y = 100y and Z = 100z. Ferroelectric materials have a
wide range of properties depending on the composition [3]. The photovoltaic
effect was first discovered in BaTiO3 material [9]. P.S. Brody first proposed
photostrictive materials as an opto-mechanical actuator, which can be used
for MEMS devices to convert optical energy to mechanical energy [10]
[11] . When ultraviolet (UV) light falls on the materials, it absorbs photons.
Absorbed photons energize electrons and holes to move in a non-random
manner; as a result, charges started to flow in an external short circuit of
PLZT material. Light illuminates from the top and along the thickness of
the material in the H direction. The photocurrent of PLZT ceramics varies
proportionally with light intensity. Photocurrent increases as the thickness of
PLZT film decreases, up to a certain limit [12]. The maximum photocurrent
is predicted at 33 μm thickness for PLZT (3/52/48) ceramic doped with
0.5% WO3 [13]. If film thickness becomes ultra-thin, then the film does not
have enough electrons to flow and photocurrent drops because of the size
effect [12]. Photocurrent of PLZT materials also depends on the wavelength
of illuminating light. The maximum photocurrent is observed at 365 nm
wavelength of light, which is out of visible light wavelength range [14]. No
significant difference in wavelength dependency has been found between
324 Mechatronics and Robotics

impurity-doped or non-doped samples. However, the valence of impurity

has a significant effect on the photovoltaic response. Impurity atoms are
classified into three categories: acceptors, donors, and remainder. K+, Na+,
Mg2+, Ba2+, Al3+ & Fe3+ atoms are in the acceptor group; Bi3+, Y3+, Nb5+,
Ta5+ & W6+ are in the donor group. Photostriction (xph) can be defined in
terms of photovoltaic voltage (Eph) and piezoelectric coefficient (d33) as
xph = d33Eph. Impurities-doping has an effect on the piezoelectric coefficient,
which improves photostriction [4] . Steady photovoltage and photocurrent
generate in non-centrosymmetric ferroelectric material when illumination
light has a wavelength corresponding to the absorption edge of the materials
[1] . Generated photovoltage is greater than the band-gap energy for some
materials. In certain ferroelectric material, photovoltage is about 1 kV/cm
and current is in the order of nA; with this photovoltage, mechanical strain
is produced due to converse piezoelectric effect [4] [13] . Several models
have been proposed to describe the photovoltaic effect [13] . Photovoltage
depends on both impurities-doping as well as crystal asymmetry.

Literature Review on the Transverse Deflection of the PLZT

Cantilever Beam with the Light Intensity
Tonokai et al. experimentally investigated PLZT on a 8 mm × 2 mm × 0.05
mm Ti wafer at 50 mW/cm2, 150 mW/cm2, 300 mW/cm2 and 450 mW/
cm2 UV light intensities; deflections were found to be 5 µm, 11 µm, 18 µm
and 25 µm respectively [15] . Liang et al. experimented on a single-layer
PLZT at 3 mW/cm2, 6 mW/cm2, 9 mW/cm2, 12 mW/cm2, and displacements
were reported to be 0.31 µm, 0.33 µm, 0.42 m and 0.57 µm respectively
[16] . Liang et al. developed a bimorph with amplificatory plate to reach
deflection 100 times that of a single layer. The bimorph is constructed by
reversing the polarization direction of a pair of PLZT unilateral wafers.
The upper PLZT wafer develops a positive charge and expands due to UV
light irradiation, and the lower PLZT wafer develops a negative charge and
contracts at the same time; as a result, the entire bimorph bends downwards.
Thakoor et al. reported that two PLZT ceramics wafers bond together to
make a bimorph. The size of the bimorph is 20 mm × 5 mm × 0.4 mm. The
maximum deflection was observed 200 µm at the tip at 80 mW/cm2 [17].
Fukuda et al. experimented on the two pieces of 21 mm × 5 mm × 0.2
mm PLZT, glued in the opposite direction of polarization; the maximum
deflections were observed to be 61 µm, 58 µm and 37 µm at 200 mW/cm2,
100 mW/cm2 and 25 mW/cm2 light intensity respectively [18].
 Experimental Investigation of Photostrictive Materials for ... 325

Literature Review on the MEMS Applications

Photostrictive effect of PLZT is useful for MEMS applications because it
can serve as a current and voltage source for optical sensors at the same time.
However, it requires some improvements: 1) improvement of output current;
and 2) improvement of response [19]. Using the electrical properties of PLZT,
an optical motor was developed as an electro-mechanical device suitable for
miniaturization [3]. The optical motor driving principle was described with
stator pads and moving pads. The PLZT electrode is connected with stator
pads. When light falls on a PLZT electrode, high voltage generates between
the stator and moving pad; then the moving pad starts to move towards
the stator side. The proposed optical motor rotor disk has four pads, and
each stator disk has three pads. The rotor disk was fixed or aligned with the
motor axis, which is on the bearing. When light falls on PLZT ceramics,
high voltage generates in the stator, and the rotor is pulled into the stator in
a clockwise direction from the top view. The optical motor could be rotated,
but its speed would be very slow. Internet traffic has grown enormously with
the development of broadband access; a PLZT optical actuator could play a
significant role in high-speed data transfer systems [20] . PLZT actuators can
be used as high speed optical switches with benefits including low-voltage,
low power consumption, low-polarization independence, noise robustness
and switching speed of less than a nano second [21] [22] . Furukawa et al.
made a 1 × 8 PLZT optical switch whose switching speed is less than 2.5 ns
[23]. Satol et al. developed an 8 × 8 channel switch comprised of a PLZT
deflector, fiber array, and micro-lens and slab waveguide [7]. Switching
speed was less than 1 µs. PLZT bimorphs could also be used as a photo
driven relay, consisting of two ceramic plates bonded together in opposite
polarization directions and a dummy PLZT plate placed adjacent to the
bimorph. Switching could be controlled by alternately irradiating light on the
bimorph and dummy plates [5]. A micro-walking machine was developed
using photostrictive bimorphs. Two PLZT ceramic legs are fixed on a
plastic board; light irradiates on the legs alternately and the micro-walking
machine moves like an inchworm at a very low speed of 10 µm/min [24].
An optical gripper was developed using a PLZT bimorph actuator. There
are two bimorph actuators and UV light is irradiated by optical fibers on
each actuator independently using two mirrors. Displacement of the gripper
at the tip was reported to be 100 µm. The gripper has strong anti-noise
characteristics and could be used for micro-surgery [18]. A photocurrent
generates due to irradiation of light on the photostrictive actuator; this
property has been used as an image comparator. One image is stored as
326 Mechatronics and Robotics

a sequence of spatially varying ferroelectric domains. The second image

is projected on the first and generates a photocurrent representing the dot
product of stored and projected images [25]. Photostrictive materials also
can be used as solar tracking shutter and photo phone [16]. Photo generated
deflection could be used to walk on the surface or to run a motor indirectly.
This kind of wireless optical control of an advanced mobile vehicle would
be used for hazardous, hard to reach locations, as long as it is in the range of
sight. Scattering effect of PLZT has been used to construct display devices
and optical shutters, optical information storage and processing [17] [26] .

Basic Photostriction Equations

When the high-energy illumination with certain wavelength irradiates the
photostrictive optical actuator, a current opposite to the polarized direction
is induced. This current generates a voltage between the two surface
electrodes, called the photovoltaic effect. The induced photovoltaic voltage
induces actuation strains due to the converse piezoelectric effect. The
induced electric field El(t) from the photovoltaic effect can be determined
from [27] :

(1)
where Es is the saturated photovoltaic field; as = La/ba (length/width)
is the aspect ratio; α is optical actuator constant and β is a constant related
to voltage leakage. The light intensity I(t) is a function of time, and El(tj)
denotes the induced electric field at time tj. While I(t) remains constant
during the interval Δt, then the electric field at tj=tj−1+Δttj=tj−1+Δt is El(tj).
The actuator temperature can be defined in a difference equation [27]:

(2)
where P is the absorbed heat from the light illumination; H is the heat
capacity of the actuator; g is the heat transfer rate from the actuator to the
structure; θ(tj) is the actuator body temperature at tj; I(tj) is an illumination
intensity at tj; and Δθ is a temperature rise in a time interval Δt.
The temperature change can trigger the pyroelectric effect from which
an additional voltage is generated. The electric field Eq(t), contributed by
pyroelectric effect, can be determined from [27] :
 Experimental Investigation of Photostrictive Materials for ... 327

(3)
where Pn is the pyroelectric constant.
The total induced electric field E(t) including both the photovoltaic
effect and the pyroelectric effect is [27] :

(4)
The magnitude of the light induced strains can be written as [18] :

(5)
where d33 is the piezoelectric-strain constant, λ is a thermal stress
coefficient, and Ya is the Young’s modulus of the optical actuator.
The electric potential of the actuator is

(6)
The objective of the current research is to investigate the influence of
continuous and pulses of light on the transverse deflection of PLZT optical
actuator. How the transverse deflection of a cantilever beam depends
on the light intensity (both continuous and pulses) will be investigated
experimentally. If the intensity of the light and the focused locations of
the beam affect the time response will also be investigated in the current
research.
METHODOLOGY
Experimentation has been performed mainly using a high intensity light
source, a good repeatability displacement measurement sensor, and a
photostrictive optical actuator.
The goal of the experiment was to observe the transverse deflection of a
PLZT thin film on a silicon wafer cantilever beam under the illumination of
continuous and pulses of high intensity light.

Experiment Method

Experimental Set-Up Design and Development

The experimental test set-up was designed to measure the photostriction
effect of a photostrictive optical actuator as shown in Figure 1.
328 Mechatronics and Robotics

The set-up consists of a high pressure short arc xenon lamp with lamp
housing, power supply with igniter, hot mirror, band pass filter, optical
chopper, photostrictive cantilever type optical actuator, and laser sensor
head with controller. Based on the design, an actual experimental set-up was
developed as shown in Figure 2.
A sample of PLZT (3/52/48) ceramic was doped with 0.5% WO3 of 1 µm
thin film deposited on top of a 550 µm silicon wafer as shown in Figure 3.
PLZT material was deposited on the silicon wafer at the MEMS Laboratory
of Georgia Institute of Technology using the spattering method. PLZT
(3/52/48) means Pb concentration of 97%, La concentration of 3% with Zr/
Ti ratio 52/48. (3/52/48) ceramic also can be represented as Pb0.97 La 0.03
(Zr0.52Ti 0.48) 0.99 O3.
A 150 watt xenon short arc lamp was provided high intensity light for
maximum photovoltaic effect on the optical actuator. The arc lamp was
placed inside the Science Tech 201 - 100 series lamp housing. A model 500
- 200 series AC power supply and a model 500-IG igniter were used to turn
on the short arc lamp. To produce pulses of high intensity light on an optical
actuator the Scitech 300CD model an optical chopper was used. A Keyence
LK-H087 laser sensor head with 0.1 µm repeatability was used to measure
the transverse deflection of actuator beam.

Figure 1. Schematic diagram of the experimental set-up for photostriction ef-

fect measurement.
 Experimental Investigation of Photostrictive Materials for ... 329

Figure 2. Experiment set-up for transverse deflection measurement of a canti-

lever beam.

Figure 3. PLZT sample manufactured at the MEMS laboratory.

Laser sensor data acquisition was controlled by the Keyence LK-G5001V
model laser sensor controller. The beam deflection data was recorded in the
computer through the laser sensor controller and data acquisition system.
A heat absorbing glass was placed in front of the lamp housing, to filter
high wavelength light that was generated by the arc lamp. In addition, a UV
band-pass filter was placed behind a heat-absorbing filter, which allowed
only 250 nm to 410 nm wavelengths of light to pass. This high intensity UV
light went through a lamp housing window lens and was used to produce
the photostriction effect on the optical actuator. A diamond-cutting blade
was used to cut the sample into the desired shape (18 mm × 8 mm) for
330 Mechatronics and Robotics

the experiment. PLZT thin film on silicon wafer sample was mounted as
a cantilever beam on a stand. An LK-H087 laser displacement sensor was
mounted on a linear actuator at the other side of the light source. The linear
actuator can travel a minimum of 0.1 mm/step in the horizontal direction,
thus allowing the laser displacement sensor head to move along the length
of the sample cantilever beam. To irradiate pulses of light on the sample
a variable frequency optical chopper was placed between the sample and
the lamp housing window lens. Source light beam has a maximum light
intensity of 250 mW/cm2 at its focal point through the lamp housing window
lens at 20 cm distance, which was measured by a UV meter. The UV meter
sensor head was 12 mm in diameter, which measured the intensity on the
average of sensor head area.
It was not experimentally possible for the whole cantilever beam to be
exposed with equal light intensity. The light intensity starts to decrease as
far as from the center of focal point on its perpendicular plane. Experiments
were performed by focusing the light on the targeted cantilever beam at
three different locations; these are the free end, middle, and fixed end as
shown in Figure 4. Using the UV meter the light intensities were measured,
at a distance of 20 cm, 25 cm, and 35 cm from the lamp housing window
lens, were 250 mW/cm2, 200 mW/cm2, and 100 mW/cm2 respectively.

Experimental Procedure
In conducting the experiment to produce photostriction effect and measuring
the transverse deflection of the actuator, the following steps have been
followed:
• A high-pressure short arc xenon lamp was placed properly inside
the lamp housing with right polarity of cathode (−) and anode (+).
This was used as a UV light source for the experiment.
• A power supply and a lamp igniter were connected to the lamp
housing; and the lamp housing was placed on a vibration isolation
table.
• A hot mirror lens and a 428 nm wavelength filter were placed after
the lamp inside the housing filter slot. The source light first passed
through the hot mirror then through the 428 nm wavelength filter.
• Output light was focused on the PLZT optical actuator cantilever
beam at different distances from the output lens to get various
light intensities on the targeted area.
 Experimental Investigation of Photostrictive Materials for ... 331

• A frequency variable optical chopper was placed in between the

light output lens and the PLZT cantilever beam to produce pulses
of light on the targeted area as shown in Figure 5.

Figure 4. UV light was focused (a) At the fixed end; (b) At the middle; and (c)
At the free end of the cantilever beam.

Figure 5. Optical chopper placement in the experiment set up to observe re-

sponse for the pulses of light.
• The laser displacement sensor head was connected with the
controller. The controller was powered by a 24V DC adapter. The
laser sensor head was placed at the backside of the target that was
the opposite side of the light source.
• The controller was connected to the computer by USB port to
collect measured data on the personal computer (PC).
• Transverse deflection of the cantilever beam was measured by the
laser displacement sensor along the length of the cantilever beam
for every 1 mm distance from the free end. All measured data was
recorded on the PC.
• Data was collected from the laser sensor at 0.1 sec intervals. An
average of 20 - 30 seconds of collected data was recorded for
analysis as shown in Figure 6.
332 Mechatronics and Robotics

• During the experiment, the photostrictive optical actuator

cantilever beam was exposed for 5 minutes to the high intensity
UV light then light was turned off and data was collected for 3
minutes.
• The average of three sets of data was plotted for the analysis.

RESULTS

Results of the Cantilever Beam for the Continuous Light

A photostrictive optical actuator was exposed under the uniform illumination
of UV light for a certain period of time. Transverse deflection of the PLZT
optical actuator cantilever beam is analyzed at various light intensities and
various focused light positions. Comparison of transverse deflection is also
presented.

Dependence of Transverse Deflection on Light Intensity at Spe-

cific Light Focused Position
Transverse deflections of the cantilever beam at 250 mW/cm2, 200 mW/
cm2, and 100 mW/cm2 light intensities were observed with light focused
at different locations on the beam. Data plotted along the length of the
cantilever beam are shown in Figures 7-9 respectively.

Figure 6. Controller output data collected in every 0.1 sec interval for 480 sec.
 Experimental Investigation of Photostrictive Materials for ... 333

Deflections were measured every 1 mm distance along the length of cantilever

beam. The negative deflection means the cantilever beam was deflected in the
opposite side of the light source. From the figures, it can be seen that the transverse
deflection near the fixed end of the beam was very small; it increased with the
increase of distance from the fixed end. The maximum transverse deflection was
observed at the free end of the beam. In addition, transverse deflection of the beam
depends on the light intensity and it increases with the increase of light intensity
at each specific focused location of the cantilever beam as shown in Figure 10.

Dependence of Transverse Deflection on Light Focused Position

at Specific Light Intensity
Transverse deflection of the beam also depends on the locations of light
focused on the beam. Deflection was maximized when the light was focused
near the fixed end of the beam and it was minimized when the light was
focused at the free end. Variation of transverse deflections due to light
focused position variations were observed for three specific light intensities
as shown in Figures 11-13. Deflections of the beam when the light was
focused at the middle and at the free end were almost the same at the 250
mW/cm2 as shown in Figure 11.

Figure 7. Comparison of transverse deflection for three different light inten-

sities when light was focused at the fixed end of the cantilever beam.

Figure 8. Comparison of transverse deflection for three different light intensi-

ties when light was focused at the middle of the cantilever beam.
334 Mechatronics and Robotics

At 200mW/cm2, when light was focused at the middle, deflection of the

beam was more than the deflection when the light was focused at the free
end but less than the deflection when the light was focused at fixed end as
shown in Figure 12.

Figure 9. Comparison of transverse deflection for three different light intensi-

ties when light was focused at the free end of the cantilever beam.

Figure 10. Transverse deflection at the free end of the beam with three differ-
ent light intensities for three different focused locations on the cantilever beam.
When the gap between the light source and the targeted area of the beam
increased the light beam diffused and allowed the cantilever beam to expose
more light to the actuator surface; this happened especially when the light
was focused at the middle position compare to the other two positions. At
100 mW/cm2, the deflections were the same when the focused position of the
light was at the middle and at the fixed end of the cantilever beam as shown
in Figure 13. More diffused light fell on the beam surface when the light
was focused at the middle of the beam compared to the other two positions.
In addition, the light focused at the fixed end position allowed the beam to
bend more near the fixed end, which ultimately resulted in these having the
same beam deflection.
 Experimental Investigation of Photostrictive Materials for ... 335

Figure 11. Variation of transverse deflection along the length of the cantilever
beam for three different light focus positions on the beam at 250 mW/cm2.

Figure 12. Variation of transverse deflection along the length of the cantilever
beam for three different light focus positions on the beam at 200 mW/cm2.

Figure 13. Variation of transverse deflection along the length of the cantilever
beam for three different light focus positions on the beam at 100 mW/cm2.
336 Mechatronics and Robotics

Time Response of the Cantilever Beam Deflection at the Free

End for Three Different Focused Positions
High intensity UV light was focused at different locations on the cantilever
beam. Deflection at the free end of the cantilever beam was observed to
measure the time needed to reach the steady state condition as shown in
Figures 14-16 respectively. Time response of the cantilever beam neither
had any major influence on the light intensity nor on the focused locations of
the actuator beam. Rising time was about 20 sec to reach 63% of the steady
state maximum deflection. The cantilever beam was approaching the steady
state after about 180 sec.

Hysteresis of the PLZT Cantilever Beam

The light was focused at the middle of the cantilever beam, and light was
turned on and off for three times to observe the hysteresis in the actuator
beam at a 200 mW/cm2 light intensity. When the light turn on and off
repeatedly, very small deflection on the cantilever beam was still observed
when the light was still turned off as shown in Figure 17.

Results for the Pulses of Light

Tests were performed by placing an optical chopper in between the PLZT
cantilever beam and the UV light source. The PLZT cantilever beam was
exposed to illumination of pulses of light intermittently by the optical
chopper. The two-slot blade optical chopper was run with a frequency of 0.5
Hz, which made the beam to be exposed to illuminations of light for 1 sec on
and 1 sec off. The experiment was conducted at light intensities of 250 mW/
cm2, 200 mW/cm2, and 100 mW/cm2. In these pulses of light case, light was
focused only at the middle of the cantilever beam.

Figure 14. Comparison of time response on transverse deflection for three dif-
ferent light intensities when light was focused at the fixed end.
 Experimental Investigation of Photostrictive Materials for ... 337

Figure 15. Comparison of time response on transverse deflection for three dif-
ferent light intensities when light was focused at the middle.

Comparison of Transverse Deflection with the Light Intensity

It has been observed that the magnitude of the corresponding transverse
deflection along the length of the cantilever beam decreases in the pulses of
light condition in comparison to the continuous light condition.

Figure 16. Comparison of time response on transverse deflection for three dif-
ferent light intensities when light was focused at the free end.

Figure 17. Hysteresis of the cantilever beam deflection at the free end.
338 Mechatronics and Robotics

Transverse deflection increases along the length of the actuator beam with
the increase of distance from the fixed end as shown in Figure 18. It is
also been observed that the deflection of the cantilever beam increases with
increase of the light intensity as shown in Figure 19.

Time Response of the Cantilever Beam Deflection at the Free

End for the Pulses of Light
The pulses of high intensity UV light were focused at the middle of the
cantilever beam; deflection at the free end of the actuator beam was
observed to measure the time needed to reach steady as shown in Figure 20.
Time response of the cantilever beam has no major influence on the light
intensities for the pulses of light condition.

Figure 18. Transverse deflection along the length of the cantilever beam at
three different light intensities for pulses of light condition.

Figure 19. Transverse deflection depends on the light intensity in pulses of light
condition.
Rising time was about 20 sec to reach 63% of the steady state maximum
deflection. The deflection of the cantilever beam was approaching a steady
state after about 180 sec.
 Experimental Investigation of Photostrictive Materials for ... 339

Comparison of Deflection Results for the Continuous and

Pulses of Lights
The transverse deflection along the length of the PLZT cantilever beam in
a continuous light condition was compared to that of the pulses of light
condition. Comparison was made between the continuous light and pulses
of light conditions at three different light intensities when light was focused
at the middle of the cantilever beam as shown in Figures 21-23. It has been
noticed that, at each light intensity, the magnitude of the free end deflection
of the beam for pulses of light was approximately 1/3rd of the free end
deflection of continuous light.

Figure 20. Comparison of time response on transverse deflection for three dif-
ferent light intensities in the pulses of light condition.

Figure 21. Comparison of transverse deflection between continuous and pulses

of light conditions at 250 mW/cm2 intensity.
340 Mechatronics and Robotics

Figure 22. Comparison of transverse deflection between continuous and pulses

of light conditions at 200 mW/cm2 intensity.

Figure 23. Comparison of transverse deflection between continuous and pulses

of light conditions at 100 mW/cm2 intensity.
In the pulses of light condition, the cantilever beam exposed to pulses
of high intensity light, could not reach the maximum photostrictive effect in
that amount of time that the continuous light condition could reach. That is
why actuator beam deflection was less in the pulses of light condition than
in continuous light condition.

CONCLUSIONS
From this experimental investigation, it has been observed that the continuous
lights and the pulses of lights have a significant influence on the transverse
deflection of PLZT optical actuator; less deflection was observed for the
pulses of light compare to the continuous light. The PLZT optical actuator
could be used in MEMS applications via a wireless control but the time
response of the actuator observed was slow.
 Experimental Investigation of Photostrictive Materials for ... 341

From the current study, analysis, and results of this research, the following
conclusions can be made:
1) Experimental transverse deflection of a cantilever beam increases
with the increase of light intensity for both continuous lights and
pulses of lights.
2) Maximum tip deflection at the free end of the cantilever beam for
pulses of light was about 1/3rd deflection of that for continuous
light.
3) The light intensity and focused locations of the cantilever beam
have no major influence on the time response. Rising time was
observed to be about 20 sec to reach 63% of the steady state the
maximum deflection for both continuous lights and pulses of light
conditions. Deflection of the cantilever beam was approaching a
steady state after about 180 sec for both continuous lights and
pulses of lights.
342 Mechatronics and Robotics

REFERENCES
1. Fridkin, V.M. (1979) Photo Ferroelectrics. Springer Series in Solid
State Sciences, Vol. 9, 85-113. https://doi.org/10.1007/978-3-642-
81351-1_6
2. Rahman, M. (2005) A Finite Element Approach to Model and Analyze
Photostrictive Optical Actuators. PhD Dissertation, University of
Alabama, Tuscaloosa.
3. Kawaguchi, K., ichiki, M., Morikawa, Y. and Nakada, T. (2002)
Electrical Properties of Photovoltaic PLZT and Its Application to
Optical Motor. Ferroelectric, 273, 41-46.
4. Tanimura, M. and Uchino, K. (1988) Effect of Impurity Doping on
Photostriction in Ferroelectric Ceramics. Sensors and Materials, 1, 47-
56.
5. Sada, T., Inoue, M. and Uchino, K. (1987) Photostriction in PLZT
Ceramics. Journal of the Ceramic Society of Japan, International
Edition, 95, 499-504.
6. Okuym, M. (1998) Micro Sensors and Micro Actuators using
Ferroelectric Thin Films. Proceedings of the International Symposium
on Micromechatronics and Human Science, Nagoya, 29-34.
7. Satol, K., Sugama, A., Akahoshi, T., Kondo, M., Aoki, S. and Kurihara,
K. (2007) Electro-Optic Effect of Lanthanum-Modified Lead Zirconate
Titanate and Application to Optical Device. 16th IEEE International
Symposium on Applications of Ferroelectrics, 827-828.
8. Huff, M.A. (2012) What Is Mems Technology? MEMS and
Nanotechnology Exchange. https://www.mems-exchange.org/MEMS/
what-is.html
9. Chynoweth, A.G. (1956) Surface Space-Charge Layers in Barium
Titanate. Physics Review, 102, 705-714. https://doi.org/10.1103/
PhysRev.102.705
10. Brody, P.S. (1975) High Voltage Photovoltaic Effect in Barium Titanate and
Lead Titanate-Lead Zirconate Ceramics. Solid State Chemistry, 12, 193-200.
11. Brody, P.S. (1983) Optomechnical Bimorph Actuator. Ferroelectrics,
50, 27-33. https://doi.org/10.1080/00150198308014428
12. Qin, M., Kui, Y. and Yung, C.L. (2007) Photo Induced Current in
(Pb0.97La0.03) (Zr0.52Ti0.48)O3 Thin Films of Different Thicknesses.
Integrated Ferroelectrics, 88, 58-67. https://doi.org/10.1080/10584580
601098647
 Experimental Investigation of Photostrictive Materials for ... 343

13. Poosanaas, P., Dogan, A., Thakoor, S. and Uchino, K. (1998)

Influence of Sample Thickness on the Performance of Photostrictive
Ceramics. Journal of Applied Physics, 84, 1508-1512. https://doi.
org/10.1063/1.368216
14. Yao, K., Gan, B.K., Chen, M. and Shannigrahi, S. (2005) Large Photo-
Induced Voltage in a Ferroelectric Thin Film with In-Plane Polarization.
Applied Physics Letters, 87, Article ID: 212906.
15. Tonokai, A., Takefumi, K., Daisuke, Y. and Koichi, S. (2010) PLZT Film
Actuators Deposited by a Hydrothermal Method. IEEE International
Conference on Robotics and Biomimetics, 1584-1589. https://doi.
org/10.1109/ROBIO.2010.5723566
16. Liang, L., Shaoping, W. and Feng, C. (2006) Characteristics of
Photo-Mechanical Device Using PLZT Wafer. Proceedings of the
2nd IEEE/ASME International Conference on Mechatronic and
Embedded Systems and Applications, 1-4. https://doi.org/10.1109/
MESA.2006.296973
17. Thakoor, S., John, M.M. and James, A.C. (1996) The Role of
Piezoceramic Microactuation for Advanced Mobility. IEEE
International Symposium on Application of Ferroelectrics, 1, 205-211.
18. Fukuda, T., Hattori, S., Arai, F., Matsuura, H., Hiramatsu, T., Ikeda,
Y. and Maekawa, A. (1993) Characteristics of Optical Actuator-
Servomechanisms using Bimorph Optical Piezoelectric Actuator.
Proceedings of 1993 IEEE Robotics and Automation Conference, 618-
623.
19. Ichiki, M., Furue, H., Kobayashi, T., Morikawa, Y., Nonaka, K.,
Maeda, R., Nakada, T. and Endo, C. (2005) Ultraviolet Sensor of Non-
Bias Type; Layered Film Structure of Photovoltaic Ferroelectrics. The
13th International Conference on Solid-State Sensors, Actuators and
Microsystems, 2, 2107-2110.
20. Mendinueta, J.M.D., Shinada, S., Furukawa, H. and Wada, N.
(2017) Ultra-High-Capacity Optical Packet Switching Networks
with Coherent Polarization Division Multiplexing QPSK/16QAM
Modulation Formats. Photonics, 4, 27.
21. Nashimoto, K., Tanaka, N., LaBuda, M., Ritums, D., Dawley, J., Raj,
M. and Kudzuma, D. (2005) High-Speed PLZT Optical Switches for
Burst and Packet Switching. IEEE 2nd International Conference on
Broadband Networks, 2, 1118-1123.
344 Mechatronics and Robotics

22. Nashimoto, K., Kudzuma, D. and Han, H. (2010) High-Speed Switching

and Filtering using PLZT Waveguide Devices. 15th Opto Electronics
and Communications Conference, 540-542.
23. Furukawa, H., Naoya, W., Hiroaki, H., Naganori, T., Keiichi, N. and
Tetsuya, M. (2008) 640 Gbit/S/Port Optical Packet Switch Prototype
with Optical Buffer using 1×8 PLZT Optical Switch and Parallel
Pipeline Buffer Manager. 34th European Conference and Exhibition
on Optical Communication, 1-2.
24. Uchino, K. (1989) Micro Walking Machine using Piezoelectric
Actuators. Robotics and Mechatronics, 124, 44-47.
25. Martin, S.J., Butler, M.A. and Lan, C.E. (1988) Ferroelectric Optical
Image Comparator using PLZT Thin Films. Electronics Letter, 24,
1486-1487. https://doi.org/10.1049/el:19881014
26. Thakoor, S., Poosanaas, P., Morookian, J.M., Yavrouian, A., Lowry,
L., Marzwell, N.I., Nelson, J.G., Neurgaonkar, R.R. and Uchino, K.
(1998) Optical Microactuation in Piezoceramics. Smart Structures
and Materials Smart Electronics and MEMS. 5th Annual International
Symposium on Smart Structures and Materials, San Diego, 20 July
1998.
27. Shih, H.R. and Tzou, H.S. (2000) Opto-Piezothermoelastic Constitutive
Modeling of a New 2-D Photostrictive Composite Plate Actuator.
ASME Control of Vibration and Noise: New Millennium, 61.
 CHAPTER

16
Theory and Simulation Analysis of the
Mode Shape and Normal Shape
Actuators and Sensors

Jwo Ming Jou

Department of Mechanical Engineering, Cheng Shiu University, Kaohsiung
City, Taiwan

ABSTRACT
In this paper, we will try to find a universal theoretical model and approximate
solutions which can be applied to both mode shape and normal shape actuators
and sensors, and which can be predicted the gain of the first three modes of
the mode shape and normal shape actuators and sensors, finally through
computer simulation analysis to validate. In order to prove the feasibility
of the theory and as well as convenient to use on the electro-mechanical

Citation:- Jou, J. (2014), “Theory and Simulation Analysis of the Mode Shape and
Normal Shape Actuators and Sensors”. Open Journal of Acoustics, 4, 184-203. doi:
10.4236/oja.2014.44019.
Copyright: © 2014 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
346 Mechatronics and Robotics

engineering, we will try to simplify the three-dimension structure problem

into an one-dimension structure problem. Furthermore we will design one
kind of bimorph type piezoelectric cantilever beam, so that it can be used
as with the actuator and sensor simultaneously, but also conducive to the
theory and simulation analysis. As for the simulation analysis, we will use
the ANSYS code.
Keywords: Mode Shape, Normal Shape, Actuator, Sensor, Gain, The
First Three Modes, ANSYS Code

INTRODUCTION
Due to the piezoelectric materials (such as PZT or PVDF) have the
advantages of fast response and high actuating force, so they are very
suitable made for sensors and actuators to sense and control the vibration
of the flexible structures, especially to sense or control the low frequency
vibration, because the high-frequency vibration is very easy to be absorbed
by the structure itself. So that the most piezoelectric sensors and actuators
are used to sense and control the low frequency vibration of flexible
structures. And since the design concept and theory of the modal sensors
and actuators have been proposed by C.K., Lee [1] [2] , making them to
sense and control the low frequency vibration of flexible structures are more
rapid, sensitive, precise, accurate and efficient. In which the main design
concepts and methods of the modal sensors and actuators are based on the
coating of the effective electrode surface and changing of the polarization
direction with specific modal or mode shape, shown as Figure 1. As for the
theoretical model, it is entirely based on the mode shape functions or spatial
distribution methods. It is successfully pulled out the transverse vibration
displacement of the coupling multi-modal by the mode shape functions
or spatial distribution methods. Next, it uses the coating methods or the
functions of effective electrode surface to sense or control the low frequency
vibration of the flexible structures. Since the relevant theoretical analysis,
computer simulation and experimental eleven were uncovered or presented
[3] - [9]. In addition, the theoretical model of modal or mode shape actuator
is also applied to the different types of piezoelectric transformers, thereby to
enhance their step-up ratio and conversion efficiency [10] [11] .
However, the theoretical model is only applicable to modal sensors and
actuators, but it can’t be applied on the normal shape sensors and actuators.
If the theoretical model of modal sensors and actuators used in the normal
shape sensors and actuators, we could not find the correct approximate
 Theory and Simulation Analysis of the Mode Shape and ... 347

solution. That is, so far, we still can’t find one kind of universal equation
to simultaneously describe or predict the behavioral of the mode shape and
normal shape sensors and actuators, or to compare the gain or to distinguish
the difference between the mode shape and normal shape sensors and
actuators from past studies.

Figure 1. (a) The side view of the polarization direction of the bimorph piezo-
electric cantilever beam; (b) The top view of the effective electrode surface of
mode shape 1; (c) The side view of the mode shape 1; (d) The top view of the ef-
fective electrode surface of mode shape 2; (e) The side view of the mode shape
2; (f) The top view of the effective electrode surface of mode shaspe 3; (g) The
side view of the mode shape 3.
In this paper, we will try from the original actuator and sensor equation,
and electromechanical boundary conditions to find one kind of universal
theoretical model and reasonable approximate solutions to compare the
gain and difference between the mode shape and normal shape sensors and
actuators. Furthermore we will also through the analysis results of computer
simulation to verify the correctness and feasibility of the universal theoretical
model and approximate solutions.

ACTUATOR EQUATION
In order to understand the differences between the mode shape and
normal shape actuators and sensors, we just start from the one-dimension
actuator equation of bimorph type piezoelectric cantilever beam. The
main contribution of this equation comes from the bending effect, so the
membrane effect can be completely ignored. Therefore, the equation can be
simplified to as follows [12]:
348 Mechatronics and Robotics

(1)
where the moment M1 can be divided into two terms of the mechanical and
electrical moment M 1 E as follow:

(2)
As for the constants in the above Equations (1) and (2) can be defined
as follow:

(3)

(4)

(5)
wherein above the symbols of c11, D11, h, w, t, x, z and are defined as the Young’s
modulus, bending stiffness constant per unit width, thickness, transverse displacement,
time, longitudinal coordinate, transverse coordinate and density.
Furthermore the electrical moment of Equation (2) can be divided into two
terms of time domain and space domain as follow:

(6)
And

(7)
wherein above the symbols of are defined as the
function of effective electrode surface of the nth mode shape, length, natural angular
velocity (natural frequency), mode shape eigenvalues, mode shape constant and mode
shape function.
Let Equations (2)-(7) substituted into Equation (1), and after finishing,
we can get a non-homogeneous partial differential equation of motion with
transverse displacement of mode shape actuator as follows:
 Theory and Simulation Analysis of the Mode Shape and ... 349

(8)
When the surface is uniformly coated and normal polarization with the
bimorph type piezoelectric cantilever beam, that is, the function of effective
electrode surface of Equation (7) can be redefined as:

(9)
Then, the second derivative of electrical moment becomes zero.

(10)
So for the normal shape actuator, Equation (8) can be simplified as:

(11)
In order for the mode shape and normal shape actuator equations can be
applied to structures of different lengths, we can try to make Equation (8)
and Equation (11) become the dimensionless equations:

(12)
And

(13)
where the dimensionless mode shape constant can be redefined as:

(14)
where the solution form of Equation (12) and Equation (13) can be
expressed as:

(15)
And
350 Mechatronics and Robotics

(16)
Furthermore let the homogeneous solution of Equation (16) and Equation
(17) are assumed to be as follows:

(17)
As for the particular solution of Equation (15) is assumed to be as
follows:

(18)
where the symbols of are defined as the undetermined coefficient and
dimensionless homogeneous transverse displacement.
We can take Equation (17) into Equation (12) and Equation (13), then
after finishing, we can get a dimensionless homogeneous four-order ordinary
differential equation, as:

(19)
where the dimensionless natural eigenvalue of mode shape is defined as:

(20)
And the i natural resonance frequency of piezoelectric beam
th

multilayer or mode shape or normal shape actuator can be derived by

Equation (20), as:

(21)
We can further get a dimensionless homogeneous solution from Equation
(19)

(22)
In addition, we can take Equation (18) into Equation (12), so get the
undetermined coefficient as:
 Theory and Simulation Analysis of the Mode Shape and ... 351

(23)
where the relationship of the fourth-order differential function and zero-
order function is:

(24)
So Equation (23) can be rewritten as:

(25)
Thus we can get the solution of the undetermined coefficient.

(26)
where in normal or natural state.
At this point, we can find the general solution of mode shape and normal
shape actuator, as:

(27)
And

(28)
Next, we can find the special solutions of the bimorph type piezoelectric
cantilever beam from the electro- mechanical boundary conditions as follow:

(29)
And

(30)
Or

(31)
352 Mechatronics and Robotics

And

(32)
where the dimensionless mode shape function of mode shape actuator
under clamped-free boundary conditions as:

(33)

And the dimensionless eigenvalues and parameters of mode shape can be

obtained from the following transcendental equation and formula:

(34)
And

(35)
According to the electro-mechanical and clamped-free boundary conditions, we
can determine the constants of Equation (27) and Equation (28), as follow:

(36)
And

(37)
where the constants of mode shape actuator are defined as

(38)
And

(39)
where the second-order and third-order derivative mode shape function on the free
boundary as follow:

(40)
 Theory and Simulation Analysis of the Mode Shape and ... 353

And

(41)
As for the constants of normal shape actuator were defined as

(42)
And

(43)
which let for the function of the effective electrode surface of normal
shape actuator.
Until now, we can get a dimensionless general solution of the mode
shape and normal shape actuator as follow:

(44)
And

(45)
Due to the high frequency vibrations can easily be absorbed by the
structure itself, so we are only interested in low-frequency vibration. And
in order to understand the difference between the mode shape and normal
shape actuators, we analyzed only for the first three resonant modes of the
structures. And according to Equation (34) and Equation (35), we can get
the eigenvalues and parameters of the first three modes, shown as Table 1.
As the same time, we can get the dimensionless mode shape functions of the
first three resonant modes by Equation (33) and Table 1, shown as Figure 2.
And we can also get the dimensionless function of the effective electrode
surface of first three modes when let Equation (33) take into Equation
(7), shown as Figure 3. In addition, we can get a dimensionless transverse
displacement of the mode shape and normal shape piezoelectric stator under
354 Mechatronics and Robotics

steady state and the same driving conditions as shown in Figure 2.

Figure 2. The dimensionless mode shape function of the first three modes of
mode shape actuator relative to the unit length of structure when i = n.

Figure 3. The dimensionless function of the effective electrode surface of the

first three modes of mode shape actuator relative to the unit length of structure
when i = n.

Table 1. The eigenvalues and parameters of the first three modes of mode shape
actuator under clamped-free boundary conditions

Another according to Equation (40) and Equation (41), we can get the second and
third derivative of mode shape function of the first three modes of mode shape actuator
relative to the unit length of structure, shown as Figure 4 & Figure 5. And we can further
that of the first three modes of mode shape actuator at free end.
Therefore in, we can let Equation (44) further simplified as follows:
 Theory and Simulation Analysis of the Mode Shape and ... 355

(46)
So far, Equation (46) of mode shape actuator appeared to be consistent
with the particular solution of previously papers [1] [10] [11] .
Furthermore, we can make Equation (45) of normal shape actuator
simplifies to

(47)
where the dimensionless mode shape function of normal shape actuator
is defined as

(48)
And the dimensionless parameters of normal shape actuator is defined as

(49)
According to Equation (46) and Equation (47), we can get the ratio of
dimensionless transverse displacement

Figure 4. The second derivative of mode shape function of the first three modes
of mode shape actuator relative to the unit length of structure when i = n.
356 Mechatronics and Robotics

Figure 5. The third derivative of mode shape function of the first three modes of
mode shape actuator relative to the unit length of structure when i = n.
of mode shape and normal shape actuator under conditions of steady
state, constant driving voltage and the same bending stiffness constant per
unit width as follows:

(50)
In order to facilitate understanding of the difference between mode shape
and normal shape actuator, we can set the above ratio at the free end, that is,
, shown as Figure 2. There fore Equation (50) can
again be expressed as

(51)

SENSOR EQUATION
For the bimorph type piezoelectric cantilever beam under condition of
constant electric potential, the sensor or current equation per unit length and
width of the mode shape and normal shape sensor can be expressed as [12] :

(52)
And we can further take Equation (46) and Equation (47) into Equation
(52) respectively, let Equation (52) be divided into two types of sensor or
current equations as follow:
 Theory and Simulation Analysis of the Mode Shape and ... 357

(53)
And

(54)
where the first derivative of dimensionless mode shape function of the
first three modes can be obtained from clamped-free boundary conditions,
or known from Figure 6. Therefore the ratio of dimensionless current of
the mode shape and normal shape sensor dependent of eigenvalues under
conditions of clamped-free, steady state, the same driving voltage and
bending stiffness constant per unit width as follows:

(55)
where the first derivative of dimensionless mode shape function of the
mode shape and normal shape sensor can be expressed as follow:

Figure 6. The first derivative of dimensionless mode shape function of the first
three modes relative to the unit length of structure when i = n.
358 Mechatronics and Robotics

(56)
And

(57)
Since the current is proportional to voltage under condition of the same
load RL, So Equation (55) can also be expressed the ratio of dimensionless
voltage of the mode shape and normal shape sensor as follows:

(58)

CASE STUDY: THEORY AND SIMULATION ANALY-

SIS
In order to understand the differences between mode shape and normal
shape actuators and sensors by theory and simulation analysis, we specially
design a series of one-dimension bimorph type piezoelectric cantilever
beams, including the mode shape and normal shape actuators and sensors,
shown as Figure 7.
Wherein the size and physical properties of the one-dimension bimorph
type piezoelectric cantilever beam can refer to Table 2.
In addition, we will use the ANSYS code to simulate analysis the mode
shape and normal shape actuators and sensors.
Where we have chosen the element type is the scalar tetragonal 98 of
couple field.
Furthermore we will do the most sophisticated cutting of mesh, so that
the analysis of the results can be quickly converged.
As for the analysis types, we were selected the modal, static and steady
state.
Wherein the step of theoretical analysis is as follows:
(1-1) Using different frequency spacing
to analyze the dimensionless
transverse displacement and current or voltage of the first three
modes of the mode shape and normal shape actuators and sensors
from Equations (46) and (47) and Equations (52) and (53).
 Theory and Simulation Analysis of the Mode Shape and ... 359

(1-2) Selecting the minimum frequency spacing to analyze

the ratio of the dimensionless transverse displacement and current
or voltage and the mode shape function of the first three modes
of the mode shape and normal shape actuators and sensors from
Equation (50) and Equation (58).
As for the step of simulation analysis is as follows:
(2-1) Modeling of the mode shape and normal shape actuators and
sensors respectively, including select element type, enter the
physical properties, as well as coordinate system conversion, as
shown Figure 8.
(2-2) Meshing of the mode shape and normal shape actuators and
sensors respectively, including select the most sophisticated
cutting of mesh or select the smart size 1, as shown Figure 9.
(2-3) Solving of the mode shape and normal shape actuators and
sensors respectively, including setting boundary conditions of
electro-mechanical, as shown Figure 10. All of which driving
voltage on the effective electrode surface is 1.0 V.
(2-4) Post-processing of the mode shape and normal shape actuators
and sensors respectively, includes processing the first three
modes, the maximum deformation or electric potential, as shown
Figures 11-13.

Figure 7. The schematic diagram of the mode shape and normal shape actuators
and sensors.
360 Mechatronics and Robotics

Table 2. The size and physical properties of one-dimension bimorph type piezo-
electric cantilever beam.

RESULTS AND DISCUSSION

According to the results of theory analysis, we found:
1) Under condition of the frequency spacing of 100 Hz, the maximum
dimensionless transverse displacement of the second modal of the
normal shape and mode shape actuators is smaller than the first
and the third modals’, as shown in Table 3 and Figure 14-(a-3). In
addition, the maximum dimensionless voltage of the normal shape
and mode shape sensors is proportional to the modal, as shown
in Table 3 and Figure 14-(b-1).
2) Under condition of the frequency spacing of 10 Hz, the maximum
dimensionless transverse displacement of the normal shape
actuators is inversely proportional to the modal, as shown in Table
4 and Figure 14-(a-2). Furthermore, the maximum dimensionless
transverse displacement of the second modal of the mode shape
actuator is smaller than the first and the third modals’, as shown
in Table 4 and Figure 14-(a-2). As for the normal shape and mode
shape sensors, the maximum dimensionless voltage is proportional
to the modal un- der condition of the frequency spacing of 10 Hz, as
shown in Table 4 and Figure 14-(b-2).

Figure 8. Modeling and setting physical properties of the mode shape and nor-
mal shape actuators and sensors. (a) The normal shape actuator and sensor; (b)
The mode shape 1 actuator and sensor; (c) The mode shape 2 actuator and sen-
sor; (d) The mode shape 3 actuator and sensor.
 Theory and Simulation Analysis of the Mode Shape and ... 361

Table 3. The maximum dimensionless transverse displacement and voltage of

the first three modes of the mode shape and normal shape actuators and sensors
under condition of the frequency spacing of 100 Hz.

Mode Actuators Sensors

NSA MS1A MS2A MS3A NSS MS1S MS2S MS3S
(m/m) (m/m) (m/m) (m/m) (V/V) (V/V) (V/V) (V/V)
1 2.20 2.39 0.09 0.03 8.62 6.17 2.11 2.01
2 1.79 0.01 2.14 0.04 40.25 0.04 48.08 2.31
3 6.86 0.00 0.01 13.30 422.45 0.00 0.30 820.33

Table 4. The maximum dimensionless transverse displacement and voltage of

the first three modes of the mode shape and normal shape actuators and sensors
under condition of the frequency spacing of 10 Hz.

Mode Actuators Sensors

N S A M S 1 A MS2A M S 3 A N S S M S 1 S M S 2 S M S 3 S
(m/m) (m/m) (m/m) (m/m) (V/V) (V/V) (V/V) (V/V)
1 217.08 188.79 0.09 0.03 487.88 764.39 2.09 2.01
2 33.59 0.01 37.77 0.04 739.25 0.04 848.82 2.29
3 21.85 0.00 0.01 44.32 1348.82 0.00 0.30 2374.72

Figure 9. Meshing the mode shape and normal shape actuators and sensors. (a)
The normal shape actuator and sensor; (b) The mode shape 1 actuator and sen-
sor; (c) The mode shape 2 actuator and sensor; (d) The mode shape 3 actuator
and sensor.
362 Mechatronics and Robotics

3) Under condition of the frequency spacing of 1 Hz, the maximum

dimensionless transverse displacement of the second modal of the
normal shape and mode shape actuators is larger than the first and
the third modals’, as shown in Table 5 and Figure 14-(a-1). As for
the maximum dimensionless voltage of the normal shape and mode
shape is proportional to the modal under condition of the frequency
spacing of 1 Hz, as shown in Table 5 and Figure 14-(b-1).
4) Overall, in addition to the modal 1 or the first modal and under
condition of the frequency spacing of 100 Hz, the gain or ratio of the
second and third modal of the mode shape actuators and sensors are
better than normal shape actuators and sensors, as shown in Table 6.
5) In terms of the resonance frequency, the approximate solutions
through theoretical derivation are consistent with the analysis
results by computer simulations, as shown in Table 7.
6) Under conditions of the simulation analysis of static, steady state
and modal types, the first modal of the normal shape and mode
shape actuators and sensors are the same gain or ratio, as shown
in Table 8, Table 9 and Figure 15.
7) Furthermore, the gain or ratio of the second and third modal of the
mode shape actuators and sensors are better than the normal shape
actuators and sensors, as shown in Table 8, Table 9 and Figure 15.

CONCLUSION
According to the results of the theory and simulation analysis, on the whole,
we found the first modal of the normal shape and mode shape actuators and
sensors are the same gain.

Figure 10. Solving Process_ Setting boundary conditions of electro-mechanical the

mode shape and normal shape actuators and sensors. (a) The normal shape actuator and
 Theory and Simulation Analysis of the Mode Shape and ... 363

sensor; (b) The mode shape 1 actuator and sensor; (c) The mode shape 2 actuator and
sensor; (d) The mode shape 3 actuator and sensor.

Table 5. The maximum dimensionless transverse displacement and voltage of

the first three modes of the mode shape and normal shape actuators and sensors
under condition of the minimum frequency spacing (Δfi = 1 Hz).

Mode Actuators Sensors

NSA MS1A MS2A MS3A NSS MS1S MS2S MS3S
(m/m) (m/m) (m/m) (m/m) (V/V) (V/V) (V/V) (V/V)
1 256.00 222.94 0.09 0.03 898.94 576.12 2.09 2.01
2 301.21 0.01 350.85 0.04 6637.87 0.04 7885.37 2.29
3 138.78 0.00 0.01 333.99 8562.97 0.00 0.30 20606.27

Table 6. The gain or ratio of dimensionless transverse displacement and voltage of the
mode shape and normal shape actuators and sensors relative to the frequency under
condition of different frequency spacing (Δfi = 1 Hz, 10 Hz & 100 Hz).
F r e q u e n c y Actuators Sensors
Spacing MS1A. MS2A. M S 3 A . MS1S. MS2S. MS3S.
N S 1 A NS2A N S 3 A N S 1 S NS2S NS3S
(m/m) (m/m) (m/m) (V/V) (V/V) (V/V)
1 Hz 0.87 1.16 2.40 0.64 1.19 2.41
10 Hz 0.87 1.12 2.03 0.64 1.15 2.03
100 Hz 1.08 1.20 1.94 0.72 1.20 1.94
364 Mechatronics and Robotics

Figure 11. Post-processing the first three modes and the maximum deformation
of the mode shape and normal shape actuators and sensors. (a) The first three
modes of the normal shape actuator and sensor; (b) The first three modes of the
mode shape 1 actuator and sensor; (c) The first three modes of the mode shape 2
actuator and sensor; (d) The first three modes of the mode shape 3 actuator and
sensor (analysis type: modal).
Or in other words, the design concept of the first modal of mode shape
actuator or sensor is not necessarily better than of the first modal of normal
shape actuator or sensor, or even worse. However, the gain of the second
and third modal of the mode shape actuators and sensors are better than the
normal shape actuators and sensors under any operating states.

Figure 12. Post-processing the first three modes and the maximum electric po-
tential of the mode shape and normal shape actuators and sensors. (a) The first
 Theory and Simulation Analysis of the Mode Shape and ... 365

three modes of the normal shape actuator and sensor; (b) The first three modes
of the mode shape 1 actuator and sensor; (c) The first three modes of the mode
shape 2 actuator and sensor; (d) The first three modes of the mode shape 3 ac-
tuator and sensor (analysis type: modal).
Most importantly, we have to find a universal theoretical model and
approximate solutions in this paper, which can predict the gain of the first
three modes of normal shape and mode shape actuators and sensors, and
through the analysis results of computer simulation to confirm.

Figure 13. Post-processing the electric potential output state of the mode shape
and normal shape actuators and sensors. (a) The normal shape actuator and sen-
sor; (b) The mode shape 1 actuator and sensor; (c) The mode shape 2 actuator
and sensor; (d) The mode shape 3 actuator and sensor (analysis type: steady
state).

Table 7. The resonance frequency and the ratio of resonance frequency of the
first three modes of the mode shape and normal shape actuators and sensors by
theory and simulation analysis (analysis type: modal).

Mode The- N S M S 1 MS2 M S 3 N S MS1 MS2 MS3

o r y (Hz) (Hz) (Hz) (Hz) Theo- Theo- Theo- Theo-
(Hz) ry ry ry ry
1 359 359 359 360 360 1.000 1.000 1.003 1.003
366 Mechatronics and Robotics

2 2253 2245 2247 2247 2247 1.001 0.999 0.999 0.999

3 6308 6256 6267 6262 6264 0.992 0.994 0.993 0.993

Table 8. The maximum deformation and the ratio of maximum deformation of the
mode shape and normal shape actuators by simulation analysis (analysis type: static &
steady state).

Anal- NSA (m) MS1A (m) MS2A (m) MS3A (m) MS1A. MS2A. MS3A.
ysis NSA NSA NSA
Type (m/m) (m/m) (m/m)
Static 5.18E−14 −1.52E−13 −9.57E−11 −1.74E−10 2.93 1846.88 3353.80
&
Steady
State

Figure 14. (a-1)-(a-3) The dimensionless transverse displacement of

the mode shape and normal shape actuators relative to the frequency under
condition of different frequency spacing; (b-1)-(b-3) The dimensionless current
or voltage of the mode shape and normal shape sensors relative to the
frequency under condition of different frequency spacing.
 Theory and Simulation Analysis of the Mode Shape and ... 367

Table 9. The maximum sensing or output voltage and the ratio of maximum
sensing or output voltage of the first three modes of the mode shape and normal
shape sensors by simulation analysis (analysis type: modal).

Mode NSS M S 1 S M S 2 S M S 3 S MS1S:NSS MS2S:NSS MS3S:NSS

(V) (V) (V) (V) (V/V) (V/V) (V/V)
1 9 45 643 686 4.82 68.70 73.37

2 330 1506 25131 26925 4.56 76.13 81.56

3 2325 11946 194692 209525 5.14 83.75 90.13

Figure 15. (a-1)-(a-3) The maximum dimensionless transverse displacement

of the first three modes of the mode shape and normal shape actuators
relative to the dimensionless length under condition of different frequency
spacing; (b-1)-(b-3) The maximum dimensionless current or voltage of
the first three modes of the mode shape and normal shape sensors relative to the
dimensionless length under condition of different frequency spacing.

ACKNOWLEDGEMENTS
Jwo Ming Jou (Project No.: MOST103-2221-E-230-007).
368 Mechatronics and Robotics

REFERENCES
1. Lee, C.-K. (1987) Piezoelectric Laminates for Torsional and Bending
Modal Control: Theory and Experiment. Ph.D. Dissertation, Cornell
University, Ithaca.  
2. Chiang, W.-W. and Lee, C.-K. (1989) Critical Active Damping Control
of a Flexible Slender Plate Using a Distributed Modal Actuator and
Sensor. American Control Conference, 700-706.   
3. Vaz, A.F. and Farr, T.J. (1997) Experimental Validation of Analytical
Models of Flexible Structures with Bonded Piezoelectric Film Actuators
and Sensors. IEEE Instrumentation and Measurement Technology
Conference, 2, 887-892.   
4. Vat, A.F. (1998) Composite Modeling of Flexible Structures with
Bonded Piezoelectric Film Actuators and Sensors. IEEE Transactions
on Instrumentation and Measurement, 47, 513-520. http://dx.doi.
org/10.1109/19.744200
5. Takagi, K., Nagase, K., Oshima, K. and Hayakawa, Y. (2001) Modal
Filtering for Simply Supported Plate Using Equally Segmented
Piezoelectric Film. American Control Conference, 2, 1593-1598.
6. Ballato, A. and Kim, Y.K. (2002) Displacements and Rotations of
Practical Vibrational Modes of Piezoelectric Bimorph Cantilever
Beams. Proceedings of IEEE Sensors, 2, 1294-1297.
7. Gou, X.K. and Tian, H.M. (2007) Active Vibration Control of a
Cantilever Beam Using Bonded Piezoelectric Sensors and Actuators.
8th International Conference on Electronic Measurement and
Instruments, 4, 85-88.
8. Rios-Gutierrez, M. and Silva-Navarro, G. (2009) Finite Element
and Modal Modeling of a Cantilever Beam with Piezoelectric Patch
Actuator for Vibration Absorption. 6th International Conference on
Electrical Engineering Computing Science and Automatic Control,
1-6.
9. Sanchez-Rojas, J.L., Hernando, J., Donoso, A., Bellido, J.C., Gutierrez,
L.M., Ababneh, A., Seidel, H. and Schmid, U. (2009) Piezoelectric
Modal Sensors/Actuators Based on Microplates Applying Surface
Electrode Patterning. Solid-State Sensors, Actuators and Microsystems
Conference, 2346-2349.   
10. Hsu, Y.-H., Lee, C.-K. and Hsiao, W.-H. (2005) Electrical and
Mechanical Fully Coupled Theory and Experimental Verification
 Theory and Simulation Analysis of the Mode Shape and ... 369

of Rosen-Type Piezoelectric Transformers. IEEE Ultrasonics,

Ferroelectrics and Frequency Control, 52, 1829-1839. http://dx.doi.
org/10.1109/TUFFC.2005.1561639   
11. Huang, Y.-T., Wu, W.-J., Wang, Y.-C. and Lee, C.-K. (2007) Multilayer
Modal Actuator-Based Piezoelectric Transformers. IEEE Ultrasonics
Ferroelectrics and Frequency Control, 52, 359-367. http://dx.doi.
org/10.1109/TUFFC.2007.249  
12. Jou, J.M. (2003) Piezoelectricity Mechanics. Chuan Hwa Book Co.
Ltd., Taipei.   
 CHAPTER

17
Experimental Study of the Response
of Transonic Diffuser Flow to a
Piezoceramic Actuator at
Diffuser Throat

Minoru Yaga1, Yusuke Uechi2, Hiroaki Ozono3, Masaaki Ishikawa1,

Isao Teruya1
1
Department of Mechanical Systems, Faculty of Engineering, University of
the Ryukyus, Okinawa, Japan
2
Mie Metal Industry Co. Ltd., Mie, Japan
Graduate School of Engineering and Science, University of the Ryukyus,
3

Okinawa, Japan

ABSTRACT
An experimental study of the response of a piezoceramic actuator set at the
throat to a transonic diffuser is carried out by measuring wall static pressure
fluctuations and by visualizing the flow field using schlieren technique. The

Citation:- M. Yaga, Y. Uechi, H. Ozono, M. Ishikawa and I. Teruya, “Experimental

Study of the Response of Transonic Diffuser Flow to a Piezoceramic Actuator at Dif-
fuser Throat,” Open Journal of Fluid Dynamics, Vol. 3 No. 2A, 2013, pp. 14-21. doi:
10.4236/ojfd.2013.32A003.
Copyright: © 2013 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
372 Mechatronics and Robotics

visualized flow fields are captured with a digital still camera and a digital
high speed video camera. The piezo ceramic actuator is attached at the
throat of the diffuser and driven by sinusoidal amplified voltage signals. The
diffuser used in this experiment is circular arc half nozzle with the height
h* and width w of 3 mm and 25 mm, respectively. The blockage factor of
the piezoceramic actuator to the diffuser throat is 9.2% assuring the effect
of change in the throat area rather than the boundary layer disturbances. The
piezoceramic actuator is driven at the frequency of 100 Hz, 200 Hz, and 300
Hz and its amplitude is about 1 mm. It is found that the wall static pressure
fluctuations and the behavior of the shock wave clearly correspond to the
vibration of the piezo ceramic actuator for all the frequency ranges whereas
the averaged shock position remains almost unchanged. All the results
mentioned above suggest that driving the piezo ceramic actuator at the
diffuser throat can be one of the promising techniques to control unsteady
transonic diffuser flow.
Keywords: Compressible Flow; Shock Wave; Transonic Diffuser;
Piezoceramic Actuator; Throat

INTRODUCTION
The unsteady flow field in a transonic diffuser has attracted a great deal
of interest not only because of the practical industrial importance but
also because of the complexity of the flow itself. It is well known that a
shock wave in a transonic flow is basically unsteady due to the interaction
between a shock wave and other flow phenomena, i.e., the shock wave
oscillation is triggered and maintained by a local interaction between the
shock foot and the boundary later developed along the wall surface or
upstream and downstream propagating disturbances. Meier [1] reported that
shock-induced separation causes the large-amplitude, unstable oscillation
of a shock wave. In addition to the shock-induced separation, the effects
of propagating disturbances toward the shock waves on the oscillations
have been analytically, numerically and experimentally investigated [2,3],
including forced oscillations by some functions or rotating rods as a forcing
function to drive the shock waves. According to previous studies, several
factors affect unsteady shock behavior in transonic diffusers. Therefore, in
addition to revealing the causes of the shock oscillations, there have also
been attempts to eliminate or reduce these unfavorable unsteady shock
oscillations both for internal and external devices. It is quite difficult to deal
with these types of oscillations unless the disturbances around the shock wave
 Experimental Study of the Response of Transonic Diffuser Flow ... 373

are attenuated or canceled by some techniques because the longitudinal and

boundary layer associated disturbances cannot be separated.The approach to
stabilize the unsteady flow fields is divided into two major controls, which
are active control with jets [4,5] and passive control with a porous cavity [6]
or vortex generators. These controls are expected to generate the opposite
phase of signals or to retard separations of boundary layers. However, the
causes of the oscillations include pressure disturbances in a core flow and
unsteady phenomena associated with the shock boundary layer interaction,
as mentioned above, the future trend for the oscillation controls might be a
combination of passive and active controls.
Although the effect of the pressure disturbances in a core flow and
disturbances associated with the boundary layer should be considered
individually in order to clarify the factors affecting the oscillations, both
effects cannot be divided because they interact with each other. Thus, in
the present paper, we focus primarily on the mechanical disturbances by
changing cross sectional area at the throat of a transonic diffuser, which is
expected to affect the location of shock waves according to the isentropic
flow relations between the throat area and local cross sectional area at a
location of the shock wave.

EXPERIMENTAL APPARATUS AND PROCEDURE

Wind Tunnel Facility

Figures 1(a)-(c) show the wind tunnel, the detail of the test section, and the
driving circuit of the piezoceramic actuator, respectively. The experimental
apparatus consists of a 0.7 MPa compressor, a settling chamber, a regulator
valve, and a circular arc transonic half diffuser in a blow down wind tunnel
discharging to the atmosphere. The throat height of the diffuser and its radius
are h* = 3 mm and R = 500 mm, respectively. The span of the diffuser is 33
mm. The flow field in the diffuser is visualized by the schlieren technique
and captured with a digital still camera or a digital high-speed video camera,
which enables images of the flow field to be captured at a rate of 41,000
frames per second. Unsteady wall static pressure fluctuations are measured
using semiconductor pressure sensors at the settling chamber, at the throat,
and at positions x = 20, 30, 40, and 50 mm downstream of the throat. The
signals from the pressure transducer are digitized at a sampling frequency of
10 kHz with 16-bit accuracy and are analyzed to evaluate the effect of the
piezoceramic actuator on the diffuser flow fields. The resonance frequency
374 Mechatronics and Robotics

of the transducers is 50 kHz assuring the accuracy of monitoring the pressure

fluctuations. The total pressure at the settling chamber is monitored during
the experiments by a computer, which controls the timing of the pressure
measurements and flow visualizations.

Piezoceramic Actuator
The piezoceramic actuator is set at the diffuser throat, as shown in Figure
1(b). In order to achieve sufficient displacement by the actuator, a bimorph
type piezoceramic actuator is adopted and driven by the circuit as shown
Figure 1(c).

Figure 1. Experimental apparatus: (a) Wind tunnel and measurement system;

(b) Detail of test section; (c) Driving circuit for piezoceramic actuator.
The length, width, and thickness of the piezoceramic actuator are 33 mm,
11 mm, and 0.8 mm, respectively. During the experiment, the piezoceramic
actuator is connected to a DC type amplified voltage source and then bent in
order to change the cross sectional area at the diffuser throat. The maximum
input voltage from the signal source, displacement, and natural resonance
frequency of the piezoceramic actuator are 60 Vp, 0.7 mm, and 400 Hz,
respectively. Accordingly, the exciting input frequency to the piezoceramic
actuator is limited to 300 Hz. From preliminary experiments, the displacement
of the piezoceramic actuator is found to be sufficiently large compared to
that with a layer type piezoceramic actuator. In this report, only sinusoidal
signals are input into the DC amplifier to achieve simplicity of the frequency
analyses.
 Experimental Study of the Response of Transonic Diffuser Flow ... 375

EXPERIMENTAL RESULTS AND DISCUSSION

Results of Flow Visualization

Figures 2(a)-(d) show typical schlieren images of the flow fields and shock
positions for the input frequency of fp = 200 Hz. It is clear that some quite
weak Mach waves are generated near the throat due to the effect of the
moving piezoceramic actuator. At the same time, at a pressure ratio of p0/
pb = 1.20, the starting shock wave appears to be very weak near x/h* =
6.0, which is followed by some weak shock waves due to the interaction
between the shock wave and the boundary layers along the diffuser walls.
Although the flow field downstream of the normal shock wave is supposed
to be subsonic, Figure 2(a) and subsequent figures illustrate the occurrence
of the shock wave downstream of the first normal shock wave.
The second shock wave or multiple shock waves are generated by the
acceleration of the main flow due to the change in the effective cross sectional
area of the diffuser and the disturbances approaching from downstream of
the shock wave.
As reported in several previous studies, the shock wave near the throat is
unstable, so that the shock wave observed in Figure 2(a) change its position,
as will be discussed later herein. Figure 2(b) shows that the slight increase
in the pressure ratio causes the downward displacement of the shock wave.
Figure 2(c) illustrates the lambda-foot type shock wave on the upper and
lower walls due to a typical moderate interaction between the shock wave
and the boundary layer. Other shock waves downstream of the first shock
position also correspond to the varying throat height.
Then, the distance between the throat and the averaged shock positions
increases with the pressure ratio. Figures 2(c) and (d) show that the clear
shock waves cause boundary layer separation followed by multiple shock
waves downstream of the interaction region.

Shock Locations for Various Frequencies

According to Figure 2, the shock waves appear to move gradually in the
previous section. Then, it is necessary to check the relation between rough
locations and shock waves for various input frequencies of the piezoceramic
actuator. Then, their positions are plotted with respect to the pressure ratio
and are shown in Figure 3 as a parameter of the input frequencies of 0 Hz,
100 Hz, 200 Hz, and 300 Hz.
376 Mechatronics and Robotics

Figure 2. Schlieren images for frequency 200 Hz: (a) p0/pb = 1.2; (b) p0/pb = 1.4;
(c) p0/pb = 1.6; (d) p0/pb = 1.8.

Figure 3. Relation between shock wave position and pressure ratio.

Figure 3 shows the relation between the positions of shock waves xs
divided by throat height h* and the pressure ratio p0/pb the shock locations
are measured in the schlieren images for various input frequencies of
0 Hz through 300 Hz. The shock wave is found to move monotonically
 Experimental Study of the Response of Transonic Diffuser Flow ... 377

downstream with the increase in the pressure ratio. However, the gradient
of the shock displacement to the pressure ratio decreases with the pressure
ratio because the shock-induced separation causes displacement delay.
Considering the effect of the piezoceramic actuator, it is easily expected that
the location of shock waves corresponds to the throat height because the
Mach number just upstream of the shock wave depends on the ratio of the
local cross sectional area to the throat area.
From the viewpoint of small-scale observation, the shock wave oscillates
around its averaged position at the fixed frequency of the piezoceramic
actuator, which will be discussed in the following section based on images
captured by a high-speed video camera. The behavior of the shock wave
is critical to the generation of large pressure fluctuations under the various
pressure ratios.

Wall Static Pressure Fluctuation

Figures 4(a)-(d) show the time series pressure fluctuations at the throat
x/h* = 0.0 and at x/h* = 10.0 for a frequency of 200 Hz. The wall static
pressure pw divided by the back pressure pb at the throat exhibits a clear
sinusoidal time history, which is considered to correspond to the behavior of
the piezoceramic actuator.
On the other hand, the pressure fluctuation at x/h* = 10.0 indicates a
severe large amplitude fluctuation. The large amplitude fluctuation suggests
that the shock wave passes back and forth over the measurement position
corresponding to the motion of the piezoceramic actuator. The increase in
amplitude can be explained by Figure 3, which shows that the shock wave
is located around x/h* = 10.0. Then, in case of a pressure ratio of 1.50,
the shock wave is expected to stand slightly downstream of the pressure
measurement position at x/h* = 10.0.
The influence of the shock wave could reach the pressure measurement
position through the subsonic layer in the boundary layer which develops
along the diffuser wall.
The pressure remains relatively low compared to that shown in Figure
4(a), which indicates that the shock wave stands further downstream than
that for p0/pb = 1.45. However, the effect of the piezoceramic actuator can be
still detected. Finally, in the case of p0/pb = 1.55, the shock wave moves far
downstream of the pressure measurement position.
378 Mechatronics and Robotics

Figure 4. Wall static pressure fluctuations at throat and x/h* = 10.0 for input
frequency 200Hz: (a) p0/pb = 1.45; (b) p0/pb= 1.50; (c) p0/pb = 1.55; (d) p0/pb=
1.55.
This is also confirmed in Figure 3, which showing xs/h* = 10.0 at a
pressure ratio of 1.55. Both the fluctuating and averaged pressure are much
lower due to the downstream displacement of the pressure measurement
position. These figures contain information about the shock position and the
shock wave behavior. In other words, Figure 4(a) shows the severe pressure
fluctuation at x/h* = 10.0, whereas the fluctuation at the throat does not. This
means that the shock wave stands between the throat and the measurement
position at x/h* = 10.0. As the pressure ratio increases, the fluctuation at x/h*
= 10.0 gradually decreases the pressure fluctuation, achieving the smallest
pressure fluctuation, as shown in Figure 4(c). The pressure fluctuation as
shown Figure 4(c) denotes the supersonic state at x/h* = 10.0.
However, the sudden change in the average pressure was observed as
shown in Figure 4(d) under the same condition, which is caused by the flow
separation downstream of the shock wave. The increase in the pressure
fluctuation indicates the shock wave displacement upstream due to the flow
separation. However, sometime after the increase in pressure, the average
value and fluctuation decrease, denoting the shock wave located downstream
of the position at x/h* = 10.0. The pressure fluctuations at the throat are
found to oscillate sinusoidally for all pressure ratios. Figure 4(d) also reveals
that the wall pressure fluctuation at a pressure ratio of 1.55 suddenly changes
its level at time t = 0.03 s, which suggests that the flow has just recovered
from an unexpected separation downstream of the shock wave.
 Experimental Study of the Response of Transonic Diffuser Flow ... 379

Root Mean Square of the Pressure Fluctuation

It is well known that wall static pressure fluctuations greatly decrease when
a flow becomes supersonic.
In other words, monitoring the wall static pressure fluctuations clarifies
whether the flow is supersonic or subsonic.
In the previous section, the pressure fluctuations also indicate the rough
shock wave positions.
Then, for a more quantitative evaluation of the pressure fluctuations, the
root mean square of pressure fluctuations is one of the indices of the shock
wave position and its behavior.
Figure 5(a) shows that the root mean square prms/pb for the input
frequency of fp = 0 Hz at the throat shows almost no increase in prms/
pb with the increase in the pressure ratio p0/pb. On the other hand,
activation occurs at the throat, the rms increases with the pressure ratio.
This indicates that the piezoceramic actuator always affects the
pressure fluctuation at the throat. When the rms reaches approximately
zero, the state of the measurement position is considered to be supersonic.
This indicates that the hock wave is located downstream of the measurement
position.
380 Mechatronics and Robotics

Figure 5. RMS of wall static pressure fluctuations: (a) x/h* = 0.0; (b) x/h* =
6.67; (c) x/h* = 10.0; (d) x/h* = 13.33.
In Figure 5(b), the clear peaks at a pressure ratio p0/pb of approximately
1.2 show that a single and relatively weak shock wave appears and moves
downstream with the increase in the pressure ratio. The sudden drop in the
rms indicates the completion of the process from the subsonic-to-supersonic
transition due to the downstream displacement of the shock wave. This
process is independent of the input frequencies of the piezoceramic actuator.
Figure 5(c) shows that the rms at a pressure ratio of approximately
1.7 sudden creases, which is caused by the unexpected separation of a
boundary layer, as mentioned in the previous section. This also indicate the
shock wave existence around the position of x/h* = 10.0. However, for a
pressure ratio greater than 1.8, the rms for all input frequencies becomes
approximately zero, which suggests that the flow is completely supersonic.
Note that the source of the signal originates at the throat, and then even if
the flow measurement position is supersonic, the signal from the throat can
be detected. Figure 5(d) indicates the same variations that is all the rms for
each input frequency suddenly decrease at the pressure ratio at 2.0 denoting
the supersonic sta this postion.
 Experimental Study of the Response of Transonic Diffuser Flow ... 381

FFT Analyses of Wall Pressure Fluctuation

The rms of the pressure fluctuation usually becomes large when the oscillating
shock wave approaches the monitoring position. It is also important to examine
the unsteady behavior of the shock wave in detail, especially in this case, in
order to confirm the response of the flow field to the piezoceramic actuator.
One of the best ways to evaluate the effect of the actuation is to check the
contributions of every frequency by means of FFT analysis of the pressure
fluctuations at x/h* = 6.67. Figures 6(a)-(d) show the results of the FFT analysis
of the pressure fluctuations for each pressure ratio. Figure 6(a) shows no
dominant frequency and almost no frequency output for pressure ratios greater
than p0/pb = 1.25, as deduced from the results for the rms shown in Figure 5(b).
This indicates that the shock wave oscillates with no dominant frequency for
all of the pressure ratios and that the state of the flow at x/h* = 6.67 becomes
supersonic at pressure ratios greater than 1.25. Moreover, the levels of each
spectrum are not so large because the shock wave at this position is not so strong.
On the other hand, Figures 6(b)-(d) clearly show the dominant frequencies for
each pressure ratio, which corresponds to the input frequency to the piezoceramic
actuator. Note that even in the supersonic state, the frequency of 200 Hz can be
observed. These figures indicate the effect of the piezoceramic actuator on the
flow fields. In addition to the results of the visualization, monitoring the wall static
pressure fluctuation is a quantitative way to confirm the effect of the actuator.

Figure 6. FFT analysis of wall static pressure fluctuations at x/h* = 6.67: (a) fp
= 0 Hz; (b) fp = 100 Hz; (c) fp = 200 Hz; (d) fp = 300 Hz.
382 Mechatronics and Robotics

Figures 7(a)-(d) show the FFT analysis of the pressure fluctuation at x/h* =
10.0 for the same input frequency as that in Figure 6. Figure 7(a) shows no
dominant frequency but only a relatively low frequency, which has the same
tendency as that in Figure 6(a)
Figures 7(b)-(d) show cleominant frequencies. This indicates the
possibility of reducing the pressure fluctuation by controlling the shock
position, which is expected to cancels the pressure fluctuation.
In the figures, a wide range of frequencies are observed at a pressure
ratio of approximately 1.7 due to the unexpected shock-induced separation
of the boundary layer.
All of the figures show a sudden decrease in the power spectrum at
a pressure ratio of approximately 1.4. This implies that the shock wave
completely passes the measurement position, resulting in a change in state
from subsonic to supersonic.

FFT Analysis of Shock Positions

The wall static pressure fluctuations provide us with only local information on
the shock wave behavior around the measurements positions. Accordingly, in
order to clarify the entire flow field, the variations in unsteady shock positions
are also important because the shock wave always contains all of the information
on its upstream and downstream conditions regardless of its position.
 Experimental Study of the Response of Transonic Diffuser Flow ... 383

Figure 8. FFT analysis of Shock positions: (a) fp = 0 Hz; (b) fp = 100 Hz; (c) fp
= 200 Hz; (d) fp = 300 Hz.
Figures 8(a)-(d) show the results of the FFT analyses of the shock
positions under the influence of the piezoceramic actuator as well as with no
control. In case of no control, there is no dominant frequency, as shown in
Figure 8(a). The relatively low-frequency spectrum is observed as long as
the pressure ratio is smaller than 1.4. This shows that the shock wave passes
the measurement position monotonically as the pressure ratio increases.
In contrast, for a shock wave frequency of 100 Hz, a dominant frequency
is observed until a pressure ratio of 1.5. The peak value of the dominant
frequency decreases as the pressure ratio increases. This decrease in the
peak value of the dominant frequency is explained by the behavior of the
shock wave, as shown in Figure 3. That is, the gradient of the shock wave
displacement decreases as the pressure ratio increase as shown in Figure 3
due to the expansion rate of the cross sectional area of the diffuser.
This implies that shock wave located downstream is insensitive to the
pressure fluctuation compared with that located upstream position. Then, the
oscillation of the shock wave gradually decreases as it moves downstream.
The diminishing of the dominant frequency for higher pressure ratios might
be due to the relative increase in other frequencies.
On the other hand, when the higher frequency is applied to the throat,
the clearly peaky dominant frequency, which is the exact same frequency as
that input to the piezoceramic actuator, can be observed. Note also that the
applied frequency remains, even for the higher pressure ratio, which suggests
that the shock wave is always under the influence of the piezoceramic
384 Mechatronics and Robotics

actuator. The tendencies observed in Figure 8(c) also appear in Figure 8(d).
In other words, the input frequency of 300 Hz is the dominant frequency for
all pressure ratios. Then, actuation at the throat is a promising method for
controlling shock wave behavior.

CONCLUSIONS
A piezoceramic actuator is applied to the throat of a circular arc diffuser with
various driving frequencies in order to clarify the response of the flow field
and shock wave behaviors to the piezoceramic actuator. The piezoceramic
actuator, which moves periodically as a reference signal, is considered to be
a driving force for oscillation phenomena. The conclusions are summarized
as follows:
1) The starting shock wave moves downstream monotonically with
the increase in the wind tunnel pressure ratio, regardless of the
input frequencies to the piezoceramic actuator.
2) The rms values of the wall pressure fluctuations decrease
suddenly just after the shock wave completely passes over the
measurement position for all driving frequencies.
3) The pressure fluctuations at the throat and downstream of the
throat correspond to the exact same frequencies of the input
frequencies to the piezoceramic actuator in both cases that the
position is supersonic and subsonic state.
4) Shock wave behaviors are also confirmed to correspond to the
piezoceramic actuator behaviors.
 Experimental Study of the Response of Transonic Diffuser Flow ... 385

REFERENCES
1. G. E. A. Meier, “Shock Induced Flow oscillations in a Laval Nozzle,”
In: K. Oswatitsch and D. Rues, Eds., Symposium Transonicum II,
Springer, Berlin, 1976, pp. 252-261.
2. T. Hsieh, T. J. Bogar and T. J. Coakley, “Numerical Simulation and
Comparison with Experiment for SelfExcited Oscillations in a Diffuser
Flow,” AIAA Journal,
3. Vol. 25, No. 7, 1987, pp. 936-943. doi:10.2514/3.9725
4. R. Bur, R. Benay, A. Galli and P. Berthouze, “Experimental and
Numerical Study of Forced Shock-Wave Oscillations,” Aerospace
Science and Technology, Vol. 10, No. 4, 2006, pp. 265-278.
doi:10.1016/j.ast.2005.12.002
5. E. Benini, R. Biollo and R. Ponza, “Efficiency Enhancement in
Transonic Compressor Rotor Blades Using Synthetic Jets: A numerical
Investigation,” Applied Energy Vol. 88, No. 3, 2011, pp. 953-962,
6. doi:10.1016/j.apenergy.2010.08.006
7. P. B. Salunkhe, J. Joseph and A. M. Pradeep, “Active Feed-Back
Control of Stall in a Axial Flow Fan under Dynamic Inflow Distortion,”
Experimental Thermal and Fluid Science, Vol. 35, No. 6, 2011, pp.
1135-1142. doi:10.1016/j.expthermflusci.2011.03.008
8. M. Yaga, T. Haga and K. Oyakawa, “Study on Passive Control in a
Transonic Diffuser,” AIAA Paper 2000-0902, 38th Aerospace Sciences
Meetings & Exhibit, Reno, January 2000, pp. 1-11.
 CHAPTER

18
An ARX-Based PID-Sliding Mode
Control on Velocity Tracking Control of
a Stick-Slip Piezoelectric-Driven
Actuator

Y. Cao, X. B. Chen
Department of Mechanical Engineering, University of Saskatchewan,
Saskatoon, Canada

ABSTRACT
Piezoelectric-driven stick slip actuators have been drawn more and more
attention in the nano- positioning application due to the high accuracy and
theoretical unlimited displacement. However, the hysteresis of piezoelectric
actuator (PEA) and the nonlinear friction force between the end- effector and
the stage make control of piezoelectric-driven stick slip actuator challenge.
This paper presents the development of an autoregressive exogenous

Citation:- Cao, Y. and Chen, X. (2015), “An ARX-Based PID-Sliding Mode Control
on Velocity Tracking Control of a Stick-Slip Pi-ezoelectric-Driven Actuator”. Modern
Mechanical Engineering, 5, 10-19. doi: 10.4236/mme.2015.51002.
Copyright: © 2015 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
388 Mechatronics and Robotics

(ARX)-based proportional-integral- derive (PID)-sliding mode control

(SMC) for the velocity tracking control of the piezoelectric-driven stick
slip actuator. Stability is guaranteed by rigorously choosing the appropriate
PID parameters and the zero steady state error is achieved. To verify the
effectiveness of the proposed method, experiments were carried out on a
commercially-available piezoelectric-driven stick slip actuator. The tracking
errors were compared with the traditional PID controller, illustrating that
in spite of existing of modeling error, the ARX-based PID-SMC is able to
better improve the velocity tracking performance of piezoelectric-driven
stick slip actuator, compared with the traditional PID controller.
Keywords: Autoregresive Progress, Sliding Mode Control, Slick-Slip,
Piezoelectric Actuator

INTRODUCTION
Piezoelectric-driven stick-slip actuators play important roles in nano-
positioning applications due to their simple configuration, high accuracy and
theoretically unlimited displacement [1] - [6]. The end-effector is supported
and guided by a movable stage, which is connected to the base through
flexible hinges and driven by a PEA, as shown in Figure 1. The end-effector
moves together with the stage under slow expansion of the PEA. Sudden
contraction of the PEA leads to the sliding motion of the end-effector since
the force due to the friction limit between the end-effector and the movable
stage. As a result, the end-effector moves a step ΔS, with respect to its
original position. Such steps can be accumulated to achieve a theoretically
unlimited displacement (actually limited by the size of the moveable stage).
The hysteresis of PEA and the friction dynamics of the end-effector make
the control of stick-slip actuator challenge. An analog electronic circuit was
developed in [7] for driving stick-slip piezoelectric linear actuators. The
key design task for the amplifier is to provide a high-voltage asymmetric
sawtooth-like signal and feed it into a capacitive load.
To compensate the hysteresis effect, combination of a charge control
scheme with switch- ing is proposed and the experiments confirm the
superiority of the proposed method over other existing techniques.
However, the charge control requires expensive hardware, which increases
the difficulty and cost of its implementation.
 An ARX-Based PID-Sliding Mode Control on Velocity Tracking ... 389

To avoid this, a voltage/frequency proportional controller was proposed in

[8] for high stroke displacement of stick-slip micro-positioning stage.
The proposed controller is essentially a combination of the sign controller,
the classical proportional control and frequency proportional control, and
the experimental results showed its priority over the basic controller.
To compensate the steady state error, integral control was employed in
addition to the aforementioned method [9] . In [10] ,
Proportional-Derivative (PD) controller was applied in the control of
the stick slip piezoelectric actuators. Its application in the sample holder for
Atomic
Force Microscope (AFM), the micro-Electrical Discharge Machining
(EDM) machine, the micro-assembly system and the micro telemanipulation
system for biological specimens was described.
All the research mentioned above related to the displacement tracking
control of the piezoelectric-driven stick-slip actuators under step or ramp
reference signal.
The velocity tracking control has not been reported yet. This work
focuses on the development of control method for the velocity tracking
control of the piezoelectric- driven stick-slip actuators.
PID/PI/PD controller has shown great potential in the control application
for nano-positioners.
The challenge of PID/PI/PD control in the velocity tracking control of
stick-slip piezoelectric-driven stage is maintaining the system stability in
the presence of uncertainty and disturbance.
It also has an issue with low gain margin in high frequency applications
[1]. Recently, sliding model control (SMC) has been drawing considerable
attention in the control research community worldwide due to its ability to
compensate for the system uncertainties and disturbance [11] -[14] .
To solve the chattering problem of the regular SMC, a PID-based SMC
(PID-SMC) was developed in [15], in which the discontinuous “bang-bang”
switching function is replaced by a PID regulator. Due to the integral effect
of the PID regulator, zero steady state tracking error can be achieved.
390 Mechatronics and Robotics

Figure 1. Working principle of the piezoelectric-driven stick- slip actuator.

As a state tracking control scheme, the PID-SMC developed in [15] has
been shown effective in the control of the second order mechanical systems,
in which the states can be readily estimated from the measured output and/or
its derivatives. In some circumstances, the system state might not be readily
or even be impossibly obtained due to system complexity. In such a case,
the application of PID-SMC is challenged due to the lack of system state
information. For this, an output tracking integrated PID-SMC is developed
[16], in which the transfer function of the plant is considered to be a cascade
of the numerator and the denominator. The numerator is compensated by an
optimal inverse feed forward controller and as such the regular state tracking
based SMC can be applied. The proposed method has been experimentally
shown effective in the control of a commercially available one-degrees-of-
freedom (DOF) piezoelectric actuator.
It is noted that if the dynamics of the plant can be represented by an auto-
regressive (ARX) model without zeros, for example, , where
u(k) and y(k) are the system input and output at time instant k respectively,
such a model will be readily transferred to a state space model with its state
being the system output and the state tracking based PID-SMC is able to be
directly applied for the dynamics compensation. Inspired by this, an ARX-
based PID-SMC is developed in this paper and applied in the velocity control
of a one-DOF stick-slip PEA. The effectiveness of the proposed method is
experimentally verified and compared with the traditional PID controller
introduced in [9].
 An ARX-Based PID-Sliding Mode Control on Velocity Tracking ... 391

ARX-BASED PID-SMC AND ITS DESIGN

The discrete nth order plant might be generally described by the following
transfer function

(1)
where are the parameters and the numerator
and denominator are relatively prime polynomials. With the use of the
system identification method, this plant model can be mathematically repre-
sented by an ARX model.

(2)
Equation (2) is identical to a discrete transfer function without zeros.
The neglect of zeros definitely leads to model errors, which is considered to
be disturbance in Equation (2).
The ARX model can be rewritten in a state space form

(3)

where is the state vector;

are the system matrixes of the discrete
state space model. The state vector is represented in terms of the outputs in
the past history, which suggest the state tracking is essentially the output
tracking. Therefore, the state tracking SMC design method can be applied.
Denote the desired output vector to be
. The objective of SMC is to force
the error state to move to the sliding surface and then
converge to zero. Equation (3) can be rewritten in terms of the dynamics of
e(k),
(4)
Similar to the general SMC design approach introduced in [15], the
input u (k) is divided into two parts and , i.e.

(5)
392 Mechatronics and Robotics

It is noted that Equation (5) is non-casual if the future desired output is

unknown. In such a case, could be estimated by the trapezoid
estimation

(6)
Substituting Equation (5) into Equation (4) yields

(7)
For the system described by Equation (3), one has
. Thus, Equation (7) is simplified as

(8)
For the sliding function that takes the following form of

(9)
The control action can be considered consisting two parts, i.e.,

(10)
where is the equivalent control and is the
switching control.
Substituting Equation (10) into Equations (8) and (9) yields

(11)

(12)
For the convenience of following discussion, Equations (9) and (11) are
rewritten as

(13)
 An ARX-Based PID-Sliding Mode Control on Velocity Tracking ... 393

where . The system described by Equation (13) can

be equivalent as a dynamic plant,
where is the input and s(k) is the output. The switching control is to
force the plant output to be zero. For this, PID controller can be employed
to generate the switching control such that the equivalent control error s(k)
is zero, which is given by

(14)
where P, I and D are parameters of the discrete PID-based SMC; T is the
sampling period.
Theorem 1: If the closed-loop system (13) is stable, the zero steady state
error can be achieved [17] .
Theorem 2: There exist some P, I and D such that the closed-loop control
system (13) is stable [16] .

EXPERIMENTS
To verify the effectiveness of the proposed method, experiments were
carried out to control the motion velocity of a stick-slip piezoelectric-
driven actuator (Figure 2). The PEA used to drive the stage generates
displacement in a range of 15 μm with a resolution of 0.5 nm. An inductive
sensor (SMU9000, Kaman, Windsor, CT) with a resolution of 10 nm was
used to measure the displacement of the end-effector. Both the actuator
and the sensor were connected to a host computer via an I/O board (PCI-
DAS1602/16, Measurement Computing Corporation) and controlled via
SIMULINK programs. The movable stage is driven by the PEA and the
end-effector on the stage can slide horizontally on the stage via a pair of V
grooves machined on them. Polished silicon wafers are attached to the V
grooves as frictional surfaces to guarantee the uniformity of the frictional
characteristics throughout the traveling range of the end-effector. All the
displacements presented in this study were measured by the inductive sensor
with a sampling interval of 0.05 ms.
Figure 3 shows the model configuration of the stick-slip piezoelectric-
driven stage developed in [18] .
394 Mechatronics and Robotics

Figure 2. Experimental settings on the piezoelectric driven stage; (a) schematic

and (b) picture.

Figure 3. Model configuration of the piezoelectric-driven stick-slip actuator

(Regenerated from [18] ).
Me is the mass of the end-effector, Ff is the friction force between the
PEA stage and the end-effector. The hysteresis model H is cascaded with the
vibration dynamics model. The output―PEA displacement is feed- forward
to the friction model, which generates the friction force and the motion
parameters of the end-effector.
To apply the ARX-based PID-SMC, the PEA and the end-effector
are considered to be an integrated system with its input being the driven
voltage u and its output being the motion velocity of the end-effector, as
shown in Figure 3. The nonlinearities such as hysteresis and the friction are
considered to be disturbances and then rejected by the proposed method.
Experiments for parameter identification were first carried out to obtain
the ARX model. Then, the ARX-based PID-SMC was applied to control
the nano-postioning stage for tracking step reference signals. Finally,
dynamic tracking control with the proposed methods was implemented and
comparison to a PID controller was made.
 An ARX-Based PID-Sliding Mode Control on Velocity Tracking ... 395

Model Identification
To drive the end-effector, a saw-tooth generator is required, as seen in Figure
4. Each saw-tooth is 20 ms wide. The amplitude of the each saw-tooth is
defined by the reference input voltage u(t). If u(t) < 0, a reverse saw- tooth
signal is demanded to move the end-effector to the backward direction, as
shown in Figure 5.

Figure 4. Saw-tooth generator for the stick-slip motion and the velocity calcu-
lator.

The driven voltage supplied to the PEA can be represented in terms

of the reference input voltage as

(15)
where T is the tooth width, mod represents the modulus after division.
The displacement of the end-effector is measured by the inductive
sensor. The average velocity v is estimated by

(16)
where is the displacement of the end-effector in one period of
the saw-tooth signal.
The dynamics of the integral system can be regarded approximately
as a second order system according to our previous research [19] [20]. As
such, a second order ARX model was employed to describe the system
dynamics, while the hysteresis exhibited by the actuator and the friction
between the stage and the end-effector are considered as extra disturbances
to the dynamic model. For the experiments, a 10 V step input reference
voltage was provided to the PEA and the corresponding output velocity was
estimated by Equation (23). With the recorded data, the parameters were
identified by using the least square method, leading to
396 Mechatronics and Robotics

(17)

Step Tracking
The ARX-based PID-SMC developed in this study was implemented in
experiments to control the velocity of the piezoelectric-driven slick-slip
actuator. The sliding surface was defined according to Equation (9), where
and m = −0.1. To minimize the rising time, the PID parameters in
the PID regulator is adjusted to be P = 0.2, I = 3, D = 0. For comparison, the
traditional PID controller was also applied, with its parameters adjusted by
trails-and-errors to be P = 0.4, I = 4, D = 0.
Figure 6 shows the velocity step response of the piezoelectric-driven
stick-slip actuator controlled by the developed control scheme and the
traditional PID controller. Table 1 shows the comparison of the rising time
and the overshoot.
It can be seen that the tracking errors approach to zero through the
use of I component in the PID regulator. The velocity step response of the
piezoelectric-driven stick-slip actuator controlled by the proposed method
is faster than that controlled by the traditional PID controller. For example,
when a 10 μm/s step reference input was provided, the rising time of the
velocity step response controlled by the ARX-based PID-SMC is 0.13 s, 0.2
ms less than that controlled by the PID controller. Faster velocity response
is also observed for the 20 μm/s step reference input. However, since the
dynamics performance of the piezoelectric-driven stick-slip actuator varies
with the amplitude of the input voltage, the model error increased when the
20 μm/s step reference input was provided to the actuator. As a result, the
overshoot increased to 10% in this case.

Figure 5. Reverse saw-tooth to driven the end-effector in the backward direc-

tion.
 An ARX-Based PID-Sliding Mode Control on Velocity Tracking ... 397

Figure 6. Velocity Step response of the piezoelectric-driven stick-slip actuator

controlled by the proposed method and the traditional PID controller (a) 10
μm/s desired velocity; (b) 20 μm/s desired velocity.

Table 1. Comparison of the velocity step response between the ARX-based

PID-SMC and the traditional PID controller.

Reference inputs Controller Rising time (s) Overshoot (%)

10 µm/s step input SMC 0.13 0
PID 0.33 0
20 µm/s step input SMC 0.11 10
PID 0.21 0
To test the control performance of the proposed method, the same velocity
step tracking experiments were also carried out with an increased mass of
the end-effector. The same parameters is applied and compared with the
traditional PID controller. Figure 7 shows the velocity step response of the
piezoelectric-driven stick-slip actuator controlled by the developed control
scheme and the traditional PID controller. Table 2 shows the comparison
of the rising time and the overshoot. It can be concluded that the velocity
step response of the piezoelectric-driven stick-slip actuator controlled by
the proposed method is faster than that controlled by the traditional PID
controller.
It is noted that the rising time of the velocity step response controlled by
the ARX-based PID-SMC is 0.09 s. This is different from the result with an
decreased mass of the end-effector, since the dynamics of the piezoelectric-
driven stick-slip actuator also changes with the weight of the end-effector.

Sinusoidal Tracking
To further show the effectiveness of the proposed control method, sinusoidal
tracking experiments with different frequencies were carried out on the
piezoelectric-driven stick-slip actuator.
398 Mechatronics and Robotics

Figure 7. 10 μm/s Step response of the piezoelectric-driven stick-slip actua-

tor controlled by the proposed method and the traditional PID controller.
Table 2. Comparison of the velocity step response between the ARX-based
PID-SMC and the traditional PID controller with an increased mass of the end-
effector.

Reference inputs Controller Rising time (s) Overshoot (%)

10 µm/s step input SMC 0.09 0
PID 0.21 0
The same parameters were applied and the control performance was
compared with the same PID controller. Figure 8 shows the sinusoidal
tracking performance of the piezoelectric-driven stick-slip actuator
controlled by the proposed method and the traditional PID controller.
Table 3 compares the tracking error evaluated in terms of the 2-norm
of the difference between the desired output and the measured output.
From Table 3, it can be seen that the developed control scheme performs
better than the traditional PID controller. As frequency increases, the priority
of applying the ARX-based PID-SMC is more obvious. For example, for a
0.25 Hz 10 µm/s sinusoidal reference input, the tracking error with ARX-
based PID-SMC is 1.12 µm, which is 29.7% of that with the PID controller.
While for a 1 Hz sinusoidal reference input, the improvement increases to
40.2%.
It is noted that the velocity tracking performance deteriorates when the
motion direction of the end-effector changes. For example, for 10 μm/s 1 Hz
sinusoidal reference velocity, the maximum tracking error in time interval
 An ARX-Based PID-Sliding Mode Control on Velocity Tracking ... 399

0.5 - 0.75 s and 1.5 - 1.75 s is 50% of the amplitude if controlled by the ARX-
based PID-SMC. The maximum tracking error even reaches 100% of the
reference amplitude of the traditional PID controller is applied. This might
be due to the nonlinear friction force between the end-effector and moving
stage when the velocity changes from positive to negative. Obviously, the
ARX-based PID-SMC partially compensates the nonlinearity. Consideration
of the nonlinear friction model and its integration in the controller design are
required for further improvement on the velocity tracking performance of
the piezoelectric-driven stick-slip actuator.

CONCLUSION
This paper presents the development of an ARX-based PID-SMC for the
velocity control of the piezoelectric- driven stick-slip actuator. Specifically,
by applying ARX model, the output-tracking problem is defined as one of
state tracking, while the “bang-bang” switching control in SMC is replaced
with PID-based one. With the developed control scheme, chattering or the
state oscillation at a high frequency can be eliminated and also the zero
steady state error can be achieved.
To verify the effectiveness of the developed control scheme, experiments
were carried out on a piezoelectric-driven stick-slip actuator, whose
dynamics was identified by experiments.
The results of velocity tracking performance with the proposed control
scheme were compared to that with the traditional PID controller. It was
shown that both the step response and the sinusoidal tracking performance
can be further improved by applying the proposed method, although the
ARX model error exists.

Table 3. Comparison of the velocity step response between the ARX-based

PID-SMC and the traditional PID controller with an increased mass of the end-
effector.

Reference inputs Frequency (Hz) Tracking error (µm/s)

ARX-based PID-SMC PID
10 µm/s sinusoidal 0.25 1.12 1.57
input 0.5 2.13 2.89
1 2.82 4.72
400 Mechatronics and Robotics

Figure 8. Comparison of sinusoidal tracking performance with different con-

trollers: displacement and tracking error for (a) (b) 0.25 Hz frequency, (c) (d)
0.5 Hz frequency and (e) (f) 1 Hz frequency.
However, the control performance of the ARX-based PID-SMC depends
on the accuracy of the identified dynamic model. This might be solved by
employing a dynamic model with adaptive parameters and an adaptive ARX-
based PID-SMC, which will be the focus of the future work. Furthermore,
nonlinear friction model should be considered in the SMC design to further
improve the velocity tracking performance of the piezoelectric-driven stick-
slip actuator.

ACKNOWLEDGEMENTS
Y. Cao, X. B. Chen (CSC) and the Natural Sciences and Engineering
Research Council (NSERC) of Canada is acknowledged.
 An ARX-Based PID-Sliding Mode Control on Velocity Tracking ... 401

REFERENCES
1. Devasia, S. and Moheimani, S.O.R. (2007) A Survey of Control Issues in
Nano-Positioning. IEEE Transactions on Control System Technology,
15, 802-823. http://dx.doi.org/10.1109/TCST.2007.903345   
2. Shiratori, H., Takizawa, M., Irie, Y., Hirata, S. and Aoyama, H. (2012)
Development of the Miniature Hemispherical Tilt Stage Driven by
Stick-Slip Motion Using Piezoelectric Actuators. Mechatronics-REM,
Paris, 21-23 November.
3. Zhang, Q.S., Chen, X.B. Yang, Q. and Zhang, W.J. (2012) Development
and Characterization of a Novel Piezoelectric-Driven Stick-Slip
Actuator with Anisotropic-Friction Surfaces. International Journal
of Advanced Manufacture Technology, 61, 1029-1034. http://dx.doi.
org/10.1007/s00170-011-3771-y
4. Eigoli, A.K. and Vossoughi, G. (2010) Dynamic Modeling of Stick-
Slip Motion in a Legged Piezoelectric Driven Microrobot. International
Journal of Advanced Robotic Systems, 7, 201-208.
5. Kang, D. (2012) Modeling of the Piezoelectric-Driven Stick-Slip
Actuators. Master Thesis, College of Graduate Studies and Research,
University of Saskatchewan, Saskatoon.
6. Bergander, A. and Breguet, J.M. (1998) Performance Improvement for
Stick-Slip Positioners. International Symposium on Micromechatronics
and Human Science, 59-66.   
7. Spiller, M. and Hurak, Z. (2011) Hybrid Charge Control for Stick-Slip
Piezoelectric Actuators. Mechatronics, 21, 100- 108. http://dx.doi.
org/10.1016/j.mechatronics.2010.09.002   
8. Rakotondrabe, M., Haddab, Y. and Lutz, P. (2008) Voltage/Frequency
Proportional Control of Stick-Slip Micropositioning Systems. IEEE
Transactions on Control System Technology, 16, 1316-1322. http://
dx.doi.org/10.1109/TCST.2008.917232   
9. Rakotondrabe, M., Haddab, Y. and Lutz, P. (2009) Development of
Modeling and Control of Mircro/Nano-Positioning 2-DOF Stick-Slip
Device. IEEE Transactions on Mechatronics, 14.   
402 Mechatronics and Robotics

10. Breguet, J.M. and Clavel, R. (1998) Stick and Slip Actuators: Design,
Control, Performances and Applications. International Symposium on
Micromechatronics and Human Sciences, Nagoya, 25-28 November
1998, 89-95.   
11. Young, K.D., Utkin, V.I. and Özgüner, Ü. (1999) A Control Engineer’s
Guide to Sliding Mode Control. IEEE Transactions on Control System
Technology, 17, 328-342. http://dx.doi.org/10.1109/87.761053   
12. Edwards, C. and Spurgeon, S.K. (1998) Sliding Mode Control. Taylor
& Francis Ltd, London.
13. Liawa, H.C., Shirinzadeh, B. and Smith, J. (2007) Enhanced Sliding
Mode Motion Tracking Control of Piezoelectric Actuators. Sensors and
Actuators, A138, 194-202. http://dx.doi.org/10.1016/j.sna.2007.04.062
14. Wang, S., Habibi, S., Burton, R. and Sampson, E. (2006) Sliding Mode
Control of an Electrohydraulic Actuator System with Discontinuous
Nonlinear Friction. Proceedings of the Institution of Mechanical
Engineers, 222, Part I:J, Systems and Control Engineering.   
15. Peng, J.Y. and Chen, X.B. (2014) Integrated PID-Based Sliding Mode State
Estimation and Control for Piezoelectric Actuators. IEEE/ASME Transactions
on Mechatronics, 19, 88-99. http://dx.doi.org/10.1109/TMECH.2012.2222428  
16. Cao, Y. and Chen, X.B. (2014) Disturbance Observer Based Sliding
Mode Control for a Three-DOF Nano-Positioning Stage. IEEE/ASME
Transactions on Mechatronics, 19, 924-931. http://dx.doi.org/10.1109/
TMECH.2013.2262802   
17. Cao, Y. and Chen, X.B. (2012) Integrated Inversion-Feedforward
and PID-Based Sliding Mode Control for Piezoelectric Actuators.
American Control Conference, Montreal, 27-29 June 2012.   
18. Peng, J.Y. and Chen, X.B. (2011) Modeling of Piezoelectric-Driven
Stick-Slip Actuators. IEEE/ASME Transactions on Mechatronics, 16,
394-399. http://dx.doi.org/10.1109/TMECH.2010.2043849   
19. Cao, Y. and Chen, X.B. (2012) A Novel Discrete ARMA-Based
Model for Piezoelectric Actuator Hysteresis. IEEE/ ASME
Transactions on Mechatronics, 17, 737-744. http://dx.doi.org/10.1109/
TMECH.2011.2128339   
20. Chen, X.B., Zhang, Q.S., Kang, D. and Zhang, W.J. (2008) On the
Dynamics of Piezoelectric Positioning Systems. Review of Scientific
Instrument, 79, Article ID: 116101. http://dx.doi.org/10.1063/1.2982238   
 INDEX

Symbols 389
Automated Monitoring 254
1-degree-of-freedom (1-DOF) 274 automatic irrigation control system
1-DOF control problems 300 65
A automatic system 70
Automation 68, 74
Acoustic and Magnetic Signals 224 Automation component 68
Actuators 320, 323, 342, 343, 344 automotive manufacturing industry
Adaptive Control 102 66
adaptive controller 106, 115 Autonomous Microbe Sensor 253,
ad-hoc relay communication. 213 254, 271
aerospace 170 Autoregresive Progress 388
agricultural production 68 auto-regressive (ARX) 390
Anisotropic Magneto Resistive autoregressive exogenous (ARX)
(AMR) 232 388
ANSYS Code 346
arc welding 82 B
Arduino Robot Control 82 bang-bang 389, 399
Artificial Intelligence 12, 14 Barbalat’s lemma 109
Atomic Force Microscope (AFM)
404 Mechatronics and Robotics

Battle Damage Assessment 226 deep learning 3, 13

Battlefield Surveillance 226 degrees-of-freedom (DOF) 390
Boxing 184, 189, 199, 200, 202 Delta robots 12, 13
broadcast warning message 206 Denavit-Hartenberg method 169
Design 81, 82, 100
C Detection and Tracking Algorithms
cameras 12 224, 247
Cartesian space analysis 120, 121, development of cloud manufactur-
122 ing 19
chain systems 10 Device management 26
changes dynamic environment 135 D-H Method 170
Chassis 69 Digital Product Process -program
closed-loop system 104, 108, 109 61
Coleman-Hodgdon model 293 Direct Drive Arm 10
commercial off-the-shelf (COTS) distributed control system 66
57 Distribution Type Sensor 184
Communication infrastructure ser- DNA binding fluorescent dyes 255
vices 26 DNA Staining 254, 271
complex contact geometry 189 Drug Administration in Hospitals
complex strategies 3, 13 226
Compressible Flow 372 Dynamic environment represents
contact friction 185, 188, 191 120
Contiki 18, 20, 21, 22, 31, 32, 33,
E
35, 37, 39
controlling flexible structures 103 Ecole Poly-technique Fédérale de
controlling of manufacturing sys- Lausanne (EPFL) 12
tems 19 economic and geopolitical situation
cost-effective devices 6 6
Cp Iot Enabler 25, 26 effective mitigation strategy 254
Cyber Iot Enabler 25 elasto-slide element 280, 281
cyber-physical component 22, 29, electric actuators 6
30, 32 Electrical Discharge Machining
Cyber-Physical Systems 18 (EDM) 389
electronic 170
D energy and transportation 170
D* Algorithm 120, 138, 139 Environmental Applications 226
Danish Technological Institute Environmental control in office
(DTI) 254 buildings 226
decision-making 19 Euler-Bernoulli 104
 Index 405

External Output Terminal 74, 75 High-speed plate 175

High-Speed Plate Carrying Manipu-
F lator 169, 170, 171, 173
Faster Response Time 69 Home and office Applications 226
Finnish Funding Agency for Tech- Home automation 226
nology and Innovation 60, 61 human body 183, 184, 201, 202
Finnish Funding Agency for Tech- Hybrid Sensor Network 205, 206
nology and Innovation hydraulic actuators 6
(TEKES) 61 hysteresis models 278, 280, 281,
flexible multi-body dynamics 103 282, 283, 284, 285, 287, 288,
flexible of whole system 207 290, 295
flexible sensor 183, 184, 185, 187,
I
189, 193, 195, 196, 197, 200,
201, 202 Impact Force 183, 184, 193, 203
Flexible Sensor 184 including biofouling 253, 254
Flood Detection 226 industrial automation systems (IAS)
Fluorescence count (AMS) 260, 263 19
Food and Agriculture Organization Industrial Automation Thing (IAT)
(FAO) 67 18, 19
force sensors 12 Industrial robotics 1
Forest Fire Detection 226 industrial robots 1, 3, 6, 7, 8, 9, 13,
four degree of freedom (4 DOF) 84 14
fourth-order differential 351 industrial workforce 82, 83
free Cartesian space (FCS) 125, 126 industry applications 121
infectivity loss 254
G information technology 170
geometry 148, 164 infrastructure services 26
Great Duck Island System 226 Injection Seawater 253, 254, 271
integrated circuits (IC) 319
H Intelligence 70
Habitat Monitoring 223, 226 internet of manufacturing things
Habitat Monitoring Applications (IoMT) 23
226 Internet of Things (IoT) 18
Health Applications 226 Intrusive sensors 227, 228
high-speed and dynamic transmis- IoT-compliant one 19, 24, 26
sion 170 IoT enabler 25
high speed pick-and-place opera- IoT manufacturing environment 21
tions 12 Irrigation System 66
406 Mechatronics and Robotics

J 62
Mechatronic component 18, 30
Japanese Robotics Association 5
Mechatronics 18, 38, 39
K micro-assembly system 389
Microbial Sensor 254
kinematics model 170, 181 microbiologically influenced corro-
Known Dynamic Environment 120 sion (MIC) 254
L microcontroller unit 225
Micro-Electro-Mechanical Systems
Laboratory calibration data 260 321
Labview interface 96, 99 micro-electro mechanical systems
Labview Robot Control 82 (MEMS) 321
Lab view software 95 microprocessors 6
Ladder logic 71, 72 Military Applications 226
Ladder logic programming 72 mobile nodes 206, 208, 210, 211,
laser scanners 12 212, 213, 214, 215, 216, 217
linear dynamic system 294 mobile sensor network 206, 207
Liqueur production mechatronic mobile sensor nodes 206, 207, 208,
system 28 210, 211, 219
logical reasoning 3, 13 Model Driven Development 44
Low-Cost 81, 82 model-driven development of com-
low-frequency spectrum 383 plex mechatronic systems
Lyapunov-based methods 103 (MTSs) 45
Lyapunov function 107, 110 modeling flexible robots 103
Lyapunov stability theory 103, 114, Model Integrated Mechatronics
115 (MIM) 44, 45, 47
M Mode Shape 345, 346
Monitoring Biodiversity 226
Magnetic sensors 232, 233 Monitoring Nuclear Reactor 226
Managing Inventory Control 226 Monitoring through Internet 226
manipulator control system 170 mono-static and mono-mobile net-
Maxwell resistive capacitor (MRC) work 217
278 Monterrey University of Technol-
mechanical design 83, 84 ogy 83
mechanical engineering 18 motors 66, 70, 73
mechanical manufacturing 170 Multicriteria optimization 142
mechanical unit 18, 29
mechatronic component 45, 49, 51, N
Nano positioning 277
 Index 407

navigation 205, 206, 207, 209, 212, Photostriction 320, 324, 326, 342
213, 215, 216, 217, 218, 219 Photostrictive effect of PLZT 325
Network communication 66 Photostrictive materials 321, 322,
non-centrosymmetric 320, 324 326
non-mapping navigation 207, 213 photostrictive optical actuator 326,
normally closed (NC) 71 327, 332
normally open (NO) 71 Physical Object and a cyber-physi-
Normal Shape 345, 346 cal IoT enabler 25
nuclear 170 PID-based SMC 389, 393
numerical simulations 152 PID control 291, 300
PID regulator 389, 396
O piezoceramic actuator 371, 373,
object- oriented (OO) 23 374, 375, 377, 379, 380, 381,
Open-loop control schemes 289 383, 384
opera- tor panel 66 Piezoelectric Actuators 273, 274,
optimal algorithm 207, 213 275, 303, 306, 307, 308, 309,
optimal design 149, 151, 154, 158, 310, 311, 315
162, 163 Piezoelectric actuators (PEAs) 273,
optimal path 121, 123, 135 274
Optimal robot path planning 142 piezoelectric cantilever 346, 347,
Optimal task placement 142 349, 351, 356, 358, 360
out-of-plane bending deformation Piezoelectric-driven stick-slip actua-
183, 185, 189 tors 388
piezoelectric film (PVDF) 185, 189
P piezo-resistance 184
painting 82 Pilot experiment 74
Particle Swarm Optimization (PSO) Platform Independent Model (PIM)
120 22
passive algorithm 207 PLC’s internal memory 69
path planning 119, 120, 121, 122, polynomial based linear mapping
123, 126, 131, 135 strategy 284
PEA-driven parallel mechanism 300 Power consumption 141
PEA hysteresis 277, 282 Power supply 69
Pediatric Research Clinical Institute practical industrial 372
xix PrandtlIshlinskii (PI) 278
penalized estimate 210 Preisach hysteresis model 278, 279,
Personalized Medicine xvi 281, 282, 283, 288, 290, 293,
petroleum chemistry 170 295
photo-ferroelectric (PFE) 321 Preisach hysteresis operator 278,
408 Mechatronics and Robotics

279, 295 Selective Compliance Assembly Ro-

Preisach weighting function 279, bot Arm 10, 16
282 sensor localization 207
preliminary experiments 374 sensor network 206, 207, 217, 219
pressure measurement 377, 378 sensors 66, 68
Preventive xvi service-oriented architecture (SOA)
Programmable Logic Controller 66, 19
68, 69, 79 service-oriented manufacturing
Proportional-Derivative (PD) 389 model 19
proportional-integral- derive (PID) Servo Motors 82, 96
388 shear forces 188
PSO optimization technique 123, Shock Wave 372
127, 128, 131, 135, 137 Slick-Slip 388
sliding mode-based observers
Q (SMOs) 298
Quality of Service (QoS) 45 Sliding Mode Control 387, 388, 402
Smart Homes 226
R solenoids 70
rapid advancement 66 solenoid valves 66
Rapidly-Exploring Random Tree spraying 82
(RRT) 122 Stanford Arm 2, 7
Related Research work 121 state of electrostatic shielding 195
reservoir plugging 254 static nodes 206, 209, 210, 211, 213,
resistance welding 82 214, 215, 216, 217
Robot Arm 81, 82 static sensor network 206, 207
robot arms 82, 83, 91, 99 statistic-learning algorithm 207
robotic 68 stereotype 53, 55
Robotics Toolbox 170, 181, 182 structural monitoring 207
Robot Navigation 205, 206 sulfate-reducing bacteria (SRB) 259
Robust Control 102 switches 66, 70
Routine Search 206 SysML Profile 44
rubbing force 188, 190, 191 System Modeling 44
run monitoring 67 Systems Engineering 44
Systems Modeling Language
S (SysML) 45, 48
safety critical mechatronic systems T
60
Sandbag 184, 199 Targeting 226
scientific literature 3 Target Tracking 226, 240
 Index 409

Tele-monitoring of Physical Data V

226
Validation 82
The First Three Modes 346
versatile transfer 4
The Internet of Things 19, 39
Vibration Control 102, 116, 117,
theoretical model 345, 346, 347,
118
365
Vicarm particularly suitable 7
Throat 371, 372
video image processing (VIP) 229
tmote node 216
virtual force field (VFF) 122
Track before Detect (TBD) 248
V-Model 44, 52, 55, 56, 58, 59, 62
Track Before Detect (TBD) 240,
247 W
Traditional sensor network 205
Trajectory Planning 170, 178, 182 wall static pressure fluctuations 371,
Transonic Diffuser 371, 372, 385 373, 379, 380, 381, 382
Two-Link Arm 120 Water Control System 66
water quality deterioration 253, 254
U Wireless Sensor network (WSN)
223
ultraviolet (UV) 322, 323
UML4IoT Profile 18 Z
unknown input observers (UIOs)
298 zero-order function 351
zirconate titanate 320, 321