FACULTY OF SCIENCE AND TECHNOLOGY

MASTER'S THESIS

Study program/specialization: Spring semester, 2016

Information Technology - Automation and
Signal Processing
Open / Confidential

Author: Ørjan Mæhre

…………………………………………
(signature author)

Instructor: Morten Mossige

Supervisors: Ståle Freyer and Karl Skretting

Title of Master's Thesis: Following Moving Objects Using Externally Guided Motion
(EGM).
Norwegian title: Følge objekter i bevegelse ved bruk av ‘’Externally Guided Motion’’ (EGM).

ECTS: 30

Subject headings: Externally Guided Motion Pages: 80

(EGM), ABB, Cognex, In-Sight, Robotics, + attachments/other: 18 + embedded file
Object Tracking, Ramp
Stavanger, June 15/2016
Date/year
This page is intentionally left blank.
Abstract

This thesis presents a sensor guided system using Externally Guided Motion (EGM)
to reduce the delay from new sensor data is available, to the robot initiates its
movement. A suggested solution to follow a disc sliding down a ramp is presented.
The position of the disc is registered by a camera and transmitted to a computer.
The data is processed on the computer before it is forwarded to the robot. Without
prediction the robot is approximately 200ms behind the disc. With prediction, the
robot is able to follow the disc. EGM has low latency, enabling the robot follow
an object moving with non-constant velocity. Even at velocities upwards to 2m/s.
Using EGM can give more eective robot cells and thereby reducing costs in the
manufacturing industry.

i
This page is intentionally left blank.
Preface

This paper entails a master's thesis carried out at the Department of Electrical
Engineering and Computer Science at the University of Stavanger. I would take
this opportunity to thank my supervisors Morten Mossige, Ståle Freyer and Karl
Skretting for valuable guidance and advice during this thesis. I would also like to
thank my fellow students for ve good years at the University of Stavanger.

iii
This page is intentionally left blank.
Contents

1 Introduction 1
1.1 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Robot Cell at the University of Stavanger . . . . . . . . . . . . . . . . 2
1.3 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Background 3
2.1 Externally Guided Motion (EGM) . . . . . . . . . . . . . . . . . . . . 3
2.1.1 Basic Approach . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1.2 Sensor Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.3 Disadvantages . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Vision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 Smart Camera . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.2 In-Sight Explorer . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.3 Edge Detection . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3 Physics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 Homogeneous Transformation . . . . . . . . . . . . . . . . . . . . . . 13

3 Using EGM to Follow Moving Objects 15

3.1 Virtual Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.1.1 Work Station . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.1.2 RAPID . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2 Real Time Sensor Data Processing . . . . . . . . . . . . . . . . . . . 18
3.2.1 Structuring of the Data, (EGM.cs) . . . . . . . . . . . . . . . . 18
3.2.2 Choosing which Position Data to use, (Program.cs) . . . . . . 18
3.2.3 Receiving Data From the Camera, (Camera.cs) . . . . . . . . 19
3.2.4 Predicting Future Positions of the Disc, (Predictor.cs) . . . 19
3.2.5 Constructing Data Packets, (Sensor.cs) . . . . . . . . . . . . 20
3.3 Vision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.4 Oine Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

4 Experiments, Results and Analysis 25

4.1 Tracing Capability of the Robot . . . . . . . . . . . . . . . . . . . . . 25
4.1.1 Constant Acceleration . . . . . . . . . . . . . . . . . . . . . . 25
4.1.2 Constant Velocity . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.2 Dierent Sample Rate . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.2.1 Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.2.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

v
vi CONTENTS

4.2.3 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.3 Ramp Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.3.1 Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.3.2 Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.3.3 Dierent Bandwidths on the Low Pass Filter . . . . . . . . . . 39
4.3.4 Movement in the X-direction . . . . . . . . . . . . . . . . . . . 41
4.3.5 Movement in all Directions . . . . . . . . . . . . . . . . . . . . 49
4.3.6 Analysis of the Ramp Experiments . . . . . . . . . . . . . . . 58

5 Discussion 59
5.1 Externally Guided Motion . . . . . . . . . . . . . . . . . . . . . . . . 59
5.2 Noise in the Discs Position . . . . . . . . . . . . . . . . . . . . . . . . 59
5.3 Friction Coecient . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

6 Conclusion and Future Work 61

6.1 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

Bibliography 63
A Experiment with Constant Acceleration A1
A.1 Without Smoothing . . . . . . . . . . . . . . . . . . . . . . . . . . . . A1
A.2 20 Degrees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A3
A.3 30 Degrees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A5
A.4 40 Degrees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A6
A.5 50 Degrees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A8
A.6 60 Degrees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A9

B Experiment with Constant Velocity B13

B.1 v100 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B13
B.2 v500 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B14
B.3 v1000 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B15
B.4 v1500 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B16
B.5 v2000 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B17
B.6 v2500 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B18

C Code C19
List of Figures

2.1 Transition between the states. . . . . . . . . . . . . . . . . . . . . . . 5

2.2 Block diagram of the Externally Guided Motion (EGM) control system. 6
2.3 Block diagram representation of the data ow using the UdpUc in-
terface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.4 The two programming environments in In-Sight Explorer. . . . . . . . 11
2.5 Forces acting on a object sliding down a ramp. . . . . . . . . . . . . . 13

3.1 The ow of position data in the system. . . . . . . . . . . . . . . . . 15

3.2 The virtual environment is setup to be as alike to the physical envi-
ronment as possible. . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.3 Overview of the simulated ramp is set up. . . . . . . . . . . . . . . . 17
3.4 The spreadsheet program used to detect the disc. . . . . . . . . . . . 21
3.5 Calibration environment after calibrating. . . . . . . . . . . . . . . . 22

4.1 Robot and discs distance and velocity. 20 degrees angle on the simu-
lated ramp. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2 Robot and discs distance and velocity. 60 degrees angle on the simu-
lated ramp. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.3 Physical setup experiment with constant velocity . . . . . . . . . . . 29
4.4 Distance and velocity with the disc travelling at 100mm/s. . . . . . . 30
4.5 Processing time camera, v100. . . . . . . . . . . . . . . . . . . . . . . 30
4.6 Distance and velocity with the disc travelling at 1000mm/s. . . . . . 31
4.7 Processing time camera, v1000. . . . . . . . . . . . . . . . . . . . . . 31
4.8 Distance and velocity with the disc travelling at 2500mm/s. . . . . . 32
4.9 Processing time camera, v2500. . . . . . . . . . . . . . . . . . . . . . 32
4.10 Distance travelled by robot and disc with two dierent sample rates,
disc velocity was 100mm/s . . . . . . . . . . . . . . . . . . . . . . . . 34
4.11 Distance travelled by robot and disc with two dierent sample rates,
disc velocity was 2500mm/s. . . . . . . . . . . . . . . . . . . . . . . . 34
4.12 Physical setup at the lab during testing. . . . . . . . . . . . . . . . . 36
4.13 Triangle with unknown angles. . . . . . . . . . . . . . . . . . . . . . . 37
4.14 Position in x-, y- and z- direction. The zoomed area is to highlight
the noise from the camera readings (red curve). . . . . . . . . . . . . 38
4.15 Noise after dierentiated the position to get the velocity and acceler-
ation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.16 Distance travelled with bandwidth low pass lter equal 3Hz . . . . . . 39
4.17 Distance travelled with bandwidth low pass lter equal 5Hz . . . . . . 40

vii
viii LIST OF FIGURES

4.18 Distance travelled with bandwidth low pass lter equal 10Hz . . . . . 40
4.19 Velocity in the x-direction with a 20 degrees angle on the ramp. . . . 42
4.20 No prediction, 20 degrees angle on the ramp. . . . . . . . . . . . . . . 43
4.21 Predicting 15 time steps ahead. Ramp angle was 20 degrees. . . . . . 43
4.22 Predicting 20 time steps ahead. Ramp angle was 20 degrees. . . . . . 44
4.23 Predicting 25 time steps ahead. Ramp angle was 20 degrees. . . . . . 45
4.24 Velocity in the x-direction with a 30 degrees angle on the ramp. . . . 46
4.25 No prediction, 30 degrees angle on the ramp. . . . . . . . . . . . . . . 47
4.26 Predicting 15 time steps ahead. Ramp angle was 30 degrees . . . . . 47
4.27 Predicting 20 time steps ahead. Ramp angle was 30 degrees. . . . . . 48
4.28 Predicting 25 time steps ahead. Ramp angle is 30 degrees . . . . . . . 49
4.29 Position and distance without prediction with a 20 degree angle on
the ramp. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.30 Velocities without prediction with a 20 degree angle on the ramp. . . 50
4.31 Position and distance predicting 15 time steps ahead. Ramp angle
was 20 degrees. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.32 Velocities predicting 15 time steps ahead. Ramp angle was 20 degrees. 51
4.33 Position and distance predicting 20 time steps ahead. Ramp angle
was 20 degrees. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.34 Velocities predicting 20 time steps ahead. Ramp angle was 20 degrees. 52
4.35 Position and distance predicting 25 time steps ahead. Ramp angle
was 20 degrees. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.36 Velocities predicting 25 time steps ahead Ramp angle was 20 degrees. 53
4.37 Position and distance without prediction with a 30 degree angle on
the ramp. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.38 Velocities without prediction with a 30 degree angle on the ramp. . . 54
4.39 Position and distance predicting 15 time steps ahead. Ramp angle
was 30 degrees. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.40 Velocities predicting 15 time steps ahead. Ramp angle was 30 degrees. 55
4.41 Position and distance predicting 20 time steps ahead. Ramp angle
was 30 degrees. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.42 Velocities predicting 20 time steps ahead Ramp angle was 30 degrees. 56
4.43 Position and distance predicting 25 time steps ahead. Ramp angle
was 30 degrees. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.44 Velocities predicting 25 time steps ahead. Ramp angle was 30 degrees. 57
List of Tables

2.1 Dierent states of the EGM process. . . . . . . . . . . . . . . . . . . 4

2.2 Basic approach for setting up EGM using an external device to give
the target of the movement. . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 Description of the dierent stages in the data ow using the UdpUc
interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.4 Example on how to set the parameters for UDPUC - communication 9
2.5 Summary of Cognex 5400 specications. . . . . . . . . . . . . . . . . 10

4.1 Dierence in time between the disc and robot at two given points,
when the distance travelled was equal to 270mm and 540mm . . . . . 27
4.2 Top velocity for the disc and robot. . . . . . . . . . . . . . . . . . . . 28
4.3 Dierence in time from the disc was in a certain point until the robot
reached the same point. . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.4 Dierence between Samplerate:=4 and Samplerate:=12. . . . . . . . 35
4.5 Dierent bandwidths on the low pass lter in the EGM controller. . . 41
4.6 Dierence between calculated velocity and measured velocity for the
disc in the x-direction. Ramp angle was 20 degrees. . . . . . . . . . . 42
4.7 Dierence in when the disc and robot reached the same point in the
x-direction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.8 Dierence in time between disc and robot had reached certain points. 49
4.9 Dierence in when the disc and robot reached the same point in all
directions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.10 Dierence in when the disc and robot reached the same point in all
directions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

ix
This page is intentionally left blank.
Code Listings

2.1 EgmHeader for the two main data structures, EgmRobot and EgmSensor. 7
2.2 Example of packet going from robot to sensor . . . . . . . . . . . . . 8
2.3 Example of a packet going from sensor to robot . . . . . . . . . . . . 9
3.1 EGMActPose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2 EGMRunPose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.3 Command to generate the EGM C# le . . . . . . . . . . . . . . . . 18
4.1 EGMActPose during ramp test . . . . . . . . . . . . . . . . . . . . . . 37
4.2 EGMRunPose during ramp test . . . . . . . . . . . . . . . . . . . . . . 37
4.3 Filtering the data using smooth . . . . . . . . . . . . . . . . . . . . . 38

xi
This page is intentionally left blank.
Acronyms

ABB Ltd. ABB robotics , Previously Asea Brown Boveri

C# C-Sharp

EGM Externally Guided Motion

FOV Field of View

POM Polyoxymethylene

ROI Region Of Interest

RRI Robot Reference Interface

RW RobotWare

TCP Tool Center Point

xiii
This page is intentionally left blank.
1. Introduction

Sensor guided robot paths become more and more usual. Where external sensors are
used to generate position data for one or several robots. As an initiative to reduce
the delay from when the robot recives new sensor data to the motion is started
ABB Ltd. have developed a new module with RobotWare (RW) 6 called Externally
Guided Motion (EGM). This module can read and write position data to the motion
system at a high rate. The robot can fetch new sensor data every 4 ms and can
1
start executing the movement within 20 ms . High rate and low delay is the main
advantage of EGM [1].
Reducing the time from new sensor data is available to the robot starts it motion
will enable robots to do more in less time. This will give more eective production
lines. More eective production can reduce costs in the manufacturing industry.
This thesis is trying to conclude if it is possible to follow an object moving in
two dimensions using EGM. A plastic disc will be used as an object. To restrain
the movement into two dimensions the disc will slide down a ramp. The velocity
is high and non -constant. The motion will be over within 1-2 s. Experiments is
conducted to test if EGM gives a quick enough response to follow the disc under the
given conditions.

1.1 Related Work

Since EGM is a new module there is no known previous work and no known pub-
lications on EGM. There have been done work with sensor guided paths. A sensor
guided system for repair of buoyancy module is presented in [2]. A vision system is
used to mark defect regions in the buoyancy module. This generates paths for the
robot to remove the defect areas.
In [3] a sensor guided system to be used in contour following is suggested. This
system uses multiple sensor. Combining vision, force and acceleration for contour
following tasks.
Luoa et al. [4] suggests an architecture for robotic seam tracing. They use o-
the self sensors to compensate the residual errors. The system were able to track
both linear and non-linear seams. TCP oset error were within ±0.1 mm with speed
equal to 100 mm/s.

1 Depending on robot type and robot conguration.

1
2 CHAPTER 1. INTRODUCTION

1.2 Robot Cell at the University of Stavanger

The robot cell which the testing has been preformed consists of two Irb 140 ABB
industrial robotic arms. Irb 140 is a six axis, small, multifunctional robotic arm.
Maximum Tool Center Point (TCP) velcocity is 2.5 m/s. Maximum TCP accelera-
2
tion is 20 m/s , with acceleration time 0-1 m/s equal to 0.15 s. Reach of 5th axis
is 810 mm [5]. To be compatible with RW 6 the CPU in the control cabined had to
be replaced.

1.3 Thesis Outline

Chapter 2 - Background:
Contains the theoretical foundation which this thesis is built on.

Chapter 3 - Using EGM to Follow Moving Objects:

Describes a suggested solution which enables an industrial robot arm with
EGM implemented to track a disc sliding down a ramp.

Chapter 4 - Experiments, Results and Analysis:

Presents all experiments conducted during this thesis. This chapter also
presents the results from these experiments and an analysis of the results.

Chapter 5 - Discussion:
Discussions around the implementation, results and problems which have oc-
curred during this thesis.

Chapter 6 - Conclusion and Future Work:

Gives the thesis conclusion based on the results from the experiments. Future
work is also presented.

Appendix A - Experiment with Constant Acceleration:

Contain remaining plots with results from the experiment conducted with
constant acceleration.

Appendix B - Experiment with Constant Velocity:

Holds the remaining plots with results from the experiments conducted with
constant velocity.

Appendix C - Code:
Gives an short introduction to the code which is embedded in this thesis.
2. Background

This chapter contains the theory used to perform the implementation and experi-
ments. EGM, Cognex In-Sight 5400 and necessary physics are some of the subjects
described in this chapter.

2.1 Externally Guided Motion (EGM)

This section is based on information found in [1]. EGM is a new module released by
ABB with RW 6 and this module gives a better response to external sensor input
1
[6]. EGM gives the possibility to adjust the the path of the robot within 10 - 20 ms
[1, p.313]. EGM goes right in to the motor reference generator, and by bypassing
the path planner the robot is able react quicker to new sensor data. All ltering,
supervision of reference and state handling is handled by EGM. In this section there
will be given an introduction to EGM and the EGM sensor protocol for UdpUc
communication. There is also possible to use analog communication between sensor
and robot, but this will not be covered in this thesis. Interessted readers can nd
out more about how to set up analog communication in [1].
The reference can either be given as joint values or as pose. Using pose there
is a need for reference frames. Measurements from the sensor and direction for
position changes can only be given relative to reference frames. Using joint values
there is no need for reference frames because both sensor values and position values
are axis angles given in degrees relative to the calibration position of each axis [1].
The setup using joint values is quite equal to pose, except for the need of reference
frames. Using joint values is covered in this thesis. Information on how to set up
the system using joint values can be found in [1].
In RW 6.x the option EGM gives access to:

ˆ Instructions which can be used to activate, set up and reset EGM.

ˆ Instructions which enables the user to initiate EGM movements and stop them.

ˆ Function which can retrieve current EGM state.

ˆ System parameters to congure EGM and assign these parameters default

values.

[1, p.314]

1 Depending on the robot type

3
4 CHAPTER 2. BACKGROUND

2.1.1 Basic Approach

With RW 6.x or higher EGM is already implemented in Robotstudio. Except for
setting up communication with the sensor there is no extra work that has needs to
be done to run EGM. Own instructions and functions to run EGM are embedded
in RAPID.
An EGM movement has to start in a ne point. Therefore it is not possible to
order parts of a movement to be conducted using EGM, e.g. have EGM to pop up
if something unexpected should happen.
The EGM process can have three dierent states: EGM_STATE_DISCONNECTED,
EGM_STATE_CONNECTED and EGM_STATE_RUNNING. The description of
these three states is given in table 2.1. The table is taken from [1, p.315].

Value Description
EGM_STATE_DISCONNECTED The EGM state of the specic process is
undened. No setup is active.

EGM_STATE_CONNECTED The specied EGM process is not activated.

Setup has been made, but no EGM move-
ment is active.

EGM_STATE_RUNNING The specied EGM process is running. The

EGM movement is active, i.e. the robot is
moved.

Table 2.1: Dierent states of the EGM process [1, p.315].

A basic approach on how to set up EGM using a external device i.e. a sensor to
give the target of the movement is shown in Table 2.2. This table is taken from [1,
p.315].
To move between the dierent states a set of specic RAPID instructions are
used. The transitions between the states can be found in Figure 2.1, this gure has
been taken from [1, p.316]. EGM can be run in RAPID by using three dierent
instructions . These three are: One  setup  instruction, one  Act  instruction and
one  Run  instruction. Two of these instructions are shown in Figure 2.1, but one
is left out. The  Act  instruction is where all the conditions of the movement is set,
like which coordinate system the movement is related to, the convergence criteria,
maximum admitted joint speed, etc. It is the  Act  instruction most of the choices
regarding the EGM movement are dened.
2.1. EXTERNALLY GUIDED MOTION (EGM) 5

Action
1 Move the robot to a ne point.

2 Register an EGM client and get an EGM identity. This identity is then used
to link setup, activation, movement, deactivation etc. to a certain EGM usage.
The EGM state is still EGM_STATE_DISCONNECTED.

3 Call an EGM setup instruction to set up the position data source

using UdpUc protocol connection. The EGM state changes to
EGM_STATE_CONNECTED.

4 Give the position convergence criteria, i.e. when the position is considered to
be reached.

5 Dene which frames are used to dene the target position and in which frame
the movement is to be applied.
6 Give the stop mode, an optional time-out and perform the movement itself.
Now the EGM state is EGM_STATE_RUNNING. This is when the robot is
moving.

7 The EGM movement will stop when the position is considered to be reached,
i.e. the convergence criteria is fullled. Now the EGM state has changed back
to EGM_STATE_CONNECTED.

Table 2.2: Basic approach for setting up EGM using an external device
to give the target of the movement. This table is taken from [1, p.315].

Figure 2.1: Transition between the states and which instructions to use
to move from one state to another. This gure is taken from[1, p.316].

A block diagram representing a simplied view of the EGM control system is

found in Figure 2.2. This gure is taken from [1, p.318]. There are two impor-
6 CHAPTER 2. BACKGROUND

tant parameters that inuence the EGM control loop, Default proportional Position
Gain and Default Low Pass Filter Bandwith Time [1]. Position gain inuences the
responsiveness moving to the target which the sensor points at, in relation to the
current robot position. The higher the value, the faster response. The low pass lter
lters the speed contribution from EGM [1].

EGM controller
Speed reference

Speed feed-forward + LP filter Servo control Robot

Position gain Sensor

Figure 2.2: Block diagram of the EGM control system. This gure is
taken from [1, p.318].

A block diagram of how the data ow is using the UdpUc protocol is found
in Figure 2.3 and Table 2.3 explain the dierent steps. Both are taken from [1,
p.317]. Something worth noticing is that in step 5. If the there is no new position
data available, motion control continuous to use the latest position data written by
EGM.

1 Motion control calls EGM

2 EGM reads feedback data from motion control

3 EGM sends feedback to the sensor

4 EGM checks the UDP queue for messages from the sensor

5 If there is a message, EGM reads the message and step 5 writes the position
data to motion control. If no position data have been sent, motion control
continues to use the latest position data previously written by EGM.

Table 2.3: Description of the dierent stages in the data ow using the
UdpUc interface. This table is taken from [1, p.317].
2.1. EXTERNALLY GUIDED MOTION (EGM) 7

1. Request (multiple of 4ms)

4. Check for
new position
2. Read feedback

Sensor EGM Motion control

Send position
(UDP 4 ms) 5. Write position

3. Send feedback

Figure 2.3: Block diagram representation of the data ow using the Ud-
pUc interface. This gure is taken from [1, p.317].

2.1.2 Sensor Protocol

The protocol is designed for high speed communication between sensor and robot
controller. The protocol is using Google Protocol Buers for encoding and UDP as
a transport protocol. The sensor protocol data structures is dened by the EGM
proto le [1]. When this le is compiled it generates serialized/de-serialized code
which can be used by the application. The robot controller and sensor uses a server
- client setup to communicate, where the sensor acts like the server. This means
that the robot controller has to initiate the communication. The sensor can not send
anything before it has received a rst message from the robot controller. After this
rst message data can ow independently of each other in both directions [1]. There
is no special connect / disconnect message and there are no built in synchronization,
request responses or supervision of lost messages. There is also no supervision over
the queue of the receiver. This means that the sender will keep sending out data
and it is the receivers job to make sure that the queue is emptied. The robot will
send and receive data every four millisecond by default. This cycle time can be
changed by using the optional argument \SampleRate of the RAPID instructions
EGMActJoint or EGMActPose[1].
In the protocol there are two main data structures, EGMRobot, which is sent from
the robot and EGMSensor, which is sent from the sensor. The header is the same for
both data structures and is shown below in Listing 2.1[1].

1 message EgmHeader
2 {
3 optional u i n t 3 2 s e q n o = 1 ; // S e q u e n c e number
4 optional u n i t 3 2 t m = 2 ; // Time stamp i n m i l l i s e c o n d s
5
6 enum M e s s a g e T y p e
7 {
8 MSGTYPE_UNDEFIND = 0;
9 M S G T Y P E _ C O M M A N D = 1 ; // For f u t u r e u s e
10 M S G T Y P E _ D A T A = 2 ; // S e n t by r o b o t c o n t r o l l e r
11 M S G T Y P E _ C O R R E C T I O N = 3 ; // S e n t by s e n s o r
12 }
13
14 optional MessageType mtype = 3 [ default = MSGTYPE_UNDEFINED ] ;
15 }

Listing 2.1: EgmHeader for the two main data structures, EgmRobot and
EgmSensor[1].
The packet being sent from sensor to robot contains less data than the packet
going the opposite way, See Listings 2.3 and 2.2. From these code snippets it clear
8 CHAPTER 2. BACKGROUND

that the robot exchange more information about its state than the sensor. This is
natural, since the sensor only tells the robot where to go, but it is good to know
for the sensor what the position of the robot is, the next planned position and that
everything is working properly.

1 header {
2 s e q n o : 385
3 tm : 4275119
4 mtype : MSGTYPE_DATA
5 }
6 feedBack {
7 joints {
8 j o i n t s : −1.1714538335800171
9 j o i n t s : 0.82521051168441772
10 j o i n t s : 0.48566141724586487
11 j o i n t s : −0.016466835513710976
12 j o i n t s : 0.27424958348274231
13 j o i n t s : −2.6035432815551758
14 }
15 cartesian {
16 pos {
17 x : 166.29362487792969
18 y : −396.09942626953125
19 z : 49.759349822998047
20 }
21 orient {
22 u0 : 0.00016568797582294792
23 u1 : −0.6623578667640686
24 u2 : 0.74915015697479248
25 u3 : 0.0074828267097473145
26 }
27 }
28 }
29 planned {
30 joints {
31 joints : −1.1378754377365112
32 joints : 0.82992005348205566
33 joints : 0.46713745594024658
34 joints : −0.013733416795730591
35 joints : 0.28822517395019531
36 joints : −2.5726895332336426
37 }
38 cartesian {
39 pos {
40 x : 166.29362487792969
41 y : −396.09942626953125
42 z : 49.759349822998047
43 }
44 orient {
45 u0 : 0.00020525175204966217
46 u1 : −0.66237151622772217
47 u2 : 0.74913811683654785
48 u3 : 0.0074862390756607056
49 }
50 }
51 }
52 motorState {
53 state : MOTORS_ON
54 }
55 mciState {
56 state : MCI_RUNNING
57 }
58 mciConvergenceMet : false
59 testSignals {
60 signals : 0
61 signals : 0
62 signals : 0
63 signals : 0
64 signals : 0
65 signals : 0
66 }
67 rapidExecState {
68 state : RAPID_RUNNING
69 }

Listing 2.2: Example of packet going from robot to sensor

2.1. EXTERNALLY GUIDED MOTION (EGM) 9

1 header {
2 s e q n o : 520
3 tm : 278273743
4 mtype : MSGTYPE_CORRECTION
5 }
6 planned {
7 cartesian {
8 pos {
9 x : 250
10 y : −400
11 z : 50
12 }
13 orient {
14 u0 : 0
15 u1 : 0
16 u2 : 0
17 u3 : 0
18 }
19 }
20 }

Listing 2.3: Example of a packet going from sensor to robot

To congure the UDP - device the following parameters has to be in order; name,
type, serial port, remote address and remote port number, see Table 2.4 These
parameters are set in Robotstudio under Controller - Conguration Editor -
Transmission Protocol.
Name Type Serial port Remote address Remote port
number
EGMsensor UDPUC N/A Computers IP-address 6510

Table 2.4: Example on how to set the parameters for UDPUC - commu-
nication [1].

2.1.3 Disadvantages
Since EGM by passes the path planner one can not expect a linear movement or
order a movement to take a specic time or with a specic speed. Other limitations
are:

ˆ Can only be used on six axis robots.

ˆ Can only be used in RAPID tasks with a robot, i.e. it is not possible to use it
in a task that contains only additional axis, i.e. in robtargets there are values
in the pose portion of the data.

ˆ An EGM movement has to start in a ne point.

ˆ For each robot a maximum of four external devices may be used to dene
position data. It is the important that each data component has one and only
one source. This is not checked by EGM.

ˆ It is not possible to perform a linear movement using EGM, due to the lack
of interpolation functionality. The actual path of the robot will depend in the
robot conguration, the start position, and the generated position data.

ˆ EGM does not support MultiMove.

10 CHAPTER 2. BACKGROUND

ˆ It is not possible to use EGM to guide a mechanical unit in a moving work

object.

These limitations are listed in [1, p.314].

2.2 Vision
As an external sensor a Cognex In-Sight 5400 smart camera is used to locate and
produce the position data of the disc. This camera communicates with the PC
through Ethernet and uses Udp as transport protocol.

2.2.1 Smart Camera

In-Sight 5400 is a smart camera produced by Cognex. The camera has a 640x480
2
resolution and can acquire new images with a frame rate up to 60 frames/second
[7]. This is a rigid camera with an IP67 grading, which means that the camera
can be deployed in rough environments. What separates a Cognex smart camera
from regular cameras is that all the image processing happens inside the camera.
Instead of outputting an image, the camera can output a string with the position
data. Cognex has its own library of nished algorithms that can be used, varying
from object tracking algorithms to mathematical- and logic- tools [7].

Resolution: 640 x 480

IP -grading: 67

Color: Greyscale

Frame rate: 60 fps

Table 2.5: Summary of Cognex 5400 specications [8].

2.2.2 In-Sight Explorer

Cognex has its own GUI based programming software called In-Sight Explorer. The
user can choose between two environments, Easy builder, a step by step environment
and Spreadsheet which is a Excel like, drag and drop environment, see Figure 2.4. An
Easy builder solution eventually get converted to spreadsheet code. In spreadsheet
the user can drag and drop desired tools into empty cells and ll in the parameters
specied for the particular tools.
Under the spreadsheet is the image produced by the camera. The spreadsheet
can be hidden by pressing the spreadsheet button in tool bar. This enable the user
to view the image captured with or without overlays. Typical overlays can be an
axis cross on the object which is to be recognize or Region Of Interest (ROI). These

2 Acquisition rate is based on minimum exposure, and full image frame capture.
2.2. VISION 11

(a) Easy Builder.

(b) Spreadsheet.

Figure 2.4: The two programming environments in In-Sight Explorer.

Easybuilder is the simplest of the two where the camera can be pro-
grammed by following the steps under Application steps. Spreadsheet
is a more drag and drop environment where tools are dragged from the
menu on the right and dropped into empty cells.

overlay can be hidden, this is done in the settings of the tool which produce the
overlay, e.g. in the Patterns tool.
If a new camera gets connected to the network and the PC have not used this
camera before it has to be added to the sensor network by In-Sight Explorer. Unless
this is done the camera is not be displayed in In-Sight Explorer.
The camera is added by going into the Add Sensor/Device to Network option
under the System menu. This will display a list over devices which are connected
to the same network, but is not connected into the In-Sight Network. The user can
here do some changes to the camera, such as name and IP-address.
12 CHAPTER 2. BACKGROUND

Cameras already connected to the In-sight network can be found in the down
left corner in Easy builder or in the left column in Spreadsheet, see Figure 2.4.

2.2.3 Edge Detection

Edges are detected by examining ROI for discontinuities in the pixel polarity. The
ROI is scanned in a systematic manner, usually line by line starting at the top.
Tracking one edge pixel to a neighbouring one and circumnavigating regions is called
edge following. Binary images is required to do this [9]. A boundary is determinated
by nding one pixel, thereafter the boundary is tracked by passing from one pixel to
one of its neighbouring pixels until a junction or an endpoint is found. Tracking must
then return to the start point and the boundary is tracked in the reversed direction
[9]. The edge strength is measured, typical the gradient magnitude. A threshold
value is used to decide if whether edges are present or not at an image point. To
link edges together and avoid marking an edge at every noisy pixel thresholding
with hysteresis can be used. This uses the assumption that edges are likely to be in
a continuous curve. The upper threshold is used to nd the start of an edge. From
the start point the edge is traced, marking an edge when the value is over the lower
threshold. The marking is stopped when the value of a pixel is below the lower
threshold [10].

2.3 Physics
This section is based on information found in [11]. Dierent forces are acting on an
object sliding down a ramp, see Figure 2.5. If the friction is constant throughout
the motion the object will slide down the ramp with constant acceleration. The
distance and velocity can than be found by applying the equations of motion for
constant acceleration, see Equations 2.1, 2.2, 2.3 and 2.4[11]. Where v is velocity, v0
is start velocity, a is acceleration, d is distance and t is time. These are to be used
when the motion is in a straight line. In cases where the motion is in a 2D-plan the
motions can be decomposed and the equations applied in each direction.

v = v0 + a · t (2.1)
v0 + v
d = ·t (2.2)
2
1
d = v0 · t + a · t2 (2.3)
2
v 2 − v02 = 2 · a · d (2.4)
2.4. HOMOGENEOUS TRANSFORMATION 13

FF FN

mg
· si
n(θ

)
s(θ
)
· co
mg

g θ

Figure 2.5: Forces acting on a object sliding down a ramp.

To nd the Euclidean distance between two points in 2D, Equation 2.5 can be
used. ∆d is the combined distance in both x- and y-direction.

p
∆d = ∆x2 + ∆y 2 (2.5)

The velocity is found by dierentiate the position with respect to time and the
acceleration is the second dierentiate of the position with respect to time. To nd
the velocity and acceleration between two points in 2D, Equations 2.6 and 2.7 can
be used [11].

p
∆v = ∆vx2 + ∆v y 2 (2.6)

p
∆a = ∆a2x + ∆ay 2 (2.7)

Friction force is dened by Equation 2.8, where µs is the friction coecient

between the object and the surface[11].

FF = µs mg(cos(θ)) (2.8)

2.4 Homogeneous Transformation

This section is based on information found in [12]. If the same point in space is
viewed from two dierent coordinate systems it is necessary to transform the point
to a common coordinate system. This is to be able to compare the two views, i.e
the same point in space is viewed from the robots - and the cameras - coordinate
system. This transformation is done by using rotation and translation matrices, see
Equations 2.9 and 2.10. When rotating it is important to keep in mind the order in
which the rotation matrices are multiplied, since this will aect the product [12].
14 CHAPTER 2. BACKGROUND

RXY Z = Rz,φ Ry,θ Rx,ψ

   
cos(φ) −sin(φ) 0 cos(θ) 0 sin(θ) 1 0 0
= sin(φ) cos(φ) 0  0 1 0  0 cos(ψ) −sin(ψ)
0 0 1 −sin(θ) 0 cos(θ) 0 sin(ψ) cos(ψ)
(2.9)

The displacement vector is the displacement between the origin of the two coor-
dinate systems in Euclidean space.

 
vx
d = vy 
 (2.10)
vz
Combining Equations 2.9 and 2.10 gives the homogeneous transformations ma-
trix which is used to transform points from one coordinate system to another, see
Equation 2.11.

 
R d
H= ; R ∈ SO(3), d ∈ R3 (2.11)
0 1
3. Using EGM to Follow Moving

Objects

This chapter describes how the system is built up to enable a robot with EGM
implemented to follow a disc sliding down a ramp. The communication is set up
like client-server communication, where the PC in both links acts like the server.
All communication happens over Ethernet with UDP as the message protocol.

Figure 3.1: The ow of position data in the system. There is a server -
client communication between camera - PC and PC - Egm controller.

15
16 CHAPTER 3. USING EGM TO FOLLOW MOVING OBJECTS

3.1 Virtual Environment

The main focus in this section is how the solution is implemented in Robotstudio.
This covers both the RAPID code and how the virtual environment is set up so to
be used for simulations. RobotStudio 6.02 and RW 6.01 were used throughout the
whole thesis.

3.1.1 Work Station

The virtual environment is set up like the lab at the University of Stavanger. With
two IRB 140 ABB robotic arms and a conveyor placed between them. The ramp is
not implemented, and there is no disc. The basis of this environment is made by
Karl Skretting. There are done some small changes in this thesis.

Figure 3.2: The virtual environment is setup to be as alike to the physical

environment as possible.

3.1.2 RAPID
In RAPID there is a program which sets up EGM and asks for new positions every
X ms.1 To enable the same rotation on all robtargets, workobjects and poses,
OrientZYX is used. This takes Euler angles as input and outputs an orient type
variable which is expressed in the form of quaternion. The x-axis is located along
the ramp, y-axis is across the ramp and z-axis is orthogonal to the ramp, see Figure
2
3.3. The coordinates for these data types are dened in world coordninates .
The workobjects x- and y-axis have the same crossing as the cameras coordinate
system. The z axis has its origin placed above the ramp. This means that the robot
moves in a plane above and parallel to the ramp. Pose is used, this means that
there is need for a reference system which all movements are relative to. The pose
data type is named posecorTopOfRamp and has the same origin and rotation as the
workobject.

1X is equal to the processing time of the camera. This is set manually in Samplerate under
EgmRunPose and depend on the camera setup.
2 Which has its origin in the robots base.
3.1. VIRTUAL ENVIRONMENT 17

Top view Side view

Y v

X
Z

Figure 3.3: Overview of the simulated ramp is set up.

The robot is moved to a fine point, named startPoint. When the robot reaches
this point, if EGM is not connected, the program sets up EGM using EGMSetupUC.
Then EGM is run by using EGMActPose and EGMRunPose, see Listings 3.1 and 3.2.

1 EGMActPose e g m I D 1 \ T o o l := t S u c t i o n C u p \ W O b j := T o p O f R a m p , p o s e c o r T o p O f R a m p ,
2 EGM_FRAME_WORLD, p o s e c o r T o p O f R a m p , EGM_FRAME_WORLD
3 \ x := e g m _ m i n m a x _ l i n 1 \ y := e g m _ m i n m a x _ l i n 1 \ z := e g m _ m i n m a x _ l i n 1
4 \ r x := e g m _ m i n m a x _ r o t 1 \ r y := e g m _ m i n m a x _ r o t 1 \ r z := e g m _ m i n m a x _ r o t 1
5 \ L p F i l t e r :=3\ S a m p l e r a t e :=12\ M a x S p e e d D e v i a t i o n := 5 0 0 0 ;

Listing 3.1: EGMActPose

Some relevant information in Listing 3.1 are:

ˆ Both poses are in relation to world coordinates.

ˆ Convergence criteria, emg_minmax_lin1 = [-1,1] and

egm_minmax_rot1 = [-2,2] is set earlier in the program.

ˆ The low-pass lters bandwidth are set to 3Hz .

ˆ The robot should request new position data every 12th ms (Samplerate)

3
ˆ Maximum admitted joint speed change is set to 5000 degrees/second.

The bandwidth of the low-pass lter is set to a small value to avoid Error: 50375:
Dynamic load too high which occurs when the speed reference changes too fast.
Using a smaller bandwidth give a smoother transition from high speed to immediate
stop and vice versa. If the bandwidth is set too small this will give a overshoot.
Through experiments 3Hz were found to be a reasonable value which eliminate the
error and gives a minimum of overshoot.

3 This is set to a random high value, because trimming the acceleration is not desired
18 CHAPTER 3. USING EGM TO FOLLOW MOVING OBJECTS

1 EGMRunPose e g m I D 1 , EGM_STOP_HOLD\ x \ y \ z \ C o n d T i m e :=20

2 \ R a m p I n T i m e :=0.05 \ R a m p O u t T i m e :=0.5;

Listing 3.2: EGMRunPose

In EGMRunPose the CondTime is set to 20 s, which means that if the convergence
criteria is not within this time the program will continue to the next instruction.
This is never a issue in this program since the motion is done within a few seconds.
The RampInTime is how fast the movement is started in seconds and RampOutTime
has no meaning since Mode is set to EGM_STOP_HOLD [1].

3.2 Real Time Sensor Data Processing

All code is written in C-Sharp (C#), which is a programming language in the .Net
family. Several dierent classes are made and they are presented in separate sections
below. All the code is placed under the solution ExternalGuidedMotion.

3.2.1 Structuring of the Data, (EGM.cs)

This class is auto generated by ProtoGen. Protobuf -csharp binaries can be down-
loaded from [13]. The zip-le should be unpacked and placed in a suitable directory.
Windows console is opened in the tools directory and the command in Listing 3.3 is
written in the console window. This generated an EGM C# (Egm.cs) le from the
egm.proto le.

1 protogen .\ egm \ egm . proto -- proto_path =.\ egm

Listing 3.3: Command to generate the EGM C# le

From the NuGet packet manager protobuf -csharp and Google.ProtocolBuers

is installed in the solution.
The egm.proto le can be obtained from the PC or the IRC5 controller.

ˆ On the PC where RobotWare is installed:

%LocalAppData%\ABB Industrial IT\Robotics IT\RobotWare\
<RobotWare_xx.xx.xxxx> utility\Template\EGM\

ˆ In the IRC5 Controller:

<SystemName>\PRODUCTS\<RobotWare_xx.xx.xxxx>\utility\
Template\EGM\

Egm.proto denes the data contract between the sensor and the robot [1].

3.2.2 Choosing which Position Data to use,

(Program.cs)
This is where the Main method is located. The arguments in Main is used to decide
which class to be used to generate the position data, Camera.cs or SimDisc.cs.
3.2. REAL TIME SENSOR DATA PROCESSING 19

Camera is given as an argument to use the camera and to use the simulated disc
Simulate is given. The argument given decide what mode is set to and this sets
which methods to be run later in the program.

3.2.3 Receiving Data From the Camera, (Camera.cs)

This class takes in the x- and y- position of the disc. Data packets from the camera
are sent as strings. This class convert the data into double to be able to do
mathematical operations. The x- and -y positions are ltered using a Median lter.
This lters out noise peaks which may be present due to false accepts in the camera.
Filtered values are stored inX and Y respectively. The class calls the NewPrediction
method in the Predictor class with X and TimeStamp as parameters. All this
happens in its own thread to minimize delay.

3.2.4 Predicting Future Positions of the Disc,

(Predictor.cs)
This predictor uses previous readings to predict future positions. The current veloc-
ity is found by taking the displacement between two consecutive samples, divided by
the change in time, see Equation 3.1. To nd the acceleration it would be possible
to derive the velocity, but this were found to give unnecessary noise. Therefore a
model of the acceleration is used, see Equation 3.2. Were θ is the angle of the ramp
in radians, g is the gravitational acceleration in mm/ms2 and µ is the frictional
coecient.

xi − xi−1
v = (3.1)
∆t
a = g · sin(θ) − µ · g · cos(θ) (3.2)

From the experiments done in Chapter 4 the delay where found to be between
200 and 300 ms. The predictor therefore has to predict the discs position 200 - 300
ms ahead in time. The predicted value is found by using Equation 3.3. Where x
is current position in mm, v is velocity in mm/ms, t is the time step in ms, c is
a constant, a 2
is the acceleration in mm/ms and x̂ is the predicted value which
is stored in the variable PredictedPosition. The time step is the time between
two consecutive samples. The predictor must predict the position around 250 ms
4
ahead in time. Since the sampling frequency is about 12 ms , c is set to a constant
between 20 and 30. The frictional coecient, µ is set equal to 0.3. This value where
found trough the experiments in Chapter 4 and conrmed by [14] where 0.3 is in
the middle scope of the friction coecient for wood - wood and plastic - plastic.

1
x̂ = x + v · t · c + a · (t · c)2 (3.3)
2
4 The time between readings depend on where the camera is placed, light conditions and other
factors. Therefore this may vary from experiment to experiment.
20 CHAPTER 3. USING EGM TO FOLLOW MOVING OBJECTS

3.2.5 Constructing Data Packets, (Sensor.cs)

The skeleton of this class is made by ABB Ltd. in EgmSensorApp.cpp which is given
as an example application to EGM. This class constructs the data packets which is
being sent to the robots motion controller. It also handles the data packets coming
from the robot. Since there are two dierent places where position data can be gen-
erated, this class uses the argument in Main to decide if the position data is coming
from the camera or the simulated disc. Further in this section all data comes from
the camera. There is a own thread, SensorThread, which sets up the communica-
tion between PC and robot. This thread handles the requests from the robot and
updates the Cartesian coordinates to the latest predicted position estimate. When
the robots current position is received the latest position estimates is sent to the
robot. Estimates are updated by the method CameraSetPos. This method fetch
the predicted x-position from PredictedPosition in the Predictor class. The y-
position comes from Camera.Y. These values are used to set the Cartesian position
in the CreateSensorMessage method. This is were packet being sent to the robot
is created. The data packet is sent to the robot using UdpClient which is a class in
the .Net framework.

3.3 Vision
The vision program is created using the spreadsheet environment in In-Sight Ex-
plorer, see gure 3.4. In the image cell (which is located by default in A0) the
parameters is set to restrict the Field of View (FOV) of the camera to only cover
the ramp. Restricting the FOV gives faster recognition, reduces noise and false ac-
cepts. Only by restricting the FOV the processing time of the camera went down
from 16ms to about 12ms.5 The Expose time is also set to a minimum to keep the
processing time down to a minimum. Depending on the light conditions in the lab
the expose time was set to 1 or 2ms.
To recognise the disc, an edge detection tool located in Patterns is used. The
calibrated image transforms pixels into mm, this is done by the Calib tool located
in C16. The calib tool is set up to use a chequerboard with 20mm spacing without
a ducial, see Figure 3.5. Four poses are used to get an accurate calibration along
the whole ramp. The origin is the top right cross seen in Figure 3.5a. The x-axis is
dened along, the y-axis is across and the z-axis is perpendicular to the ramp.
Device located in C21 sets up the UDP communication between the camera and
the PC. Host Name is the computers IP-address, Port is which port to use, in this
case it is set to 3000. Timeout is set to a high value, because it is not desired that the
camera times out if there is no communication. Binary Packet is left unchecked.
Read in D21 points to the Device structure in C21. The Write cell points to
the image in A0, the device structure in C21, and the to the data structure in E22.
The data structure in E22 is the position in both x-and y-direction, timestamps and
sequence number in a String format. It is this data which is being sent to the PC.
The timestamps is the processing time of the camera. This is used in the predictor

5 Depending on the setup

3.3. VISION 21

and for debugging. Sequence number is only used for debugging. The position is in
mm and the timestamps are in ms, therefore the data is restricted to zero decimal
points since mm and ms gives an acceptable precision.

Figure 3.4: The spreadsheet program used to detect the disc. Behind the
spreadsheet live image of what is being recorded is found. In this image
it is possible to see the disc behind the spreadsheet, with an overhead
displaying the axis cross for the disc.

There are two events locate in D30 and D31. D30 is used to update the sequence
number for each iteration. D31 is used to reset the sequence number when the camera
goes from online to oine. So that each session starts with sequence number 0.
22 CHAPTER 3. USING EGM TO FOLLOW MOVING OBJECTS

(a) Calibrated image.

(b) Result after calibrating.

Figure 3.5: Calibration environment after calibrating. After calibrating

In-Sight Explorer gives a score on how good the calibration was. In this
case the score was 0.114 and this is dened as Excellent.

3.4 Oine Processing

A Matlab script was created to do oine processing. This script uses the position
and time data from the .txt -les generated by the C# program. Processing this
data creates plots of the position, velocity and acceleration with respect to time.
The position data from the robot is in relation to the base, and the position data
of the disc is related to the workobject. This script transforms the position data to
a common coordinate system, so the data from the two sources can be compared.
This transformation is done by rotation and translation matrices.
The distance that the disc and has travelled is found by using the Algorithm 1
(based on Equation 2.5).
3.4. OFFLINE PROCESSING 23

Algorithm 1 How to calculate distance in a 3D-space.

for i = 1: Step =
p 1: i ≤ length position of data do
distance[i] = (x0 − xi )2 + (y0 − yi )2 + (z0 − zi )2
end for

The velocity is found from the derivative of the position. First the derivative is
found in all three axis, and then to nd the total velocity an algorithm similar to
Algorithm 1 is used.
And last the acceleration is found by using the second derivative of the position.
As for the velocity a similar algorithm to Algorithm 1 is used to nd the combined
acceleration.
With these data in place it is possible compare travelled distance, velocity and
acceleration for both disc and robot.
This page is intentionally left blank.
4. Experiments, Results and

Analysis

This chapter describes the setup for all experiments conducted during this thesis.
Results and an analysis of the results are also given. All experiments were conducted
in the robotics lab at the University of Stavanger, see Section 1.2.

4.1 Tracing Capability of the Robot

These two experiment were done to test the tracing capability of the robot. This was
to test how far the robot was behind the disc with constant velocity and constant
acceleration. Results from these experiments gave a pinpoint on how far ahead the
predictor had to predict.

4.1.1 Constant Acceleration

The purpose of this experiment was to see how well the robot could follow the path
of a simulated disc. The disc moved with constant acceleration. Position data for
both disc and robot were stored in a .txt - le for oine processing. This experiment
was conducted to try and answer the following questions:

ˆ Can the robot follow an object with non- constant velocity?

ˆ Is the dierence in the robot and disc's position constant throughout the mo-
tion?

ˆ Will the dierence in the robot and disc's position be dierent with high
acceleration versus low acceleration?

Dierence in position means how long after the disc have travelled a certain
distance have the robot travelled the same distance, or visa versa.

Setup
The disc was created in C#. This disc slides frictionless down a ramp according
to the laws of physics, see Equations 4.1, 4.3 and 4.2. The robot tried to copy the
1
motion of the disc. The robots start point was [496, -526, 650]. . The disc moved

1 In world coordinates

25
26 CHAPTER 4. EXPERIMENTS, RESULTS AND ANALYSIS

in the work objects x-direction. The movement was started by pressing s in the
console window. This started the timer which was used to update the x-position
of the disc and assigned the y-position to a constant equal −125mm. Before s
were typed in the position data were set to [0,0,0] which gave the origin of the work
object as the target.

a = g · sin(θ) (4.1)

Where g is the gravitational force and θ is the angle of the ramp.

v =a·t (4.2)

1
s= ·v·t (4.3)
2
Five dierent angles were tested, 20, 30, 40, 50 and 60 degrees. Distance, velocity
and acceleration were calculated using Algorithm 1, found in Section 3.4.

Results
This section presents the results from the experiments done with 20 degrees and
60 degrees. Plots from the remaining experiments can be found in Appendix A.
This appendix also holds plots displaying the movement and velocity in all three
directions. A summary of the results can be found in Tables 4.1 and 4.2.
By setting the simulated ramps angle to 20 degrees, the theoretical acceleration
for the disc is 3.36 · 10−3 mm/ms2 , which gives a top velocity equal to 2.78m/s. The
whole motion took 828ms. Measured top velocity for the disc was 2.29m/s. The
disc had a higher top velocity than the robot, the dierence was 0.56m/s, see Figure
4.1b. The disc pulled away from the robot during the motion, but not by much.
At distance equal 270mm the robot was 190ms behind the disc. At distance equal
540mm the dierence was 201ms, see Figure 4.1a.

Distance disc Distance robot Velocity disc Velocity robot

900 2.5

800

2
700
Velocity [mm/ms]

600
Position [mm]

1.5
500

400
1
300

200
0.5

100

0 0
2000 2500 3000 3500 4000 4500 5000 5500 6000 2000 2500 3000 3500 4000 4500 5000 5500 6000
Time [ms] Time [ms]

(a) Distance. (b) Velocity.

Figure 4.1: Robot and discs distance and velocity. 20 degrees angle on
the simulated ramp.
4.1. TRACING CAPABILITY OF THE ROBOT 27

8.50·10−3 mm/ms2 ,
Angle equal to 60 degrees will give a theoretical acceleration equal
which gives a top velocity equal to 4.84m/s. The whole motion was over after 569ms.
Measured top velocity for the disc was 3.48m/s, and for the robot 2.21m/s, see Fig-
ure 4.2b. Dierence in top velocity was 1.27m/s. During the motion the disc pulled
away from the robot. At distance equal 270mm the dierence in time was 183ms
and at 540mm the robot was 209ms behind the disc, see Figure 4.2a.

Distance disc Distance robot Velocity disc Velocity robot

900 3.5

800
3
700
2.5

Velocity [mm/ms]
600
Position [mm]

500 2

400 1.5

300
1
200
0.5
100

0 0
7000 7500 8000 8500 9000 9500 10000 10500 11000 11500 7000 7500 8000 8500 9000 9500 10000 10500 11000 11500
Time [ms] Time [ms]

(a) Distance. (b) Velocity.

Figure 4.2: Robot and discs distance and velocity. 60 degrees angle on
the simulated ramp.

Angle 270 [mm] 540 [mm]

20 190 [ms] 201 [ms]

30 189 [ms] 203 [ms]

40 186 [ms] 203 [ms]

50 184 [ms] 205 [ms]

60 183 [ms] 209 [ms]

Table 4.1: Dierence in time between the disc and robot at two given
points, when the distance travelled was equal to 270mm and 540mm

Analysis
The peaks at the start of the velocity plots were because the y- position was set
to −125mm. Dierence in theoretical and measured top velocity can be due to the
smoothing of the plots. Example of unltered plots are found in Appendix A. The
robot is able to follow the disc. Distance between disc and robot were not constant.
From the numbers it looks like the acceleration aected how far behind the robot
was. Already at 30 degrees the discs top velocity were higher than the robots max
TCP velocity. This is why the robot fell farther and farther behind the disc as the
ramps angle was increased.
28 CHAPTER 4. EXPERIMENTS, RESULTS AND ANALYSIS

Top velocity
Angle Disc Robot
20 2.29 [m/s] 1.73 [m/s]

30 2.70 [m/s] 1.90 [m/s]

40 3.07 [m/s] 2.06 [m/s]

50 3.27 [m/s] 2.14 [m/s]

60 3.48 [m/s] 2.21 [m/s]

Table 4.2: Top velocity for the disc and robot.

4.1.2 Constant Velocity

The purpose of this experiment was to see if the robot could follow an object mov-
ing with constant velocity. The experiment was conducted to try and answer the
following questions:

ˆ Can the robot follow an object moving with constant velocity?

ˆ Is the dierence in the discs and robots position constant throughout the
motion?

ˆ Does dierent velocities inuence the distance between disc and robot?

ˆ What is the processing time of the camera? And is this aected by the objects
velocity?

Setup
This test were done with the setup seen in Figure 4.3. The robot on the right had
a disc mounted to its tool. This robot moved linear and with constant velocity over
the white area seen in Figure 4.3. This robot was programmed to move to a point,
wait for 2s before returning to the start point with the same velocity. Looking down
on the white area was a Cognex 5400 smart camera. This camera was set up to
locate the disc and send its position over to a program in C#. The PC forwarded
the position data to the robot on the left. This robot copied the discs motion. The
disc was set up to move only in the x-direction (there were some movement in the
y-direction as well due to inaccuracy in the calibration). There were no movement
in the z-direction of the disc. Six dierent velocities were tested, from 100mm/s to
2500mm/s.
The camera used an edge detection algorithm to locate the disc. This algorithm
is built into the Cognex In-Sight Explorer software. A detailed description on how
the camera was programmed is found in Section 2.2.2.
4.1. TRACING CAPABILITY OF THE ROBOT 29

Figure 4.3: Physical setup experiment with constant velocity. The robot
on the right moved the disc linear and with constant velocity in the FOV
of the camera. The robot on the left copied the discs motion.

Results
The plots presented in this section is for the velocity equal 100mm/s, 1000mm/s and
2500mm/s. The remaining plots are found in Appendix B. Table 4.3 summarizes
the results for all velocities tested.
In all results there were a lot of noise. Reason for this noise is discussed in
Section 5.2. All data presented is smoothed using smooth in Matlab.
When the discs velocity was 100mm/s the robot was approximately 200ms be-
hind the disc, except for in the acceleration and deceleration phase, see Figure 4.4.
How far behind the robot was compared to the disc was approximately constant
throughout the motion. 300mm the robot
When distance travelled was equal to
was 209ms behind the disc and at 700mm the dierence was 210ms, see Figure 4.4.
The cameras processing time was not constant. 1032 out of 3461 samples the camera
used 9ms and the remaining 2429 samples the camera used 10ms, see Figure 4.5.
30 CHAPTER 4. EXPERIMENTS, RESULTS AND ANALYSIS

Distance disc Distance robot Velocity disc Velocity robot

1000 0.14

0.12
800
0.1

Velocity [mm/ms]
Position [mm]

600
0.08

0.06
400

0.04
200
0.02

0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
4
Time [ms] ×10 Time [ms] ×10 4

(a) Distance. (b) Velocity.

Figure 4.4: Distance and velocity with the disc travelling at 100mm/s.

Time delay, number of samples: 3461

2500

2000
Number of occurences

1500

1000

500

0
9 10
Delay[ms]

Figure 4.5: Processing time camera, v100.

With disc velocity equal to 1000mm/s the robot was approximately 200ms be-
hind the disc throughout the motion, except during acceleration and deceleration.
This dierence was constant. When distance travelled was 300mm the robot was
208ms behind the disc and at 700mm the dierence was 207ms, see Figure 4.6. The
cameras processing time was not constant. 344 out of 1246 samples the camera used
9ms and the remaining 902 samples it used 10ms, see Figure 4.7.
4.1. TRACING CAPABILITY OF THE ROBOT 31

Distance disc Distance robot Velocity disc Velocity robot

1000 1.2

1
800

Velocity [mm/ms]
0.8
Position [mm]

600

0.6

400
0.4

200
0.2

0 0
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Time [ms] Time [ms]

(a) Distance. (b) Velocity.

Figure 4.6: Distance and velocity with the disc travelling at 1000mm/s.

Time delay, number of samples: 1246

1000

900

800

700
Number of occurences

600

500

400

300

200

100

0
9 10
Delay[ms]

Figure 4.7: Processing time camera, v1000.

With the discs velocity equal to 2500mm/s, which is the top velocity of the robot
[5], the dierence from when the disc had travelled a certain distance until the robot
had travelled the same distance was not constant. At 300mm the robot was 175ms
behind and at 700mm the dierence was 233ms, see Figure 4.8. The processing
time of the camera was not constant. 281 out of 1274 samples the camera used 9ms
and 993 samples 10ms, see Figure 4.9.
32 CHAPTER 4. EXPERIMENTS, RESULTS AND ANALYSIS

Distance disc Distance robot Velocity disc Velocity robot

1000 2.5

800 2

Velocity [mm/ms]
Position [mm]

600 1.5

400 1

200 0.5

0 0
2000 3000 4000 5000 6000 7000 8000 9000 10000 11000 2000 3000 4000 5000 6000 7000 8000 9000 10000 11000
Time [ms] Time [ms]

(a) Distance. (b) Velocity.

Figure 4.8: Distance and velocity with the disc travelling at 2500mm/s.

Time delay, number of samples: 1274

1000

900

800

700
Number of occurences

600

500

400

300

200

100

0
9 10
Delay[ms]

Figure 4.9: Processing time camera, v2500.

Velocity 300 [mm] 700 [mm]

100 [mm/s] 209 [ms] 210 [ms]

500 [mm/s] 210 [ms] 210 [ms]

1000 [mm/s] 208 [ms] 207 [ms]

1500 [mm/s] 200 [ms] 208 [ms]

2000 [mm/s] 179 [ms] 227 [ms]

2500 [mm/s] 175 [ms] 233 [ms]

Table 4.3: Dierence in time from the disc was in a certain point until
the robot reached the same point.
4.2. DIFFERENT SAMPLE RATE 33

Analysis
The time from the disc was in a certain point to the robot reached that same point
was dependant of the discs velocity. Higher velocity meant that the robot was
further behind the disc compared with lower velocities. There was a lot of noise in
the samples of the discs position. This made it impossible to use unltered data.
There may therefore be some inaccuracy in the numbers presented due to smoothing.
The robot was able to follow an object travelling with constant velocity as long at
the velocity was under 2000mm/s. The dierence in the discs and robots position
was constant through out the motion with velocity ≤ 1500mm/s.
±1ms.
The cameras processing time was not constant varying with The cameras
processing time was not aected by the discs velocity. Approximately 1/3 of the
times the camera used 9ms and the remaining times it used 10ms.

4.2 Dierent Sample Rate

In RAPID it is possible to set how often the robot should request new position
estimates from the PC. This value is set as a parameter in EgmActPose and has to
be a multiple of 4. To nd out this aect the delay in the system, two dierent
values were tested. Samplerate:=4, since this is the smallest allowed value and
Samplerate:=12, since this is the multiple of 4 closest to the cameras processing
time.

4.2.1 Setup
The setup is identical to the setup in Section 4.1.2, only dierence is the value of
the parameter Samplerate in RAPID.

4.2.2 Results
In Section 4.1.2 Samplerate:=4 was used. Therefore the numbers in this section
was the same. Only results using Samplerate:=12 and the dierence between the
two sample rates is presented in this section.
When the discs velocity was 100mm/s the robot was 213ms behind the disc
when distance travelled was 300mm. When distance travelled was 700mm the
robot was 210ms behind, see Figure 4.10. This means that the dierence in us-
ing Samplerate:=4 and Samplerate:=12 was 4ms at 300ms and 0ms at 700mm.
34 CHAPTER 4. EXPERIMENTS, RESULTS AND ANALYSIS

Distance disc Distance robot

Distance disc Distance robot
1000
1000
900

800
800
700

Position [mm]
600
Position [mm]

600
500

400
400
300

200 200

100

0 0
0 0.5 1 1.5 2 2.5 3
0 0.5 1 1.5 2 2.5 3
Time [ms] ×10 4
Time [ms] ×10 4

(a) Samplerate:=12. (b) Samplerate:=4.

Figure 4.10: Distance travelled by robot and disc with two dierent
sample rates, disc velocity was 100mm/s

With disc velocity equal 2500mm/s the robot was 196ms behind the disc when
distance travelled was 300mm. When distance travelled was 700mm the robot
were 238ms behind, see Figure 4.11. This means that the dierence between
Samplerate:=4 and Samplerate:=12 with disc velocity equal 2500mm/s was 21ms
at 300mm and 5ms at 700mm.
Results from all velocities tested are found in Table 4.4.

Distance disc Distance robot

Distance disc Distance robot
1000
1000
900

800
800
700
Position [mm]

600
Position [mm]

600
500

400
400
300

200 200

100

0 0
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 11000
Time [ms] 2000 3000 4000 5000 6000 7000 8000 9000 10000 11000
Time [ms]

(a) Samplerate:=12 (b) Samplerate:=4

Figure 4.11: Distance travelled by robot and disc with two dierent
sample rates, disc velocity was 2500mm/s.
4.3. RAMP EXPERIMENTS 35

Velocity 300 [mm] 700 [mm]

100 [mm/s] 4 [ms] 0 [ms]

500 [mm/s] 5 [ms] 4 [ms]

1000 [mm/s] 5 [ms] 6 [ms]

1500 [mm/s] 7 [ms] 6 [ms]

2000 [mm/s] 18 [ms] 4 [ms]

2500 [mm/s] 21 [ms] 5 [ms]

Table 4.4: Dierence between Samplerate:=4 and Samplerate:=12.

With Samplerate:=12 the robot is 4 − 21ms further behind than with
Samplerate:=4.

4.2.3 Analysis
The results show that using Samplerate:=12 gives slower response than using
Samplerate:=4. Since the sample rate of the robot changed, the rate at which the
data was written to le were changed. The data is written every time the robot ask
for new positions. This may be a explanation for why there is an dierence. Since
the cameras processing time is 10ms, ±1ms, the rate at which the robot is fed with
new position data is not changed using the two sample rates. With Samplerate:=4
the robot revives new position data every third sample (every 12thms), which is
the same as using Samplerate:=12. When disc velocity equals 2000mm/s and
2500mm/s, the dierence is quite huge at the rst reading (300mm). This is most
likely a coincidence.

4.3 Ramp Experiments

This section presents all the experiments conducted on the ramp. The purpose of
these experiments was to see if the robot could follow a disc sliding down a ramp.
Dierent predictor values were tested. These experiments tried to nd an answer to
the following questions:

ˆ Can the robot follow a disc sliding down a ramp?

ˆ How far ahead does the predictor have to predict?

ˆ Is the distance between disc and robot constant throughout the motion?

4.3.1 Setup
During the ramp experiments the lab was setup like in Figure 4.12. The Cognex
camera was looking down at the ramp and most of the ramp were within the range
of the robot. Due to the limited work range of the Irb140 the start point of the
36 CHAPTER 4. EXPERIMENTS, RESULTS AND ANALYSIS

disc was moved down the ramp, by about 300 mm. By limiting the sliding surface
typical errors like 50449: Mechanical unit close to joint bound and 50456: Close
to singularity while External Guided Interface active were eliminated.

Figure 4.12: Physical setup at the lab. The camera was mounted so the
FOV covers the whole ramp, and the ramp was placed within the work
area of the robot.

The camera was placed on the opposite side of the ramp in relation to the robot.
With a small side view of the ramp to avoid the robot blocking the cameras view
of the disc. The camera was also placed high above the ramp to cover the whole
2
sliding surface.
The ramp consisted of wood, with aluminium edges to avoid the disc from sliding
out on one of the sides. Each level (the bolts in Figure 4.12) represented a 10 degrees
change in the ramps angle. The hight of these levels were calculated using the Sine
Rule, see Equation 4.4 and Figure 4.13. Due to friction the disc would not slide
with an angle of 10 degrees and therefore the application was not tested with a
ramp angle lower than 20 degrees.

a b c
= = (4.4)
sin(A) sin(B) sin(C)
2 The lens mounted on the camera had limited wide angle. Other lenses where not tested.
4.3. RAMP EXPERIMENTS 37

Figure 4.13: Triangle with unknown angles.

The disc was made out of Polyoxymethylene (POM) which is a hard types of
plastic [15]. Robot used during the tests was an Irb140 ABB industrial robotic arm,
with an working range of 810 mm [5].
Two dierent angles were tested, 20 and 30 degrees. During the experiments
the cameras processing time was 15ms when ramp angle was 20 degrees and 18ms
when the angle was 30 degrees. EGMActPose and EGMRunPose used the parameters
in Listing 4.1 and 4.2, unless otherwise stated. Due to the dierence in the cameras
processing time, Samplerate:= depended on the ramps angle. When the ramp angle
was 20 degrees, Samplerate was equal to 16, and 30 degrees it was set equal to 20.

1 EGMActPose e g m I D 1 \ T o o l := t S u c t i o n C u p \ W O b j := T o p O f R a m p , p o s e c o r T o p O f R a m p ,EGM_FRAME_WORLD,
2 p o s e c o r T o p O f R a m p , EGM_FRAME_WORLD \ x := e g m _ m i n m a x _ l i n 1 \ y := e g m _ m i n m a x _ l i n 1 \ z := e g m _ m i n m a x _ l i n 1
3 \ r x := e g m _ m i n m a x _ r o t 1 \ r y := e g m _ m i n m a x _ r o t 1 \ r z := e g m _ m i n m a x _ r o t 1 \ L p F i l t e r :=3\ S a m p l e r a t e :=16
4 \ M a x S p e e d D e v i a t i o n := 5 0 0 0 ;

Listing 4.1: EGMActPose during ramp test

1 EGMRunPose e g m I D 1 , EGM_STOP_HOLD\ x \ y \ z \ C o n d T i m e :=20 \ R a m p I n T i m e : = 0 . 0 5 \ R a m p O u t T i m e : = 0 . 5 ;

Listing 4.2: EGMRunPose during ramp test

4.3.2 Noise
In the position data from the camera there were some noise, see Figure 4.14. Samples
from the robot were not as noisy as the samples from the camera. The noise showed
better when the position was dierentiated to produce the velocity and even more
when the velocity was dierentiated to display the acceleration, see Figure 4.15.
38 CHAPTER 4. EXPERIMENTS, RESULTS AND ANALYSIS

600

400

200

Disc X position
Position [mm]

0 Robot X position
Disc Y position
Robot Y position
-200 Disc Z position
Robot Z position

-400

-600

-800
5000 6000 7000 8000 9000 10000 11000 12000
Time [ms]

Figure 4.14: Position in x-, y- and z- direction. The zoomed area is to

highlight the noise from the camera readings (red curve).

0.5 1.5

0 1
Acceleration [mm/ms 2 ]

-0.5 0.5
Velocity [mm/ms]

Disc X velocity Disc X acceleration

Robot X velocity Robot X acceleration
Disc Y velocity Disc Y acceleration
-1 0
Robot Y velocity Robot Y acceleration
Disc Z velocity Disc Z acceleration
Robot Z velocity Robot Z acceleration
-1.5 -0.5

-2 -1

-2.5 -1.5
5000 6000 7000 8000 9000 10000 11000 12000 4000 6000 8000 10000 12000
Time [ms] Time [ms]

(a) Velocity. (b) Acceleration.

Figure 4.15: Noise after dierentiated the position to get the velocity
and acceleration.

Due to the noise, all data presented were smoothed using the smooth in Matlab,
see Listing 4.3. 'loess' is local regression using weighted linear least squares and
a 2nd degree polynomial [16].

1 for i = 1: 1: 3
2 r o b o t X Y Z i ( 1 : end , i ) = s m o o t h ( r o b o t X Y Z i ( 1 : end , i ) , 2 0 0 , ' l o e s s ' ) ;
3 end

Listing 4.3: Filtering the data using smooth

4.3. RAMP EXPERIMENTS 39

4.3.3 Dierent Bandwidths on the Low Pass Filter

This experiment tested dierent bandwidths on the low pass lter which is between
the EGM controller and the robots motion controller, see Figure 2.2. This exper-
iment was conducted to see how dierent bandwidths inuenced the movement of
the robot.

Setup
This experiment was conducted with the setup described in Section 4.3.1. All tests
were run with a 30 degree angle on the ramp.

Results
With bandwidth equal 3Hz the robot started its motion 181ms after the disc. Mid-
way, when distance travelled was equal to 382mm, the robot was 253ms behind.
At the end, from the disc reached its stationary position to the robot stopped its
movement the dierence was 322ms, see Figure 4.16.

Distance disc Distance robot

900

800

700

600
Position [mm]

500

400

300

200

100

0
5000 6000 7000 8000 9000 10000 11000 12000
Time [ms]

Figure 4.16: Distance travelled with bandwidth low pass lter equal 3Hz .

When the bandwidth was equal to 5Hz , the robot started its motion 202ms after
the disc. Midway the dierence was 250ms and at the end 444ms, see Figure 4.17.
40 CHAPTER 4. EXPERIMENTS, RESULTS AND ANALYSIS

Distance disc Distance robot

900

800

700

600

Position [mm]
500

400

300

200

100

0
1000 2000 3000 4000 5000 6000 7000
Time [ms]

Figure 4.17: Distance travelled with bandwidth low pass lter equal 5Hz .

When the bandwidth was equal to 10Hz the robot started its motion 100ms
after the disc. Midway the dierence was 240ms. At the end, the robot reached a
stationary position 680ms after the disc, see Figure 4.18.
A summary of the numbers are presented in Table 4.5.

Distance disc Distance robot

900

800

700

600
Position [mm]

500

400

300

200

100

0
0.8 0.9 1 1.1 1.2 1.3 1.4
Time [ms] 4
×10

Figure 4.18: Distance travelled with bandwidth low pass lter equal
10Hz .

Bandwidth equal 15Hz was also tested, but triggered a Dynamic load too high
when the robot moved from the start point to the start point of the disc.
4.3. RAMP EXPERIMENTS 41

Bandwidth [Hz] Start [ms] Midway [ms] End [ms]

3 181 253 322

5 202 250 444

10 100 240 680

Table 4.5: How long behind the robot is compared to the disc at certain
distances using dierent bandwidths on the low pass lter in the EGM
controller.

Analysis
The results show that the bandwidth of the low pass lter aects the deceleration
of the robot, higher bandwidth gives longer deceleration. From the numbers it may
look like it also aect the start of the motion, with a bandwidth equal to 10Hz the
robot started its motion faster than with the other bandwidths tested. Since the
robot started its motion slower using 5Hz compared with 3Hz , the quicker response
with 10Hz is most likely a coincident.

4.3.4 Movement in the X-direction

Most of the discs movement were in the work objects x-direction. In this section
the results in x-direction are presented. Dierent time steps in the predictor was
tested to nd the best t, see Section 3.2.4. Time step is the time between two
consecutive samples, i.e. 15 time steps means the time between two consecutive
samples multiplied with 15. Time behind is the time from when the disc had travelled
a certain distance until the robot had travelled the same distance or visa versa.
Distance travelled gives a good pinpoint on the distance between robot and disc.

20 Degrees Angle on the Ramp

With 20 degrees angle on the ramp the theoretical velocity for the disc after 0.5s in
the x-direction is equal to 0.29m/s, while the true velocity was measured to 0.34m/s,
see Figure 4.19.
42 CHAPTER 4. EXPERIMENTS, RESULTS AND ANALYSIS

Results

0.4
Disc
Robot
0.2
X: 6390
Y: 4.101e-05
0
Velocity [m/s]

-0.2

-0.4 X: 6890
Y: -0.334

-0.6

X: 7390
-0.8
Y: -0.6805

-1
5000 6000 7000 8000 9000 10000 11000
Time [ms]

Figure 4.19: Velocity in the x-direction with a 20 degrees angle on the

ramp. The cursors are placed at the start of the movement, 0.5 seconds
after the movement started and 1 second after the movement started.

The measured velocity was not equal to the theoretical velocity. The dierence is
presented in Table 4.6.

Time Theoretical Measured Dierence

0.5 sec 0.29[m/s] 0.34[m/s] 0.05[m/s]
1.0 sec 0.59[m/s] 0.69[m/s] 0.1[m/s]

Table 4.6: Dierence between calculated velocity and measured velocity

for the disc in the x-direction. Ramp angle was 20 degrees.

With no prediction the robot is behind the disc throughout the motion. 0.5s
into the motion the robot was 238ms behind the disc, after 1s the dierence was
241ms and 1.25s into the motion the dierence was 240ms. The dierence in the
robots and discs position was approximately constant throughout the motion, except
during acceleration and deceleration, see Figure 4.20.
4.3. RAMP EXPERIMENTS 43

500
Disc
400 Robot

300

200
Position [mm]

100

-100

-200

-300

-400
5000 6000 7000 8000 9000 10000 11000
Time [ms]

Figure 4.20: No prediction, 20 degrees angle on the ramp.

Predicting 15 time steps ahead the robot was behind the disc throughout the
motion. 0.5s in the motion the robot was 81ms behind the disc. 1s the dierence
was 66ms and after 1.25s the dierence was 60ms. How far the robot was behind
the disc were not constant, see Figure 4.21. There was an overshoot at the end of
the robots motion.

500
Disc
400 Robot

300

200
Position [mm]

100

-100

-200

-300

-400

-500
500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500
Time [ms]

Figure 4.21: Predicting 15 time steps ahead. Ramp angle was 20 degrees.
44 CHAPTER 4. EXPERIMENTS, RESULTS AND ANALYSIS

The robot caught up with the disc and pass it when predicting 20 time steps
ahead. 0.5s into the motion the robot was 42ms behind the disc. After 1s the
dierence was 12ms and 1.25s into the motion the robot had passed the disc and
was 16ms in front of the disc, see Figure 4.22. There was an overshoot at the end
of the robots motion.

500
Disc
400 Robot

300

200
Position [mm]

100

-100

-200

-300

-400

-500
4000 5000 6000 7000 8000 9000 10000
Time [ms]

Figure 4.22: Predicting 20 time steps ahead. Ramp angle was 20 degrees.

Predicting 25 time steps ahead the robot caught up with and passed the disc.
0.5s into the motion the robot was 10ms behind the disc. After 1s the robot had
passed the disc and was 31ms in front of the disc. 1.25s into the motion the robot
was 46ms ahead of the disc, see Figure 4.23. There was an overshoot at the end of
the robots motion.
A summary of all numbers presented in this section are found in Table 4.7.
4.3. RAMP EXPERIMENTS 45

500
Disc
400 Robot

300

200
Position [mm]

100

-100

-200

-300

-400

-500
5000 6000 7000 8000 9000 10000 11000 12000
Time [ms]

Figure 4.23: Predicting 25 time steps ahead. Ramp angle was 20 degrees.

Time steps in the predictor

Time 0 15 20 25
0.5 sec −238[ms] −81[ms] −42[ms] −10[ms]
1.0 sec −241[ms] −66[ms] −12[ms] 31[ms]
1.25 sec −240[ms] −60[ms] 16[ms] 46[ms]

Table 4.7: Dierence in when the disc and robot reached the same point
in the x-direction. Negative sign means the robot was behind and positive
means it was in front. Ramp angle was 20 degrees.

30 Degrees Angle on the Ramp

Changing the angle of the ramp to 30 degrees will give an theoretical velocity after
0.5s equal to 1.18m/s in the x-direction, measured value was equal to 1.06m/s, see
Figure 4.24.
46 CHAPTER 4. EXPERIMENTS, RESULTS AND ANALYSIS

Results

0.5
Disc
Robot

0
X: 7450
Y: 0.004037
Velocity [m/s]

-0.5

-1
X: 7950
Y: -1.049
-1.5

-2
5000 6000 7000 8000 9000 10000 11000 12000
Time [ms]

Figure 4.24: Velocity in the x-direction with a 30 degrees angle on the

ramp. The cursors are placed at the start of the movement and 0.5
seconds after the movement started

Without prediction the robot was behind the robot throughout the motion. How
far behind the robot were was not constant. 250ms into the motion the robot was
234ms behind. 500ms into the motion the dierence was 250ms and 750ms after
the motion had started the dierence was 255ms, see Figure 4.25.
4.3. RAMP EXPERIMENTS 47

500
Disc
400 Robot

300

200
Distance [mm]

100

-100

-200

-300

-400
5000 6000 7000 8000 9000 10000 11000 12000
Time [ms]

Figure 4.25: No prediction, 30 degrees angle on the ramp.

When predicting 15 time steps ahead the robot was behind the disc throughout
the motion. The dierence was not constant. 250ms the robot was 105ms
After
behind the disc. 500ms into the motion the dierence was 91ms and after 750ms
the dierence was 85ms, see Figure 4.26. There was an overshoot at the end of the
robots motion.

500
Disc
400 Robot

300

200
Distance [mm]

100

-100

-200

-300

-400

-500
1000 2000 3000 4000 5000 6000 7000 8000
Time [ms]

Figure 4.26: Predicting 15 time steps ahead. Ramp angle was 30 degrees
48 CHAPTER 4. EXPERIMENTS, RESULTS AND ANALYSIS

Predicting 20 time steps ahead the robot was behind the disc throughout the
motion. The dierence was not constant. After 250ms the robot was 82ms behind
the disc. 500ms into the motion the robot was 54ms behind and 750ms into the
motion the robot was 35ms behind, see Figure 4.27. There was an overshoot at the
end of the robots motion.

500
Disc
400 Robot

300

200
Distance [mm]

100

-100

-200

-300

-400

-500
6000 7000 8000 9000 10000 11000 12000 13000
Time [ms]

Figure 4.27: Predicting 20 time steps ahead. Ramp angle was 30 degrees.

When predicting 25 time steps ahead in time the robot caught up with and
passed the disc. After 250ms the robot was 72ms behind the disc. 500ms after the
movement started the dierence was 20ms and 750ms into the motion the robot
had passed the disc and was 2ms in front of the disc.
All numbers presented in this section are summarized in Table 4.8.
4.3. RAMP EXPERIMENTS 49

500
Disc
400 Robot

300

200
Distance [mm]

100

-100

-200

-300

-400

-500
5000 6000 7000 8000 9000 10000 11000
Time [ms]

Figure 4.28: Predicting 25 time steps ahead. Ramp angle is 30 degrees

Time 0 15 20 25
250[ms] −234[ms] −105[ms] −82[ms] −72[ms]
500[ms] −250[ms] −91[ms] −54[ms] −20[ms]
750[ms] −255[ms] −85[ms] −35[ms] 2[ms]

Table 4.8: Dierence in time between disc and robot had reached certain
points. Negative sign means the robot is behind and positive means it is
in front. Ramp angle was 30 degrees

4.3.5 Movement in all Directions

In this section the results from the movement in all three axis are presented. The
results come from the same experiments as the results presented in Section 4.3.4.
Looking at distance travelled gives a good indication on the dierence in position.
How far the robot was behind the disc and visa versa. Velocities are also presented.

20 Degrees Angle on the Ramp

This section presents the results and analysis from the movement in all three direc-
tions. Midway means that the distance travelled is equal to 380mm and at the end
of the motion means that distance travelled equals 800mm. This is right before the
robot starts its deceleration.
50 CHAPTER 4. EXPERIMENTS, RESULTS AND ANALYSIS

Results
Without prediction the robot was 245ms behind the disc midway in the motion.
At the end of the motion the robot was 256ms behind the disc, see Figure 4.29.
Distance between disc and robot was not constant throughout the motion. The disc
reached a top velocity at 1.03m/s, while the robots top velocity was 0.99m/s, see
Figure 4.30.

600
Distance disc Distance robot

900
400
800
200
700
Disc X position
Position [mm]

0 Robot X position 600

Position [mm]
Disc Y position
Robot Y position 500
-200 Disc Z position
Robot Z position 400

-400 300

200
-600
100

-800 0
5000 6000 7000 8000 9000 10000 11000 5000 6000 7000 8000 9000 10000 11000
Time [ms] Time [ms]

(a) Position. (b) Distance.

Figure 4.29: Position and distance without prediction with a 20 degree

angle on the ramp.

0.4
Velocity disc Velocity robot

1.2
0.2

0 1
Velocity [mm/ms]

Disc X velocity
Velocity [mm/ms]

-0.2 Robot X velocity 0.8

Disc Y velocity
Robot Y velocity
-0.4 Disc Z velocity 0.6
Robot Z velocity

-0.6 0.4

-0.8 0.2

-1 0
5000 6000 7000 8000 9000 10000 11000 5000 6000 7000 8000 9000 10000 11000
Time [ms] Time [ms]

(a) Velocity in all directions. (b) Combined velocity.

Figure 4.30: Velocities without prediction with a 20 degree angle on the

ramp.

Predicting 15 time steps ahead the robot was 71ms behind the disc midway. At
the end of the motion the robot was 56ms behind the disc, see Figure 4.31. There
was an overshoot equal to 58mm at the end of the robots motion. The robot reached
a top velocity equal to 1.04m/s and the discs top velocity was 1.04m/s, see Figure
4.32.
4.3. RAMP EXPERIMENTS 51

600
Distance disc Distance robot

1000
400

200 800

Disc X position
Position [mm]

0 Robot X position

Position [mm]
Disc Y position 600
Robot Y position
-200 Disc Z position
Robot Z position
400
-400

200
-600

-800 0
0 1000 2000 3000 4000 5000 6000 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500
Time [ms] Time [ms]

(a) Position. (b) Distance.

Figure 4.31: Position and distance predicting 15 time steps ahead. Ramp
angle was 20 degrees.

0.4
Velocity disc Velocity robot

1.2
0.2

0 1
Velocity [mm/ms]

Disc X velocity
Velocity [mm/ms]

-0.2 Robot X velocity 0.8

Disc Y velocity
Robot Y velocity
-0.4 Disc Z velocity 0.6
Robot Z velocity

-0.6 0.4

-0.8 0.2

-1 0
0 1000 2000 3000 4000 5000 6000 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500
Time [ms] Time [ms]

(a) Velocity in all directions. (b) Combined velocity.

Figure 4.32: Velocities predicting 15 time steps ahead. Ramp angle was
20 degrees.

When predicting 20 time steps ahead the robot caught up with the disc. The
robots position was approximately equal to the discs throughout the motion. Mid-
way the robot was 17ms behind the disc. At the end of the motion the robot was
5ms in front of the disc, see Figure 4.33. There was an overshoot at the end of the
robots motion equal to 79mm. The top velocity of the disc was 0.95m/s and the
robots was 0.89m/s, see Figure 4.34.
52 CHAPTER 4. EXPERIMENTS, RESULTS AND ANALYSIS

600
Distance disc Distance robot

1000
400

200 800

Disc X position
Position [mm]

0 Robot X position

Position [mm]
Disc Y position 600
Robot Y position
-200 Disc Z position
Robot Z position
400
-400

200
-600

-800 0
4000 5000 6000 7000 8000 9000 10000 4000 5000 6000 7000 8000 9000 10000
Time [ms] Time [ms]

(a) Position. (b) Distance.

Figure 4.33: Position and distance predicting 20 time steps ahead. Ramp
angle was 20 degrees.

0.4
Velocity disc Velocity robot

1
0.2

0 0.8
Velocity [mm/ms]

Disc X velocity
Velocity [mm/ms]

-0.2 Robot X velocity

Disc Y velocity 0.6
Robot Y velocity
-0.4 Disc Z velocity
Robot Z velocity
0.4
-0.6

0.2
-0.8

-1 0
4000 5000 6000 7000 8000 9000 10000 4000 5000 6000 7000 8000 9000 10000
Time [ms] Time [ms]

(a) Velocity in all directions. (b) Combined velocity.

Figure 4.34: Velocities predicting 20 time steps ahead. Ramp angle was
20 degrees.

The robot caught up and passed the disc when predicting 25 time steps ahead.
Midway the robot was 17ms in front of the disc. At the end of the motion the robot
was 47ms in front of the disc, see Figure 4.35. There was an overshoot at the end
of the robots motion equal to 105mm. The discs top velocity was 1.25m/s and the
robots top velocity was 1.22m/s, see gure 4.36.
A summary of the dierence between the robot and disc's position is found in
Table 4.9.
4.3. RAMP EXPERIMENTS 53

600
Distance disc Distance robot

1000
400

200 800

Disc X position
Position [mm]

0 Robot X position

Position [mm]
Disc Y position 600
Robot Y position
-200 Disc Z position
Robot Z position
400
-400

200
-600

-800 0
5000 6000 7000 8000 9000 10000 11000 12000 5000 6000 7000 8000 9000 10000 11000 12000
Time [ms] Time [ms]

(a) Position. (b) Distance.

Figure 4.35: Position and distance predicting 25 time steps ahead. Ramp
angle was 20 degrees.

0.4
Velocity disc Velocity robot

0.2 1.4

0 1.2
Velocity [mm/ms]

-0.2 Disc X velocity 1

Velocity [mm/ms]

Robot X velocity
Disc Y velocity
-0.4 0.8
Robot Y velocity
Disc Z velocity
-0.6 Robot Z velocity 0.6

-0.8 0.4

-1 0.2

-1.2 0
5000 6000 7000 8000 9000 10000 11000 12000 5000 6000 7000 8000 9000 10000 11000 12000
Time [ms] Time [ms]

(a) Velocity in all directions. (b) Combined velocity.

Figure 4.36: Velocities predicting 25 time steps ahead Ramp angle was
20 degrees.

Time steps in the predictor

Distance 0 15 20 25
380 [mm] −245[ms] −71[ms] −17[ms] 17[ms]
800 [mm] −256[ms] −56[ms] 5[ms] 47[ms]

Table 4.9: Dierence in when the disc and robot reached the same point.
Negative sign means the robot was behind and positive means it was in
front. Ramp angle was 20 degrees.
54 CHAPTER 4. EXPERIMENTS, RESULTS AND ANALYSIS

30 Degrees Angle on the Ramp

This section presented the results in from movement in all three directions with a 30
degree angle on the ramp. Midway means that distance travelled was equal 390mm
and at the end of the motion, distance travelled was equal 700mm. This was before
the robot started its deceleration.

Results
Without prediction the robot laid behind the disc throughout the motion. Midway
the robot was 253ms behind the disc. At the end of the motion the robot was
262ms behind, see Figure 4.37. The discs top velocity was 1.94m/s and the robots
top velocity was 1.68m/s, see Figure 4.38.

600
Distance disc Distance robot

900
400
800
200
700
Disc X position
Position [mm]

0 Robot X position 600

Position [mm]

Disc Y position
Robot Y position 500
-200 Disc Z position
Robot Z position 400

-400 300

200
-600
100

-800 0
5000 6000 7000 8000 9000 10000 11000 12000 5000 6000 7000 8000 9000 10000 11000 12000
Time [ms] Time [ms]

(a) Position. (b) Distance.

Figure 4.37: Position and distance without prediction with a 30 degree

angle on the ramp.

0.5
Velocity disc Velocity robot

1.5
Velocity [mm/ms]

Disc X velocity
-0.5
Velocity [mm/ms]

Robot X velocity
Disc Y velocity
Robot Y velocity
Disc Z velocity 1
-1
Robot Z velocity

-1.5 0.5

-2 0
5000 6000 7000 8000 9000 10000 11000 12000 5000 6000 7000 8000 9000 10000 11000 12000
Time [ms] Time [ms]

(a) Velocity in all directions. (b) Combined velocity.

Figure 4.38: Velocities without prediction with a 30 degree angle on the

ramp.
4.3. RAMP EXPERIMENTS 55

Predicting 15 time steps ahead the robot was behind the disc throughout the
motion. Midway the robot was 86ms behind. At the end of the motion the robot
was 81ms behind. There was an overshoot in the robots motion at the end equal
to 63mm , see Figure 4.39. The discs top velocity was 1.93m/s and the robots top
velocity was 1.76m/s, see Figure 4.40.

600
Distance disc Distance robot

1000
400

200 800

Disc X position
Position [mm]

0 Robot X position

Position [mm]
Disc Y position 600
Robot Y position
-200 Disc Z position
Robot Z position
400
-400

200
-600

-800 0
1000 2000 3000 4000 5000 6000 7000 8000 1000 2000 3000 4000 5000 6000 7000 8000
Time [ms] Time [ms]

(a) Position. (b) Distance.

Figure 4.39: Position and distance predicting 15 time steps ahead. Ramp
angle was 30 degrees.

0.5
Velocity disc Velocity robot

1.5
Velocity [mm/ms]

Disc X velocity
-0.5
Velocity [mm/ms]

Robot X velocity
Disc Y velocity
Robot Y velocity
Disc Z velocity 1
-1
Robot Z velocity

-1.5 0.5

-2 0
1000 2000 3000 4000 5000 6000 7000 8000 1000 2000 3000 4000 5000 6000 7000 8000
Time [ms] Time [ms]

(a) Velocity in all directions. (b) Combined velocity.

Figure 4.40: Velocities predicting 15 time steps ahead. Ramp angle was
30 degrees.

Predicting 20 time steps ahead the robot was behind the disc throughout the
motion. Midway the robot was 40ms behind the disc. At the end of the motion the
robot was 30ms behind the disc. There was an overshoot in the robots motion at
the end equal to 87mm, see Figure 4.41. The discs top velocity was 1.98m/s and
the robots top velocity was 1.91m/s, see Figure 4.42.
56 CHAPTER 4. EXPERIMENTS, RESULTS AND ANALYSIS

600
Distance disc Distance robot

1000
400

200 800

Disc X position
Position [mm]

0 Robot X position

Position [mm]
Disc Y position 600
Robot Y position
-200 Disc Z position
Robot Z position
400
-400

200
-600

-800 0
6000 7000 8000 9000 10000 11000 12000 13000 6000 7000 8000 9000 10000 11000 12000 13000
Time [ms] Time [ms]

(a) Position. (b) Distance.

Figure 4.41: Position and distance predicting 20 time steps ahead. Ramp
angle was 30 degrees.

0.5
Velocity disc Velocity robot

1.5
Velocity [mm/ms]

Disc X velocity
-0.5
Velocity [mm/ms]

Robot X velocity
Disc Y velocity
Robot Y velocity
Disc Z velocity 1
-1
Robot Z velocity

-1.5 0.5

-2 0
6000 7000 8000 9000 10000 11000 12000 13000 6000 7000 8000 9000 10000 11000 12000 13000
Time [ms] Time [ms]

(a) Velocity in all directions. (b) Combined velocity.

Figure 4.42: Velocities predicting 20 time steps ahead Ramp angle was
30 degrees.

Predicting 25 time steps ahead in time the robot caught up and passed the disc.
The dierence in time midway in the motion was 14ms. At the end of the motion
the robot was 11ms in front of the disc. There was an overshoot at the end of the
robots motion equal to 102mm, see Figure 4.43. The discs top velocity was 1.88m/s
and the robots top velocity was 1.98m/s, see Figure 4.44.
A summary of the dierence between the robot and disc's position is found in
Table 4.10.
4.3. RAMP EXPERIMENTS 57

600
Distance disc Distance robot

1000
400

200 800

Disc X position
Position [mm]

0 Robot X position

Position [mm]
Disc Y position 600
Robot Y position
-200 Disc Z position
Robot Z position
400
-400

200
-600

-800 0
5000 6000 7000 8000 9000 10000 11000 5000 6000 7000 8000 9000 10000 11000
Time [ms] Time [ms]

(a) Position. (b) Distance.

Figure 4.43: Position and distance predicting 25 time steps ahead. Ramp
angle was 30 degrees.

0.5
Velocity disc Velocity robot

1.5
Velocity [mm/ms]

Disc X velocity
-0.5
Velocity [mm/ms]

Robot X velocity
Disc Y velocity
Robot Y velocity
Disc Z velocity 1
-1
Robot Z velocity

-1.5 0.5

-2 0
5000 6000 7000 8000 9000 10000 11000 5000 6000 7000 8000 9000 10000 11000
Time [ms] Time [ms]

(a) Velocity in all directions. (b) Combined velocity.

Figure 4.44: Velocities predicting 25 time steps ahead. Ramp angle was
30 degrees.

Time steps in the predictor

Distance 0 15 20 25
390 [mm] −253[ms] −86[ms] −40[ms] −14[ms]
700 [mm] −262[ms] −81[ms] −30[ms] 11[ms]

Table 4.10: Dierence in when the disc and robot reached the same point.
Negative sign means the robot was behind and positive means it was in
front. Ramp angle was 30 degrees.
58 CHAPTER 4. EXPERIMENTS, RESULTS AND ANALYSIS

4.3.6 Analysis of the Ramp Experiments

This section holds the analysis for the experiments presented in Section 4.3.4 and
4.3.5.

20 Degree Angle
Without prediction the dierence in the disc's and robot's position were constant
throughout the motion, except during acceleration and deceleration.With prediction
the robot was able to follow the disc, but the dierence in disc and robot's position
were not constant throughout the motion. This indicates that the predictor was not
optimal. Further discussion around the predictor is found in Chapter 5. Predicting
20 time steps ahead gave the best result. The overshoots were due to the predictor.
At the end of the disc's movement the predictor sent targets to the robot which were
beyond the stop point of the disc.

30 Degree Angle
With no prediction the distance between the robot and disc were not constant
throughout the motion. The dierence was not constant using a predictor. This
points to the fact that the predictor was not optimal. The robot is able to follow
the disc, even though the dierence was not constant. The overshoots were due to
the predictor. At the end of the disc's movement the predictor sent targets to the
robot which were beyond the stop point of the disc. Predicting 25 time steps ahead
gave the best results.
With a 20 degree angle predicting 20 time steps ahead gave the best results and
with a 30 degree angle predicting 25 time steps ahead gave the best results. This is
another indication that the predictor was not optimal.
5. Discussion

This chapter contains some discussions around the system and the results presented
in this thesis.

5.1 Externally Guided Motion

This thesis suggest a solution on how to implement EGM in a system using a camera
to give the robot targets. The solution presented gives acceptable results. The
robot is able to follow the disc with velocities upwards to max TCP velocity for
the robot. This makes the results even more impressive. EGM can potentially be
a game changer giving more eective production lines. Even though EGM gives
good results there are some limitations:

ˆ Lack of interpolation. There is possible to interpolate the sensor data before

it reaches the robot, but this may cause jagged movements.

ˆ Without path planning the robot can not plan a path to avoid awkward po-
sitions. Positions which may trigger errors and stop the motion e.g. 50449:
Mechanical unit close to joint bound at the end of the movement.

ˆ EGM can trigger a lot of the robots safety mechanisms, which stops of the
motion. Error messages like Dynamic load too high and Mechanical unit
close to joint bound popped up frequently during this thesis, e.g. if two
consecutive targets are far from one another (about 400mm) this triggered a
Dynamic load too high because the robot tried to move to fast. There is
possible to set a maximum admitted joint speed change, but this will trim the
acceleration/deceleration.

5.2 Noise in the Discs Position

The noise in the camera readings may come from the lack of time synchronization.
The frequency which the robot asks for a new position estimate and the time the
camera uses to take a new picture, nd the disc and send the position vary with
±1ms. The times are not synchronized. The position of the robot and disc is
written to le every time the robot request a new position estimate. Lets say that
the camera uses 12ms to take a new picture, nd the disc and send the position
over to the PC. The robot requests a new position from the PC with a frequency

59
60 CHAPTER 5. DISCUSSION

of 12ms. The disc position which is being written may be 11ms old. In the next
iteration, lets say that the camera uses 12ms to send new data, and that the robot
asks for a new position after 13ms. Then the discs position which are written to
le would be 0ms old. There is also the risk of skipping one position estimate. The
fact that there is noise in the data from the simulated disc experiment, presented in
Section 4.1, helps to substantiate this theory.
The other more obvious explanation is that there is noise in the readings it self.
In In-Sight Explorer it is possible to see the value that represents the position of the
disc in mm. When the program ran there where some oscillations in this values, but
not by more than 1mm at the most. But this was just by visual registration while
the program ran, and since this value changed every 10 − 20ms it is not possible
to conclude if the value varied more than the naked eye registered. A method to
register the positions as they came from the camera with timestamps should have
been implemented to test if this was the reason for the noise.

5.3 Friction Coecient

Friction coecient chosen in Section 3.2.4 is most likely a bit o, since the dierence
between disc and robot is not constant in the results presented in Section 4.3. The
coecient is not o by much, but small adjustments in the value have a big aect
on the acceleration, e.g. with an angle equal to 20 degrees, the acceleration will be
2 2
equal to 0.59m/s with 0.3 and 0.68m/s if the friction coecient is equal to 0.29.
Wrong friction coecient is most likely one of the reasons that the predictor is not
optimal.
A better way might have been to calculate the acceleration from the second
derivative of the position. To be able to do this the noise has to be minimized or
ltered.
6. Conclusion and Future Work

EGM have been introduced and presented theoretical and experimentally. The use
of EGM in sensor guided paths minimizes the time from new sensor data is sent, to
the robot initiates the movement. A suggested solution have been presented. This
solution enables the robot to follow a disc sliding down a ramp. With prediction the
robot was able to follow the disc with velocities upwards to 2m/s and with a sliding
surface equal to 800mm. Without prediction the robot was able to follow, but was
200 − 300ms behind the disc throughout the motion. With more eective code
and better tuned EGM parameters it should be possible to bisect this dierence.
EGM gives a quick enough response to new sensor data, enabling the robot to
follow an object moving with high, non-constant velocity in a 2D-space. Using a
better predictor will produce even better results. Implementing a predictor in the
y-direction can enable the disc to bounced in the edges of the ramp and give an
more complex motion down the ramp.

6.1 Future Work

ˆ Improve the predictor: The predictor implemented is not optimal. The
dierence between the robot and discs position is not constant throughout the
motion. A better predictor will improve the results.

ˆ Implement a predictor on the y-axis: Predicting the discs path in the

y-axis can enable more complex motions down the ramp.

ˆ Pick up the disc: This is the next natural step of this application. A way
to do this might be to have an vacuum tool turned on from the start of the
movement. When the dierence between the discs - and robots - position is
lower than a certain threshold, lower the robots height over the ramp until it is
at the same height as the disc. After a certain time, when the disc is attached
1
to the tool , command the robot on another path which leads out of the ramp.

ˆ Compare with Robot Reference Interface (RRI): In this thesis it has

been proven that EGM gives fast response to external sensor data, but how
fast is it compare to other modules? It might be useful to develop develop two
equal applications, one which uses RRI, which is the predecessor to EGM and
one which use EGM. Compare the responsiveness of the two applications to
see if EGM gives better response, and if so, how much better response.

1 How long this takes have to be found by experimenting

61
62 CHAPTER 6. CONCLUSION AND FUTURE WORK

ˆ Implement EGM in an existing application using external sensors:

To test if the system can perform its task's faster and if it aects the quality.
Bibliography

[1] ABB, Application manual - controller software IRC5. RobotWare 6.01,

revision a. ed., 2015. Document ID: 3HAC050798-001.

[2] D. Deng, W. Sam, P. Tao, and T. Lubecki1, Sensor guided robot path
2015 IEEE International
generation for repair of buoyancy module, in
Conference on Advanced Intelligent Mechatronics (AIM), pp. 1613  1618,
IEEE, 2015.

[3] H. Koch, A.König, A.Weigl-Seitz, K.Kleinmann, and J.Suchý, Multisensor

contour following with vision, force, and acceleration sensors for an industrial
IEEE TRANSACTIONS ON INSTRUMENTATION AND
robot, in
MEASUREMENT, vol. 62, pp. 268  280, IEEE, 2013.

[4] Z. Luoa, J. Daia, C.Wangc, F.Wangd, Y.Tiana, and M.Zhaoc, Predictive

seam tracking with iteratively learned feedforward compensation for
Advances in Manufacturing Machines
high-precision robotic laser welding, in
and Systems  Selected Papers from the 9th International Conference of
Frontiers of Design and Manufacturing, vol. 31, p. 27, Elsevier Ltd, 2012.

[5] ABB., Irb 140. small, powerful and fast 6-axes robot..
http://new.abb.com/products/robotics/industrial-robots/irb-140. Accessed
January 24. 2016.

[6] ABB., Latest robot controller software from abb combines exibility and
productivity for developers..
http://www.abb.com/cawp/seitp202/2a21e772c71016d1c1257daf004046bc.aspx.
Accessed January 24. 2016.

[7] Cognex., In-sight 5000 datasheet..

http://www.cognex.com/downloads/LiteratureDelivery.aspx?id=12966.
IS5DS-1309. Visited March 04. 2016.

[8] Cognex., In-sight standard vision system specications..

http://www.cognex.com/support/downloads/ns/1/11/35/51xx_54xx_ss.pdf .
Accessed March 04. 2016.

[9] T. Morris., Computer vison and image processing. Palgrave Macmillan., 2004.
ISBN : 0-333-99451-5.

63
64 BIBLIOGRAPHY

[10] M. Sonka, V. Hlavac, and R. Boyle., Image Processing, Analysis, and Machine
Vision. Brooks/Cole Publishing Company., 1999. ISBN : 0-534-95393-X.

[11] j. John W. Jewett and R. A.Serway, Physics for scientists and engineers,
vol. 1. Brooks/Cole Cengage learing, 8th. international ed., 2010. ISBN10:
1-4390 -4876-2.

[12] M. W. Spong, Hutchinson, and Vidyasagar., Robot Modeling and Control.

John wiley and Sons, INC., 2006. ISBN : 978-0-471-64990-8.

[13] J. Skeet., protobuf-csharp-port..

https://github.com/jskeet/protobuf-csharp-port. IS5DS-1309. Visited May 16
2016.

[14] Unknown., Coecient of friction, rolling resistance and aerodynamics..

http://www.tribology-abc.com/abc/cof.htm. Accessed May 29 2016.

[15] S. Ore and A. Stori, Plast.. https://snl.no/plast. Accessed May 26 2016.

[16] MathWorks., Curve tting toolbox - smooth..

http://se.mathworks.com/help/curvet/smooth.html. Accessed June 08. 2016.
A. Experiment With Constant

Acceleration

In this appendix plots showing all the results from the experiment in Chapter 4.1.1
is presented. There is also given an example which shows the dierence between
smoothed and not smoothed plots.

A.1 Without Smoothing

Angle of the simulated ramp was 60 degrees. This section displays the noise
present in the readings.

Distance disc Distance robot Distance disc Distance robot

900
800
800
700
700
600
600
Position [mm]
Position [mm]

500
500
400
400
300
300
200
200

100
100

0
0
7000 7500 8000 8500 9000 9500 10000 10500 11000 11500 7600 7800 8000 8200 8400 8600 8800 9000 9200 9400
Time [ms] Time [ms]

(a) Distance without smoothing (b) Distance without smoothing, zoomed in

A1
A2 APPENDIX A. EXPERIMENT WITH CONSTANT ACCELERATION

Velocity disc Velocity robot Velocity disc Velocity robot

35 5

30
4

25

Velocity [mm/ms]
Velocity [mm/ms]

3
20

2
15

1
10

5 0

0
7000 7500 8000 8500 9000 9500 10000 10500 11000 11500 7500 8000 8500 9000 9500
Time [ms] Time [ms]

(a) Velocity without smoothing (b) Velocity without smoothing, zoomed in

Acceleration disc Acceleration robot

Acceleration disc Acceleration robot
5
45

40 4
Acceleration [mm/ms 2 ]

35
Acceleration [mm/ms 2 ]

3
30

25 2

20
1
15

10 0

5
-1
0 7600 7800 8000 8200 8400 8600 8800 9000 9200 9400
7000 7500 8000 8500 9000 9500 10000 10500 11000 11500 Time [ms]
Time [ms]
(b) Acceleration without smoothing, zoomed
(a) Acceleration without smoothing
in
A.2. 20 DEGREES A3

A.2 20 Degrees

800

600

400

200 Disc X position

Position [mm]

Robot X position
Disc Y position
0
Robot Y position
Disc Z position
-200 Robot Z position

-400

-600

-800
2000 3000 4000 5000 6000
Time [ms]

Figure A.4: Positions in all three directions

0.5

-0.5
Velocity [mm/ms]

Disc X velocity
Robot X velocity
Disc Y velocity
-1
Robot Y velocity
Disc Z velocity
Robot Z velocity
-1.5

-2

-2.5
2000 3000 4000 5000 6000
Time [ms]

Figure A.5: Velocity in all three directions

A4 APPENDIX A. EXPERIMENT WITH CONSTANT ACCELERATION

0.06

Acceleration [mm/ms 2 ] 0.04

0.02
Disc X acceleration
Robot X acceleration
Disc Y acceleration
0
Robot Y acceleration
Disc Z acceleration
Robot Z acceleration
-0.02

-0.04

-0.06
2000 3000 4000 5000 6000
Time [ms]

Figure A.6: Accelerations in all three directions

Acceleration disc Acceleration robot

0.08

0.07

0.06
2
Acceleration [mm/ms ]

0.05

0.04

0.03

0.02

0.01

0
2000 2500 3000 3500 4000 4500 5000 5500 6000
Time [ms]

Figure A.7: Acceleration combined

A.3. 30 DEGREES A5

A.3 30 Degrees
800
Distance disc Distance robot

600 900

800
400
700
200 Disc X position
Position [mm]

Robot X position 600

Position [mm]
Disc Y position
0 500
Robot Y position
Disc Z position
-200 Robot Z position 400

300
-400
200
-600
100

-800 0
5000 6000 7000 8000 9000 10000 5500 6000 6500 7000 7500 8000 8500 9000 9500 10000
Time [ms] Time [ms]

(a) Positions (b) Distance

Figure A.8: The positions and distance with 30 degree angle on the ramp

0.5
Velocity disc Velocity robot

3
0

2.5
-0.5
Velocity [mm/ms]

Disc X velocity
Velocity [mm/ms]

Robot X velocity 2
Disc Y velocity
-1
Robot Y velocity
Disc Z velocity 1.5
Robot Z velocity
-1.5
1

-2
0.5

-2.5 0
5000 6000 7000 8000 9000 10000 5500 6000 6500 7000 7500 8000 8500 9000 9500 10000
Time [ms] Time [ms]

(a) Velocity in all three directions (b) Combined velocity.

Figure A.9: Velocities with 30 degree angle on the ramp

A6 APPENDIX A. EXPERIMENT WITH CONSTANT ACCELERATION

0.06
Acceleration disc Acceleration robot

0.08
0.04
0.07
Acceleration [mm/ms 2 ]

0.02 0.06

Acceleration [mm/ms 2 ]
Disc X acceleration
Robot X acceleration
0.05
Disc Y acceleration
0
Robot Y acceleration
Disc Z acceleration 0.04
Robot Z acceleration
-0.02 0.03

0.02
-0.04
0.01

-0.06 0
5000 6000 7000 8000 9000 10000 5500 6000 6500 7000 7500 8000 8500 9000 9500 10000
Time [ms] Time [ms]

(a) Accelerations in all three directions (b) Acceleration combined

Figure A.10: Accelerations with 30 degree angle on the ramp

A.4 40 Degrees
800
Distance disc Distance robot

600 900

800
400
700
200 Disc X position
Position [mm]

Robot X position 600

Position [mm]

Disc Y position
0 500
Robot Y position
Disc Z position
-200 Robot Z position 400

300
-400
200
-600
100

-800 0
7000 8000 9000 10000 11000 12000 7000 7500 8000 8500 9000 9500 10000 10500 11000 11500
Time [ms] Time [ms]

(a) Positions (b) Distance

Figure A.11: The positions and distance with 40 degree angle on the
ramp
A.4. 40 DEGREES A7

0.5
Velocity disc Velocity robot

3.5
0

-0.5
Velocity [mm/ms]

Disc X velocity 2.5

Velocity [mm/ms]
Robot X velocity
Disc Y velocity
-1 2
Robot Y velocity
Disc Z velocity
Robot Z velocity 1.5
-1.5

-2
0.5

-2.5 0
7000 8000 9000 10000 11000 12000 7000 7500 8000 8500 9000 9500 10000 10500 11000 11500
Time [ms] Time [ms]

(a) Velocity in all three directions (b) Velocities combined

Figure A.12: Velocities with 40 degree angle on the ramp

0.06
Acceleration disc Acceleration robot

0.08
0.04
0.07
Acceleration [mm/ms 2 ]

0.02 0.06
Acceleration [mm/ms 2 ]

Disc X acceleration
Robot X acceleration
0.05
Disc Y acceleration
0
Robot Y acceleration
Disc Z acceleration 0.04
Robot Z acceleration
-0.02 0.03

0.02
-0.04
0.01

-0.06 0
7000 8000 9000 10000 11000 12000 7000 7500 8000 8500 9000 9500 10000 10500 11000 11500
Time [ms] Time [ms]

(a) Accelerations in all three directions (b) Acceleration combined

Figure A.13: Accelerations with 40 degree angle on the ramp

A8 APPENDIX A. EXPERIMENT WITH CONSTANT ACCELERATION

A.5 50 Degrees
800
Distance disc Distance robot

600 900

800
400
700
200 Disc X position
Position [mm]

Robot X position 600

Position [mm]
Disc Y position
0 500
Robot Y position
Disc Z position
-200 Robot Z position 400

300
-400
200
-600
100

-800 0
7000 8000 9000 10000 11000 7000 7500 8000 8500 9000 9500 10000 10500 11000
Time [ms] Time [ms]

(a) Positions (b) Distance

Figure A.14: The positions and distance with 50 degree angle on the
ramp

0.5
Velocity disc Velocity robot

3.5
0

3
-0.5
Velocity [mm/ms]

Disc X velocity 2.5

Velocity [mm/ms]

-1 Robot X velocity
Disc Y velocity
2
Robot Y velocity
-1.5 Disc Z velocity
Robot Z velocity 1.5

-2
1

-2.5
0.5

-3 0
7000 8000 9000 10000 11000 7000 7500 8000 8500 9000 9500 10000 10500 11000
Time [ms] Time [ms]

(a) Velocity in all three directions (b) Velocities combined

Figure A.15: Velocities with 50 degree angle on the ramp

A.6. 60 DEGREES A9

0.06
Acceleration disc Acceleration robot

0.08
0.04
0.07
Acceleration [mm/ms 2 ]

0.02 0.06

Acceleration [mm/ms 2 ]
Disc X acceleration
Robot X acceleration
0.05
Disc Y acceleration
0
Robot Y acceleration
Disc Z acceleration 0.04
Robot Z acceleration
-0.02 0.03

0.02
-0.04
0.01

-0.06 0
7000 8000 9000 10000 11000 7000 7500 8000 8500 9000 9500 10000 10500 11000
Time [ms] Time [ms]

(a) Accelerations in all three directions (b) Acceleration combined

Figure A.16: Accelerations with 50 degree angle on the ramp

A.6 60 Degrees
800

600

400

200 Disc X position

Position [mm]

Robot X position
Disc Y position
0
Robot Y position
Disc Z position
-200 Robot Z position

-400

-600

-800
7000 8000 9000 10000 11000 12000
Time [ms]

Figure A.17: Positions in all three directions

A10 APPENDIX A. EXPERIMENT WITH CONSTANT ACCELERATION

0.5

-0.5

Velocity [mm/ms]
-1 Disc X velocity
Robot X velocity
Disc Y velocity
-1.5
Robot Y velocity
Disc Z velocity
-2 Robot Z velocity

-2.5

-3

-3.5
7000 8000 9000 10000 11000 12000
Time [ms]

Figure A.18: Velocity in all three directions

0.06

0.04
Acceleration [mm/ms 2 ]

0.02
Disc X acceleration
Robot X acceleration
Disc Y acceleration
0
Robot Y acceleration
Disc Z acceleration
Robot Z acceleration
-0.02

-0.04

-0.06
7000 8000 9000 10000 11000 12000
Time [ms]

Figure A.19: Accelerations in all three directions

A.6. 60 DEGREES A11

Acceleration disc Acceleration robot

0.08

0.07

0.06

Acceleration [mm/ms 2 ]
0.05

0.04

0.03

0.02

0.01

0
7000 7500 8000 8500 9000 9500 10000 10500 11000 11500
Time [ms]

Figure A.20: Acceleration

This page is intentionally left blank.
B. Experiment with Constant

Velocity

B.1 v100
600

400

200
Position [mm]

0 Disc X position
Robot X position
Disc Y position
Robot Y position
Disc Z position
-200 Robot Z position

-400

-600

-800
0 0.5 1 1.5 2 2.5 3
Time [ms] ×10 4

Figure B.1: Position in all directions, v100.

0.15

0.1

0.05
Velocity [mm/ms]

Disc X velocity
Robot X velocity
0 Disc Y velocity
Robot Y velocity
Disc Z velocity
Robot Z velocity

-0.05

-0.1

-0.15
0 0.5 1 1.5 2 2.5 3
Time [ms] 4
×10

Figure B.2: Velocity in all directions, v100.

B13
B14 APPENDIX B. EXPERIMENT WITH CONSTANT VELOCITY

B.2 v500
600 Distance disc Distance robot

1000

400
900

800
200

700
Position [mm]

0 Disc X position 600

Position [mm]
Robot X position
Disc Y position
Robot Y position
Disc Z position 500
-200 Robot Z position

400

-400 300

200
-600
100

-800 0
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Time [ms] ×10 4 Time [ms] ×10 4

(a) Position in all three directions, v500. (b) Distance, v500.

Figure B.3: Position and distance for robot and disc, v500.

0.6 Velocity disc Velocity robot

0.6

0.4

0.5

0.2
0.4
Velocity [mm/ms]

Velocity [mm/ms]

Disc X velocity
Robot X velocity
0 Disc Y velocity
Robot Y velocity
Disc Z velocity 0.3
Robot Z velocity

-0.2
0.2

-0.4
0.1

-0.6 0
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Time [ms] ×10 4 Time [ms] ×10 4

(a) Velocities in all three directions, v500. (b) Combined velocity, v500.

Figure B.4: Velocities for robot and disc, v500.

B.3. V1000 B15

B.3 v1000
600

400

200
Position [mm]

0 Disc X position
Robot X position
Disc Y position
Robot Y position
Disc Z position
-200 Robot Z position

-400

-600

-800
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Time [ms]

Figure B.5: Position in all directions, v1000.

1.5

0.5
Velocity [mm/ms]

Disc X velocity
Robot X velocity
0 Disc Y velocity
Robot Y velocity
Disc Z velocity
Robot Z velocity

-0.5

-1

-1.5
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Time [ms]

Figure B.6: Velocity in all directions, v1000.

B16 APPENDIX B. EXPERIMENT WITH CONSTANT VELOCITY

B.4 v1500
600 Distance disc Distance robot

1000

400
900

800
200

700
Position [mm]

0 Disc X position 600

Position [mm]
Robot X position
Disc Y position
Robot Y position
Disc Z position 500
-200 Robot Z position

400

-400 300

200
-600
100

-800 0
0 2000 4000 6000 8000 10000 12000 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 11000
Time [ms] Time [ms]

(a) Position in all three directions, v1500. (b) Distance, v1500.

Figure B.7: Position and distance for robot and disc, v1500.

2 Velocity disc Velocity robot

1.6
1.5

1.4

1
1.2

0.5
Velocity [mm/ms]

1
Velocity [mm/ms]

Disc X velocity
Robot X velocity
0 Disc Y velocity
Robot Y velocity
Disc Z velocity 0.8
Robot Z velocity

-0.5
0.6

-1
0.4

-1.5
0.2

-2 0
0 2000 4000 6000 8000 10000 12000 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 11000
Time [ms] Time [ms]

(a) Velocities in all three directions, v1500. (b) Combined velocity, v1500.

Figure B.8: Velocities for robot and disc, v1500.

B.5. V2000 B17

B.5 v2000
600 Distance disc Distance robot

1000

400
900

800
200

700
Position [mm]

0 Disc X position 600

Position [mm]
Robot X position
Disc Y position
Robot Y position
Disc Z position 500
-200 Robot Z position

400

-400 300

200
-600
100

-800 0
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Time [ms] Time [ms]

(a) Position in all three directions, v2000. (b) Distance, v2000.

Figure B.9: Position and distance for robot and disc, v2000.

2.5 Velocity disc Velocity robot

2.5
2

1.5
2
1
Velocity [mm/ms]

0.5
Velocity [mm/ms]

Disc X velocity 1.5

Robot X velocity
0 Disc Y velocity
Robot Y velocity
Disc Z velocity
Robot Z velocity
-0.5
1

-1

-1.5
0.5

-2

-2.5 0
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Time [ms] Time [ms]

(a) Velocities in all three directions, v2000. (b) Combined velocity, v2000.

Figure B.10: Velocities for robot and disc, v2000.

B18 APPENDIX B. EXPERIMENT WITH CONSTANT VELOCITY

B.6 v2500
600

400

200
Position [mm]

0 Disc X position
Robot X position
Disc Y position
Robot Y position
Disc Z position
-200 Robot Z position

-400

-600

-800
2000 3000 4000 5000 6000 7000 8000 9000 10000 11000
Time [ms]

Figure B.11: Position in all directions, v2500.

2.5

1.5

1
Velocity [mm/ms]

0.5
Disc X velocity
Robot X velocity
0 Disc Y velocity
Robot Y velocity
Disc Z velocity
Robot Z velocity
-0.5

-1

-1.5

-2

-2.5
2000 3000 4000 5000 6000 7000 8000 9000 10000 11000
Time [ms]

Figure B.12: Velocity in all directions, v2500.

C. Code

All the code written during this thesis are embedded in the document in the le
ExternallyGuidedMotion.7z. All code is place under the folder src.

ˆ Opening the ExternalGuidedMotion.sln will open Visual Studio and

display all the classes written in C#.

ˆ The Matlab scripts are placed in the Matlab folder. RotateTranslate.m is

the script which is used to nd distance travelled, velocity and acceleration
for the disc and robot. DelayDistribution.m plots the processing time in
histograms.

ˆ In the Robotstudio folder there is an PackAndGo le holding the program

written in Robotstudio.

ˆ The program for the camera is found in the Cognex folder.

All code is available on Github:

https://github.com/orjanmehre/ExternallyGuidedMotion.git

C19