Received: 23 October 2017

DOI: 10.1002/cae.21917
| Accepted: 16 January 2018

SHORT COMMUNICATION

A low-cost feedback control systems laboratory setup via

Arduino–Simulink interface

Ismail Uyanik1 | Bahadir Catalbas2

1 Laboratory of Computational Sensing and

Robotics, Johns Hopkins University, Abstract
Baltimore, Maryland Control theory education, when supported by practice, becomes more comprehend-
2 Department of Electrical and Electronics
ible for students and useful for their professional career. This paper presents low-cost
Engineering, Bilkent University, Ankara,
Turkey experiments for laboratory sessions of a feedback control systems course, which
introduces them modeling feedback control systems, proportional-integral-derivative
Correspondence
(PID) controller design, root locus and Bode plots. The experiments are organized
Ismail Uyanik, Laboratory of
Computational Sensing and Robotics, Johns around the Arduino-based identification and control of a DC motor via Matlab/
Hopkins University, Baltimore, 21218 MD. Simulink. The objective of this laboratory session is to support teaching feedback
Email: [email protected]
control systems via experimental investigations on a low-cost laboratory kit. The built
Funding information in-house setups support Arduino–Simulink interface, so that students can download
Scientific and Technological Research their control diagrams in Simulink to the Arduino board directly. This interface
Council of Turkey (TÜBİTAK)
allows students to utilize high level control design tools, such as Matlab/Simulink
while working on a low-cost hardware laboratory setup. Students’ performance in the
written exams before and after the laboratory setup were reported to evaluate the
instructional effectiveness. Besides, student feedback for four semesters are also
presented to evaluate the effectiveness of the laboratory experiments.

KEYWORDS
control education, education, laboratory education, laboratory techniques, learning control

1 | INTRODUCTION developing countries such as Turkey due to the high student

counts and expensive prices of the commercial educational
The importance of hands-on experience for engineering experimental kits. This is why we used to perform simulation-
education has long been established [9,20]. Literature studies based laboratory sessions in Department of Electrical and
suggest that practice is fundamental for complementing Electronics Engineering of Bilkent University, Turkey.
strong theory background [15] as well as developing new There are three common ways to introduce practical
theories [6], since working with tangible components experience to students in engineering education; simulation,
supports active and sensory learning for students [12,22]. emulation, and use of physical test hardware. Among these
Besides its substantial advantages in learning, the laboratory three, the highest educational benefit can be achieved by
modules became a must for engineering colleges after using pilot plants which means using real size pilot hardware
ABET's (Accreditation Board of Engineering Technology) for experiments [15]. However, this choice is not feasible for
requirement for adequate laboratory practice for students [1]. many colleges in terms of cost and large area require-
However, designing and integrating such laboratory sessions ments [15]. At this point, the use of small-scale plants (such
to the undergraduate curriculum is very challenging in newly as [11,13,23] can be considered as simple ad-hoc solutions.
718 | © 2018 Wiley Periodicals, Inc. wileyonlinelibrary.com/cae Comput Appl Eng Educ. 2018;26:718–726.
UYANIK AND CATALBAS
| 719

For example, the use of a small-scale wind turbine for control these studies are capable of teaching special applications of
theory education has been presented in [13]. The authors control theory, these methods do not provide the feeling of
developed an educational software tool to familiarize working with physical components and do not help students for
graduate level students to the concept of decoupling control gaining experience to work on a real hardware system [15].
strategies for wind turbines via small-scale plant models [13]. There are some companies offering commercial products
There are some studies which develop experiments based on for feedback control systems laboratories. However, such kits
control and monitoring of an experimental biomass combus- are generally quite expensive, especially when the goal is to
tor via LabVIEW on a small-scale experimental plant [23]. make sure that all students interact with the hardware. For
Similarly, some PC-based electro-hydraulic system control instance, Quanser develops DC motor control and vertical
experiments has also been proposed in [11]. This paper take-off kits for use in feedback control systems laboratories.
presents a good example of training students/researchers with However, these kits are very expensive for newly developing
PC-based control system design tools and fluid power countries when the student counts are high to give a lab setup to
systems. The mathematical model of the electro-hydraulic each one or two students. In order to solve this problem [7]
system is derived from first principles and the model is proposes two sets of experiments which include eight hands on
implemented in Matlab/Simulink environment. The control- experiments that benefit from a wide range of equipment such
ler designed on the Matlab/Simulink environment has been as use of circuit simulations, circuit realization and control
applied on a physical test bench by which students’ experience laboratory kits in order to teach control theory notions. Thus,
with the physical hardware has also been ensured [11]. students learn overshoot, rise time, settling time concepts, and
However, [13] and [23] present graduate level experiments use of PSpice and Matlab simulation programs. Besides, there
involving deeper theoretical insight in control theory which are some experimental setups, which are capable of conducting
goes beyond the scope of introductory level feedback control PID controller and frequency response experiments with a real
theory courses for undergraduate education that we focus on DC Motor module [19,29]. Unfortunately, most of these
this paper. On the other hand, despite the fact that [11] presents examples become expensive for high number of purchases
experiments for undergraduate students (in addition to despite their clear advantages on teaching.
graduate students), it is still challenging (in terms of cost) to Considering the state of the art techniques in the literature,
reproduce this kind of small scale test benches when the goal is some design decisions are determined for the desired
to provide a test setup for each student (or pairs). experimental feedback control system setup as
Different than these studies, remote laborato-
ries [10,17,18,24,31] have also widely been used in the control
The students should use high level design tools for
theory education literature. Remote laboratory concept controller design.
becomes a cost effective solution for practice needs and it
Students should not deal with the details of hardware and
may satisfy the feeling of working with a real hardware up a focus on the principles of feedback control theory.
certain level. For instance, a Wiki-based remote laboratory
Laboratory setups should be low-cost, so that many of them
platform has been developed to provide experience to students can be simply built and replaced easily in case of failures.
about characterization of variations of PID control techniques
and remote tuning [31]. There is also web-based remote Motivated by these design requirements, Matlab/Simulink
learning systems, such as NeTRe-LAB [10], which provides an environment was used due to its wide spread use in both
opportunity to reach DC motor exercises in conventional academia and industry [21,30]. Matlab provides simple and
laboratories without place and time requirements. Even though efficient programming environment, especially for educational
there are favorable features of remote laboratories such as low purposes, for developers from various experience levels.
cost as compared to physical test benches, there are some Sometimes, Matlab may be used as sole programming tool for
important drawbacks. It has been shown that majority of the most engineering studies due to its capability to mathematical
students prefer hand-on experiments with respect to remote programming philosophy and integrating different program-
ones [31]. One reason of this tendency may be related with lack ming platforms [20,21]. In addition, Simulink provides a
of physical interaction in remote laboratories. graphical editor for block diagram programming, which eases
There are also some studies that design simulators for the use of custom blocks and reduces the possibility of
teaching control theory [3,25,26,28]. For instance, a special- implementation errors. Currently, Simulink has a wide
ized simulation infrastructure called ASPEN HYSYS has application area including control systems, digital signal
been used for control theory education to the undergraduate processing and communication fields and there are various
chemical engineering students. Besides, virtual laboratory built-in block modules that provide simplified and time
tools also provide similar experience of working with a DC efficient working environment for professionals as well
motor and controller circuit in simulation environment [3]. as students [2]. Among all, Matlab/Simulink provides
However, although the proposed simulation experiments in hardware support for numerous microcontrollers, motor
720
| UYANIK AND CATALBAS

controllers and sensors. The hardware interface of Matlab/ which introduce the principles of feedback control, are
Simulink will be the key functionality that is being used during supported by three laboratory sessions. Prior to proposed lab
kit's development stage. kits, students performed Matlab/Simulink based simulation
Another key property of using Matlab/Simulink is that experiments for the identification and control of a DC motor
students can develop their control diagrams in Simulink and model for several years using a sample DC motor data. They
directly download them to a supported hardware without also submitted a post lab report after the lab to discuss and
dealing with details of hardware design or micro-controller comment on the results they obtained during the lab session.
programming. Among various alternative microcontroller Student evaluations showed that the simulation based lab
boards that have Matlab/Simulink support, Arduino Uno was sessions did not help for the course topics. This result raised
used due to its low cost regarding to our third design decision. two fundamental problems. First, students cannot integrate
In addition to its low price, Arduino offers open source the theoretical courses to the practice. Second, the lab sessions
hardware and software option for developers with varying do not contribute to their understanding of course topics.
backgrounds. This feature makes Arduino a fast prototyping
platform for various projects [5,14,16,32]. Also, there are
2.2 | Proposed hardware setup
some example uses for education such as laboratory
equipment for students [27] and educational mobile robot [4]. This section focuses on the proposed lab kit for the feedback
Motivated by these, a new laboratory setup has been control systems experiments (an exploded view is illustrated
proposed that is focused on identification and control of a DC in Figure 1). As briefly mentioned in section 1, hardware
motor using Matlab/Simulink environment and Arduino setup concentrates on Arduino-based identification and
microcontroller. Despite there are a lot of successful systems control of a DC motor. The proposed kit consists of an
that are currently used in the literature, the proposed setup Arduino Uno microcontroller, Arduino motor driver shield, a
differs from the available literature in the following manners: DC motor with encoders, a power adapter and a USB cable for
computer (Matlab/Simulink) interface. Arduino Uno micro-

The proposed setup is very low cost as compared to controller and associated motor driver shield are commercial
literature (below 100 USD), which allows providing one off-the-shelf products. The motor driver shield supports up to
hardware setup for each student (or pairs) during laboratory 12 V and 2 A supply voltage and current for the target motor
experiments. The Department of Electrical Engineering at with a power adapter, respectively. The motor is 12 V Pololu
Bilkent University built 100 of these setups and they have DC motor with gearhead and integrated quadrature encoder.
been successfully used for the last 4 years. 12 V power adapter with maximum 3 A current supplies

Students program their controllers in Simulink environ- necessary power for the lab kit. Finally, a standard USB cable
ment and download them to Arduino microcontroller. The is used to interface Arduino with computer and hence Matlab/
control diagram works in real-time with minimal delays Simulink through serial communication. In addition, several
(10 ms delay due to 100 Hz sampling frequency), which are consumables are used for the preparation of the lab kit.
generally higher for low-cost hardware in the loop systems. Table 1 lists all equipment used in the developed

The experimental investigation of very abstract concepts experimental kit with their current prices. Notice that total
such as root locus, Fourier analysis, frequency domain cost of a single kit is only about 97 USD, which is very cheap
identification, Bode plots and time delays (detailed when compared to commercial laboratory kits. This is very
investigation on frequency domain analysis has initially important when the goal is to build many of these kits to make
reported in [8]) has been demonstrated on this low-cost setup. sure that all students interact with the hardware during the lab

Most importantly, the paper reports both exam results that sessions. To accomplish this, 100 kits are built assuming that
illustrate the instructional effectiveness of the course as well as the students will work in pairs to complete the lab sessions
students' surveys that assess the practicality of the lab setups. (for a maximum student count of 200). Note that although the
students work in pairs to collect data and build controllers in
lab, they write individual post lab reports to make sure that
they interpret the results based on their own understanding.
2 | DESIGN AND DEVELOPMENT OF
On the software side, Matlab/Simulink is used as
EXPERIMENTAL KIT
fundamental programming interface to program the experi-
mental kit. In addition, these computers should have a
2.1 | Prior status of the lab course
C compiler compatible with Matlab, Arduino USB driver
Feedback Control Systems is one of the must courses in and Arduino target installer. Fortunately, Matlab/Simulink is
the curriculum of the Bachelor of Science program of the a key programming tool for most electrical and electronics
Electrical and Electronics Engineering (EEE) Department of engineering departments. Having a licensed Matlab, Arduino
Bilkent University, Ankara, Turkey. The theoretical lectures, target installer can be freely obtained from Simulink support
UYANIK AND CATALBAS
| 721

FIGURE 1 Proposed lab setup for feedback control systems course laboratory sessions

TABLE 1 Equipment list On the other side, the disadvantage of building own lab kits
Equipments Price (USD) is that it requires some man power. The nice thing about the
1. Arduino Uno 15,49 proposed setup is that it uses off-the-shelf components and do
not require so much design and production processes. Based on
2. Arduino Motor Shield 21,99
the experience for building 100 kits for Bilkent University, two
3. DC Motor with Encoder 39,99
teaching assistants can develop the software for actuation and
4. Power Adaptor 2,3 sensing blocks by spending ten hours each. One teaching
5. Mechanical Components 10,29 assistant spent four hours for 3D-CAD design of the main body
6. Plexiglass Base 6 frame and motor inertia. The most time-consuming part was
7. Consumables 1 the integration and test of the whole setup before giving them to
Total Cost: 97,06 students. Five teaching assistants spent eight hours each for
integrating and testing all 100 setups. There is also a plenty of
time spent for the design of the lab questions. However, only
packages library. C compilers and Arduino USB drivers can
the time spent required for preparing the lab kits is reported,
also be obtained as free versions from web. Therefore,
since each instructor will most probably prepare different lab
proposed setup does not require additional paid software
questions once they built this setup.
when Matlab is available.
In addition to having a computer with the software
2.3 | Intended outcomes
mentioned above installed, some previously compiled code
blocks are supplied to students to make sure that they don't get The feedback control theory topics, modeling control systems,
distracted with the details of microcontroller level programming. computing performance characteristics, root locus, Bode plots,
The first code block supplied is the Actuation block which and Nyquist stability are taught in class as a single condensed
converts the input voltage command generated by the controller course in one semester. The lectures are explained in classroom
to direction and pulse-width-modulation signals, so that Arduino by the course instructor and students worked on several
and the motor driver shield feeds desired supply voltage to the theoretical questions as well as some Matlab simulations in
motor terminals. The second block supplied is the Sensing block class. The three simulation-based lab experiments were aimed
that measures the speed of the motor by counting the quadrature to support students’ learning. However, the simulation-based
encoder pulses. The two software blocks are combined in a single experiments did not motivate the students to integrate their
Simulink block named as DC Motor Plant (illustrated in theoretical knowledge with real life problems. Hence, students
Figure 2), so that students can use this block as a plant model in were not able to grasp the fundamentals of the course.
their Simulink diagrams. This way, complexity of dealing with With this in mind, the intended course outcomes with the
the details of hardware for the lab course is reduced significantly. new laboratory experiments are designed as follows.
Interested students can later go into the details of these code
blocks to modify them for further use in their different projects. 1. The lab setups should allow high level control design
One advantage of using Arduino and a DC motor in the without distracting students’ attention with the details of
lab sessions encourages students to work with hardware and microcontroller level programming.
notice that controlling a DC motor via feedback has various 2. The lab experiments should support teaching fundamental
applications in different robotics and control theory projects. control theory topics such as system modeling and
722
| UYANIK AND CATALBAS

FIGURE 2 A sample feedback control system block diagram that is downloaded to Arduino to work inside the microcontroller. DC Motor
Plant block is supplied to students to be used as a plant transfer function block in their control diagrams. A serial transmitter block is attached to
the output signal to transmit measured outputs back to the computer via serial communication

characterization, root locus and Bode plots and they should achieve zero steady state error with maximum percent-
increase students’ understanding of these topics by utilizing age overshoot 10% and settling time less than 1.5 s.
them on a physical system. ii. Record the velocity response of the controller in
3. The simple nature of proposed lab kits should encourage simulation (shown with blue in Figure 3 (b))
students for working with hardware. 3 Test the controller on hardware.
i. Implement the same controller for DC Motor Plant and
download the code to Arduino and record the velocity
3 | DETAILS OF INDIVIDUAL response (shown with red in Figure 3 (b)).
EXPERIMENTS ii. Compare the velocity response of the same controller
both for the estimated transfer function and real hardware.
This section details the three individual lab experiments, each
of which is designed to support teaching of different feedback
control theory concepts. 3.2 | Lab 2: Controller design via root locus
This lab experiment aims to present the concepts of controller
3.1 | Lab 1: DC motor identification and design via root locus as an example of adjusting control
velocity control parameters with respect to a design criterion without manual
tuning. Students use the proposed setup to design position
This lab experiments aims to present the concept of controller for the DC motor considering the sampling delay in
identification and control for a DC motor. To this end,
the system. The lab experiment requires the completion of the
students are given the DC Motor Plant block explained in following tasks:
section 2.2 to interface their control diagrams in Simulink to the
hardware. Note that students program Arduino by designing
1 Use Pade approximation to model the sampling delay.
Simulink blocks and downloading to hardware. Since the i. Use Pade approximation transfer function to approxi-
blocks run on Arduino, Serial Communication block is used to mate 10 ms delay due to 100 Hz clock frequency.
transmit the measured output to computer (Matlab/Simulink). 2 Plot the root locus of a closed-loop system.
A sample feedback control diagram (designed in Simulink and
installed to hardware) is illustrated in Figure 2. The lab
experiments require the completion of the following tasks:

1 Perform parametric system identification for a DC motor to

obtain its transfer function from input–output data.
i. Design a Simulink block diagram to apply 10V step
input to DC Motor Plant and record the output velocity
(shown with red in Figure 3 (a)).
ii. Fit a first order transfer function model to velocity
response (shown with blue in Figure 3 (a)) and obtain the
transfer function via Laplace transform.
2 Design a controller to regulate the output of the identified
plant in simulation.
i. Design a PI controller for the estimated transfer function FIGURE 3 Step response and velocity control results for
in Simulink and tune the control parameters manually to simulation and hardware
UYANIK AND CATALBAS
| 723

i. Add an integrator block to convert angular velocity 2 Observe the effect of time delay on frequency response
output of the DC Motor Plant to angular position. i. Comment on the difference in phase plots.
ii. Use a PD controller in the form Gc ðsÞ ¼ K 0:0064sþ1
0:064sþ1
with ii. Add a Pade approximation to the estimated transfer
a single, tunable gain. function for 10 ms sampling delay and compare the new
iii. Derive the characteristic equation of the closed-loop Bode plot (see Figure 5) with the previous one.
system and plot the pole locations on the complex plane
for different values of K (a sample is shown in Figure 4).
3 Find the stability range to choose optimal control parameters 4 | ASSESSMENT ON THE
i. Find the stability range for K using the Routh-Hurwitz INSTRUCTIONAL EFFECTIVENESS
and validate on the root locus that poles beyond this
gain cross real axis. In this part, some results regarding students' performances in
ii. Choose optimal pole location yielding a stable system written exams before and after the use of proposed lab setups
with maximum damping for the complex poles. are presented. To accomplish this, the exams for the last four
iii. Compute the gain to achieve desired pole locations. course offerings before and after the use of new lab setups were
iv. Apply the resulting controller to first simulation plant collected. In order to obtain a fair understanding of the effects
and then to real hardware the compare the controller of lab setups, students' marks (which were previously graded
performance. by the course instructors) for the topics mathematical modeling
and performance characteristics, root locus and Bode plots
associated to each lab session are noted.
3.3 | Lab 3: Frequency domain system Of course, each of these topics were not mentioned in all
identification exams and average student marks for each exam was also different
based on the difficulty level of the exam. Considering these
The goal of this lab experiment is to introduce frequency
observations, students’ success for each topic as a normalized
domain system identification of a linear time-invariant system
mean across different exams before and after the use of lab setups
from input–output data. Students use the proposed setup to
has been computed. Therefore, for each topic there was a related
perform data-driven system identification in frequency domain
question in M exams before the lab setups such as E1 ; E2 ; :::; EM .
by utilizing principles from linearity and time-invariance.
Let's also define the number of students took each exam as
Students complete the following tasks to achieve these goals:
N E1 ; N E2 ; :::N EM . Student's mark on the related course topic as αEj i
and on the exam as βEj i for each student and each exam was noted.
1 Use single-sine excitations to obtain Bode plots of an LTI
Note that α and β is normalized, such that full mark for each
system with input-output data
question and exam corresponds to 1. Hence, students' performance
i. Apply 15 sinusoids with each having different
for a course topic in each exam is computed as
frequencies to DC Motor Plant and record the angular
velocity data. E
ii. Use Fourier transform to find magnitude and phase 1 Ni
μ :¼ Ei ∑ αEj i
Ei
ð1Þ
changes corresponding to each frequency. N j¼1
iii. Plot the Bode plot by using obtained magnitude and
phase changes (a sample is illustrated in Figure 5). Similarly, students’ performance on each exam is
iv. Plot the Bode plot of the estimated transfer function computed as
obtained in Lab 1 using Matlab's built-in bode function
and compare it with the experimental one (see Figure 3).

FIGURE 4 Root locus with a template PD controller FIGURE 5 Bode plots of the velocity transfer function
724
| UYANIK AND CATALBAS

E
1 N i Ei the survey yielded an average of 4.58 for four questions,
ΨEi :¼ ∑β ð2Þ
N Ei j¼1 j which corresponds to somewhere in the middle of “Good” and
“Excellent” in our qualitative survey. For the following
Now, the results, μEi to obtain ΨEi ¼ 0:5 for each exam semesters, 93, 99, and 58 students yielded an average of 4.07,
are normalized to consider the effects of exam's difficulty 4.20, and 4.41, respectively.
levels. Hence, the normalized students' success for each exam
is computed as
5.2 | Difficulty level of lab components
ηEi :¼ 0:5μEi =ΨEi ð3Þ In this part, students are asked to evaluate the difficulty level
of different components in the lab session. These questions
Finally, student’ performance on each topic across all specifically ask the difficulty level of “Working with
exams before the lab setups is computed as hardware components,” “Working with Matlab/Simulink
environment,” and “Laboratory Assignments.”
M Note that, student answer about difficulty level of
∑ ηE i N E i working with hardware components for the first semester
P :¼ i¼1M ð4Þ with 90 evaluations yielded an average result of 3.92 that
∑ N Ei almost corresponds to easy (4.0) in our survey. When the
i¼1
other semesters are considered, student evaluation averages
are 3.56, 3.93, and 3.55. There are two important results that
Now, using (4), one can compute the students' perfor-
needs to be discussed here. First of all, students find it almost
mance for each course topic before and after the use of lab
easy (somewhere between fair and easy but close to easy with
setups. Figure 6 demonstrates the results of this analysis. As
a weighted average of all semesters as 3.76) to work with
seen in the Table 2, students normalized marks on these
hardware components. This supports our goal for building a
course topics increased in a noticeable manner. These results
simple hardware, where students do not deal with the details
show the instructional effectiveness of the proposed setups for
of hardware and use it as simple as possible for their
control theory courses.
experiments. The second thing that needs to be comment
on is that 3.92 and 3.93 results belong to Spring semesters,
when the course is offered in its regular semester for
5 | ASSESSMENT BASED ON
regular students. On the other hand, 3.56 and 3.55 belong to
STUDENT SURVEYS
Fall semesters when the course is offered for irregular
students. Evaluation results show that regular students find it
In this section, student surveys four the last course offerings
easier to work with hardware components. Also, only twelve
regarding the usefulness of the lab setups are presented in
students responded very difficult for this item during all
terms of relating theory to experiments and difficulty level of
semesters.
lab components.
The students answer for evaluating difficulty level of
working with Matlab/Simulink environment yielded an
average result of 4.00 for the first semester and 3.53, 3.89,
5.1 | Relating theory to experiments
and 3.52 for the following semesters. Evaluations show that
In this part, students are asked to evaluate if the lab sessions working with Matlab/Simulink is also not a big deal for
were helpful to understand the applications of theory they students. Additionally, bias between regular and irregular
learned in class. The four questions in this set specifically asks semester students are also observed in this item.
whether the experiments performed in different lab sessions
helped them to understand the use of “System Identification
via Mathematical Modeling,” “Controller Design via Root
Locus,” and “Frequency Response of Linear Time-Invariant
(LTI) Systems.” Finally, as an overall assessment, students
are asked to evaluate lab sessions' contribution to their
understanding of “Identification and Control of a DC motor.”
In order to avoid the subjectivity of students in a
quantitative analysis, students are asked to give their answers
in a qualitative manner, which are later mapped to grades 1–5,
respectively (as shown in Table 2). Interestingly, during the FIGURE 6 A comparative illustration of students’ performance
first semester of the lab sessions, where 90 students answered on each topic before and after using the lab setups
UYANIK AND CATALBAS
| 725

TABLE 2 Student Evaluations: Part I: Poor, 1; Needs Improvement, 2; Fair, 3; Good, 4; Excellent, 5. Part II: Very Difficult, 1; Difficult, 2; Fair, 3;
Easy, 4; Very Easy, 5
Questions Mean 1 (%) 2 (%) 3 (%) 4 (%) 5 (%)
PART I: Relating theory to experiments
a. System identification via mathematical modeling 4,39 1,18 1,18 8,85 35,40 53,39
b. Controller design via root locus 4,32 0,89 2,36 11,80 33,92 51,03
c. Frequency response of LTI systems 4,14 1,78 6,21 16,57 27,22 48,22
d. Identification and control of a DC motor 4,36 1,82 1,21 9,39 34,55 53,03
PART II: Difficulty level of lab components
a. Working with hardware components 3,76 2,65 7,35 29,12 32,94 27,94
b. Working with Matlab/Simulink 3,76 2,95 5,31 31,27 34,22 26,25
c. Laboratory assignments 3,52 2,35 9,41 40,59 28,82 18,83

Finally, students are asked to evaluate the difficulty level ACKNOWLEDGMENTS

of the laboratory assignments regardless of the hardware and
We would like to thank Hitay Özbay, Ömer Morgül, Orhan
Matlab/Simulink to understand their perception of difficulty.
Arıkan, A. Bülent Özgüler, Hasan Hamzaçebi, Ali Nail İnal,
Student evaluations for the four semesters yielded average
Mustafa Gül, H. Eftun Orhon, and A. Safa Öztürk for their
scores of 3.78, 3.21, 3.72, and 3.29 for the same qualitative
ideas and support. The authors appreciate the financial
difficulty measures. First of all, the bias between regular
support of the Scientific and Technological Research
and irregular students is still there for a validation. Second,
Council of Turkey (TÜBİTAK).
students find the assignments more difficult than working
with hardware or Matlab/Simulink.
ORCID

6 | CONCLUSION Ismail Uyanik http://orcid.org/0000-0002-3535-5616

Bahadir Catalbas http://orcid.org/0000-0002-8684-5754
In this paper, a novel low-cost experiments setup for feedback
control systems course is proposed. The proposed experiment
kit utilizes simple Arduino micro-controllers with Matlab/ REFERENCES
Simulink interface. The setup can be programmed easily via
1. ABET. (2012). Criteria for accrediting engineering programs,
Simulink without distracting student focus with the details of
available at http://www.abet.org/wp-content/uploads/2015/04/
micro-controller programming.
etac-criteria-2013-2014.pdf.
Students’ performance in written exams are reported 2. A. Albagul, O. O. Khalifa, and M. Wahyudi, Matlab and simulink in
to illustrate the instructional effectiveness of the proposed mechatronics education. Int. J. Eng. Educ. 21 (2005), 896–905.
lab setups. Besides, student assessments suggest that the 3. I. H. Altas and H. Aydar, A real-time computer-controlled
proposed setup is easy to use with plug and play features to simulator: For control systems. Comput. Appl. Eng. Educ. 16
Simulink. Students also find it useful for relating theory to (2008), 115–126.
experiments, since the setup supports experimental validation 4. A. Araujo et al., Integrating Arduino-based educational mobile
robots in ROS. In 2013 13th International Conference on
of some challenging tasks such as modeling and control of a
Autonomous Robot Systems (Robotica), (2013), 1–6.
DC motor, controller design via root locus and frequency 5. J. A. Arízaga et al., Automatic control for laboratory sterilization
domain system identification for an LTI system. process based on arduino hardware. In Electrical Communications
The results show that the proposed lab kit supports both and Computers (CONIELECOMP), 2012 22nd International
theoretical and practical goals of the lab sessions. In addition, Conference on (2012) (pp. 130–133). IEEE.
its low price (below 100 USD) makes it an excellent choice 6. D. S. Bernstein, Control experiments and what I learned from them:
for the universities with high student counts, especially A personal journey. IEEE Control Systems Magazine, 18 (1998),
when the goal is to give a setup to each one or two students. 81–88.
7. Y. Buğday and M. Ö. Efe, A set of computer aided automatic
The Matlab/Simulink interface supports easy programming
control experiments for undergraduate students. Comput. Appl.
in a block-diagram based environment. Also, this interface Eng. Educ. 21 (2013), 300–312.
eases the transition to proposed hardware, since most 8. B. Çatalbaş and İ. Uyanık, A low-cost laboratory experiment setup
engineering colleges use Matlab/Simulink in their education for frequency domain analysis for a feedback control systems
programs. course. IFAC-PapersOnLine, 50 (2017), 15704–15709.
726
| UYANIK AND CATALBAS

9. Y. C. Chen and J. M. Naughton, An undergraduate laboratory 29. E. E. Topcu, Z. Kamis, and I. Yuksel, Automatic control training
platform for control system design, simulation, and implementa- on a DC motor control module. IFAC Proc. Vol. 36 (2003),
tion. IEEE Cont. Syst. 20 (2000), 12–20. 269–274.
10. I. Colak et al., A novel web-based laboratory for DC motor 30. S. Uran and K. Jezernik, MATLAB web server and M-file
experiments. Comput. Appl. Eng. Educ. 19 (2011), 125–135. application. In 2006 12th International Power Electronics and
11. E. Erzan Topçu, PC-based control and simulation of an electro- Motion Control Conference, IEEE, 2006, pp. 2088–2092.
hydraulic system. Comput. Appl. Eng. Educ. 25 (2017), 706–718. 31. N. Wang et al., Integration of a remote PID motor speed control
12. R. M. Felder, Reaching the Second Tier: Learning and Teaching experiment with teaching in engineering education. Eng. Edu. Lett.
Styles in College Science Education. J. Col. Sci. Teach. 23 (1993), 2017 (2017), 1.
286–290. 32. M. Zolkapli et al., High-efficiency dual-axis solar tracking
13. S. Fragoso et al., Educational software tool for decoupling control developement using Arduino. 2013 International Conference
in wind turbines applied to a lab-scale system. Comput. Appl. Eng. on Technology, Informatics, Management, Engineering and
Educ. 24 (2016), 400–411. Environment, IEEE, 2013, pp. 43–47.
14. F. D. Gonzalez, J. G. Guarnizo, and G. Benavides, Emulation
system for a distribution center using mobile robot, controlled by
artificial vision and fuzzy logic. IEEE Latin America Transactions, I. UYANIK is a postdoctoral researcher
12 (2014), 557–563. in Laboratory of Computational Sen-
15. G. C. Goodwin et al., Emulation-based virtual laboratories: A low- sing and Robotics (LCSR) at Johns
cost alternative to physical experiments in control engineering Hopkins University. He received
education. IEEE Transactions on Education. 54 (2011), 48–55. his PhD degree in Electrical and
16. G. Guo and W. Yue, Autonomous platoon control allowing Electronics Engineering from Bilkent
range-limited sensors. IEEE Trans. Vehi. Technol. 61 (2012),
University in May 2017. He also re-
2901–2912.
17. E. Guzmán-Ramírez et al., An educational tool for designing DC
ceived his BSc and MSc degrees from
motor control systems through FPGA-based experimentation. Int. J. the same department in June 2009 and August 2011, resp-
Elec. Eng. Educ. 52 (2015), 22–38. ectively. His research focuses on discovering the principles
18. M. Kalúz et al., ArPi lab: A low-cost remote laboratory for control of animal locomotion by developing novel techniques in the
education. IFAC Proc. Vol. 47 (2014), 9057–9062. areas of system identification theory, computational neuro-
19. Z. Kamis, E. E. Topcu, and I. Yuksel, Computer-aided automatic science and robotics. He is also the recipient of the Aselsan
control education with a real-time development system. Comput. PhD Fellowship.
Appl. Eng. Educ. 13 (2005), 181–191.
20. N. A. Kheir et al., Control systems engineering education.
B. CATALBAS is a PhD candidate in
Automatica. 32 (1996), 147–166.
Electrical and Electronics Engineer-
21. T. Michałowski, Applications of MATLAB in Science and
Engineering. Published by InTech Janeza Trdine, 9 (2011), 51000.
ing Department at Bilkent Univer-
22. S. Papert, Mindstorms: Children, computers, and powerful ideas. sity. He also received his BSc and
Basic Books, Inc. (1980). MSc degrees from the same depart-
23. A. Regueiro et al., A practice for engineering students based on the ment in June 2013 and September
control and monitoring an experimental biomass combustor using 2015, respectively. He is working
labview. Comput. Appl. Eng. Educ. 25 (2017), 392–403. in developing novel deep learning
24. P. Reguera et al., A low-cost open source hardware in control methods and he also interests in legged locomotion
education. case study: Arduino-Feedback MS-150. IFAC-Papers-
applications of deep learning. He is the recipient of the
OnLine. 48 (2015), 117–122.
25. A. Rmilah et al., A PC-based simulation platform for a quadcopter
Scientific and Technological Research Council of Turkey
system with self-tuning fuzzy PID controllers. Comput. Appl. Eng. (TÜBİTAK) Graduate Scholarship.
Educ. 24 (2016), 934–950.
26. J. Sánchez, S. Dormido, and F. Esquembre, The learning of control
concepts using interactive tools. Comput. Appl. Eng. Educ. 13 (2005), How to cite this article: Uyanik I, Catalbas B.
84–98.
A low-cost feedback control systems laboratory
27. J. Sarik and I. Kymissis, Lab kits using the Arduino prototyping
setup via Arduino–Simulink interface. Comput Appl
platform. In Frontiers in Education Conference (FIE), 2010 IEEE
(2010) (pp. T3C-1). IEEE. Eng Educ. 2018;26:718–726.
28. A. A. Taimoor, Virtualization of the process control laboratory using https://doi.org/10.1002/cae.21917
ASPEN HYSYS. Comput. Appl. Eng. Educ. 24 (2016), 887–898.