Journal of Mechatronics, Electrical Power, and Vehicular Technology 13 (2022) 15-23

Journal of Mechatronics, Electrical Power,

and Vehicular Technology
e-ISSN: 2088-6985
p-ISSN: 2087-3379
mev.lipi.go.id

Design and application of models reference adaptive control

(MRAC) on ball and beam
Muhammad Zakiyullah Romdlony a, Muhammad Ridho Rosa a, *,
a b
Edwin Muhammad Puji Syamsudin , Bambang Riyanto Trilaksono ,
Agung Surya Wibowo a, c
a
School of Electrical Engineering, Telkom University
Jl. Telekomunikasi Terusan Buah Batu, Bandung, 40257, Indonesia
b
School of Electrical Engineering and Informatics, Bandung Institute of Technology
Jl. Ganesha, Bandung, 40132, Indonesia
c
Department of Electronics Engineering, NSCL Laboratory, Jeonbuk National University
567 Baekje-daero, Deokjin-gu, Jeonju-si, Jeollabuk-do, South Korea

Received 29 September 2021; Revised 29 November 2021; Accepted 8 December 2021; Published online 29 July 2022

Abstract
This paper presents the implementation of an adaptive control approach to the ball and beam system (BBS). The dynamics
of a BBS are non-linear, and in the implementation, the uncertainty of the system's parameters may occur. In this research, the
linear state-feedback model reference adaptive control (MRAC) is used to synchronize the states of the BBS with the states of
the given reference model. This research investigates the performance of the MRAC method for a linear system that is applied
to a non-linear system or BBS. In order to get a faster states convergence response, we define the initial condition of the
feedback gains. In addition, the feedback gains are limited to get less oscillation response. The results show the error
convergence is improved for the different sets of the sinusoidal reference signal for the MRAC with modified feedback gains.
The ball position convergence improvement of MRAC with modified feedback gains for sinusoidal reference with an amplitude
of 0.25, 0.5, and 0.75 are 35.1 %, 36 %, and 52.4 %, respectively.
Copyright ©2022 National Research and Innovation Agency. This is an open access article under the CC BY-NC-SA license
(https://creativecommons.org/licenses/by-nc-sa/4.0/).
Keywords: model reference adaptive control; modified feedback gains; ball and beam system.

I. Introduction Adaptive control is an advanced control method

with parameter adjustments that can regulate the
In designing the control of a system, one can use states or output of the uncertain system to track a
the simulation and experiment to see the certain value [3]. In adaptive control, the unknown
effectiveness of the proposed method. However, this system is expected to converge its states or output to
approach may cost a lot due to the miscalculation in reference model states or output. The model
the control system design. Hardware in-the-loop reference adaptive control (MRAC) makes the
(HIL) is a simulation technique performed by unknown system dynamics similar to the reference
combining hardware and software in the process [1]. model dynamics [4][5]. The MRAC can be
Implementing HIL will facilitate the testing process categorized into two types based on how its
and reduce the level of errors or failures that occur estimates the unknown parameters, indirect and
as well as the costs required in the design of the direct MRAC [6][7]. The author uses the state-
control system [2]. In this research, we use a real feedback direct MRAC control method with modified
plant BBS and the controller in MATLAB, which is the feedback gains to get fast adaptation. Recent
opposite of the HIL scheme. Using this approach, one research on fast adaptation in adaptive control can
can directly implement the proposed control be found in [8].
method that is designed in MATLAB to a real plant. The BBS is one of the most widely used examples
of control systems application in control engineering.
The primary purpose of BBS is to track the ball to the
* Corresponding Author. Phone: +62-81802228261 commanded position by designing a particular
E-mail address: [email protected]

doi: https://dx.doi.org/10.14203/j.mev.2022.v13.15-23
2088-6985 / 2087-3379 ©2022 National Research and Innovation Agency
This is an open access article under the CC BY-NC-SA license (https://creativecommons.org/licenses/by-nc-sa/4.0/)
MEV is Scopus indexed Journal and accredited as Sinta 1 Journal (https://sinta.kemdikbud.go.id/journals/detail?id=814)
How to Cite: M.Z. Romdlony et al., “Design and application of models reference adaptive control (MRAC) on ball and beam,” Journal of
Mechatronics, Electrical Power, and Vehicular Technology, vol. 13, no. 1, pp. 15-23, July 2022.
 16 M.Z. Romdlony et al. / Journal of Mechatronics, Electrical Power, and Vehicular Technology 13 (2022) 15-23

control system [9]. This research uses the direct

method MRAC for a linear system that is
implemented to BBS, a non-linear system. In other
studies, various adaptive methods for BBS systems
have been carried out. The control for the ball and
beam system that combines the conventional
dynamic surface control and the adaptive fuzzy
scheme is proposed for the equilibrium balance of
the ball [10]. There are also designing a model
reference adaptive control system using the MIT rule
to control a ball and beam system so the plant could
track the reference model [11]. The comparison
between the integer and the fractional controller for
BBS has been discussed and tested in [12]. It is
known that a mechatronic system has limited
control input due to actuator limitations. The
parameter projection algorithm is used to solve the
control saturation that may lead to undesirable
results in adaptive controllers [13]. Compared with
most adaptive control of BBS literature, this work Figure 1. Ball and beam dynamics, where m is ball mass, 𝑀 is rod
mass, 𝑑 is offset arm length, 𝑔 is gravitational acceleration, 𝑅 is ball
shows the effectiveness of direct state-feedback radius, 𝑟 is ball position, 𝜃 is angle of servo, and L is length beam
MRAC both in simulation and in an experiment as a
𝑑
real-time controller of the BBS. In addition, we It is known that 𝛼 = 𝜃 , so that we have the
𝐿
proposed predefined feedback gains to have a faster following BBS dynamics equation (3)
convergence rate and feedback gains saturation to
𝑗𝑏 𝑑
get a less oscillated response shown by a small state � + 𝑚� 𝑟̈ = 𝑚𝑚 θ (3)
𝑅2 𝐿
error value.
Using the differential equation in equation (3),
the dynamics of the second-order BBS modeling
II. Materials and Methods system in the state-space form can be defined as
A. Ball and beam system equation (4)

The BBS objective is to keep the ball's position in 0

𝑟̇ 0 1 𝑟
the desired location by connecting the beam to a � �=� � � � + � 𝑚𝑚𝑚 �𝜃 (4)
𝑟̈ 0 0 𝑟̇ 𝐽
𝐿� 2 +𝑚�
servo motor. The ball's position is determined from 𝑅

the edge of the beam. The BBS configuration can be where 𝑟̇ is velocity of the ball and 𝑟̈ is acceleration of
seen in Figure 1. the ball.
The motion of the ball can be found by using
B. Model reference adaptive control (MRAC)
Newton’s law that satisfies the following equation
(1) The model reference adaptive control (MRAC) is
one of the adaptive control methods which aims to
𝑚𝑚 sin 𝛼 − 𝐹𝑓 = 𝑚𝑟̈ (1)
solve control problems with limited parameters to
𝑗
where 𝐹𝑓 = 𝑏2 𝑟̈ is the frictional force, 𝑗𝑏 is the compensate for unknown system parameters by
𝑅
adapting the characteristics of the stable reference
moment of inertia of the ball, and 𝛼 is the deflection
model. Thus the system has the same characteristics
of the beam. Here we are assuming that 𝛼 is very
similar to the reference model. In this study, direct
small. Thus we have linearized BBS motion in
MRAC was used [14]. Figure 2 shows the structure of
equation (2)
the direct MRAC.
𝑚𝑚 𝛼 −
𝑗𝑏
𝑟̈ = 𝑚𝑟̈ (2) In direct state-feedback MRAC, the following
𝑅2
equations are used:

Figure 2. Direct state-feedback MRAC

 M.Z. Romdlony et al. / Journal of Mechatronics, Electrical Power, and Vehicular Technology 13 (2022) 15-23 17

Figure 3. Block diagram system

Unknown system using equation (5) • The states response from the BBS is measured
using an infrared sensor and sent back to
𝑥̇ = 𝐴𝐴 + 𝐵𝐵 (5)
MATLAB via Arduino Uno. The results are the
Reference model using equation (6) states of the BBS, the position and the speed of
the ball.
𝑥̇ 𝑚 = 𝐴𝑚 𝑥𝑚 + 𝐵𝑚 𝑟 (6)
• The ball moves on the beam according to the
Lyapunov equation using equation (7) given control and adapts the reference model
that was designed previously.
𝐴′𝑚 𝑃 + 𝑃𝐴𝑚 = −𝑄, 𝑄 > 0 (7) • This process will continue until the SIMULINK
Adaptive Laws using equation (8) and (9) running time finish.

𝑘�̇ = 𝛾1 𝐵′ 𝑚 𝑃𝑃(𝑥 − 𝑥𝑚 )𝑥 ′ sgn(𝑙 ∗ ) (8) D. System flowchart

The system starts by initializing its parameters
𝑙̃̇ = −𝛾2 𝐵′ 𝑚 𝑃𝑃(𝑥 − 𝑥𝑚 )𝑟 sgn(𝑙 ∗ ) (9)
and adjusting the beam's position in its equilibrium
Control Law using equation (10) state. After the setup has been set, we run the
program that triggers the infrared sensor to detect
𝑢 = −𝑘𝑘 + 𝑙𝑙 (10) the ball's position. The measured states are used as
where 𝐴 is state matrix of unknown system, 𝐴𝑚 is the inputs for the MRAC besides the reference signal.
state matrix of model reference, 𝐵 is input matrix of The adaptive laws (8) and (9) will adaptively
unknown system, 𝐵𝑚 is input matrix of model calculate the feedback gain. The output of the MRAC
reference, r is input reference, u is control law, x is or the control law will be sent to the servo motor.
system’s states, 𝑥𝑚 is reference model’s states, Figure 4 shows the diagram of the entire BBS system.
𝑄 & 𝑃 is matrix positive definite, 𝑒 is error in state E. State feedback direct MRAC setup
feedback, 𝛾1 & 𝛾2 is adaptive gains, and 𝑘 & 𝑙 is
feedback gains. The BBS parameters are defined in MATLAB to
facilitate the simulation of the BBS. The state-space
C. System setup modeling of BBS is designed in SIMULINK, where the
Unlike the standard hardware in-the-loop (HIL) BBS parameters are defined as follows:
scheme that uses a real controller to control the • m = 0.148 kg
2
virtual system, we proposed the opposite of the HIL • g = 9.8 m/s
scheme. In this research, the controller is MATLAB, • L = 0.35 m
and the system/BBS is a real plant. Figure 3 shows Then the state-space form of the BBS system in
the detailed system block diagram proposed in this equation (4) can be rewritten in equation (11)
research. 𝑟̇ 0 1 𝑟 0
The following is an explanation of the block � �=� �� � + � �𝜃 (11)
𝑟̈ 0 0 𝑟̇ 0.75
diagram in Figure 3:
and the stable model reference dynamics in the
• The initial condition is given by the user as an
state-space form is defined as equation (12)
initial state of the BBS in SIMULINK MATLAB.
• The initial information is processed in 0 1 𝑥1 0
𝑥̇ 𝑚 = � � � 𝑚� + � � 𝑟 (12)
SIMULINK MATLAB, in which MRAC control is −0.1 −0.2 𝑥2 𝑚 1
designed. Equation (12) is the reference model dynamics
• The resulting control signal is sent to the ball where the poles are on the left half plane and located
and beam system via Arduino Mega. Then the near the origin. We used the adaptive law in
servo motor will move the beam according to 0.00025
equations (8) and (9), where 𝛾1 = � � , 𝛾2 =
the control signal command. 0.0025
[−0.0025], and Q = �1 0�.
0 1
 18 M.Z. Romdlony et al. / Journal of Mechatronics, Electrical Power, and Vehicular Technology 13 (2022) 15-23

III. Results and Discussions converge to the states of the model reference with
small errors at different timescales marked by
A. State feedback direct MRAC on ball and beam dotted lines. Therefore, we can take the values the
system experiment feedback gains, 𝑘 and 𝑙, at that time range as the
First, we define the initial ball position at 17.5 cm initial and saturation values.
from the edge of the beam, the sinusoidal frequency B. The influence of the Initial definition and
is 0.005 rad/sec, the sinusoidal bias is 1.75, and the saturation value on the BBS system
SIMULINK run time is 10,000 seconds. The responses
based on the simulation and experiment of the From the previous results, we get the information
proposed MRAC design for BBS can be seen in to define the initial value and the saturation value of
Figure 5. In the experiment (blue line), the state the feedback gains 𝑘 and 𝑙. For sinusoidal reference
error between simulation and experiment is 7.65 % signal with an amplitude of 0.25, we have 𝑘1 ∈ [5,10] ,
for an amplitude of 0.25, 13.41 % for an amplitude of 𝑘2 ∈ [50,100], and 𝑙 ∈ [40,80]. For sinusoidal reference
0.5, and 15.14 % for an amplitude of 0.75. It can be signal with an amplitude of 0.5, we have 𝑘1 ∈ [8,10],
seen in Figure 6 that the states of the BBS can 𝑘2 ∈ [35,98], and 𝑙 ∈ [40,80]. For sinusoidal reference
signal with an amplitude of 0.75, we have 𝑘1 ∈ [5,12],
𝑘2 ∈ [40,140], and 𝑙 ∈ [40,100].
The responses, simulation and experiment, of the
proposed MRAC with modified feedback gains 𝑘 and
𝑙 can be seen in Figure 6. The yellow signal color
indicates the reference signal, the blue color
indicates the BBS experimental signal, and the red
dotted line indicates the simulation signal. In the
experiment (blue line), the states error between
simulation and experiment is 4.97 % for an
amplitude of 0.25, 8.57 % for an amplitude of 0.5,
and 7.21 % for an amplitude of 0.75.
Table 1 shows the state error (error position of
the ball) value before and after we define the initial
and the saturation values of the feedback gains 𝑘 and
𝑙. It can be concluded that the modified feedback
gains 𝑘 and 𝑙 decreased the error value by 2.68 % in
the case of sinusoidal with an amplitude of 0.25,
4.84 % in the case of sinusoidal with an amplitude of
0.5, and 7.93 % in the case of sinusoidal with an
amplitude of 0.75. Figure 7 shows the comparison of
the ball position when using the MRAC with
modified feedback gains and standard MRAC.
Table 2 shows the performance improvement in
terms of ball position error after the modification of
the initial and the saturation values of feedback
gains 𝑘 and 𝑙. It can be concluded that the modified
feedback gains 𝑘 and 𝑙 gives the biggest
improvement in terms of ball position error when
the sinusoidal amplitude is 0.75. The smaller
amplitude gives a lower performance improvement.

Table 1.
Comparison of the difference of ball position error in the
experiment

𝒌 and 𝒍 (% Modified 𝒌 The difference

Amplitude
error) and 𝒍 (% error) (% error)
0.25 7.65 % 4.97 % 2.68 %
0.5 13.41 % 8.57 % 4.84 %
0.75 15.14 % 7.21 % 7.93 %

Table 2.
Performance improvement MRAC with modified feedback gains 𝒌
and 𝒍 in the experiment

Performance improvement with modified 𝒌

Amplitude
and 𝒍 (in percentage)
0.25 35.1 %
0.5 36 %
0.75 52.4 %
Figure 4. BBS flowchart
 M.Z. Romdlony et al. / Journal of Mechatronics, Electrical Power, and Vehicular Technology 13 (2022) 15-23 19

Figure 5. Comparison of simulations and experiments of BBS without initial definition and saturation values at the amplitude of (a) 0.25; (b) 0.5;
and (c) 0.75

Figures 5 and 6 show the simulation responses disturbances. In the experiment, the non-linear
are better than the experimental responses. This dynamics cannot be neglected, but it can be seen
happens because the BBS model in simulation is that the MRAC with modified feedback gains gives a
defined as a linear system and is unaffected by any better response or synchronization.
 20 M.Z. Romdlony et al. / Journal of Mechatronics, Electrical Power, and Vehicular Technology 13 (2022) 15-23

Figure 6. Comparison of simulations and experiments of BBS with initial definition and saturation values at the amplitude of (a) 0.25; (b) 0.5;
and (c) 0.75

Table 3 shows the ball position error both in different response error between the simulation and
simulation and experiment. It is shown that the experiment of less than 10 %. We have different
designed MRAC with modified feedback gain gives a responses error between the simulation and
 M.Z. Romdlony et al. / Journal of Mechatronics, Electrical Power, and Vehicular Technology 13 (2022) 15-23 21

Figure 7. Comparison of the ball position when using the MRAC with modified feedback gains and standard MRAC in the experiment at the
amplitude of (a) 0.25; (b) 0.5; and (c) 0.75

experiment in the case of sinusoidal with an into account the input saturation [15], where we
amplitude of 0.25, 0.5, and 0.75 equal to 4.37 %, know that the BBS system has limited servo
6.37 %, and 6.56 %, respectively. Future work may actuation.
include implementing adaptive control that takes
 22 M.Z. Romdlony et al. / Journal of Mechatronics, Electrical Power, and Vehicular Technology 13 (2022) 15-23

Table 3.
Comparison of the ball position error values in simulations and experiments BBS

Amplitude Simulation (% error) Experiment (% error) Difference between simulation and experiment (% error )
0.25 0.004 % 7.65 % 7.646 %
0.25-modified 𝒌 and 𝒍 0.6 % 4.97 % 4.37 %
0.5 0.0036 % 13.41 % 13.4064 %
0.5-modified 𝒌 and 𝒍 2.2 % 8.57 % 6.37 %
0.75 0.004 % 15.14 % 15.136 %
0.75-modified 𝒌 and 𝒍 0.65 % 7.21 % 6.56 %

Additional information
IV. Conclusion
Reprints and permission: information is available at
In this research, we have shown that the https://mev.lipi.go.id/.
modified feedback gains are able to make the system Publisher’s Note: National Research and Innovation
performance better, which is shown by a smaller Agency (BRIN) remains neutral with regard to jurisdictional
error value. This work shows the experiment result claims in published maps and institutional affiliations.
of the MRAC and we proposed the modified feedback
gains 𝑘 and 𝑙 . The state-feedback MRAC with References
modified feedback gains 𝑘 and 𝑙 experiment resulted [1] P. Kotsampopoulos et al., "A benchmark system for hardware-
in a smaller ball position error with lower error in-the-loop testing of distributed energy resources," IEEE
percentage value of 2.68 % for the sinusoidal Power and Energy Technology Systems Journal, vol. 5, no. 3, pp.
94-103, Sept. 2018.
amplitude of 0.25, 4.84 % for the sinusoidal
[2] K. S. Amitkumar, R. S. Kaarthik, and P. Pillay, "A versatile
amplitude of 0.5, and 7.93 % for the sinusoidal power-hardware-in-the-loop-based emulator for rapid testing
amplitude of 0.75. The performance improvement of transportation electric drives," IEEE Transactions on
with modified 𝑘 and 𝑙 (in percentage) is 35.1 % for Transportation Electrification, vol. 4, no. 4, pp. 901-911, Dec.
2018.
the sinusoidal amplitude of 0.25, 36 % for the [3] J. E. Gaudio, A. M. Annaswamy, E. Lavretsky, and M. Bolender,
sinusoidal amplitude of 0.5, and 52.4 % for the "Parameter estimation in adaptive control of time-varying
sinusoidal amplitude of 0.75. Comparison of the systems under a range of excitation conditions," IEEE
Transactions on Automatic Control, 2021.
simulation ball position errors is better in the [4] G. Lymperopoulos and P. Ioannou, "Model reference adaptive
simulation. This happens due to the simplified BBS control for networked distributed systems with strong
model and the absence of any disturbances. The interconnections and communication delays," J Syst Sci
Complex 31, pp. 38–68, 2018.
modification of the feedback gains gives better ball [5] T. Yucelen and W. M. Haddad, "Low-frequency learning and
position convergence to the reference model, but a fast adaptation in model reference adaptive control," IEEE
testing case is required to get the information on the Transactions on Automatic Control, vol. 58, no. 4, pp. 1080-
initial values and saturation values of the feedback 1085, April 2013.
[6] C. Yuquan, W. Yiheng, L. Shu, W. Yong, "Indirect model
gains. reference adaptive control for a class of fractional order
systems," Communications in Nonlinear Science and
Numerical Simulation, Volume 39, pp. 458-471, 2016, ISSN
Acknowledgment 1007-5704.
[7] I. Barkana, "Simple adaptive control–a stable direct model
The authors express their gratitude to the reference adaptive control methodology–brief survey",
Ministry of Research and Technology of the Republic International Journal of Adaptive Control and Signal Processing,
of Indonesia (Ristekdikti) for providing the Post Volume 28, Issue 7-8, 2014, pp. 567-604.
doctoral research at Bandung Institute of Technology [8] Joseph E. Gaudio, Anuradha M. Annaswamy, Eugene Lavretsky
and Michael A. Bolender. "Fast parameter convergence in
(ITB). adaptive flight control," AIAA 2020-0594. AIAA Scitech 2020
Forum. January 2020.
[9] M.F. Rahmat, H. Wahid, and N.A. Wahab, "Application of
Declaration intelligent controller in a ball and beam control system,"
International journal on smart sensing and intelligent systems,
Author contribution 3(1), pp. 45-60, 2017.
[10] Y. H. Chang, W. S. Chan, and C. W. Chang, “T-S fuzzy model-
All authors contributed equally as the main contributor
based adaptive dynamic surface control for ball and beam
of this paper. All authors read and approved the final paper.
system,” IEEE Trans. Ind. Electron., vol. 60, no. 6, pp. 2251–
2263, 2013.
Funding statement
[11] P. Jain and M. J. Nigam, “Real time control of ball and beam
This research did not receive any specific grant from system with model reference adaptive control strategy using
funding agencies in the public, commercial, or not-for- MIT rule,” 2013 IEEE Int. Conf. Comput. Intell. Comput. Res.
(IEEE ICCIC 2013), no. 4, pp. 4–7, 2013.
profit sectors.
[12] I. M. Mehedi, U. M. Al-Saggaf, R. Mansouri, and M. Bettayeb,
"Two degrees of freedom fractional controller design:
Competing interest
application to the ball and beam system," Measurement,
The authors declare that they have no known Volume 135, pp. 13-22, 2019, ISSN 0263-2241.
[13] S. S. Tohidi, Y. Yildiz, and I. Kolmanovsky, "Adaptive control
competing financial interests or personal relationships that
allocation for constrained systems," Automatica, Volume 121,
could have appeared to influence the work reported in this 109161, 2020, ISSN 0005-1098.
paper. [14] P. Ioannou and B. Fidan, "Adaptive control tutorial," Advance in
design and control, 2006.
 M.Z. Romdlony et al. / Journal of Mechatronics, Electrical Power, and Vehicular Technology 13 (2022) 15-23 23

[15] M. R. Rosa, “Leader-follower synchronization of uncertain

Euler-Lagrange dynamics with input constraints,” Aerospace,
vol. 7, no. 9, 2020.