1
979-8-3503-1576-9/23/$31.00 ©2023 IEEE
Tracking Control of an Inverted Pendulum System:
A Novel Radial basis function Neural Network
supervisory control approach
Narayan Nayak
Department of EE
SIT, Bhubaneswar
narayan@silicon.ac.in
Binayak Ghosal
Department of EEE
SIT, Bhubaneswar
binayak5766@gmail.com
Dipak Ranjan Nayak
Department of EEE
SIT, Bhubaneswar
deepak.ranjan@silicon.ac.in
Ambarish G. Mohapatra
Department of EE
SIT, Bhubaneswar
ambarish.mahapatra@silicon.ac.in
Abstract – The control of an Inverted Pendulum (IP) is one
of the benchmark problems in control systems. It is very
difficult to control due to its innate instability and
extremely nonlinear properties. Here, a novel supervisory
control method based on Radial Basis Function Neural
Networks (RBFNN) approach aims to enhance the
tracking performance of an IP system. The RBFNN is
used as a dynamic compensator to improve the
performance of the typical Proportional-Integral-
Derivative (PID) controller. The network is trained to
approximate the unknown nonlinear dynamics of the IP
system. The corrective control signal is then produced by
the trained network and added to the PID controller's
output. An adaptive tuning algorithm is created to
continuously update the RBFNN's parameters to achieve
accurate tracking control. As a result, the system can
instantly adapt to modifications in the dynamics of the
plant. The effectiveness of the suggested approach in
achieving reliable tracking control of the IP system is
demonstrated by simulation results. When compared to
the traditional PID and FUZZY-PID controller, the
RBFNN supervisory control strategy significantly
enhances the system's tracking performance with
minimum overshoot and settling time. The method offers
a promising framework for the control of complex
nonlinear systems.
Keywords –Inverted Pendulum(IP), PID, RBFNN, Fuzzy-PID,
RBFNN-PID
I. INTRODUCTION
Control engineering has paid a lot of attention to the difficult
task of designing and controlling an inverting pendulum(IP)
system[1-2]. An IP is a pendulum system whose objective is
to keep the pendulum in an inverted position, that is, upright
or vertically aligned, despite outside disturbances. Engineers
can learn more about the fundamentals of control theory and
apply them to a variety of real-world situations by
successfully designing a controller that can stabilize an IP.
The pendulum, when left to itself, will naturally fall due to
gravity. However, through the application of external forces
or torques, The inverted pendulum concept finds numerous
applications across various fields such as robotics, control
system design, gyroscopic stabilization systems,
transportation systems, etc. Control engineers and
researchers use it to test and compare different control
strategies, such as PID controllers [3],robust PID[1][4], fuzzy
logic [5-9], adaptive control, and advanced optimization-
based methods. The inverted pendulum provides a practical
and intuitive platform for studying feedback control, state
estimation, and stability analysis. Multiple PID controllers
are employed to stabilize the angle of the inverted pendulum
[2]. The pendulum can move in a straight line, horizontal
plane as well as vertical plane. However, the results are not
desirable due to the high non-linearity of the system. A fuzzy
parallel distributed compensation (PDC) controller is
proposed for the angle stabilization of the IP [4]. The settling
time was greatly reduced with the use of these controllers.
PID and fuzzy controllers are used for the stabilization and
control of an IP moving on an inclined surface [7]. The angle
of the pendulum and the position of the cart were controlled
by using Linear Quadratic Regulator (LQR) and Pole
Placement control strategy [8].Linearization of the inverted
pendulum dynamics is done with the use of Taylor series
approximation. RBFNN are used for the position tracking of
an inverted pendulum system [10-11]. In this paper, we have
proposed a control strategy for the inverted pendulum
problem through the use of RBFNN along with a
Proportional-Integral-Derivative (PID) controller to control
the IP system. The main objective of this study is to give a
thorough overview of the controller design methods for
inverting pendulum systems and to discuss the advantages
and disadvantages of various methods. The paper will also
highlight the trade-offs between simplicity, robustness, and
performance as well as the difficulties in applying these
control strategies.
This paper is divided into five sections. An
introduction and a literature review are included in Section I.
In section II, inverting pendulum systems with transfer
functions and state space models were mathematically
analyzed. Section III covered the various control strategies.
Results and discussion were completed in section IV.
Concluding remarks were provided in section V at the end.
II. MATHEMATICAL ANALYSIS OF INVERTED PENDULUM
SYSTEM
An inverted pendulum (IP) system is a simple model for
understanding the complex and dynamic systems that are
available in the real world. Classical control techniques are
not employed for the control of an IP system due to its
nonlinear control. This is a result of having 2 degrees of
freedom, namely the cart position and angle of the inverted
2023 1st International Conference on Circuits, Power and Intelligent Systems (CCPIS) | 979-8-3503-1576-9/23/$31.00 ©2023 IEEE | DOI: 10.1109/CCPIS59145.2023.10291515
Authorized licensed use limited to: Silicon Institute of Technology. Downloaded on October 30,2023 at 05:20:40 UTC from IEEE Xplore. Restrictions apply.