Tuning of pid controller of inverted pendulum using genetic algorithm

IJRET: International Journal of Research in Engineering and Technology ISSN: 2319-1163
__________________________________________________________________________________________
Volume: 01 Issue: 03 | Nov-2012, Available @ http://www.ijret.org 359
TUNING OF PID CONTROLLER OF INVERTED PENDULUM USING
GENETIC ALGORITHM
P. Kumar1
, O.N. Mehrotra
2
, J. Mahto3
Abstract
The paper presents two different ways of mathematical modeling of Inverted Pendulum. A Proportional-Integral-Derative (PID)
controller is designed for its stabilization. Some reference stable system is selected after designing of PID controller to optimize
different types of error using Genetic Algorithms. The proposed system extends classical inverted pendulum by incorporating two
moving masses. A tuning mechanism is implemented by genetic algorithm for optimizing different gains of controller parameter. Also,
different performance indices are calculated in MATLAB environment. This paper exhibits to demonstrate the capability of genetic
algorithm to solve complex and constraint optimization problems and as a general purpose optimization tool to solve control system
design problems.
Keyword-Inverted pendulum,Mathematical modelling,swing up control ,PID controller,Tuning,Genetic
Algorithm,Performance Indeces,Error minimization.
--------------------------------------------------------------------*****--------------------------------------------------------------------
1.INTRODUCTION
The inverted pendulum may be viewed as a classical problem in
dynamics and control theory[1,9] and is widely used as a
benchmark[16] for testing control algorithms(PID controllers,
Linear Quadratic Regulator (LQR), neural networks, fuzzy logic
control, genetic algorithms, etc)[7,8].The inverted pendulum is
unstable[11] in the sense that it may fall any time in any
direction unless a suitable control force is applied. The control
objective of the inverted pendulum is to swing up[4] the
pendulum hinged on the moving cart by a linear motor[12] from
stable position (vertically down state) to the zero state(vertically
upward state) and to keep the pendulum in vertically upward
state in spite of the disturbance[5,13].
In the field of engineering and technology the importance of
benchmark [14,7] needs no explanation. They make it easy to
check whether a particular algorithm [6] yields the requisite
results.
Several work has been reported on the inverted pendulum for its
stabilization. Attempts have been made in the past to control it
using classical control [3]. In this paper, the controller is
proposed to be tuned using Genetic Algorithm(GA)[15]
technique.. Genetic Algorithm[2,10] have been shown to be
capable of locating high performance areas in complex domains
without experiencing difficulties, even associated with high
dimensionality or false optima as many occurring with some
other optimization method.
2. MECHANICAL CONSTRUCTION
The system comprises of a horizontal plate connected to two
wheels through a connecting rod. The wheels are independent of
each other and are placed in the centre of the rail. The platform
thus can move on a horizontal surface and is able to rotate about
the axis of wheels. There are two masses on top of the system
that can slide along the horizontal rail, the masses being on both
sides of the rail. The system is shown in fig (1)
3. MATHEMATICAL MODEL OF THE PLANT
Defining the angle of the rod from the vertical (reference) line
as θ , displacement of the cart as x, assuming the force applied
to the system be F , centre of gravity of the pendulum rod is at
its geometric centre and l be the half length of the pendulum rod,
the physical model of the system is shown in fig (2).

__________________________________________________________________________________________
The Lagrangian of the entire system is given as,
L=
1
2
(mx2 +2mlx cos�+ml2
θ2+Mx2)+
1
2
Iθ2 ]-mglcos�
The Euler-Lagrange’s equation for the system is :
d
dt
δL
δx
−
δL
δx
+ bx = F
d
dt
δL
δθ
−
δL
δθ
+ dθ = 0
The dynamics of the entire system using above equation is
I + ml2
θ + ml cos θ x − mgl sin θ + dθ = 0 ……….(1)
M + m x + ml cos θ θ − ml sin θ θ2
+ bx = F ……..(2)
𝑥1 = 𝑥 + 𝑙 sin 𝜃
𝑦1 = 𝑙 cos 𝜃 𝜃 𝑚𝑔 𝑙
F 𝑀 𝑥
Fig 2 : The Inverted Pendulum System
In order to derive the linear differential equation model, the non
linear differential equation obtained need to be linearized. For
small angle deviation around the upright equilibrium (fig.2)
point, assumption made
sin θ = θ, cos θ = 1, θ2
= 0
Using above relation , equation (3) and (4) is derived.
rθ+qx-kθ+d =0………..(3)
px + qθ + bx = F …….(4
)
Where, (M + m)= p, mgl=k, ml=q, I + ml2
= r
Eq (3&4) is the linear differential equation model of the entire
system.
Laplace transform model of the system is obtained substituting
the parameter value (table 1),
θ s
F s
=
−qs2
rs2 −k+ ds
θ s
F s
=
−0.04283097s2
0.1539s4 + 0.01265 s3 − 0.6167s2 − 0.02099 s
θ s
F s
=
−0.2783 S2
s (s + 2.026) (s − 1.978) (s + 0.03402)
and
X s
F s
=
rs2−k+ds
pr−q2 s4 + pd +br s3+ bd −pk s2−kbs
X(s)
F(s)
=
0.106 S2 + 0.005 S− 0.4197
0.1539 S4 + 0.01265 S3 − 0.6167 S2− 0.02099 S
X s
F s
=
0.68843 (s + 2.014) (s − 1.967)
s (s + 2.026) (s − 1.978) (s + 0.03402)
The system poles lies on R.H plane confirming the system to be
unstable.
Table 1. Parameters of the system from feedback instrument
.U.K.
Parameter Value unit
Cart mass(𝑀) 1.206 Kilo gram
Mass of the pendulum(𝑚) 0.2693 Kilo gram
Half Length of
pendulum(𝑙)
0.1623 meter
Coefficient of frictional
force(𝑏)
0.005 Ns/m
Pendulum damping
coefficient(q)
0.005 𝑀𝑚/rad
Moment of inertia of
pendulum(𝐼)
0.099 𝑘𝑔/𝑚2
Gravitation force(𝑔) 9.8 𝑚/𝑠2
4. PERFORMANCE INDICES
The design of a control system is an attempt to meet a set of
specifications which define the overall performance of the
system in terms of certain measurable quantities. In the normal
way, some specific parametric values of the system are assumed
and the control system is designed accordingly to meet desired
performance of the system.. Four most commonly mathematical
functions are used as a performance index associated with error
of a closed loop system. A performance index is a number
which indicates goodness of system performance. The objective
is to design an optimal system by proper choice of its parameters
such that the specified performance index is optimum. A
performance index must be a single positive number or zero, the
latter being obtained if and only if the measure of the deviation
becomes identically zero.

__________________________________________________________________________________________
The commonly used performance indices (PI) are:
 Integral of squared error (ISE),
J = 𝑒2∞
0
𝑡 𝑑𝑡
 Integral of time multiplied squared error (ITSE),
J = 𝑡𝑒2
𝑡 𝑑𝑡
∞
0
 Integral of absolute error (IAE),
J = 𝑒(𝑡)
∞
0
𝑑𝑡
 Integral of time multiplied absolute error (ITAE),
J= 𝑡 𝑒(𝑡)
∞
0
𝑑𝑡
Here the error is define as e t = x t − y t . The stable
reference model has been taken for angle whose transfer
function is
The transfer function of the system with angle as the output is
θ s
F s
=
−0.04283097s2
0.1539s4 + 0.01265 s3 − 0.6167s2 − 0.02099 s
Table 2. The different performance indices with the angle of
the pendulum as output.
Performance Indexes PID
ISE
−50.87𝑠2
− 58.99𝑠 − 149.5
𝑠
ITSE
−50.72𝑠2
− 58.95𝑠 − 150.8
𝑠
IAE
−50.66𝑠2
− 58.87𝑠 − 149.1
𝑠
ITAE
−50.05𝑠2
− 58.41𝑠 − 151.4
𝑠
Table 3 The different performance indices with position of the
cart as output
Performance Indexes PID
ISE
45.66𝑠2
− 58.99𝑠 − 149.5
𝑠
ITSE
42.78𝑠2
+ 174.6𝑠 + 22.42
𝑠
IAE
44.56𝑠2
+ 187.9𝑠 + 22.4
𝑠
ITAE
43.09𝑠2
− 175.7𝑠 − 38.28
𝑠
Transfer function of stable reference model with position of the
cart is:
Reference Model
X1(s)
F1(s)
=
s + 2
s4 + 9s3 + 43s2 + 143s + 204
The transfer function of the system with position as the output
is:
X(s)
F(s)
=
0.106s2
+ 0.005s − 0.4197
0.1539s4 + 0.01265 s3 − 0.6167s2 − 0.02099 s
5. GENETIC ALGORITHM
Genetic algorithms (GA) are search procedures inspired by the
laws of natural selection and genetics. They can be viewed as a
general-purpose optimization method and have been
successfully applied to search, optimization and machine
learning tasks. GA has the ability to solve difficult, multi
dimensional problems with little problem-specific information
and hence has been chosen as the optimization technique to
solve various problems in control systems.
It has been shown that compared with other traditional heuristic
optimization method, Genetic Algorithm is likely to be more
computationally efficient. The controller parameters are usually
determined by trial-and-error through simulation. In such case,
the paradigm of GA appear to offer an effective way for
automatically and efficiently searching for a set of control
performance.
6.SIMULATION & RESULTS
Fig. 3.ISE GA PID Controller of angle
Fig. 4.ITSE GA PID Controller of angle

__________________________________________________________________________________________
Fig. 5.IAE GA PID Controller of angle
Fig. 6.ITAE GA PID Controller of angle
Fig. 7.ISE GA PID Controller of cart
Fig. 8.ITSE GA PID Controller of cart
Fig. 9.IAE GA PID Controller of cart
Fig. 10.ITAE GA PID Controller of cart
CONCLUSIONS
Modeling of inverted pendulum shows that system is unstable
with non-minimum phase zero. Results of applying PID
controllers show that the system can be stabilized. while PID
controller method is cumbersome because of selection of
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Step Response
Time (sec)
Amplitude
PID
GA PID

__________________________________________________________________________________________
constants of controller, Constant of the controllers can be tuned
by some Genetic Algorithm technique for better result. Results
with GA tuned controller are better in respect of rise time and
overshoot wgen the angle is measured. The choice of the
reference model like that of our system yields comparatively
better result. The use of Walsh function in equation (1&2)
would help finding out the solution of non-linear differential
equations thus helping towards the design of non-linear
controller.
REFERENCES
[1] Elmer P. Dadias, Patrick S. Fererandez, and David
J,”Genetic Algorithm on Line Controller For The Flexible
Inverted Pendulum Problem”, Journal Of Advanced
Computational Intelligence and Intelligent Informatics
[2] ] R. Murillo Garcia1, F. Wornle1, B. G. Stewart1 and D. K.
Harrison1, “Real-Time Remote Network Control of an Inverted
Pendulum using ST-RTL”, 32nd ASEE/IEEE Frontiers in
Education Conference November 6 - 9, 2002, Boston, MA.
[3] DONGIL CHOI and Jun-Ho Oh “Human-friendly Motion
Control of a Wheeled Inverted Pendulum by Reduced-order
Disturbance Observer” 2008 IEEE International Conference on
Robotics and Automation Pasadena, CA, USA, May 19-23,
2008.
[4] W. Wang, “Adaptive fuzzy sliding mode control for inverted
pendulum,” in Proceedings of the Second Symposium
International Computer Science and Computational
Technology(ISCSCT ’09) uangshan, P. R. China, 26-28, Dec.
pp. 231-234, 2009.
[5] Berenji HR. A reinforcement learning-based architecture for
fuzzy logic control. International Journal of Approximate
Reasoning 1992;6(1):267–92.
[6] I. H. Zadeh and S. Mobayen, “ PSO-based controller for
balancing rotary inverted pendulum, ” J. AppliedSci., vol. 16,
pp. 2907-2912 2008.
[7] I. H. Zadeh and S. Mobayen, “ PSO-based controller for
balancing rotary inverted pendulum, ” J. AppliedSci., vol. 16,
pp. 2907-2912 2008.
[8] Mohd Rahairi Rani,Hazlina Selamat,Hairi Zamzuri,”Multi
Objective Optimization For PID Controller Tuning Using The
Global Ranking Genetic Algorithm”,International Journal Of
Innovative Computing, Information and Control, VOL-8,
Number 1(A),January-2012
[9] Ohsumi A, Izumikawa T. Nonlinear control of swing-up and
stabilization of an inverted pendulum. Proceedings of the 34th
Conference on Decision and Control, 1995. p. 3873–80.
[10] Eiben, A.E., Hinterding, R. and Michalewicz, Z. Parameter
Control in Evolutionary Algorithms. IEEE Transactions on
Evolutionary Computation, 3, 2 (1999), 124-141.
[11] Kumar,P, Mehrotra, O.N, Mahto, J, Mukherjee, Rabi
Ranjan,”Modelling and Controller Design of Inverted
Pendulum”, National Conference on Communication,
Measurement and Control, Vol-I, 14th August, 2012,in press.
[12]Kumar,P, Mehrotra, O.N, Mahto, J, Mukherjee, Rabi
Ranjan,”Stabilization of Inverted Pendulum using LQR”,
National Conference on Communication, Measurement and
Control, Vol-I, 14th August, 2012 in press.
[13] Behra Laxmidhar & Kar Indrani; Intelligent Systems and
Control Principals and Applications; Oxford University Press
[14]Ogata, K.; System Dynamics, 4th Edition Englewood Cliffs,
NJ: Prentice-Hall, 2003.
[15]Goldberg, D.E. Genetic Algorithms in search, optimization,
and machine learning. Reading, Mass. : Addison-Wesley, 1989.
[16] feedback instrument,U.K
BIOGRAPHIES:
Mehrotra, a Gold Medalist at B.Sc.
Engineering(B.U), M.E.(Hons)(U.O.R) and
Ph.D. (R.U) all in Electrical Engineering,
has the industrial exposure at SAIL as
Testing & Commissioning Engineer. Served
Department of Science & Technology, Govt.
of Bihar & Govt. of Jharkhand for 35 years
and retired as Professor in Electrical Engineering. Served as
coordinator of various projects sanctioned through MHRD and
AICTE, including TEQIP, a World Bank Project. His research
interests include control and utilization of renewable energies,
power quality and power system.
Jagdeo Mahto was born in Madhubani,
Bihar, India, in 1943. He obtained the B.Sc
(Engg) degree in Electrical Engineering
from Bhagalpur University in 1964,
M.Tech. in Control System from IIT
Kharagpur, India in 1970 and Ph.D in
Control System in 1984 from IIT Delhi,
India. He served MIT Muzaffarpur from 1964 to 1971 in the
capacity of Lecturer and Assistant Professor. From 1971 to
1980 he served as Asst. Professor, from 1980 to 1985 as
Associate Professor and from 1985 to 1988 as Professor in the
Department of Electrical Engineering at BIT Sindri, India. He
taught at Bright Star University, Brega (Libya) from 1988 to
1989. From 1989 to 2003 he was again at BIT Sindri. From
2004 till date he is Professor at Asansol Engineering College,
India.
Pankaj Kumar was born in Muzaffarpur,
India, in 1970 and received the B.Sc and
M.Sc. degree in Electronics Honours and
Electronic Science respectively from
Magadh University and Gauhati
University Assam. He received M.Sc
Engineering in Control System
Engineering from Patna University in 2004. He began his
career as Lecturer in Bihar University Muzaffarpur. Currently
he is an Assistant Professor in the Department of Electrical
Engineering, Asansol Engineering College, Asansol, India.

Tuning of pid controller of inverted pendulum using genetic algorithm

More Related Content

What's hot (19)

Viewers also liked (20)

Similar to Tuning of pid controller of inverted pendulum using genetic algorithm (20)

More from eSAT Publishing House (20)

Recently uploaded (20)

Tuning of pid controller of inverted pendulum using genetic algorithm