17-03-2014, 10:58 AM
Abstract- In this paper, a new method for adaptive
control of nonlinear systems using neural networks
and Proportional-Integral-Derivative (PID)
methodology is proposed. In this method, a PID
control and adaptive linear neural network is used
to control a inverted pendulum with uncertainities
and external disturbances. The system consists of an
inverted pole hinged on a cart which is free to move
in the x direction. The performance of the proposed
method is demonstrated via simulations.
I. INTRODUCTION
Control of uncertain systems deserves a careful study
because of the operational safety and performance
considerations. Existence of noise on the measured
quantities and time varying parameters of the plant under
control constitute prime difficulties in the design of a
suitable controller. Therefore the alleviation of the
uncertainties and impreciseness encourages the use of
intelligent controllers having the capabilities of fault
tolerance and improving the future performance by
adjusting the design parameters through a learning
process, by which the autonomous behavior is acquired.
[1]
The proportional-integral-derivative (PID) controller is
widely used in many control applications because of its
simplicity and effectiveness. Though the use of PID
control has been a long history in the field of control
engineering, the three controller gain parameters,
proportional gain KP, integral gain KI, and derivative
gain KD, are usually fixed. The disadvantage of PID
controller is poor capability of dealing with system
uncertainty, i.e., parameter variations and external
disturbance. Robustness has gained more and more
attention.
In recent years, there has been extensive interest in selftuning
these three controller gains. For examples, the
PID self-tuning methods based on the relay feedback
technique were presented for a class of systems [2,3]. An
adaptive PID control tuning was proposed to cope with
the control problem for a class of uncertain chaotic
systems with external disturbance [4]. A genetic
algorithm was used to find the optimum tuning
parameters of the PID controller by taking integral
absolute error as fitting function [5]. Sliding mode
control (SMC) is one of the popular strategies to deal
with uncertain control systems [6-8]. Neural networks
(NN) are used for modeling and control of complex
physical systems because of their ability to handle
complex input-output mapping without detailed
analytical models of the systems.
Adaptive PID control systems were initially proposed by
Åström in the early 1980s [9]. Since then, considerable
work has been done on the adaptive tuning of gains and
related parameters of PID controllers. Roughly speaking,
the approaches to self-tuning of PID controller gains can
be classified into two categories -- model-based tuning
[10] and knowledge-based tuning [11,12]. Normally,
these approaches require the knowledge of either the
plant models or the plant behaviors. For example, in
[12], a fuzzy logic controller was designed based on a
plant model obtained by using a neural network. In such
a case, the convergence of a multilayer neural network to
the plant model becomes acritical issue when the
environment parameters, such asfrictions, temperature
and pressures, are subject to continuous changes.[1]
control of nonlinear systems using neural networks
and Proportional-Integral-Derivative (PID)
methodology is proposed. In this method, a PID
control and adaptive linear neural network is used
to control a inverted pendulum with uncertainities
and external disturbances. The system consists of an
inverted pole hinged on a cart which is free to move
in the x direction. The performance of the proposed
method is demonstrated via simulations.
I. INTRODUCTION
Control of uncertain systems deserves a careful study
because of the operational safety and performance
considerations. Existence of noise on the measured
quantities and time varying parameters of the plant under
control constitute prime difficulties in the design of a
suitable controller. Therefore the alleviation of the
uncertainties and impreciseness encourages the use of
intelligent controllers having the capabilities of fault
tolerance and improving the future performance by
adjusting the design parameters through a learning
process, by which the autonomous behavior is acquired.
[1]
The proportional-integral-derivative (PID) controller is
widely used in many control applications because of its
simplicity and effectiveness. Though the use of PID
control has been a long history in the field of control
engineering, the three controller gain parameters,
proportional gain KP, integral gain KI, and derivative
gain KD, are usually fixed. The disadvantage of PID
controller is poor capability of dealing with system
uncertainty, i.e., parameter variations and external
disturbance. Robustness has gained more and more
attention.
In recent years, there has been extensive interest in selftuning
these three controller gains. For examples, the
PID self-tuning methods based on the relay feedback
technique were presented for a class of systems [2,3]. An
adaptive PID control tuning was proposed to cope with
the control problem for a class of uncertain chaotic
systems with external disturbance [4]. A genetic
algorithm was used to find the optimum tuning
parameters of the PID controller by taking integral
absolute error as fitting function [5]. Sliding mode
control (SMC) is one of the popular strategies to deal
with uncertain control systems [6-8]. Neural networks
(NN) are used for modeling and control of complex
physical systems because of their ability to handle
complex input-output mapping without detailed
analytical models of the systems.
Adaptive PID control systems were initially proposed by
Åström in the early 1980s [9]. Since then, considerable
work has been done on the adaptive tuning of gains and
related parameters of PID controllers. Roughly speaking,
the approaches to self-tuning of PID controller gains can
be classified into two categories -- model-based tuning
[10] and knowledge-based tuning [11,12]. Normally,
these approaches require the knowledge of either the
plant models or the plant behaviors. For example, in
[12], a fuzzy logic controller was designed based on a
plant model obtained by using a neural network. In such
a case, the convergence of a multilayer neural network to
the plant model becomes acritical issue when the
environment parameters, such asfrictions, temperature
and pressures, are subject to continuous changes.[1]