TUNING OF PID CONTROLLERS USING EVOLUTIONARY OPTIMIZATION TECHNIQUES

Abstract

The basis of any problem in control engineering is the design of the controller which meets user requirements. Performance of any controller depends on the optimal selection of its parameter values. In this work evolutionary optimisation algorithms were used for tuning PID controllers. Here two different evolutionary optimisation techniques namely CA and DE were used for tuning PID controllers. The above controllers were designed for the cart control problem of inverted pendulum and the results obtained were verified on hardware. As an extension to the above work a two PID controller structure was tuned using multi objective optimisation for stabilisation of inverted pendulum system .Here also the results obtained were verified experimentally. The advantages of evolutionary techniques are that they deal with a set of solutions at any one time. Initial selection of these solutions is random and hence they are spread throughout the entire solution space which increases the chance of finding the global - optimum. Since these techniques are basically search methods they use information from fitness function only and do not depend on any other factors. These techniques can be applied for a wide range of applications where the task can be formulated as an optimisation problem. As in any optimisation problem the most important factor here is how the fitness function is defined. If output performance is the sole criteria then generally one of the time domain error indices are defined as the fitness function. In this work tuning of PID controllers is set up as an optimisation problem with Integral Square Error as the fitness function to be minimised.