Specification of a control system fitness function using constraints for genetic algorithm based design methods

Abstract

This paper proposes a fitness function for genetic algorithms used in control system design based on derived response shape. A technique is presented that allows control system performance to be specified as constraints on the system response rather than an error function. A genetic algorithm is used to solve the resulting constrained optimisation problem with the nonlinearity of the cost function. A design example of practical PID control of alift system is given, showing the flexibility of this type of control system specification and the performance achieved.