Constrained genetic algorithm-based computer-aided control system design-fixed versus variable size population

This paper shows how the desired shape of the system response can be used to define the fitness of a design. The paper also shows that the GA can be used to solve practically constrained optimization problems. The technique is applied to the design of proportional and/or integral and/or derivative behavior controllers for a DC servomechanism system, which is also used as an example to demonstrate this approach of generalized fitness function specification. GAs allow more realistic and more easily understood definitions of constrained optimization algorithms. The constraints are expressed in a general explicit language. The optimal solution obtained by combining partial information from all the population is the foundation of GA theory, therefore, the infeasible solutions should provide information and not just be thrown away. Yahiaoui and Hamam (1997) present some results with a fixed population. In this paper we propose a new extension using a variable population. The first results of this extension are very promising despite some problems resulting from unregulated increase of our population in certain cases.