Feedback-control operators for evolutionary multiobjective optimization

Abstract

New operators for Multi-Objective Evolutionary Algorithms (MOEA's) are presented here, including one archive-set reduction procedure and two mutation operators, one of them to be applied on the population and the other one on the archive set. Such operators are based on the assignment of "spheres" to the points in the objective space, with the interpretation of a "representative region". The main contribution of this work is the employment of feedback control principles (PI control) within the archive-set reduction procedure and the archive-set mutation operator, in order to achieve a well-distributed Pareto-set solution sample. An example EMOA is presented, in order to illustrate the effect of the proposed operators. The dynamic effect of the feedback control scheme is shown to explain a high performance of this algorithm in the task of Pareto-set covering. © Springer-Verlag 2009.