A cluster-based evolutionary algorithm for multi-objective optimization

In this paper a new evolutionary algorithm is described for multiobjective optimization. The new method handles non-linear objective functions and constraints and supports the decision-maker with an estimation of the Pareto set. This cluster-based method applies the Pareto-dominance principle. It approximates the Pareto set with the prototypes for each cluster and alternative prototypes as secondary population. The non-dominated set is continuously being up-dated: based on the Pareto ranking, the poorest clusters are regularly deleted, and the new ones are set. The method solves the usual test problems with a satisfactory level of accuracy.