Co-Evolutionary Multi-Agent System with Speciation and Resource Sharing Mechanisms

keywords: Multi-agent systems, evolutionary algorithms, co-evolution, niching, speciation, multi-modal optimization
Niching techniques for evolutionary algorithms are used in order to locate basins of attraction of the local minima of multi-modal fitness functions. Co-evolutionary techniques are aimed at overcoming limited adaptive capabilities of evolutionary algorithms resulting from the loss of useful population diversity. In this paper the idea of niching co-evolutionary multi-agent system (NCoEMAS) is introduced. In such a system the species formation phenomena occurs within one of the preexisting species as a result of co-evolutionary interactions. The results of experiments with Rastrigin and Schwefel multi-modal test functions aimed at the comparison of NCoEMAS to other niching techniques are presented. Also, the influence of resource sharing mechanism's parameters on the quality of speciation processes in NCoEMAS are investigated.
reference: Vol. 25, 2006, No. 4, pp. 305–331