A new approach on particle swarm optimization for multimodal functions

Abstract

This paper describes a technique that extends PSO to locate multiple optima on a multimodal functions. In this paper, we present a new algorithm based on clustering particles to identify niches. For that we employ the standard k-means clustering algorithm which can identify the number of clusters adaptively. In each niche we used artificial immune system algorithm to determine the true members of it. Experimental results show that the proposed algorithm can successfully locate all optimum solutions on a small set of test functions during all simulation runs.