Locating multiple optima using particle swarm optimization

Abstract

Many scientific and engineering applications require optimization methods to find more than one solution to multimodal optimization problems. This paper presents a new particle swarm optimization (PSO) technique to locate and refine multiple solutions to such problems. The technique, NichePSO, extends the inherent unimodal nature of the standard PSO approach by growing multiple swarms from an initial particle population. Each subswarm represents a different solution or niche; optimized individually. The outcome of the NichePSO algorithm is a set of particle swarms, each representing a unique solution. Experimental results are provided to show that NichePSO can successfully locate all optima on a small set of test functions. These results are compared with another PSO niching algorithm, lbest PSO, and two genetic algorithm niching approaches. The influence of control parameters is investigated, including the relationship between the swarm size and the number of solutions (niches). An initial scalability study is also done.