Multi-objective PSO algorithm based on fitness sharing and online elite archiving

Abstract

A new technique for multi-objective PSO (Particle Swarm Optimization) based on fitness sharing and online elite archiving is proposed. Global best position of particle swarm is selected from repository by fitness sharing, which guarantees the diversity of the population. At the same time, in order to ensure the excellent population, the elite particles from the repository are introduced into next iteration. Three well-known test functions taken from the multi-objective optimization literature are used to evaluate the performance of the proposed approach. The results indicate that our approach generates a satisfactory approximation of the Pareto front and spread widely along the front. © Springer-Verlag Berlin Heidelberg 2006.