Multi-swarm particle swarm optimization with a center learning strategy

Abstract

This paper proposes a new variant of particle swarm optimizers, called multi-swarm particle swarm optimization with a center learning strategy (MPSOCL). MPSOCL uses a center learning probability to select the center position or the prior best position found so far as the exemplar within each swarm. In MPSOCL, Each particle updates its velocity according to the experience of the best performing particle of its partner swarm and its own swarm or the center position of its own swarm. Experiments are conducted on five test functions to compare with some variants of the PSO. Comparative results on five benchmark functions demonstrate that MPSOCL achieves better performances in both the optimum achieved and convergence performance than other algorithms generally. © 2013 Springer-Verlag Berlin Heidelberg.