Elastic boundary for particle swarm optimization

Abstract

Standard particle swarm optimization (PSO) introduced in 2007, here called 2007-sPSO, is chosen as a starting algorithm in this paper. To solve the problems of the swarm's velocity slowing down towards zero and stagnant phenomena in the later evolutionary process of 2007-sPSO, elastic boundary for PSO (EBPSO) is proposed, where search space boundary is not fixed, but adapted to the condition whether the swarm is flying inside the current elastic search space or not. When some particles are stagnant, they are activated to speed up in the range of the current elastic boundary, and personal cognition is cleared. Experimental results show that EBPSO improves the optimization performance of 2007-sPSO, and performs better than comparison algorithms. © 2012 Springer-Verlag.