About selecting the personal best in multi-objective particle swarm optimization

Abstract

In particle swarm optimization, a particle's movement is usually guided by two solutions: the swarm's global best and the particle's personal best. Selecting these guides in the case of multiple objectives is not straightforward. In this paper, we investigate the influence of the personal best particles in Multi-Objective Particle Swarm Optimization. We show that selecting a proper personal guide has a significant impact on algorithm performance. We propose a new idea of allowing each particle to memorize all non-dominated personal best particles it has encountered. This means that if the updated personal best position be indifferent to the old one, we keep both in the personal archive. Also we propose several strategies to select a personal best particle from the personal archive. These methods are empirically compared on some standard test problems. © Springer-Verlag Berlin Heidelberg 2006.