Multiobjective Particle Swarm Optimization with Predatory Escaping Behavior

Abstract

Due to the fast convergence, Particle swarm optimization (PSO) has been advocated to be especially suitable for multiobjective optimization. However, there is no information-sharing of with other particles in the population, except that each particle can access the global best. Thus, the premature convergence and lacks of intensification around the local best locations are inevitable during extending PSO to solve multiobjective optimization problems. In this paper, we propose a method of information-sharing by offering particle the predation escaping behavior in order to provide the necessary selection pressure to propel the population moving towards the true Pareto front. To demonstrate the efficiency of the proposed approach based on NSPSO, experimental results obtained on benchmark test functions are compared with NSPSO, and show that the modified NSPSO can find out the better Pareto Front.