Multi-group particle swarm optimization with random redistribution

Abstract

Particle Swarm Optimization (PSO) is fast and popular algorithm to find the optimum value of non-linear and multi-dimensional function. However, it often easily trapped into local optima because the particles move closer to the best particle quickly. This paper purposes a new algorithm called Multi-Group Particle Swarm Optimization with Random Redistribution (MGRR-PSO) that tried to solve the weakness of standard PSO. MGRR-PSO combines two groups of PSO with opposite acceleration coefficients. In addition, some particles are redistributed when they are trapped in local optima. Experimental studies on 5 benchmark functions with 50-dimensions and 100-dimensions show that the MGRR-PSO can solve the problems that can't be solved by original PSO with better performance.