A Global-Crowding-Distance Based Multi-objective Particle Swarm Optimization Algorithm

Abstract

A global-crowding-distance based multi-objective particle swarm optimization (GCDMOPSO) algorithm is proposed in this paper. The concept of global crowding distance is introduced into the proposed algorithm to estimate the density of the non-dominated solutions in the external archive, and a dynamic global-crowding-distance based maintenance scheme is used to prune the external archive. Meanwhile, a chaotic mutation operator, which is related to the iteration number, is utilized in the approach to avoid being trapped into local minimum. In order to extend the search ability of the algorithm, not the particles in the swarm should be mutated, but part of non-dominated solutions in the external archive should also be mutated. To improve the global search ability of the algorithm and to make every non-dominated solution in the archive have a higher probability to be selected as the global best position, an improved Roulette Gambling selection strategy is designed to select global best for every particle in the swarm. The experiment results show that, the GCDMOPSO algorithm can get better Pareto optimality solution sets over almost all the benchmark test functions when compared with the other two classical algorithms.