Abstract:

In order to improve the convergence of multi-objective particle swarm optimization (MOPSO) while ensuring well distribution, a new method of MOPSO based on auxiliary fitness value is proposed. By establishing an angular coordinate, the angular coordinate parameters of the objective vector are ascertained as well as angular reference line’s parameters in various dimensional spaces. And an auxiliary fitness value is defined  to compare non-dominating individuals. Simulation results indicate that IMOPSO commendably balances the conflicts of convergence and distribution even for the small size of archive. Moreover, its pareto solutions would not get crowded in a small region along the Pareto front in test functions. Consequently, IMOPSO is validated and proven effective while its runtime is less than NSGA2, SPEA2 and MOEA/D.