FMPSO: fuzzy-dominance based many-objective particle swarm optimization

The performance of multiobjective evolutionary algorithms, based on Pareto-dominance, deteriorates when applied to many-objective problems which are the problems having higher than three objectives. In view of this, various alternative dominance-based approaches have been suggested in the literature. Another major concern in solving these problems is the selection of proper variation operators for exploring the problem's solution space. This work proposes a new algorithm, fuzzy-dominance based many-objective particle swarm optimization (FMPSO) that addresses these two concerns. It uses the existing SMPSO as base algorithm and replaces the Pareto-dominance technique used in it with fuzzy-dominance as optimization criterion. Additionally a novel archive updation scheme is introduced which provides improved convergence along with the archive refinement scheme which gives enhanced diversity of solutions. We have conducted a number of experiments on the proposed method using standard benchmark problems and have got improved results for FMPSO over other competing approaches.