Enhance the Convergence and Diversity for epsilon-MOPSO by Uniform Design and Minimum Reduce Hypervolume

To overcome the conflict of convergence and diversity for MOPs, ϵ-MOEA is an efficient method based on ϵ-dominance concept. When two candidate solutions in the same hyper-box and no-dominate each other, ϵ-MOEA chooses the solution nearer to box corner. In this paper, we firstly propose an accept rule Minimum Reduce Hypervolume, it chooses the solution which make larger contribute to the hypervolume. A novel Particle Swarm Optimization is presented as ϵ-MOPSO_MRV ^UD: (I) The first population is generated by Uniform Design, it can get evenly distribute solutions and provide good information for next offspring; (II) The update strategy of archive naturally combine the ϵ-dominance concept and Minimum Reduce Hypervolume. Experiment on different multi-Objective problems include ZDTx and DTLZx in jMetal 2.0, the results show that the new algorithm not only enhance the convergence and diversity, but also improve the hypervolume, reduce NFFEs than ϵ-MOEA.