A Novel Diversity Maintenance Scheme for Evolutionary Multi-objective Optimization

Recently, decomposition-based multi-objective evolutionary algorithm (MOEA/D) has received increasing attentions due to its simplicity and decent optimization performance. In the presence of the deceptive optimum, the weight vector approach used in MOEA/D may not be able to prevent the population traps into local optimum. In this paper, we propose a new algorithm, namely Diversity Preservation Multi-objective Evolutionary Algorithm based on Decomposition (DivPre-MOEA/D), which uses novel diversity maintenance scheme to enhance the performance of MOEA/D. The proposed algorithm relaxes the dependency of the weight vector approach on approximated ideal vector to maintain diversity of the population. The proposed algorithm is evaluated on CEC-09 test suite and compared the optimization performance with MOEA/D. The experiment results show that DivPre-MOEA/D can provide better solutions spread along the Pareto front.