NSGA-II with Local Perturbation

To improve the overall performance of one algorithm, most researchers focus on the fitness assignment, preserving diversity and hybridizing different search methods. Different from the above strategies, this paper focuses on the convergence. According to the analysis of the convergence of NSGA-II, local perturbation strategy is introduced to improve the efficiency of NSGA-II in the paper. Local perturbation strategy is able to enlarge the search space and more optimal solutions can be found with large probability. To illustrate the effect of local perturbation, the proposed LPNSGA-II with other three outstanding algorithms is tested on six test instances. Experimental results illustrate that the proposed LPNSGA-II outperforms the three algorithms and the convergence of NSGA-II is improved greatly using local perturbation strategy.