Improving NSGA-II by a Local Search Strategy with Gaussian Mutation
Zhijun Zhang, Baiquan Lü
Abstract
An improved non-dominated sorting genetic algorithm NSGA-II-GLS is proposed in this paper, which intends to enhance the local search ability of the original NSGA-II and reduce the likelihood of falling into local optimal. The proposed algorithm NSGA-II-GLS introduces a Gaussian mutation operator into the genetic operation in the algorithm, which makes the individuals focus on exploiting the space near themselves. Also, the improved algorithm adopts a jiggling local search strategy in order to jump out of non-global Pareto fronts. The performance of NSGA-II-GLS is compared with NSGA-II and NSGA-II-DE through five DTLZ test problems. The simulation results indicate that the NSGA-II-GLS has a good local search capability and the convergence and diversity of Pareto optimal solutions from NSGA-II-GLS are better than the original NSGA-II. In addition, solutions generated by NSGA-II-GLS suggest that NSGA-II-GLS is more versatile than NSGA-II-DE.