Enhancing diversity for NSGA-II in evolutionary multi-objective optimization
Jinghua Zheng, Ruimin Shen, Juan Zou · 2012
The NSGA-II method has been shown highly effective to provide sufficient selection pressure searching towards Pareto optimal set in multi-objective optimization. However, an important drawback in NSGA-II is that the diversity of resulting populations is not satisfactory due to the shortcoming of crowding distance. In this paper, we propose a diversity maintenance strategy for NSGA-II to enhance diversity during evolution process. We employ sphere to define a neighborhood for each individual. Moreover, a diversity maintenance strategy integrates into the critical selection scheme. It picks out extreme individuals and prohibits or postpones the archive of adjacent individuals. From an extensive comparative study with original NSGA-II and two other MOEAs, the proposed method shows a good balance among convergence, uniformity and spread.