一种改进的非支配排序多目标遗传算法  

Improved non-dominated sorting genetic algorithm for multi-objective optimization

作　　者：陈静[1] 伍军[1] 郑金华[1] 

机构地区：[1]湘潭大学信息工程学院,湖南湘潭411105

出　　处：《计算机工程与应用》2009年第29期60-63,71,共5页Computer Engineering and Applications

基　　金：国家自然科学基金(No.60773047);湖南省教育厅科研计划重点项目(No.06A074);湖南省教育厅一般科研项目(No.07c752)~~

摘　　要：多目标进化算法的研究目标主要是使算法快速收敛,并且广泛而均匀分布于问题的非劣最优域。在NSGA-II算法的基础上,提出了一种新的构造种群的策略——按照聚集距离选取部分非支配个体,并选取部分较好的支配个体形成下一代种群。该策略与原算法相结合后的算法(NSGA-II+IMP)与原NSGA-II进行比较,结果表明新算法较好地改善了分布性和收敛性。The main goal for research on MOEAs is to make the algorithms converge rapidly,and gain solutions that are widely and uniformly scattered in the non-dominated feasible areas of the problems.This paper,which is on the basis of NSGA2,proposes a new strategy for generating new population,that is not only selecting a certain proportion of non-dominated individuals according to the crowding distance,but also choosing some other dominated but potential individuals to form the next generation. The new strategy-combined algorithm（NSGA-Ⅱ＋IMP） is compared with the original NSGA2,and the result shows that the new one can better improve the diversity and the convergence of the solution set.

关 键 词：多目标进化算法 多目标优化问题 种群维护 聚集距离 分布性 保持策略 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]