一种基于DCD和a-tDX改进的NSGA-II算法  

An Improved NSGA-II Algorithm Based on Dynamic Crowding Distance and Adaptive t-Distribution Crossover

作　　者：时思思 张新燕[1] 王志浩 SHI Si-si;ZHANG Xin-yan;WANG Zhi-hao(College of Electrical Engineering,Xinjiang University,Urumqi Xinjiang 830047,China)

机构地区：[1]新疆大学电气工程学院

出　　处：《计算机仿真》2019年第12期257-262,共6页Computer Simulation

基　　金：国家自然科学基金项目(51667018)

摘　　要：带精英策略的非支配排序遗传算法(NSGA-Ⅱ)是目前较常见的多目标优化算法。但上述方法基于固定拥挤度筛选中间种群,并未考虑筛选过程中拥挤度的动态变化,且算法不能适应不同进化时期对于解空间的不同搜索需求,搜索效率较低。提出一种基于动态拥挤度和自适应t分布交叉算子(a-tDX)改进的NSGA-Ⅱ算法,在每次删除拥挤度最低的解后更新非支配解的拥挤度,直至选出所需数量的非支配解,以此维护解集的多样性,并在算法迭代过程中采用自适应进化时期的t分布交叉以满足算法对搜索空间的动态需求。原始NSGA-Ⅱ算法、正态分布交叉算子改进的NSGA-Ⅱ算法和所提算法在5个基准函数上的测试结果证明所提算法有更好的收敛性和多样性。Non-dominated sorting genetic algorithm(NSGA-Ⅱ)with elite strategy is the most common multi-objective optimization algorithm.However,this method screens the intermediate population based on the fixed crowding distance,and does not consider the dynamic change of the crowding distance during the screening process;and the algorithm cannot adapt to the different search requirements of the solution space in different evolutionary periods,and the search efficiency is low.In this regard,an improved NSGA-Ⅱalgorithm based on dynamic crowding distance and adaptive t-distribution crossover operator(a-tDX)is proposed to update the crowding distance of non-dominated solutions after each deletion of the least crowding distance solution until the selection.The required number of non-dominated solutions are used to maintain the diversity of the solution sets;the t-distribution of the adaptive evolutionary period is used in the iterative process of the algorithm to satisfy the dynamic requirements of the algorithm for the search space.The traditional NSGA-Ⅱalgorithm,the improved NSGA-Ⅱalgorithm with normal distribution crossover operator,and the test results of the proposed algorithm on the five benchmark functions prove that the proposed algorithm has better convergence and diversity.

关 键 词：遗传算法 动态拥挤度 精英保留策略 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]