基于正交设计的自适应ε占优MOEA/D算法研究  

RESEARCH ON ADAPTIVE EPSILON-DOMINATION BASED ORTHOGONAL MOEA/D FOR MULTI-OBJECTIVE OPTIMIZATION

作　　者：周攀[1] 张冬梅[1] 龚文引[1] 李阳[1] 刘凯伟[1] 

机构地区：[1]中国地质大学(武汉)计算机学院,湖北武汉430074

出　　处：《计算机应用与软件》2013年第2期58-64,124,共8页Computer Applications and Software

基　　金：国家自然科学基金项目(40972206;61075063);中央高校基本科研业务费专项资金项目(1323520909)

摘　　要：MOEA/D是一种简单、高效的多目标优化算法,但在更新子问题时,会丢失部分优良个体,降低算法的收敛速度。针对上述不足,提出一种基于正交设计的自适应ε占优算法。新算法改进如下:(1)采用正交试验设计和连续空间量化初始化种群,使初始化群体能均匀分布;(2)设计一种自适应调整松弛变量改进的ε占优机制,并用它来更新Archive种群保存非劣解;(3)将精英策略引入到MOEA/D中,加快收敛速度。实验结果表明新算法较好地改善了MOEA/D算法的收敛性以及非劣解的分布性。MOEA/D is a simple,effective multi-objective optimization algorithm,however,when MOEA/D updates subproblems,it may lost some good individuals,and thus lead to reducing the convergence speed of the algorithm.In this paper,in order to remedy this pitfall,an improved MOEA/D based on the orthogonal design adaptive epsilon-domination is proposed.The new algorithm can be characterized as：（1） Using the orthogonal experimental design with quantization to initialize the population.（2）Proposing an improved epsilon-domination which can be self-adaption and Using it to update Archive population which retain the obtained non-dominated solutions.（3）The elitist strategy is introduced into MOEA/D,speeding up the convergence speed.The simulation results show that the new algorithm improves the speed of original MOEA/D＇s convergence and the distribution of non-inferior solutions in multi-objective problem.

关 键 词：MOEA D 自适应ε占优 正交实验 多目标演化算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]