求解多目标优化问题的改进布谷鸟搜索算法  

Improved cuckoo search algorithm for multi-objective optimization problems

作　　者：杨辉华[1,2] 谢谱模 张晓凤[1] 马巍[1] 刘振丙[3] 

机构地区：[1]桂林电子科技大学广西信息科学实验中心,广西桂林541004 [2]北京邮电大学自动化学院,北京100876 [3]桂林电子科技大学电子工程与自动化学院,广西桂林541004

出　　处：《浙江大学学报（工学版）》2015年第8期1600-1608,共9页Journal of Zhejiang University：Engineering Science

基　　金：国家自然科学基金资助项目(21365008;61105004);广西自然科学基金资助项目(2012GXNSFAA053230;2013GXNSFBA019279);广西信息科学实验中心重点资助项目(20130103)

摘　　要：针对求解多目标优化问题,提出一种改进的多目标布谷鸟搜索算法(IMOCS).相比于标准多目标布谷鸟搜索算法(MOCS),IMOCS在莱维飞行中使用动态自适应的步长控制量,并基于层级和拥挤度距离选择下一次莱维飞行的种群.为了验证算法的有效性,通过在测试实例(SCH,ZDT系列,LZ)计算所求Pareto前沿与真实Pareto前沿的广义距离和所求Pareto前沿的多样性来测试IMOCS的性能.结果表明,与MOCS,NSPSO,NSGA-II比较,IMOCS所求的广义距离更小,即由IMOCS所求Pareto前沿更加接近于真实Pareto前沿,同时IMOCS的Pareto前沿分布更加广泛和均匀,即多样性更好.In trying to solve multi-objective optimization problems,an improved multi-objective cuckoo search algorithm(IMOCS)was introduced.Compared with the standard multi-objective cuckoo search algorithm(MOCS),the IMOCS had two improvements,used a dynamic adaptive step-size control amount in Lévy flight;chose the next Lévy flight population based on the level and crowding distance.To verify the effectiveness of the IMOCS,test instances(SCH,ZDT series,LZ)were used to evaluate the performance:the generalized distance between the obtained Pareto front and the true Pareto front,and the diversity of the obtained Pareto front.Our results and comparison with the MOCS,NSPSO and NSGA-II showed that the IMOCS obtains a smaller generalized distance,which meant the IMOCSs Pareto front is closer to the true Pareto front,and at the same time its Pareto front distributes broader and more uniform,that diversity is better.

关 键 词：多目标优化 布谷鸟搜索算法 自适应步长控制量 非支配集排序 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]