带基因交换和动态网格的多目标粒子群优化算法  

Multi-objective Particle Swarm Optimization with Gene Exchange and Dynamic Mesh

作　　者：夏禹[1] 吴朋[1,2] 

机构地区：[1]重庆大学自动化学院,重庆400030 [2]重庆川仪自动化股份有限公司,重庆400000

出　　处：《信息与控制》2017年第5期593-600,共8页Information and Control

基　　金：国家2015年战略性新兴产业发展专项资金资助项目

摘　　要：网格策略在对多目标粒子群优化算法档案的多样性分析时具有简单快捷的特点,但由于其分辨率的限制,无法对含有相同粒子数量的超体进行多样性判断,对此,提出了动态网格的策略,通过目标空间的动态划分、超体的动态调节等手段,对多样性相同的超体进行密度排序,方便算法对档案进行管理和选择领导粒子,同时利用了以往被忽视的档案解,对其进行基因交换,提高算法的收敛速度.通过DTLZ系列函数的验证,表明了算法在高维多目标优化中仍具有良好的多样性和更快的收敛速度,能有效解决高维多目标问题.The grid strategy conducts simple and fast analysis of the diversity of the multi-particle swarm optimization algorithm.However,because of limited resolution,this strategy is unable to determine the diversity of a hypercube with the same number of particles.Thus,we propose a dynamic grid strategy.Through dynamic partitioning of the target space and dynamic adjustment of the hypercube,the same density of the hypercube is sorted.This approach is better for archive management and selection of leading particles.We exchange genes in the archival solution,which was previously neglected,to improve the convergence rate of the algorithm.Verification of DTLZ function proved that the algorithm has good diversity and faster convergence speed in high-dimension multi-objective optimization,thereby effectively solving such problem.

关 键 词：基因交换 档案 动态网格 多目标粒子群优化算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]