一种分阶段的改进多目标粒子群优化算法  

An Improved Multistage Multi-objective Particle Swarm Optimization Algorithm

作　　者：尤嘉兴 陈基漓[1,2] 董明刚[1] 

机构地区：[1]桂林理工大学信息科学与工程学院,广西桂林541004 [2]桂林理工大学广西空间信息与测绘重点实验室,广西桂林541004

出　　处：《小型微型计算机系统》2015年第4期792-796,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61203109)资助;广西空间信息与测绘重点实验室开(桂科能1103108-16)资助;广西研究生教育创新计划资助项目(YCSZ2014157)资助

摘　　要：在多目标优化问题求解上,粒子群优化算法存在所得最优解集精度不足、分布不够均匀的缺点,针对上述问题,提出了一种多种群分阶段的多目标粒子群优化算法.算法对外部档案个体采取多种算子进行处理以提高解集的收敛精度,引入简化粒子群优化模型使算法更适应多目标优化问题的求解,通过分阶段选取领导个体以及分阶段采取不同策略对非支配解集进行维护以维持解分布均匀性的同时提高收敛速度,重点改善高维多目标优化问题的解集分布均匀性.实验结果表明,改进算法所得的非支配解集具有更好的分布均匀性和收敛精度.In multi-objective optimization, particle swarm algorithm exist insufficient precision in solution set and the distribution is not uniform enough,to address the problem,in view of this problem,a multi-population in multistage multi-objective particle swarm opti- mization algorithm was presented. To improve the convergence accuracy of solution set, algorithm take a variety of operators to handle the external archive, a simplified model of particle swarm optimization is introduced to make this algorithm could be adapted to solving multi-objective optimization problems. Through selecting individual leadership in stages and a phased in implementation of different strategies for maintain non-dominated solutions to keep the uniformity of the solution set and enhance the convergence rate at the same time and improve the uniformity of the solution set of high-dimensional multi-objective optimization problem. A typical multi-objective functions test shows that the improved algorithm could be able obtain a non-dominated solutions with better uniformity and conver- gence precision.

关 键 词：多目标优化 粒子群算法 拥挤距离 外部档案 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]