正在加载图片...
D0L:10.13374.issn1001-053x.2013.09.001 第35卷第9期 北京科技大学学报 Vol.35 No.9 2013年9月 Journal of University of Science and Technology Beijing Sep.2013 基于精英重组的混合多目标进化算法 吴迪凶,李苏剑,李海涛 北京科技大学机械工程学院,北京100083 ☒通信作者,E-mail:potato851124@163.com 摘要针对多目标进化算法搜索效率低和收敛性差的问题,提出了基于精英重组的混合多目标进化算法,将多目标优 化问题分解为多个单目标优化问题单独求解,并采用基于遗传算法的精英重组策略将多个相异解重组生成唯一的精英 解.提出区域化的种群初始化方法,改进局部搜索及群体选择机制,采用以优化子群为核心的分组交叉策略及自适应多 位变异算子,并引入基于混沌优化的重启机制,有效克服了精英保存的固有缺陷,以及现有多目标进化算法存在的目标 空间解拥挤、收敛慢、易早熟等问题。多目标测试函数的数值仿真和关键步骤的性能分析证明了本文算法的有效性和优 越性. 关键词多目标优化:精英重组:遗传算法:混沌理论 分类号TP18 Elite-recombination-based hybrid multi-objective evolutionary algo- rithm WUDi凶,LISu-jian,LI Hai-tao School of Mechanical Engineering,University of Science and Technology Beijing,Beijing 100083,China Corresponding author,E-mail:potato851124@163.com ABSTRACT Considering the bad efficiency and convergence of multi-objective evolutionary algorithms,this article introduces an elite-recombination-based hybrid multi-objective evolutionary algorithm (ERHMEA).In the algorithm, the multi-objective optimization problem was decomposed into multiple single-objective optimization problems and generated the only elite solution with the genetic-algorithm-based elite recombination strategy.Strategies such as regional population initialization,improved local search and selection mechanisms,optimized subgroup based packet crossover and adaptive multiple mutation operator,and chaos optimization based restart mechanism effectively overcome the inherent defects of elite preservation,as well as the multi-objective evolutionary algorithm (MEA)existing target space solution crowding,slow convergence,prematurity,and other issues.Multi-objective test functions analysis and experimental simulation prove the effectiveness and superiority of the proposed algorithm. KEY WORDS multi-objective optimization;elite recombination;genetic algorithms;chaos theory 现实问题往往由多个相互作用相互冲突的目 标优化问题转化为一个单目标优化问题,然而由 标组成,由此产生了多目标优化问题(multi-objec- 于权重参数的不确定性导致此方法性能较差.多 tive optimization problem,MOP).多目标优化问 目标进化算法(multi-objective evolutionary algo- 题由于目标的不一致性而难以产生唯一的最优解, ithm,MEA)作为一种基于群体的智能搜索方法, 它的非劣解是由一组互相矛盾的解所构成的折中 能够并行搜索大规模空间并生成多个Pareto非 解集合.传统的求解方法是基于各目标权重将多目 劣解,在运行速度及解的质量上均优于传统方法, 收稿日期:2012-08-04第 35 卷 第 9 期 北 京 科 技 大 学 学 报 Vol. 35 No. 9 2013 年 9 月 Journal of University of Science and Technology Beijing Sep. 2013 基于精英重组的混合多目标进化算法 吴 迪 ,李苏剑,李海涛 北京科技大学机械工程学院, 北京 100083 通信作者,E-mail: potato851124@163.com 摘 要 针对多目标进化算法搜索效率低和收敛性差的问题,提出了基于精英重组的混合多目标进化算法,将多目标优 化问题分解为多个单目标优化问题单独求解,并采用基于遗传算法的精英重组策略将多个相异解重组生成唯一的精英 解. 提出区域化的种群初始化方法,改进局部搜索及群体选择机制,采用以优化子群为核心的分组交叉策略及自适应多 位变异算子,并引入基于混沌优化的重启机制,有效克服了精英保存的固有缺陷,以及现有多目标进化算法存在的目标 空间解拥挤、收敛慢、易早熟等问题. 多目标测试函数的数值仿真和关键步骤的性能分析证明了本文算法的有效性和优 越性. 关键词 多目标优化;精英重组;遗传算法;混沌理论 分类号 TP18 Elite-recombination-based hybrid multi-objective evolutionary algo￾rithm WU Di , LI Su-jian, LI Hai-tao School of Mechanical Engineering, University of Science and Technology Beijing, Beijing 100083, China Corresponding author, E-mail: potato851124@163.com ABSTRACT Considering the bad efficiency and convergence of multi-objective evolutionary algorithms, this article introduces an elite-recombination-based hybrid multi-objective evolutionary algorithm (ERHMEA). In the algorithm, the multi-objective optimization problem was decomposed into multiple single-objective optimization problems and generated the only elite solution with the genetic-algorithm-based elite recombination strategy. Strategies such as regional population initialization, improved local search and selection mechanisms, optimized subgroup based packet crossover and adaptive multiple mutation operator, and chaos optimization based restart mechanism effectively overcome the inherent defects of elite preservation, as well as the multi-objective evolutionary algorithm (MEA) existing target space solution crowding, slow convergence, prematurity, and other issues. Multi-objective test functions analysis and experimental simulation prove the effectiveness and superiority of the proposed algorithm. KEY WORDS multi-objective optimization; elite recombination; genetic algorithms; chaos theory 现实问题往往由多个相互作用相互冲突的目 标组成,由此产生了多目标优化问题 (multi-objec￾tive optimization problem, MOP). 多目标优化问 题由于目标的不一致性而难以产生唯一的最优解, 它的非劣解是由一组互相矛盾的解所构成的折中 解集合. 传统的求解方法是基于各目标权重将多目 标优化问题转化为一个单目标优化问题,然而由 于权重参数的不确定性导致此方法性能较差. 多 目标进化算法 [1](multi-objective evolutionary algo￾rithm, MEA) 作为一种基于群体的智能搜索方法, 能够并行搜索大规模空间并生成多个 Pareto 非 劣解,在运行速度及解的质量上均优于传统方法, 收稿日期:2012-08-04 DOI:10.13374/j.issn1001-053x.2013.09.001
向下翻页>>
©2008-现在 cucdc.com 高等教育资讯网 版权所有