基于正交设计的元胞多目标遗传算法  

Cellular Genetic Algorithm for Multiobjective Optimization Based on Orthogonal Design

作　　者：张屹[1] 万兴余 郑小东[1] 孙莉莉[1] 

机构地区：[1]三峡大学机械与动力学院,湖北宜昌443002

出　　处：《电子学报》2016年第1期87-94,共8页Acta Electronica Sinica

基　　金：国家自然科学基金(No.51275274);三峡大学研究生科研创新基金(No.2013CX029)

摘　　要：元胞多目标遗传算法在求解两目标优化问题时是比较高效的.但是,初步实验显示其在求解三目标优化问题(例如DTLZ系列)时,表现不是十分令人满意.为了进一步提高算法的性能,引入了正交设计的思想,提出了基于正交设计的多目标元胞遗传算法.在改进算法的迭代过程中,先对父代个体进行分段,之后按照正交表来对这些片段进行重新组合产生多个子代个体,然后从这些子代个体中找出适应度较优的进入下一代种群.实验结果表明,引入正交设计思想能够提高算法性能,与其他优秀算法进行比较的结果说明,改进算法求解三目标问题(DTLZ系列)也是具有竞争力的.Multi-objective cellular genetic algorithm has proven to be effective in solving bi-objective MOPs. However,preliminary experiments have revealed that it has difficulties when dealing with three-objective MOPs（ the DTLZ problem family）. In order to enhance the performance,the orthogonal design idea is introduced and a newcellular genetic algorithm called cellular genetic algorithm for multi-objective optimization based on orthogonal design is proposed. In the progress of iteration of improved algorithm,the parent individuals are divided into many segments,and then several offsprings are produced by recombining the segments according to the orthogonal table,finally,choose the individuals which have better fitness value from the offsprings to the next population. The experiments showthat the performance is improved after introducing the orthogonal design. Compared with several state-of-the-art multi-objective metaheuristics,the obtained results showthat the improved algorithm is competitive for DTLZ problem family,too.

关 键 词：多目标 元胞遗传算法 正交设计 函数优化 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]