进化超多目标优化研究进展及展望  

Research progress and prospect of evolutionary many-objective optimization

作　　者：肖人彬[1,2] 李贵[1,2] 陈峙臻 XIAO Ren-bin;LI Gui;CHEN Zhi-zhen(School of Artificial Intelligence and Automation,Huazhong University of Science and Technology,Wuhan 430074,China;Institute of Artificial Intelligence,Huazhong University of Science and Technology,Wuhan 430074,China;Business School,University of Greenwich,London SE109LS,UK)

机构地区：[1]华中科技大学人工智能与自动化学院,武汉430074 [2]华中科技大学人工智能研究院,武汉430074 [3]格林威治大学商学院,伦敦SE109LS

出　　处：《控制与决策》2023年第7期1761-1788,共28页Control and Decision

基　　金：科技创新2030——“新一代人工智能”重大项目(2018AAA0101200);国家自然科学基金项目(52275249).

摘　　要：近年来,超多目标优化逐渐成为多目标优化研究的热点之一,由于超多目标优化问题具有难以寻优的高维目标空间,其研究颇有挑战性,因此受到广泛关注.现有综述性文献通常只是针对某个特定方面,缺乏系统性考察.鉴于此,首先从问题定义出发,综合考虑超多目标优化问题范畴,进行超多目标优化问题的概念辨析;其次通过对近些年的相关文献整理,系统分析超多目标优化问题进展并对其中部分经典方法加以介绍,通过对基准测试函数和性能指标的说明,围绕超多目标优化研究方法展开综合性论述;接着选取5个典型的超多目标进化算法,在2组基准测试函数和4个实际问题上分别展开仿真实验,通过性能指标和非参数检验对不同类别的算法进行理论分析;最后在明确超多目标优化研究领域的若干前沿问题的基础上,对今后的研究工作进行展望.In recent years,many-objective optimization has gradually become one of the research hotspots of multi-objective optimization.Due to the high-dimensional objective space is difficult to optimize,the research on many-objective optimization problems(MaOPs)is quite challenging and has received extensive attention.The existing surveys usually only focus on a specific aspect and lacks systematic investigation.Therefore,this paper firstly starts from the problem definition,considers the category of MaOPs,and makes the concept analysis of MaOPs.Secondly,the progress of MaOPs is systematically analyzed and some classical methods are introduced by collating the relevant works in recent years.Through the explanation of benchmark functions and performance indicators,the research method of many-objective optimization is comprehensively discussed.Then,five typical many-objective evolutionary algorithms(MaOEAs)are selected.The simulation experiments are carried out on two groups of benchmark functions and four practical problems.The different algorithms are analyzed theoretically by performance indicators and nonparametric tests.Finally,the future research work is prospected based on identifying some frontier problems in many-objective optimization.

关 键 词：超多目标优化 高维多目标 超多目标应用 进化算法 性能指标 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]