基于动态ε约束处理机制的双种群约束多目标优化算法  

A two-population constrained multi-objective optimization algorithm based on a dynamicεconstraint processing mechanism

作　　者：涂继伟 汪镭[1] 蔡振翔 耿绍晋 李东洋 TU Jiwei;WANG Lei;CAI Zhenxiang;GENG Shaojin;LI Dongyang(School of Electronics and Information Engineering,Tongji University,Shanghai 201804,China;Sino-German College of Applied Science,Tongji University,Shanghai 201804,China)

机构地区：[1]同济大学电子与信息工程学院,上海201804 [2]同济大学中德工程学院,上海201804

出　　处：《南昌工程学院学报》2024年第1期82-92,共11页Journal of Nanchang Institute of Technology

摘　　要：约束多目标优化问题(CMOPs)除了需要解决多个相互冲突的目标之外,还需要满足一定的约束条件。针对约束造成CMOPs的Pareto前沿被分为多个部分,同时不可行区域的扩张进一步阻碍种群的探索,使种群陷入局部最优及其多样性急剧下降等问题,提出了一种基于动态ε约束处理机制的双种群约束多目标优化算法。该算法使用双种群协同进化策略,主种群考虑约束,通过改进的动态ε约束处理机制,充分利用不可行解提供的有效信息;而辅助种群不考虑约束,在平衡多样性的基础上向无约束Pareto前沿(UPF)快速收敛,并及时向主种群提供可行域外的有效信息,指导主种群的探索方向。实验结果表明所提出的算法在MW测试问题上相比其他算法更具竞争力。Constrained multi-objective optimization problems(CMOPs)need to meet certain constraints in addition to solving multiple conflicting objectives.In view of the fact that the existence of constraints causes the Pareto front of CMOPs to be divided into multiple parts,while the expansion of infeasible regions further hinders the exploration of the population,causing the population to fall into a local optimum and its diversity to decrease dramatically,this paper proposes a dual-population-based optimization algorithm for constrained multi-objective optimization based on dynamic epsilon constraints processing mechanism.The algorithm uses a dual-population co-evolutionary strategy,in which the main population takes constraints into consideration and makes full use of the effective information provided by infeasible solutions through the improved dynamic epsilon constraint processing mechanism,while the auxiliary population does not consider constraints and converges rapidly to the Unconstrained Pareto Front(UPF)based on balancing the diversity,and provides effective information outside the feasible domain to the main population in a timely manner,and also provide the main population with effective information outside the feasible domain in time to guide the exploration direction of the main population.The experimental results show that the proposed algorithm is more competitive than other algorithms in the MW testing problems.

关 键 词：约束处理机制 约束多目标优化 双种群 进化算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]