考虑全局和局部帕累托前沿的多模态多目标优化算法  

Multimodal Multi-objective Evolutionary Algorithm Considering Global and Local Pareto Fronts

作　　者：李文桦 明梦君 张涛[1,2] 王锐 黄生俊[1,2] 王凌 LI Wen-Hua;MING Meng-Jun;ZHANG Tao;WANG Rui;HUANG Sheng-Jun;WANG Ling(College of Systems Engineering,National University of Defense Technology,Changsha 410073;Hunan Key Laboratory of Multi-energy System Intelligent Interconnection Technology,Changsha 410073;Department of Automation,Tsinghua University,Beijing 100084)

机构地区：[1]国防科技大学系统工程学院,长沙410073 [2]多能源系统智慧互联技术湖南省重点实验室,长沙410073 [3]清华大学自动化系,北京100084

出　　处：《自动化学报》2023年第1期148-160,共13页Acta Automatica Sinica

基　　金：国家优秀青年科学基金(62122093);国家自然科学基金(72071205,62273193)资助。

摘　　要：多模态多目标优化问题(Multimodal multi-objective optimization problems,MMOPs)是指具有多个全局或局部Pareto解集(Pareto solution sets,PSs)的多目标优化问题(Multi-objective optimization problems,MOPs).在这类问题中,Pareto前沿(Pareto front,PF)上相距很近的目标向量,可能对应于决策空间中相距较远的不同解.在实际应用中全局或局部最优解的缺失可能导致决策者缺乏对问题的整体认识,造成不必要的困难或经济损失.大部分多模态多目标进化算法(Multimodal multi-objective evolutionary algorithms,MMEAs)仅关注获取尽可能多的全局最优解集,而忽略了对局部最优解集的搜索.为了找到局部最优解集并提高多模态优化算法的性能,首先提出了一种局部收敛性指标(并设计了一种基于该指标和改进种群拥挤度的环境选择策略.基于此提出了一种用于获取全局和局部最优解集的多模态多目标优化算法.经实验验证,该算法在对比的代表性算法中性能较好.Multimodal multi-objective optimization problems(MMOPs)refer to problems with multiple global or local Pareto solution sets(PSs).Different solutions far apart in the decision space may correspond to objective vectors in the Pareto front(PF)that are closed.The lack of global or local optimal solutions in practical applications may lead to the lack of overall understanding of the problem for decision-makers,resulting in unnecessary difficulties or economic losses.Most of the multimodal multi-objective evolutionary algorithms(MMEAs)mainly focus on obtaining the global optimal solution sets and pay little attention to the local optimal solutions.In order to find the local optimal solution sets and improve the performance of MMEAs,this paper proposes a local convergence indicator(gorithm for obtaining global and local optimal solution sets is proposed.Experiments show that the performance of the proposed algorithm is better than that of the compared representative algorithms.

关 键 词：多模态多目标优化 局部收敛性 进化算法 种群多样性 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]