基于多种群分区辅助的用于解决约束多目标优化问题的协同进化算法

A Cooperative Evolutionary Algorithm for Solving Constrained Multi-Objective Optimization Problems Based on Multi-Population Partition Assistance

在解决约束多目标优化问题时,需要同时考虑多个优化目标以及问题的约束条件,这给问题的解决带来了很大的挑战,主要体现在解的可行性、收敛性和多样性三个方面。为了应对这些挑战,本文提出了一种新的约束多目标进化优化算法,该算法采用了三个阶段和多个辅助种群的设计。具体来说,第一阶段中,算法首先利用不受约束的种群引领其他种群进行搜索,在第二阶段,每个辅助种群致力于解决单个约束条件下的多目标优化问题。最后,本文设计了一种分区辅助策略,根据不同种群在每个分区的表现来调整辅助种群的参与度。为了评估算法的表现和有效性,我们将其与当前较为优秀的四个算法在DASCMOPs测试基上进行了比较。实验结果表明,我们的算法具有出色的表现。

In tackling constrained multi-objective optimization problems,it is necessary to consider multiple optimization objectives as well as the constraints of the problem,which poses significant challenges to finding solutions.These challenges are mainly reflected in the feasibility,convergence,and diversity of solutions.To address these challenges,this paper proposes a novel constrained multi-objective evolutionary optimization algorithm,which employs a three-stage approach and multiple auxiliary populations.Specifically,in the first stage,the algorithm utilizes an unconstrained population to guide the search of other populations.In the second stage,each auxiliary population focuses on solving the multi-objective optimization problem under a single constraint.Finally,a partition-driven strategy is designed to adjust the participation of auxiliary populations in each partition based on their performance,thereby enhancing the efficiency of the main population in obtaining superior solutions.To evaluate the performance and effectiveness of the algorithm,we compare it with four currently classic and outstanding algorithms on the DASCMOPs test suite.The experimental results demonstrate the excellent performance of our algorithm.