基于分解的约束多目标进化算法

Decomposition-Based Constrained Multi-Objective Evolutionary Algorithm

摘要: 现实世界中为平衡多方利益，需要进行多目标优化的研究，然而实际问题中多数问题往往带有约束条件，富有挑战性。在研究中，首要解决的问题为如何处理在优化过程中出现的非可行解。我们认为非可行解在优化过程中是重要的，其包含着种群进化方向的信息，需在优化过程中保存适当比例的非可行解。为此本文提出了一种基于分解的约束多目标进化算法，根据进化过程中种群变化来自适应产生权重保留一部分非可行解引导种群收敛。为验证所提算法的性能，选取了24个测试问题，4个对比算法，经数值实验证明了所提算法的有效性。

Abstract: In the real world, it is necessary to study multi-objective optimization in order to balance mul-ti-interests. However, most of the practical problems are often challenging with constraints. How to deal with the infeasible solution in the optimization is the most important problem to be solved in the research. We think that the infeasible solution is important in optimization, and it contains the information of population evolution direction, so it is necessary to keep an appropriate proportion of the infeasible solution in optimization. In this paper, a decomposition-based constrained multi-objective evolutionary algorithm is proposed, which adaptively generates weights according to population changes in evolution and preserves some infeasible solutions to guide population convergence. In order to verify the performance of the proposed algorithm, 24 test problems and 4 comparison algorithms are selected. Numerical experiments show the effectiveness of the proposed algorithm.