Penalty-Fuzzy Logic Controller-Based Dual Population Evolutionary Algorithm

针对约束多目标优化问题的复杂性，本文提出了一种基于模糊惩罚的双种群进化算法。首先，设计了一种基于模糊逻辑控制器的动态约束惩罚机制，该机制通过模拟人类思维自适应地调整惩罚因子，有效平衡了种群在可行解与不可行解之间的探索。其次，针对不同种群的进化需求，设计了差异化的子代生成机制，以增强种群的探索和开发能力。最后，利用信息共享机制，使得不同种群间能够共享优质解和有效的约束信息，进一步提高解的质量。将所提出的算法与目前最先进的六种约束多目标优化算法进行实验对比。实验结果表明，所提出的算法在复杂约束环境下展现出较强的鲁棒性，尤其能够有效处理决策变量耦合的问题。In response to the complexity of constrained multi-objective optimization problems, this paper proposes a dual population evolutionary algorithm based on fuzzy penalties. Firstly, a dynamic constraint penalty mechanism based on a fuzzy logic controller is designed. This mechanism adaptively adjusts the penalty factor by simulating human thought processes, effectively balancing the exploration of feasible and infeasible solutions within the population. Secondly, a differentiated offspring generation mechanism is designed to address the evolutionary needs of different populations, enhancing the exploration and exploitation capabilities of the population. Finally, different populations can share high-quality solutions and effective constraint information by utilizing the information-sharing mechanism, thereby further improving the solution quality. The proposed algorithm is compared experimentally with six state-of-the-art constrained multi-objective optimization algorithms. The experimental results demonstrate that the proposed algorithm exhibits significant robustness in complex constrained environments, particularly in its ability to effectively handle the coupling of decision variables.