Dynamic Constrained Boundary Method for Constrained Multi-Objective Optimization
Qiuzhen Wang,
Zhibing Liang,
Juan Zou,
Xiangdong Yin,
Yuan Liu,
Yaru Hu,
Yizhang Xia
Affiliations
Qiuzhen Wang
Key Laboratory of Intelligent Computing and Information Processing, Ministry of Education, School of Computer Science and School of Cyberspace Science of Xiangtan University, Xiangtan 411100, China
Zhibing Liang
Key Laboratory of Intelligent Computing and Information Processing, Ministry of Education, School of Computer Science and School of Cyberspace Science of Xiangtan University, Xiangtan 411100, China
Juan Zou
Key Laboratory of Intelligent Computing and Information Processing, Ministry of Education, School of Computer Science and School of Cyberspace Science of Xiangtan University, Xiangtan 411100, China
Xiangdong Yin
Faculty of Informational Engineering, Hunan University of Science and Engineering, Yongzhou 411201, China
Yuan Liu
Key Laboratory of Intelligent Computing and Information Processing, Ministry of Education, School of Computer Science and School of Cyberspace Science of Xiangtan University, Xiangtan 411100, China
Yaru Hu
Key Laboratory of Intelligent Computing and Information Processing, Ministry of Education, School of Computer Science and School of Cyberspace Science of Xiangtan University, Xiangtan 411100, China
Yizhang Xia
Key Laboratory of Intelligent Computing and Information Processing, Ministry of Education, School of Computer Science and School of Cyberspace Science of Xiangtan University, Xiangtan 411100, China
When solving complex constrained problems, how to efficiently utilize promising infeasible solutions is an essential issue because these promising infeasible solutions can significantly improve the diversity of algorithms. However, most existing constrained multi-objective evolutionary algorithms (CMOEAs) do not fully exploit these promising infeasible solutions. In order to solve this problem, a constrained multi-objective optimization evolutionary algorithm based on the dynamic constraint boundary method is proposed (CDCBM). The proposed algorithm continuously searches for promising infeasible solutions between UPF (the unconstrained Pareto front) and CPF (the constrained Pareto front) during the evolution process by the dynamically changing auxiliary population of the constraint boundary, which continuously provides supplementary evolutionary directions to the main population and improves the convergence and diversity of the main population. Extensive experiments on three well-known test suites and three real-world constrained multi-objective optimization problems demonstrate that CDCBM is more competitive than seven state-of-the-art CMOEAs.
