.png){kind=icon}
.png){kind=icon}
Return
Dual-stage dual-population diversity maintenance for global and local exploration of constrained multiobjective optimization
.png){kind=icon}
PRE
AI
H
DOI：10.1016/j.eswa.2024.126229.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In the field of constrained multiobjective optimization, some constrained multiobjective optimization problems (CMOPs) have large infeasible regions and discrete small feasible regions. Solutions to these problems are increasingly challenging due to the necessity for the population to traverse extensive infeasible regions while preserving diversity. To address this challenge, this paper introduces a constrained multiobjective evolutionary algorithm with diversity maintenance for global and local exploration, termed DMGLE. Specifically, the global exploration stage aims to converge to an unconstrained Pareto front, and the two populations focus on convergence and diversity without considering any constraints. In this stage, diversity is maintained through a global search operator that integrates an affinity propagation algorithm and a global differential evolutionary algorithm. In the local exploration stage, the two populations focus on feasibility and diversity. A global search operator and a local search operator are employed to sustain diversity and prevent premature convergence. An improved e-constraint handling technique is also introduced to guide populations gradually towards the true constrained Pareto front. Additionally, the exchange of information between the two populations bolsters the diversity of the algorithm. Compared to seven current state-of-the-art multiobjective evolutionary algorithms on three benchmark test suites and 10 real-world CMOPs, the proposed DMGLE achieved superior or highly competitive performance, especially for CMOPs with large infeasible regions and discontinuous small feasible regions.
Keywords:
Diversity maintenance
Affinity propagation
Constrained multiobjective optimization
Evolutionary algorithm
Journal
IF:
7.5
Papers:
2.9W
Citations:
10.2W
Organization
No organization information available
Cited Papers