Scholarlink

<p>Evolutionary multitasking (EMT) frameworks have attracted widespread attention in constrained multi-objective optimization problems (CMOPs). However, the challenges persist in task formulation and knowledge transfer. This paper proposes a three-stage multitasking collaborative optimization framework, which divides the optimization process into three stages: unconstrained exploration, dynamic constraint allocation, and full-constraint convergence. In the dynamic constraint allocation stage, constraint priorities are determined based on the distribution of constraint violations in the current population and are dynamically assigned to the main and auxiliary tasks. To further enhance the effectiveness of knowledge transfer, a boundary-guided individual transfer strategy is designed. The effectiveness of the proposed algorithm has been confirmed by experiments performed on four standard CMOP benchmark suites and 11 real-world CMOPs. The experimental results indicate the superiority of the proposed algorithm over several state-of-the-art algorithms in handling CMOPs.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A three-stage dynamic constraint allocation algorithm with boundary guided search for constrained multi-objective optimization

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月2日

查看全文
  • Yujia Wang,
  • Jingzhi Zhang,
  • Jiayi Wang

摘要

Evolutionary multitasking (EMT) frameworks have attracted widespread attention in constrained multi-objective optimization problems (CMOPs). However, the challenges persist in task formulation and knowledge transfer. This paper proposes a three-stage multitasking collaborative optimization framework, which divides the optimization process into three stages: unconstrained exploration, dynamic constraint allocation, and full-constraint convergence. In the dynamic constraint allocation stage, constraint priorities are determined based on the distribution of constraint violations in the current population and are dynamically assigned to the main and auxiliary tasks. To further enhance the effectiveness of knowledge transfer, a boundary-guided individual transfer strategy is designed. The effectiveness of the proposed algorithm has been confirmed by experiments performed on four standard CMOP benchmark suites and 11 real-world CMOPs. The experimental results indicate the superiority of the proposed algorithm over several state-of-the-art algorithms in handling CMOPs.