Scholarlink

<p>This paper proposes a hybrid search algorithm that integrates random flight, follow leader policy, and reinforcement learning, aiming to efficiently solve the flexible job shop scheduling problems. The algorithm adopts a two-stage encoding policy and a random-key-based encoding conversion mechanism, effectively establishing a mapping relationship between individual positions and the flexible job shop scheduling problem solutions. By introducing a reinforcement learning mechanism, the flexible job shop scheduling problem is transformed into a Markov decision process. Furthermore, a carefully designed system of state space, action space, and reward is utilized to achieve precise and efficient exploration of the local search space. This algorithm framework combines the extensive exploration capabilities of global search with the fine optimization capabilities of local search, significantly enhancing solution efficiency and algorithm performance. Empirical analysis demonstrates that the results on multiple authoritative benchmark datasets outperform current state-of-the-art algorithms, verifying the algorithm's outstanding performance and broad applicability in solving the flexible job shop scheduling problems.</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 random flight–follow leader and reinforcement learning approach for flexible job shop scheduling problem

  • 期刊论文

  • 发表日期：

    2025年2月10日

查看全文
  • Changshun Shao,
  • Zhenglin Yu,
  • Hongchang Ding,
  • Guohua Cao,
  • Jingsong Duan,
  • Bin Zhou

摘要

This paper proposes a hybrid search algorithm that integrates random flight, follow leader policy, and reinforcement learning, aiming to efficiently solve the flexible job shop scheduling problems. The algorithm adopts a two-stage encoding policy and a random-key-based encoding conversion mechanism, effectively establishing a mapping relationship between individual positions and the flexible job shop scheduling problem solutions. By introducing a reinforcement learning mechanism, the flexible job shop scheduling problem is transformed into a Markov decision process. Furthermore, a carefully designed system of state space, action space, and reward is utilized to achieve precise and efficient exploration of the local search space. This algorithm framework combines the extensive exploration capabilities of global search with the fine optimization capabilities of local search, significantly enhancing solution efficiency and algorithm performance. Empirical analysis demonstrates that the results on multiple authoritative benchmark datasets outperform current state-of-the-art algorithms, verifying the algorithm's outstanding performance and broad applicability in solving the flexible job shop scheduling problems.