Scholarlink

We propose DRL_HGSANN, a novel deep reinforcement learning framework for the NP-hard flexible job-shop scheduling problem. Our approach combines: (1) heterogeneous graph representation of operations, machines and constraints, (2) self-attention based relationship modeling, and (3) integrated state embedding for policy training. Through stacked HGSANN layers with mean pooling, the method achieves superior solution quality and computational efficiency compared to existing approaches, as demonstrated across multiple problem scales in manufacturing applications.

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

Flexible Job-Shop Scheduling via Graph Neural Network and Deep Reinforcement Learning

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Wan-Ze Li,
  • Zhi-Qiang Wang,
  • Jing-Qiu Peng,
  • Zi-Qi Zhang,
  • Kun Li,
  • Rong Hu,
  • Zi-Ming Ma,
  • Ming-Yao Li,
  • Bin Qian

摘要

We propose DRL_HGSANN, a novel deep reinforcement learning framework for the NP-hard flexible job-shop scheduling problem. Our approach combines: (1) heterogeneous graph representation of operations, machines and constraints, (2) self-attention based relationship modeling, and (3) integrated state embedding for policy training. Through stacked HGSANN layers with mean pooling, the method achieves superior solution quality and computational efficiency compared to existing approaches, as demonstrated across multiple problem scales in manufacturing applications.