Scholarlink

The Job Shop Scheduling Problem (JSSP) is a well-known combinatorial optimization problem in operations research, commonly found in fields like manufacturing and transportation. There are two main challenges: 1) a limited number of data instances, and 2) the inconsistent distributions in the scheduling process. This paper develops a domain adaptive-based reinforcement learning algorithm (DA-L2D) for job shop scheduling problems. In particular, it comprises two main parts: a domain adaptation module is designed to learn and comprehend the characteristics of data instance distribution, and an L2D model is employed to understand and manage the dynamic graph relationships. The combination of these two components enables the DA-L2D model to adeptly address intricate scheduling problems by mining the data’s characteristics. We conduct experiments on the Taillard and DMU datasets. Experimental results demonstrate the superiority of the domain adaptation to the scheduling problems.

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 Domain Adaptive Based Reinforcement Learning Algorithm for Job Shop Scheduling Problems

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Kuan Miao,
  • Lilan Peng,
  • Wuyang Zhang,
  • Jibao Pan,
  • Chongshou Li,
  • Tianrui Li

摘要

The Job Shop Scheduling Problem (JSSP) is a well-known combinatorial optimization problem in operations research, commonly found in fields like manufacturing and transportation. There are two main challenges: 1) a limited number of data instances, and 2) the inconsistent distributions in the scheduling process. This paper develops a domain adaptive-based reinforcement learning algorithm (DA-L2D) for job shop scheduling problems. In particular, it comprises two main parts: a domain adaptation module is designed to learn and comprehend the characteristics of data instance distribution, and an L2D model is employed to understand and manage the dynamic graph relationships. The combination of these two components enables the DA-L2D model to adeptly address intricate scheduling problems by mining the data’s characteristics. We conduct experiments on the Taillard and DMU datasets. Experimental results demonstrate the superiority of the domain adaptation to the scheduling problems.