Scholarlink

<p>In intelligent manufacturing industry, automatic guided vehicle (AGV) is becoming the main transportation tool in production workshop. At the same time, in order to quickly respond to market demand, dynamic job arrival gradually become the norm in job processing. Therefore, the dynamic job shop scheduling problem with AGV (DJSP-AGV) and random job arrival has important significance and practical value. To minimize the makespan of DJSP-AGV, firstly, the DJSP-AGV model is established. Subsequently, the DJSP-AGV model is converted into Markov decision process (MDP) model, in which the state features of the job, machine and AGV are defined as state space according to characteristic of DJSP-AGV, job dispatching rules and new AGV dispatching rules are designed as the action space, average machine utilization rate and the time of job waiting for AGV are designed as reward functions, and the agent is constructed. Moreover, DQN training algorithm for the DJSP-AGV is devised. Finally, the experiments of dynamic instances are conducted, and the proposed DQN algorithm obviously outperforms the combination dispatching rules, and can deal with dynamic scheduling in DJSP-AGV.</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

Research on dynamic job shop scheduling problem with AGV based on DQN

  • 期刊论文

  • 发表日期：

    2025年2月25日

查看全文
  • Zhengfeng Li,
  • Wengpeng Gu,
  • Huichao Shang,
  • Guohui Zhang,
  • Gaofeng Zhou

摘要

In intelligent manufacturing industry, automatic guided vehicle (AGV) is becoming the main transportation tool in production workshop. At the same time, in order to quickly respond to market demand, dynamic job arrival gradually become the norm in job processing. Therefore, the dynamic job shop scheduling problem with AGV (DJSP-AGV) and random job arrival has important significance and practical value. To minimize the makespan of DJSP-AGV, firstly, the DJSP-AGV model is established. Subsequently, the DJSP-AGV model is converted into Markov decision process (MDP) model, in which the state features of the job, machine and AGV are defined as state space according to characteristic of DJSP-AGV, job dispatching rules and new AGV dispatching rules are designed as the action space, average machine utilization rate and the time of job waiting for AGV are designed as reward functions, and the agent is constructed. Moreover, DQN training algorithm for the DJSP-AGV is devised. Finally, the experiments of dynamic instances are conducted, and the proposed DQN algorithm obviously outperforms the combination dispatching rules, and can deal with dynamic scheduling in DJSP-AGV.