Scholarlink

In Industry 4.0, Unmanned Vehicles play significant roles in smart manufacturing as they can replace human workers and improve the efficiency of manufacturing. Therefore, in this paper, we aim to address the allocation problem of Unmanned Vehicles in Industry 4.0. The Deep Reinforcement Learning algorithms are utilized to solve the allocation problem in both the single agent and multiple agents model. The agent, as the task assigner, observes the current Unmanned Vehicles utilization condition and predicts the usage patterns of these Unmanned Vehicles in the next time slot. In the next time slot, the agent can directly utilize the idle Unmanned Vehicles to complete the tasks. Traditional Deep Reinforcement Learning, especially Deep Q-Networks cannot address the large action space, therefore we propose to use the Deep Deterministic Policy Gradient algorithms to determine the optimal action for the agents. The simulation results show the superiority of the Deep Deterministic Policy Gradient algorithm in solving both the centralized and the distributed optimization problems for each agent.

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

Time Series Prediction in UAV-UGV Integrated Automation System by Deep Reinforcement Learning in Industry 4.0

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Haowen Pan,
  • Dongfang Hou,
  • Yuan Xing,
  • Jason Liu,
  • Abhishek Verma

摘要

In Industry 4.0, Unmanned Vehicles play significant roles in smart manufacturing as they can replace human workers and improve the efficiency of manufacturing. Therefore, in this paper, we aim to address the allocation problem of Unmanned Vehicles in Industry 4.0. The Deep Reinforcement Learning algorithms are utilized to solve the allocation problem in both the single agent and multiple agents model. The agent, as the task assigner, observes the current Unmanned Vehicles utilization condition and predicts the usage patterns of these Unmanned Vehicles in the next time slot. In the next time slot, the agent can directly utilize the idle Unmanned Vehicles to complete the tasks. Traditional Deep Reinforcement Learning, especially Deep Q-Networks cannot address the large action space, therefore we propose to use the Deep Deterministic Policy Gradient algorithms to determine the optimal action for the agents. The simulation results show the superiority of the Deep Deterministic Policy Gradient algorithm in solving both the centralized and the distributed optimization problems for each agent.