Scholarlink

In many military and engineering scenarios, systems are required to perform a series of missions. For a multi-unit system, a portion of the units is often needed to perform the missions, while the other units are on standby. As working units during a mission cannot be changed, selecting the optional units to work at the beginning of each mission is significant for mission success. Meanwhile, the remaining units being on standby can be maintained according to the strategy of maintenance. Condition on limited maintenance resources, such as personnel, devices, and budget, the strategy of maintenance should cooperate with the usage scheduling of units to enhance the system’s performance. Existing studies mainly focus on the optimization of maintenance or usage independently. In this paper, a novel framework for joint optimization combining maintenance and unit usage for multi-unit systems is presented. In the framework, the system performs a limited series of missions. The objective of this paper is to maximize the mission success. A deep reinforcement learning (DRL) approach is presented for solving the resulting dynamic optimization problem. Finally, a numerical example proves the effectiveness of our method.

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 DRL-Based Usage and Maintenance Policy Optimization Approach for Multi-unit Systems

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Shihan Tan,
  • Qiwei Hu,
  • Yongshen Bai,
  • Zhifeng You,
  • Mingchang Song,
  • Hao Yan,
  • Quan Shi

摘要

In many military and engineering scenarios, systems are required to perform a series of missions. For a multi-unit system, a portion of the units is often needed to perform the missions, while the other units are on standby. As working units during a mission cannot be changed, selecting the optional units to work at the beginning of each mission is significant for mission success. Meanwhile, the remaining units being on standby can be maintained according to the strategy of maintenance. Condition on limited maintenance resources, such as personnel, devices, and budget, the strategy of maintenance should cooperate with the usage scheduling of units to enhance the system’s performance. Existing studies mainly focus on the optimization of maintenance or usage independently. In this paper, a novel framework for joint optimization combining maintenance and unit usage for multi-unit systems is presented. In the framework, the system performs a limited series of missions. The objective of this paper is to maximize the mission success. A deep reinforcement learning (DRL) approach is presented for solving the resulting dynamic optimization problem. Finally, a numerical example proves the effectiveness of our method.