Scholarlink

In contemporary applications of Unmanned Systems (US), Task Allocation (TA) is one of the pivotal technologies enabling efficient operation. Traditional task allocation methodologies encounter numerous challenges in dynamic and complex settings, particularly when handling multiple tasks concurrently. This paper introduces a novel task allocation approach for unmanned systems that integrates Multi-task Learning (MTL) with Deep Reinforcement Learning (DRL), specifically through the application of the Deep Q-Network (DQN) algorithm, to optimize task distribution in dynamic and complex environments. We validated the effectiveness of this unmanned system intelligent task allocation model on the Miaosuan Land Commander platform. Experimental results demonstrate that our model significantly outperforms traditional methods and other DRL algorithms across several evaluation metrics, such as win rates, reward values, and training efficiency. This research not only enhances the efficiency and adaptability of task allocation in complex environments for unmanned systems but also offers new perspectives and practical evidence for the application of deep reinforcement learning in multi-task learning.

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

Efficient Task Allocation in Unmanned Systems via Multi-Task Deep Reinforcement Learning

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Meng Yang,
  • Wenxin Wang,
  • Mingfeng Lin,
  • Kai Wang,
  • Wenfei Huang,
  • Ling Ye

摘要

In contemporary applications of Unmanned Systems (US), Task Allocation (TA) is one of the pivotal technologies enabling efficient operation. Traditional task allocation methodologies encounter numerous challenges in dynamic and complex settings, particularly when handling multiple tasks concurrently. This paper introduces a novel task allocation approach for unmanned systems that integrates Multi-task Learning (MTL) with Deep Reinforcement Learning (DRL), specifically through the application of the Deep Q-Network (DQN) algorithm, to optimize task distribution in dynamic and complex environments. We validated the effectiveness of this unmanned system intelligent task allocation model on the Miaosuan Land Commander platform. Experimental results demonstrate that our model significantly outperforms traditional methods and other DRL algorithms across several evaluation metrics, such as win rates, reward values, and training efficiency. This research not only enhances the efficiency and adaptability of task allocation in complex environments for unmanned systems but also offers new perspectives and practical evidence for the application of deep reinforcement learning in multi-task learning.