Scholarlink

With the proliferation of various mobile smart devices, edge computing and computational offloading technologies have emerged as pivotal support mechanisms, enhancing the service quality of these devices. To facilitate the learning of task offloading strategies across diverse and complex scenarios, this study introduces a federated learning strategy algorithm based on multi-agent deep reinforcement learning. This algorithm aims to aggregate the training strategies of multiple edge computing devices, thereby enabling the synthesis of superior task offloading strategies tailored to a wide array of environments while also providing safeguards against malicious nodes. This paper specifically examines the reward inversion attack, illustrating the algorithm’s capability to identify and counteract such malicious threats effectively. Experimental results validate that the proposed algorithm not only robustly defends against these attacks but also adeptly learns the task offloading strategies pertinent to each node.

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

Robust Multi-agent Federated Reinforcement Learning for Task Offloading

  • 会议论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年

查看全文
  • Dibao Yan,
  • Yongfeng Wang,
  • Wenjing Hou,
  • Huanhuan Song,
  • Hong Wen,
  • Wendi Ma,
  • Fan Sun

摘要

With the proliferation of various mobile smart devices, edge computing and computational offloading technologies have emerged as pivotal support mechanisms, enhancing the service quality of these devices. To facilitate the learning of task offloading strategies across diverse and complex scenarios, this study introduces a federated learning strategy algorithm based on multi-agent deep reinforcement learning. This algorithm aims to aggregate the training strategies of multiple edge computing devices, thereby enabling the synthesis of superior task offloading strategies tailored to a wide array of environments while also providing safeguards against malicious nodes. This paper specifically examines the reward inversion attack, illustrating the algorithm’s capability to identify and counteract such malicious threats effectively. Experimental results validate that the proposed algorithm not only robustly defends against these attacks but also adeptly learns the task offloading strategies pertinent to each node.