Scholarlink

This paper proposes a federated unmanned aerial vehicle (UAV)-assisted dueling double deep Q-network (FedUAV-D3QN) for mobile edge computing (MEC) to address task offloading and resource allocation challenges in dynamic environments. By formulating a joint optimization problem that minimizes the weighted sum of delay and energy consumption, the proposed method integrates federated learning to enable distributed model training while preserving data privacy. Simulation results demonstrate that when the cache size is \(300\,\text {kbits}\) , the proposed method reduces the system cost by approximately \(9.76\%\) compared to DQN. These results validate the effectiveness and scalability of the proposed FedUAV-D3QN in enhancing system performance in UAV-assisted MEC environments.

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

Federated D3QN-Based Offloading and Resource Optimization in UAV-Assisted Integrated Communication and Computing Networks

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ziyu Ma,
  • Chunyu Pan,
  • Dongdong Zhou

摘要

This paper proposes a federated unmanned aerial vehicle (UAV)-assisted dueling double deep Q-network (FedUAV-D3QN) for mobile edge computing (MEC) to address task offloading and resource allocation challenges in dynamic environments. By formulating a joint optimization problem that minimizes the weighted sum of delay and energy consumption, the proposed method integrates federated learning to enable distributed model training while preserving data privacy. Simulation results demonstrate that when the cache size is \(300\,\text {kbits}\) , the proposed method reduces the system cost by approximately \(9.76\%\) compared to DQN. These results validate the effectiveness and scalability of the proposed FedUAV-D3QN in enhancing system performance in UAV-assisted MEC environments.