Scholarlink

This chapter focuses on an unmanned aerial vehicle (UAV) assisted FEEL system, where the UAV is in charge of coordinating distributed model training among ground devices. The UAV’s high altitude and mobility are leveraged to establish short-distance line-of-sight links with devices, preventing any single device from becoming a communication bottleneck. This approach accelerates model aggregation and reduces cumulative model loss due to device scheduling, thereby decreasing the overall completion time. We illustrate the impact of device scheduling on the convergence of the FEEL system and formulate a training time minimization problem with respect to device scheduling and UAV trajectory, followed by developing an alternating Lagrange dual ascent algorithm.

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 Edge Learning via Unmanned Aerial Vehicle

  • 章节

  • 发表日期：

    2026年

查看全文
  • Yong Zhou,
  • Wenzhi Fang,
  • Yuanming Shi,
  • Khaled B. Letaief

摘要

This chapter focuses on an unmanned aerial vehicle (UAV) assisted FEEL system, where the UAV is in charge of coordinating distributed model training among ground devices. The UAV’s high altitude and mobility are leveraged to establish short-distance line-of-sight links with devices, preventing any single device from becoming a communication bottleneck. This approach accelerates model aggregation and reduces cumulative model loss due to device scheduling, thereby decreasing the overall completion time. We illustrate the impact of device scheduling on the convergence of the FEEL system and formulate a training time minimization problem with respect to device scheduling and UAV trajectory, followed by developing an alternating Lagrange dual ascent algorithm.