Optimizing Energy Efficiency for Federated Learning in Rotary-Wing UAV Air-to-Ground Communications

Abstract

In the 5G and upcoming 6G eras, federated learning (FL) plays a crucial role in enabling intelligent, privacy-preserving networks by training models locally on user devices. However, FL faces challenges such as unstable wireless channels and limited device energy. This paper proposes a UAV-assisted framework to reduce energy consumption in FL by allowing simultaneous user transmissions over shared spectrum. We consider realistic air-to-ground (A2G) communication, including both LoS and NLoS links between the UAV and UEs, while accounting for the UAV’s limited energy. The problem of minimizing user energy consumption during FL is formulated as a nonconvex optimization, which is efficiently solved using alternating optimization and inner approximation techniques. Simulation results demonstrate the effectiveness of our approach in optimizing UAV placement and improving energy efficiency.