On the Optimization of UAV-Assisted Wireless Networks for Hierarchical Federated Learning

This paper considers an unmanned aerial vehicle (UAV)-assisted Hierarchical Federated Learning (HFL), where UAVs act as intermediate aggregators. We formulate an optimization problem that aims to find the best UAV placements, user-UAV associations, channel assignments, and user selection to minimize the time needed for HFL to achieve a target learning accuracy. We propose a two-phase approach to solve the problem. The first one deals with the UAV placement using the K-means algorithm. For the second phase, we propose a user association and selection algorithm that prioritizes selecting users who can make a time-efficient and significant contribution to the FL training process. Simulation results show the proposed algorithm’s efficiency and ability to reach the target learning accuracy faster than the considered benchmarks.