Joint User Selection and Resource Allocation for Efficient Federated Learning in UAV-Assisted Wireless Networks

Federated Learning(FL) empowers numerous terrestrial users to implement distributed learning. However, enabling FL in unmanned aerial vehicle(UAV)-assisted wireless networks is still challenging due to the heterogeneity of participating terrestrial users and limited communication resources. In this paper, we propose a FL framework to improve the training efficiency. Firstly, incorporating the idea of clustering, we explore a semi-supervised clustering ant colony algorithm to alleviate the issue of low training efficiency caused by the heterogeneity of terrestrial users. Subsequently, we introduce a convex optimization-based strategy to allocate resource blocks for scheduled users. Comprehensive experimental results demonstrate that, compared with benchmarking strategies, our solution can reduce training time by up to 36.4% for reaching the specific accuracy, and achieve up to 13.21% accuracy improvement in a relaxed scenario.