Personalized Federated Learning based Joint Latency and Power Optimization for UAV-assisted C-V2X Communications

This paper investigates the system performance and communication delays between vehicles and unmanned aerial vehicles (UAVs) in UAV-assisted cellular vehicle-to-everything (C-V2X) environment. In order to minimize the weighted total energy consumption of a UAV, we propose to optimize the transmission window length and maximize the UAV’s available transmit power in each transmission window. We also propose to efficiently control the power consumption of the UAV while considering the effects of orthogonal time frequency space modulation. We formulate a multi-objective optimization problem and implement a personalized federated learning (PFL) based solution. The simulation results reveal that with PFL based distributed and co-operative learning, the weighted total energy consumption of the UAV decreases by 10-15% compared with traditional federated learning algorithms such as federated averaging and federated stochastic gradient decent. Moreover, PFL achieves approximately 40% delay reduction in UAV-vehicle communication compared to traditional federated learning.