Mobility-aware and Dynamic Resource Allocation for Federated Learning in Vehicular Edge Computing

In Vehicular Edge Computing (VEC), the Mobile Edge Computing (MEC) servers at the edge of the vehicular network provide communication, computation, storage, and data resources for nearby vehicles to meet the low-latency and high-reliability requirements of V2X (Vehicle-to-Everything) applications. In order to reduce communication costs and protect data privacy of vehicle users, Federated Learning (FL) is introduced. In VEC, designing an efficient resource allocation scheme for FL is crucial to improve the quality of service and user experience of real-time applications. However, most of the existing resource allocation studies lack sufficient consideration of the high mobility of vehicles and data offloading to assist FL to meet the demands of highly dynamic scenarios in VEC. Therefore, this paper proposes a dynamic resource allocation method based on vehicle mobility and data offloading to minimize FL training delay and model training loss. Simulation results show that the proposed algorithm is well adapted to VEC scenarios and has higher model training performance, compared to the existing schemes.