Federated Learning Empowered V2V Resource Allocation in IRS-assisted Vehicular Networks

Abstract

To solve the issue of the low transmission rate and energy efficiency in vehicle-to-vehicle (V2V) communication, a periodic-federated-adjusted double deep Q-network (PFADDQN) algorithm is proposed for intelligent reflecting surface (IRS)assisted vehicular networks. Considering the constraints in channel and power allocation in V2V communication, a joint benefit, which is defined as the combination of transmission success rate and energy consumption, is maximized. By using ajoint federated learning and double deep Q-network approach, the original NPhard optimization problem is solved. Simulation results demonstrate the superiority of our proposed PFADDQN algorithm over other baselines in IRS-assisted vehicular networks.