Resource Allocation in V2V Communication via Federated Multi-Agent Reinforcement Learning

With the increasing number of vehicles, the Internet of Vehicles (IoV) exhibits significant dense deployment characteristics, resulting in severe overlaps of communication areas among vehicles and complex interference. To address these issues, the resource allocation problem under complex interference in densely deployed IoV is investigated in this paper. Specifically, an interference hypergraph model is constructed to simultaneously analyze the complex interference relationships among multiple vehicles. Subsequently, the resource allocation problem is transformed into a vertex coloring problem on the hypergraph. Furthermore, a federated double dueling deep Q-Network algorithm is proposed to achieve conflict-free resource allocation while maximizing network throughput. Simulation experiments demonstrate that the proposed method achieves an average network throughput improvement of more than 17.93% compared to the baseline algorithms, showcasing superior network performance in densely deployed IoV scenarios.