Federated Deep Reinforcement Learning-Based Spectrum Access Algorithm With Warranty Contract in Intelligent Transportation Systems

Cognitive radio (CR) provides an effective solution to meet the huge bandwidth requirements in intelligent transportation systems (ITS), which enables secondary users (SUs) access idle spectrum of primary (PUs). However, high mobility and real-time service result additional transmission collisions interference, degrades rate quality (QoS) ITS. This paper proposes a algorithm (Feilin) based on federated deep reinforcement learning (FDRL) improve rate, maximizes QoS reward function with considering hybrid benefits delay, power utility SUs. To guarantee SUs, warranty contract is designed for SUs obtain compensation data failure, promotes compete more resources. ITS, model called FDQN-W proposed Q-network (DQN), adopts asynchronous weighted (AFWLA) share update weights DQN multiple agents decrease time cost accelerate convergence. Detailed simulation results show that, multiuser scenario, compared existing methods, Feilin increases success by 15.1%, reduces collision PUs 46.4% 6.8%, respectively.