Dynamic Spectrum Access for C-V2X via Imitating Indian Buffet Process

In dense traffic cases, the channels among vehicles and infrastructures in cellular vehicle-to-everything (C-V2X) network are likely to be congested. Vehicles can select spectrum resources autonomously by adopting sensing-based semipersistent scheduling scheme, which may result in frequent packet collisions due to the random selection process, especially when spectrum resources are limited. So far, there still lacks a definite stochastic expression to characterize the channel selection process for vehicles in C-V2X. In this article, we propose a novel deep reinforcement learning (DRL)-based dynamic spectrum access (DSA) algorithm for C-V2X via imitating Indian buffet process (IBP), aiming to meet the vehicles' strict communication requirements on high-transmission rate and low-collision probability. First, we explore the correlations among historical spectrum access data to acquire the channel availability list. Then, we exploit DRL to achieve distributive DSA among vehicles. Specifically, the channel state prediction for the distributive DSA is facilitated via imitating the classic IBP, and the spectrum access decisions are made by leveraging the deep Q -learning network combined with long short-term memory technique. Finally, comprehensive simulation results are presented to show that the proposed algorithm can improve the transmission rate by 15% and reduce the collision probability by 12% with a false alarm probability of 0.1 compared with other spectrum access methods.