Resource Allocation of Full-Duplex Vehicular Networks Based on Multi-Agent Deep Reinforcement Learning

With the development of vehicular networks, spectrum resources are becoming increasingly scarce, and improving spectrum efficiency (SE) has become more and more important. Compared with common half-duplex (HD) technology, wireless full-duplex (FD) technology can effectively improve SE. Therefore, in this paper, we introduce FD into vehicle-to-vehicle (V2V) links, investigate the problem of spectrum resource reuse in vehicular networks, and realize the reuse of FD V2V links for the spectrum occupied by vehicle-to-infrastructure (V2I) links. At the same time, we also adopt a multi-agent deep reinforcement learning method to solve the problem that the rapid changes in the vehicular environment. The simulation results show that under 120 dB self-interference cancellation, the transmission rate of FD V2V links is significantly improved compared with that of HD V2V links.