User Pairing in Full-Duplex Enabled Vehicular Networks Based on Deep Reinforcement Learning

This paper investigates the user pairing in vehicular networks where a full-duplex (FD) base station simultaneously communicates with multiple pairs of uplink (UL) and downlink (DL) half-duplex (HD) users. Specifically, we formulate the user pairing problem to maximize the total sum rate of the system. Traditional pairing methods, characterized by high computational complexity and limited adaptability to the dynamic nature of vehicular networks, are not suitable for practical vehicular scenarios. To address this challenge, we employ deep reinforcement learning to solve the user pairing problem. Our experiments show that the proposed algorithm outperforms random pairing methods approaching optimal performance with less time consumption.