Joint Medium Access and Beam Training Using Sub-6 GHz Channels for mmWave NR-V2X Sidelink Communication

The millimeter wave (mmWave) vehicle-to-everything (V2X) sidelink communication is a promising approach to support the exchange of raw sensors between autonomous vehicles at Gbps data rates. However, the frequent high overhead of mmWave beamforming and high collision of directional sensing in dynamic environments bring great challenges to vehicular wireless access, and the two processes are coupled and influenced by each other. In this paper, we propose a joint directional channel access and beam management scheme for distributed V2X sidelink communications. Specifically, in the initial access stage, an autonomous distributed mode under mmWave and sub-6 GHz cooperated scheduling scheme is proposed to reduce directional signaling collisions and latency, which is modeled and analyzed theoretically. Then, in the connected stage, we design a dual-layer deep reinforcement learning method to solve the joint optimization problem of mmWave beam selection and resource allocation for a dynamic vehicular environment. The learning network includes a communication module and an adaptive learning module, and the spatial correlation between mmWave and sub-6 GHz channels is exploited. Simulation results demonstrate that the proposed sub-6 GHz and mmWave cooperative access scheme reduces control signaling overhead by 30% compared to traditional mmWave access schemes while improving transmission reliability. Moreover, the proposed dual-layer reinforcement learning-based optimization method in the connection stage can converge quickly in a dynamic environment and increase the average effective achievable rate by 17%.