A Deep Reinforcement Learning Framework to Combat Dynamic Blockage in mmWave V2X Networks

Millimeter Wave (mmWave) systems are considered as one of the key technologies in future wireless systems due to the abundant spectrum resources in mmWave band. With the aim of achieving the capacity requirements in vehicular networks, large antenna arrays can be deployed at both the road side units (RSUs) side and the vehicles side. However, dynamic blockage caused by mobile obstacles in mmWave bands may hinder the system reliability. In this work, we study the temporal effects of dynamic blockage in vehicular networks and propose a deep reinforcement learning framework to overcome dynamic blockage. By dynamically adjusting blockage detection parameters and making intelligent handover decisions according to the observed states, system reliability can be significantly improved. Simulation results based on ray-tracing channel data show that the proposed scheme reduces the violation probability by 28.9% over conventional schemes.