A DDQN-based Energy-Efficient Resource Allocation Scheme for Low-Latency V2V communication

Abstract

The emerging new radio vehicle-to-everything (NR-V2X) based on 5G cellular network has developed rapidly. In this paper, we focus on the sidelink (SL) resource allocation for NR-V2X, especially the Vehicle-to-Vehicle (V2V) sidelink outside the signal coverage of the 5G base station. Based on network slicing technology, V2V sidelinks communicate through spectrum sliced for 5G ultra-reliable and low latency communications (URLLC) scenario. To ensure the energy efficiency of V2V sidelink and meet the delay requirements of URLLC communication, a deep reinforcement learning (DRL) architecture using centralized training and distributed execution is proposed, and a model that meets the above requirements is trained with the help of the Double Deep Q-Network(DDQN). Simulation results indicate our DDQN-based algorithm outperforms the compared algorithm in the performance of energy efficiency and latency.