QoS based Deep Reinforcement Learning for V2X Resource Allocation

Abstract

The 3rd generation partnership project (3GPP) standard has introduced vehicle to everything (V2X) communication in Long Term Evolution (LTE) to pave the way for future intelligent transport solutions. V2X communication envisions to support a diverse range of use cases for e.g. cooperative collision avoidance, infotainment with stringent quality of service (QoS) requirements. The QoS requirements range from ultra-reliable low latency to high data rates depending on the supported application. This paper presents a QoS aware decentralized resource allocation for V2X communication based on a deep reinforcement learning (DRL) framework. The proposed scheme incorporates the independent QoS parameter, i.e. priority associated to each V2X message, that reflects the latency required in both user equipment (UE) and the base station. The goal of the approach is to maximize the throughput of all vehicle to infrastructure (V2I) links while meeting the latency constraints of vehicle to vehicle (V2V) links associated to the respective priority. A performance evaluation of the algorithm is conducted based on system level simulations for both urban and highway scenarios. The results show that incorporating the QoS parameter (i.e. priority) pertaining to the type of service supported is crucial in order to meet the latency requirements of the mission critical V2X applications.