MARL-based Resource Allocation for Heterogeneous Traffic in V2X Communications

New Radio Vehicle-to-Everything (NR-V2X) has been recognized as a key technology for advanced driving applications such as autonomous driving, platooning, and Intelligent Transportation Systems (ITS). However, this also has led to the emergence of a heterogeneous traffic environment, where packets with diverse purposes, formats, and priorities are transmitted over the V2X network. Therefore, in order to meet the strict Quality of Service (QoS) requirements of NR-V2X in a heterogeneous traffic environment with limited shared resources, developing appropriate resource allocation methods for NR-V2X became one of the major problems. This paper proposes a Multi-Agent Reinforcement Learning (MARL) approach to address this issue. We propose to improve the overall Packet Reception Ratio (PRR) performance of the V2X network. Through simulations, we demonstrate a comparison of the PRR of the proposed approach with random/optimal resource allocation methods. We confirm that our proposed method performed almost similarly to an optimal resource allocation scheme.