MSRA: Mode Selection and Resource Allocation for Cooperative Vehicle-Infrastructure System

Cellular vehicle-to-everything (C-V2X) communication is essential for supporting diverse vehicle applications in cooperative vehicle-infrastructure systems (CVISs). However, various factors, such as building obstructions, signal blockages, and network conditions, severely affect the latency and reliability of V2V communication. To address this challenge, we propose a novel dynamic method for communication mode selection and resource allocation in C-V2X, termed MSRA, which enables vehicles to dynamically select and utilize different communication modes, such as V2V, Vehicle-to-Infrastructure (V2I), and Vehicle-to-Network (V2N) for direct, relay, or forwarding communication, based on real-time conditions. To this end, we formulate the joint optimization problem of mode selection and resource allocation as a Markov decision process (MDP) and propose a solution based on multiagent reinforcement learning (MARL) with centralized training and decentralized execution. Specifically, each vehicle acts as an agent, independently selecting communication modes and resources based on real-time network status and communication link quality, aiming to satisfy the latency and reliability requirements of V2V communication while maximizing the capacity of V2N communication based on heterogeneous Quality-of-Service (QoS) requirements. Simulation results indicate that the proposed algorithm significantly outperforms other decentralized baselines and demonstrates superior performance under various conditions.