Joint Design of Communication Efficiency and Resource Allocation in V2X Networks Based on Multi-agent Deep Reinforcement Learning^*

As wireless networks evolve, Vehicle-to-Everything (V2X) communications, including vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I), and vehicle-to-pedestrian (V2P), have gradually become more sophisticated, enabling information exchange between vehicles and their surrounding environment, thereby enhancing road safety and traffic efficiency. However, ensuring service quality when vehicles are moving at high speeds remains a significant challenge that cannot be overlooked.Due to the rapid changes in channels caused by the high mobility of vehicles, this paper models resource allocation as a multi-agent deep reinforcement learning problem. It analyzes multiple V2V links and V2I links and proposes a resource allocation algorithm that considers V2V communication efficiency based on the Deep Deterministic Policy Gradient (DDPG).Each agent interacts with the V2X network environment to obtain a common reward function and aggregates actions from other agents for training the critic network collectively. By designing the reward function, a balance between communication efficiency and power control can be achieved, thereby effectively increasing the transmission rate of V2V links.