Efficient Correlated Sources Scheduling in Vehicular Networks via Deep Reinforcement Learning

Selecting the optimal vehicle for communication in cooperative sensing scenarios of vehicular networks is crucial for minimizing costs and ensuring timely information updates, especially when multiple vehicles are monitoring the same event. This paper proposes a Proximal Policy Optimization with Attention (PPO-A) algorithm to address the vehicle selection problem, taking into account the correlation of information among vehicles. A comprehensive network resource model is developed to analyze the relationship between information correlation and the Age of Information (AoI). Compared to baseline algorithms, the AoI is reduced by 12% to 25%, the transmission delay is reduced by 10% to 25%, and the overall communication efficiency is enhanced through the application of the proposed PPO-A algorithm. These improvements demonstrate its effectiveness in enhancing communication efficiency within vehicular networks.