Deep Reinforcement Learning-Based AoI Minimization in UAV-Assisted Multi-Platoon Vehicular Networks

In the current landscape of intelligent transportation, vehicle platooning has become a key strategy for improving traffic efficiency and safety. However, as multiple platoons move at high speeds, the platoon encounters an unpredictable communication environment, which poses a major challenge for coordinated control of multi-platoon vehicles. Meanwhile, the freshness of information plays a pivotal role in ensuring road traffic safety, quantified by the Age of Information (AoI). This study focuses on a UAV-assisted platooning system in Manhattan scenarios, aiming to minimize vehicle AoI while improving the communication success rate between vehicles and UAV nodes. To tackle this challenge, we model this problem as a Partially Observable Markov Decision Process (POMDP). We propose a trajectory and scheduling strategy based on the Multi-Agent Twin Delayed Deep Deterministic Policy Gradient (MATD3) approach. Additionally, to better adapt to the dynamic nature of multi-platoon vehicular networks, we integrate the Prioritized Experience Replay (PER) technique to prioritize learning and processing of critical experience samples. Through simulation experiments and comparisons with baseline methods, the superiority of the proposed algorithm is verified.