Next-Generation Autonomous Vehicle Navigation: A DDPG Reinforcement Learning Approach to Dynamic HD Map Caching.

This work presents a solution to the challenges of real-time access and processing of High-Definition maps (HD-maps) in vehicular networks with limited bandwidth and high latency. Leveraging the Deep Deterministic Policy Gradient (DDPG) algorithm, the proposed approach dynamically caches relevant map tiles in Roadside Units (RSUs). By intelligently determining which map tiles to cache, considering factors like popularity, recency, and anticipated demand, RSUs can serve a significantly portion of map requests from vehicles. Empirical results demonstrate the superiority of proposed DDPG-based HD map caching over traditional methods leading to improved map retrieval times and alleviating network congestion.