Data-Aware Path Planning for Autonomous Vehicles Using Reinforcement Learning

This paper addresses the challenge of optimizing path planning for autonomous vehicles in urban environments by considering both traffic and bandwidth variability on the road. Traditional path planning methods are inadequate for the needs of interconnected vehicles that require significant real-time data transfer. We propose a reinforcement learning approach for path planning, formulated to use road traffic conditions and bandwidth availability. This approach optimizes routes by minimizing travel time while maximizing data transfer capability. We create a realistic simulation environment using GraphML, incorporating real-world map data and vehicle mobility patterns to evaluate the effectiveness of our approach. Through comprehensive testing against various baselines, our reinforcement learning model demonstrates the ability to adapt and find optimal paths that significantly outperform conventional strategies. These results emphasize the feasibility of using reinforcement learning for dynamic path optimization and highlight its potential to improve both the efficiency of travel and the reliability of data-driven decisions in autonomous vehicular networks.