A Traffic Intersection Branch Flow Prediction Method Based on Reinforcement Learning

The increasing complexity of modern transportation networks, coupled with the need for efficient traffic management, necessitates the development of intelligent and adaptive prediction systems. Accurate traffic flow prediction, especially at intersections is crucial for improving road network design, and providing real-time travel information to commuters. To address the issue of mutual correlation and delay in traffic time series at multi branch intersections, we have introduced a new and simple regression method based on reinforcement learning to compensate the delay, making the intersection branches' prediction flow close to the actual situation. Experiments demonstrate the method's simplicity, ease of use, and robustness.