Cooperative Multi-Intersection Traffic Signal Control Based on Deep Reinforcement Learning

Intersections are the key to improve traffic efficiency. For intersections with complex traffic conditions, if we want to improve traffic efficiency effectively, we should make traffic signals adjust adaptively according to different traffic status. Obviously, the traditional fixed timing strategy is hard to achieve this. In addition, cooperative control of multiple intersections will maximize their overall interests and reduce the contradictions between different intersections. Therefore, in this paper, we propose an adaptive traffic signal control method for multiple intersections based on deep reinforcement learning. We build a new model, focusing on the coordination between multiple intersections and carry out experiments on the simulation of urban mobility (SUMO) platform. In contrast, we find that the model based on deep reinforcement learning is very effective, and its control strategy for multiple intersections is much better than the fixed timing strategy.