EMGLight: A Joint Optimization Method for Emergency Signal Priority and Recovery Based on DDPG

The rapid response of emergency vehicles (EVs) is crucial in safeguarding lives and property during emergencies. However, conventional traffic signal control methods for EV priority often disrupt normal traffic flow, leading to significant delays for general vehicles and decreased overall traffic efficiency. This study proposes EMGLight, a novel traffic signal control framework based on Deep Deterministic Policy Gradient (DDPG), to optimize EV priority and signal recovery jointly. By leveraging DDPG's ability to handle continuous action spaces, EMGLight achieves fine-grained control over traffic signals, adapting dynamically to real-time traffic conditions. Additionally, a dynamic reward mechanism is introduced, balancing EV priority with the recovery needs of general traffic. Simulation results demonstrate that EMGLight outperforms traditional fixed-cycle and greedy preemption methods, significantly reducing EV travel time while minimizing the adverse impact on general traffic flow. This approach highlights the potential of reinforcement learning to enhance emergency urban traffic resilience.