Emergency vehicle signal priority control method for arterial intersections in an intelligent connected environment

Emergency vehicle (EV) progression at consecutive arterial intersections in an intelligent connected environment is often disrupted by upstream queues, downstream spillback, and mismatches in multi-intersection signal coordination. To address this problem, this study proposes a spatiotemporal graph multi-agent proximal policy optimization (STG-MAPPO) method for EV signal priority control. The proposed method models consecutive signalized intersections as cooperative agents, incorporates EV arrival prediction, target-lane yielding capacity, and downstream blockage risk into the state representation, and uses a spatiotemporal graph encoder to capture both topological coupling and traffic-state propagation among intersections. Heterogeneous yielding responses of connected and automated vehicles (CAVs) and human-driven vehicles (HDVs) are explicitly considered in the target-lane yielding model. Simulation experiments based on the Zhengzhou Longhu Autonomous Driving Test Zone show that STG-MAPPO achieves better EV progression efficiency, lower general-traffic delay, and improved safety indicators than fixed-time control, actuated control, rule-based EV priority, DDQN, IPPO, and MAPPO. Under the specified SUMO-based single-EV simulation conditions, the proposed method reduces EV travel time, EV delay, and EV stop frequency by 22.28%, 51.02%, and 60.00%, respectively, compared with DDQN, and by 8.54%, 22.58%, and 33.33%, respectively, compared with MAPPO.