Trajectory Data-Based Model Predictive Control of Adjacent Intersection Signal Control Under Traffic Accident Scenarios

When traffic accidents occur on connecting roadways at adjacent intersections, vehicles often overflow into upstream intersections, causing congestion to diffuse. In traffic management systems involving either autonomous or traditional vehicles, effectively responding to unexpected traffic incidents and ensuring smooth road operations is crucial. With the development of connected vehicles technologies, acquiring vehicle trajectory data over a wide area has become possible. The real-time acquisition of vehicle trajectory data during traffic accidents offers new possibilities for intersection signal control schemes to effectively manage accidents. However, studies applying trajectory data to adjacent intersections in the context of accidents are limited. Therefore, this study proposes a highly scalable model predictive control strategy for traffic accident scenarios that can be applied to a wide range of realistic high-density urban road networks. The proposed model takes vehicle trajectory data generated during an accident as input, quantifies the impact of the accident, and outputs an optimal signal control scheme to reduce delays and prevent congestion diffusion. The model was validated by performing simulations in Python and accessing the component object model interface of VISSIM. Comparisons were made with the Decentralized-MPC method and the fixed-time control scheme. The applicability of the Decentralized-MPC method in traffic accident scenarios with different traffic volumes was also examined. The proposed MPC-Incident method reduced the average vehicle delay by 7.8% compared with the Decentralized-MPC method. The proposed signal control optimization scheme can offer technical support for maintaining traffic flow stability and smoothness when autonomous and human-driven vehicles coexist.