A Dynamic Risk Assessment Method for Vehicles at Intersections Considering Imminent and Collision Risks

Abstract

Most of the traditional dynamic risk assessment methods for vehicles use parameters such as distance and speed to describe the current motion relationship between a traveling vehicle and surrounding objects, which cannot well describe the risk associated with potential future collisions of multidirectional traffic participants at intersections. To better describe the collision possibilities of moving vehicles and surrounding objects in an intersection, this study proposes a dynamic risk assessment method for vehicles considering imminent and collision risks. Firstly, the imminent risk is calculated based on the positional relationship between the surrounding objects and the target vehicle and the speed of the target vehicle, which characterizes the movement trend of the surrounding objects and the target vehicle; the trajectories of the surrounding objects and the target vehicle are predicted, and the collision risk is calculated, which characterizes the possibility of collision between the surrounding objects and the target vehicle. Then, the conflict time between the surrounding objects and the target vehicle is calculated, and the imminent risk and collision risk are selected; finally, the risks posed by other environmental elements to which the target vehicle is exposed are combined to determine the overall risk faced by the vehicle in real-time. This information can serve as a robust foundation for issuing safety warnings and making assisted driving decisions. In terms of experimental validation, the braking events were screened using the actual intersection data in Suzhou, and the distributional differences of the risks were examined using statistical methods, it was found that the differences in the risk levels before and after braking were significant at the 0.1% confidence level, which indicated that the model could accurately measure the driving risks of the vehicles, and the validity of the model was verified.