A collision-probability-based vehicle risk field model based on rectangular contours and its application in vehicle lane-change process

This study proposes a vehicle collision risk field model that represents risk values based on collision probability and uses the vehicle’s rectangular contour as the risk field boundary. The model matches the distance that would be traveled over the time-to-collision (TTC) at the current relative speed with the statistical or driver-perceived probability of a collision, so that the calculated risk value can be consistent with the actual collision probability or the driver’s perceived probability of collision. Constructing the risk field based on the rectangular contour also allows a more accurate representation of the vehicle’s planar shape. Additionally, for vehicles traveling stably in different lanes and gradually approaching each other longitudinally, the model resolves the misjudgment of oblique collision risks common in traditional elliptical-frustum risk fields. Moreover, an adaptive mechanism for adjusting risk sensitivity was introduced. Finally, the study presented application demonstrations of the collision risk field in a human-driving simulator lane-change experiment and in a simulated lane-change scenario for an autonomous vehicle. The proposed model and its application demonstrations can provide a reference for the application of risk field models in human-driving and autonomous-driving scenarios.