• An Oriented Bounding Box based method is developed to detect potential collision boundaries and predict omnidirectional collision risk for vehicles. • A novel concept of vehicle collision avoidance extreme decision point is put forward to quantify the limit time of vehicle emergency collision avoidance. • A unified vehicle collision risk assessment and early warning framework is constructed to evaluate real-time in-transit relative risk and output graded warning signals. Accidents involving large vehicles on highways can lead to severe outcomes. This paper presents an integrated framework for real-time collision prediction and warning, specifically designed for large vehicles. The proposed method combines a vehicle approach collision prediction (VACP) model based on an oriented bounding box, enabling accurate estimation of minimum distance, time-to-collision, and collision location. Building on driver reaction capacity, the concept of collision avoidance extreme decision points (CEDP) is defined to determine critical thresholds for turning or maintaining lanes under varying speeds and road friction conditions. These components are unified within a collision risk assessment and warning (CRAW) framework that evaluates surrounding traffic risk and provides graded, proactive alerts. The effectiveness of CRAW was validated through simulations and empirical data analysis. Simulation scenarios quantified safe limits under different speeds and road friction coefficients, revealing critical thresholds of CEDP. The multi-vehicle scenario further confirmed the framework’s capacity to capture dynamic risk evolution and provide timely warnings. In addition, analysis of the HighD dataset, comprising 1,882 large-vehicle trajectories, verified consistency with real-world traffic dynamics. Critical cases such as rapid lane changes and overtaking maneuvers were accurately identified, with the system escalating warnings appropriately. Overall, the framework provides effective technical support for proactive safety management and prevention of severe accidents in complex multi-vehicle environments.
Li et al. (Mon,) studied this question.