Detecting Vehicle – Two Wheeler Collisions Based on Extreme Value Theory and Single Vehicle Data

With the growth of vehicle ownership, road traffic crashes are gradually increasing. Among them, the conflict between cars and two-wheelers cannot be ignored. The vehicle-road cooperation technology provides new solution for traffic safety. Therefore, this paper builds the typical intersection crash scenarios, collects the risk avoidance behavior data of drivers’ conditions through driving simulation, and establishes a collision detecting model based on extreme value theory and surrogate safety indicator. The results show that using deceleration rate as an alternative safety indicator and taking the threshold T=6.5m/s3 can effectively detect collision events. Meanwhile, this paper analyses the detection results and warning strategy, warning time was found to be too early (4.0s) or too late (2.5s) to cause collision easier, and visual warning was more likely to collision than audio-visual warning. Research provides new ideas for collision detection and vehicle warning in the context of vehicle-road cooperation.