A Study on the Development of Delta-V Prediction Model for Rear-end Collision Accidents Using Machine Learning

Abstract

With the increasing number of vehicles equipped with ADAS(Advanced Driver Assistance Systems), passenger injury characteristics are changing in the event of a collision. AEBS(Autonomous Emergency Braking System) is the representative ADAS. It is a system that activates the brake to avoid collision, or mitigate impact in a collision risk situation. Recent rear-end collisions tend to be low-speed collisions because collisions are completely unavoidable in all accident situations. Low-speed collisions have a relatively higher risk of causing neck injuries than other types of injuries. The characteristics of neck injuries vary from person to person. Neck injuries are generally known to occur at an effective collision speed of 8 km/h or higher. In this study, actual crash test data were programmed as machine learning techniques to derive effective collision speeds under collision conditions. As a result, we have developed a model that could induce effective collision speeds from vehicle collisions. The developed model can calculate an effective collision speed by taking into account the speed, weight, angle, and offset of the vehicle. Using the developed model, it is possible to estimate the seriousness of a passenger's neck injuries in traffic accidents without using any other analysis program.