An Architecture for Predicting Two-dimensional Traffic Collision Considering Shape

In this decade, the improvement of vehicular network gave impetus to vehicle collision warning systems (CWS), which serve to curb the wounded and fatalities in crashes. Many position-based CWS and corresponding collision avoidance algorithms were developed but in the scenarios of two-dimension collision and multiple entities collision, shape of all participants is usually ignored. In this paper, shape of all manner of participant are subdivided to lots of squares based on their profile and filled in gridded road (we call this process gridding) and exploited to detect collision. In the simulated intersection collision scenario, the proposed CWS which we integrate the gridding method shows more accuracy in spacing between participants (affected by the size of cells in gridding road), and we find that the overlapped cells - which depict the coincidence of the predicted trajectories of participants - show closely relation between collision probability evaluation in simulation result. The results indicate that gridding method can served as other metric of collision prediction and is validated to measure and predict collision more precisely.