Rough Set Based Method for Vehicle Collision Risk Assessment Through Inferring Driver's Braking Actions in Near-Crash Situations

Driving information and data under potential vehicle crashes create opportunities for extensive real-world observations of driver behaviors and relevant factors that significantly influence the driving safety in emergency scenarios. Furthermore, the availability of such data also enhances the collision avoidance systems (CASs) by evaluating driver?s actions in near-crash scenarios and providing timely warnings. These applications motivate the need for heuristic tools capable of interpreting the correlations of driving risk with driver/vehicle characteristics and incidental traffic factors. In this paper, we acquired amount of real-world field data and built a comprehensive "driver-vehicle-road" dataset for actual driver behavior evaluation. The proposed method works in two steps. In the first step, a variable precision rough set (VPRS) based classifier derives a simplified decision rules from field driving dataset, which presents the essential attributes relevant to driving safety. In the second step, we quantify the mutual information entropy of each attribute to evaluate the significance of different factors on happening a vehicle crash, then an accumulation of weighted "driver-vehicle-road" is calculated to achieve an index reflecting the driving safety level. The performance of the proposed method is demonstrated in an offline analysis of the driving data collected from field trials, where the goal is to infer the emergency braking actions in next short term. The results indicate that our proposed model is a good alternative for providing drivers immediate warnings with high prediction accuracy and stability.