Injured probability assessment in frontal pedestrian-vehicle collision counting uncertainties in pedestrian movement

Abstract

In probabilistic approaches, where uncertainties in pedestrian motion are counted, the distance model is important for the accurate and robust collision risks evaluation. Focusing on the typical frontal pedestrian-vehicle collision, the conflict distance model with the merit to distinguish front impacts from side ones is presented. Based on stochastic pedestrian model and Unscented Transformation (UT) method, the time-to-collision (TTC) and conflict distance are sampled, and their probability density functions (pdf) as well as the collision probability are deduced. The presented distance model and probabilistic method are verified with Monte Carlo (MC) as the reference, and show the high accuracy and potential for real-time application. Utilizing the estimated collision speed and its probability distribution, a unified model assessing the injured probabilities of pedestrian in front collision is proposed. A pedestrian-vehicle conflict scenario is constructed to evaluate the effectiveness of the probabilistic injury assessment. Simulation results show that the proposed method is sensitive to evasive maneuvers, and provides more useful cues for the optimization of collision avoidance control than the deterministic approaches.