Hazard Anticipatory Autonomous Braking Control System Based on 2-D Pedestrian Motion Prediction

Abstract

This paper discusses 2-dimensional (2-D) pedestrian motion prediction and autonomous braking control for enhancing the collision avoidance performance of an active safety system. The paper targets a typical scenario involving a pedestrian walking toward a parked vehicle on a crowded urban road. The pedestrian is not expected to continue walking in a straight line. Conventional first-order motion prediction accuracy alone is not enough to predict the pedestrian motion because prediction is based on the pedestrian’s current position and velocity within a finite time. We formulated a 2-D pedestrian motion model of the parked vehicle based on learning the measured trajectory of pedestrians in the same scenario. We then designed an autonomous braking control system based on whether the vehicle will overtake a pedestrian. We evaluated the validity of the proposed autonomous braking control system in simulation experiments.