Steering Collision Avoidance Control of Intelligent Vehicles for Crossing Pedestrians at Unsignalized Intersections

An active steering collision avoidance control strategy is proposed to cope with pedestrian-vehicle collision avoidance of intelligent vehicles at unsignalized intersections. Based on the multi-layer model predictive control method, the local planning layer controller and the global tracking layer controller are designed by the hierarchical control strategy. On this basis, the remaining time of pedestrian-vehicle collision is calculated according to the trajectory characteristics of vehicles and pedestrians at intersections, and the modified traditional artificial potential field method is employed to construct the collision avoidance function. The local collision avoidance path is planned to minimize the deviation of tracking the global reference path and avoid any pedestrian-vehicle collision at intersections. By CarSim/Simulink co-simulation platform, various scenarios are designed for simulation analysis. The factors that have significant effects on pedestrian-vehicle collision at intersections are selected from the traffic accident database of Guangdong Province from 2006 to 2018. The results show that intelligent vehicles can track the reference path at different initial points, and the controllers have good robustness subjected to different speeds and adhesions. In a high-speed and low-adhesion scenario, the lateral acceleration of intelligent vehicles is less than 0.4 g; the side-slip angle is less than 2°; the front wheel yaw angle is less than 2.5°. At the four typical unsignalized intersections, intelligent vehicles at different speeds can steer to avoid collision for crossing pedestrians when going straight or turning through the intersection.