Motion Planning Based on Steering Obstacle Avoidance Under Emergency Conditions

Abstract

The interpolating curves of the discrete optimization-based path planner are too conservative, which will lead to the failure to plan a safe collision avoidance path under emergency conditions. Hence, a novel segmented path planner for emergency obstacle avoidance conditions is designed. In the first stage of path planning, the smoothest and obstacle avoidance path is obtained taking advantage of the automatic parking path planning. The second stage finds a suitable path to stabilize the vehicle. However, different from the automatic parking system running at low speed, the vehicle obstacle avoidance operation in emergency conditions often makes the vehicle approach its limit. In the past, the conservative vehicle stability criteria would cause the vehicle to fail to find the safest trajectory to complete collision avoidance. A vehicle motion stability classifier is modeled by an SVM model and assembled in the dynamic programming to judge the driving safety of the vehicle according to the inputs of longitudinal acceleration, speed, and road curvature. The simulation results show that the proposed collision avoidance motion planning method has a safer trajectory in the same longitudinal distance.