Collision avoidance method of autonomous vehicle based on improved artificial potential field algorithm

Both emergency braking and active steering are possible choices for collision avoidance manoeuvres, and any obstacle avoidance strategy aims to design a control algorithm preventing accidents. However, the real-time path needs to consider the motion state of surrounding participants on the road. This work presents a collision avoidance algorithm containing the path-planning and the tracking controller. Firstly, the lateral lane-changing spacing model and the longitudinal braking distance model are presented, describing the vehicle to reactively process dynamic scenarios in real environments. Then, we introduce the safety distance into the artificial potential field algorithm (APF), thereby generating a safe path in a simulated traffic scene. Redesigning the influence range of obstacles based on the collision areas and corresponding safety distance compared with the classic APF. Besides, based on the threat level, the repulsion is divided into the force of the position repulsion and the speed repulsion. The former is related to the relative position and prevents the vehicle from approaching the obstacle. The latter is opposite to the relative speed vector and decelerates the ego vehicle. Simultaneously, the attraction is improved to apply a dynamic environment. Finally, we design a model predictive control (MPC) to track the lateral motion through steering angle and a Fuzzy-PID control to track the longitudinal speed, turning the planned path into an actual trajectory with stable vehicle dynamics. To verify the performance of the proposed method, three cases are simulated to obtain the vehicle responding curves. The simulation results prove that the active collision avoidance algorithm can generate a safe path with comfort and stability.