Frenet-Artificial Potential Field Method for Vehicle Obstacle Avoidance Path Planning

Artificial potential field (APF) based vehicle obstacle avoidance path planning shows promise in improving traffic safety. However, traditional APF encounters challenges when applied in autonomous driving scenarios due to its inherent limitations and the realistic constraints of traffic dynamics. To solve the problem, this paper proposes the Frenet-artificial potential field (Frenet-APF) method. Firstly, to overcome traditional APF limitations, we introduce a regulatory factor into the repulsive potential field to handle unreachable target points issues, and constructs a heading-adjustable safe ellipse model to resolve local optimal issues. Secondly, for adaptation to multi-lane structured road, we construct a structured road repulsive potential field to meet route constraints. Thirdly, to enhance safety against dynamic obstacles, we design a virtual target point-guided lane change strategy into APF, enabling proactive lane change. Fourthly, for application in large curvature paths, we combine the improved APF with Frenet coordinates to simplify curve into straight scenarios. Finally, simulation results demonstrate that the proposed algorithm successfully generates an efficient, smooth and collision-free path in structured road scenarios with large curvature, multi-lanes and multi-dynamic obstacles.