Structured road-oriented motion planning and tracking framework for active collision avoidance of autonomous vehicles

This paper proposes a novel motion planning and tracking framework based on improved artificial potential fields (APFs) and a lane change strategy to enhance the performance of the active collision avoidance systems of autonomous vehicles on structured roads. First, an improved APF-based hazard evaluation module, which is inspired by discrete optimization, is established to describe driving hazards in the Frenet-Serret coordinate. Next, a strategy for changing lane is developed in accordance with the characteristics of the gradient descent method (GDM). On the basis of the potential energy distribution of the target obstacle and road boundaries, GDM is utilized to generate the path for changing lane. In consideration of the safety threats of traffic participants, the effects of other obstacles on safety are taken as additional safety constraints when the lane-changing speed profile for ego vehicles is designed. Then, after being mapped into the Cartesian coordinate, the feasible trajectory is sent to the tracking layer, where a proportional-integral control and model predictive control (PI-MPC) based coordinated controller is applied. Lastly, several cases composed of different road geometrics and obstacles are tested to validate the effectiveness of the proposed algorithm. Results illustrate that the proposed algorithm can achieve active collision avoidance in complex traffic scenarios.