Path planning refers to the process of generating a collision-free traveling path within a specified time and space range, which is used to guide an autonomous vehicle. However, common path planning methods seldom consider the uncertainty of environmental information and struggle to accurately describe dynamic obstacles like moving vehicles. Therefore, this paper proposes a path planning method based on spline network roads and spatio-temporal legacy risk fields. First, in the Frenet coordinate system, a cubic spline function is used to network the road, aiming to generate obstacle avoidance paths with high smoothness and continuous curvature. Meanwhile, the spatio-temporal legacy risk field is established based on dynamic environment information. This involves considering the positional changes of obstacles over a forward-looking timeframe and describing their spatial legacy risk level based on the temporal factor, in order to enhance the ability to avoid dynamic obstacles. Also, the problem of unreasonable return of the controlled vehicle to the main lane during continuous obstacle avoidance is solved by improving the gravitational potential of the main lane. Finally, a trajectory tracking controller based on MPC and PID is designed and validated through hardware-in-the-loop simulation. The findings demonstrate that the proposed method can generate collision-free and smooth paths in dynamic environments.
Jin et al. (Mon,) studied this question.