This paper focuses on the lane-changing trajectory planning (LTP) process in the automatic driving environments. Existing studies about LTP algorithms are primarily proposed on the straight roads. However, in real-world traffic, the curvature of different road sections is different, and the lane-changing vehicle needs to adjust its position and velocity according to the plane geometric alignment of the roads in real-time. To address such limitations, this paper describes a novel LTP model that can be applied on the straight and curved roads simultaneously. The model employs the regression functions to represent the road information and then the information can be directly applied to the generation of the lane-changing trajectory. The time-dependent polynomial trajectory curve is adopted by the LTP model to avoid the discontinuity of the velocities and accelerations. Moreover, collision detection is mapped into a parameter space by adopting infinite dynamic circles. Finally, the numerical simulations under a variety of lane-changing scenarios are conducted. The results demonstrate that the lane-changing trajectory planned by proposed model not only can be tracked by automated vehicles safely, but also satisfies the requirements of vehicle kinematics and comfort control.

CICTP