Autonomous Vehicle Path Tracking with Unreliable Reference Path

Path Tracking Controller (PTC) is a key subsystem of Autonomous Vehicle’s (AV) operation. PTC is responsible for guiding the vehicle on the provided reference path. During practical applications, the reference path may not be reliable or even unavailable in some cases. In this work, we propose a low-level motion control-based solution when a reliable reference path is not generated by the path planner. The proposed methodology relies on the higher-level perception subsystem which provides lane line positions based on onboard sensors such as camera and Lidar. A novel Nonlinear Model Predictive Controller (NMPC) is proposed to facilitate driving the vehicle on its current lane safely. This includes the case when only the position of one lane line is available, and it is not possible to calculate the reference path from two lane lines. From the performance of the proposed controller, it is evident that the vehicle is maintained in the current lane safely. Besides, the controller provides a smooth steering angle even in the absence of a reference yaw angle.