Design of a vision-based autonomous path-tracking control system and experimental validation

This paper describes an automated path-tracking system using a vision sensor. A lateral control law for path tracking is especially emphasized which is developed by using back-stepping control design methodology. To establish the proposed control system, the lateral offset to the reference path, the heading angle of the vehicle relative to the tangent line to the path, and the path curvature are required. Those inputs to the controller have been calculated through a Kalman filter which is frequently utilized for the purpose. Lane mark detection has been achieved in an electronic control unit platform with a vision sensor. The yaw rate and side-slip angle also needed in the controller are estimated by the observer. To demonstrate the performances of the proposed controller at different speeds, an experiment has been conducted on a proving ground having a straight section and a curved section with a curvature of about 260 m.