Optimal Preview Distance Control Using Model Prediction for Autonomous Vehicle

The performance of the autonomous driving control system is the key to the stability and safety of autonomous vehicles. As part of the control system, path tracking is responsible for tracking the desired trajectory. This article proposes a multi-point preview method based on the model prediction, which uses the preview distance obtained by the Stanley path tracking method to interpolate back and forth to obtain candidate preview points, which are converted into steering angles through geometric relations. Finally, selecting the optimal steering angle is selected through the designed optimal function. Different from the traditional kinematics model, this paper also predicts the steering angle at the next moment to achieve more accurate vehicle state prediction. In this way, the proposed method is well compared and analysed with the traditional Stanley controller scheme. Comparing with the proposed method, the traditional Stanley controller is not capable of keeping the steering stable as it faces the road with changeable curvature. The results show that the implementation of model predictive multi-point preview control strategy for Stanley tracking effectively improves the stability and accuracy of tracking.