Nonlinear Model Predictive Controller Based on Data and Mechanism Hybrid Model for Autonomous Vehicle Path Tracking Control

In the process of path tracking, autonomous vehicles often encounter modeling errors in the controller’s vehicle model, which can affect the path tracking performance, especially in high-speed turning scenarios. To address this issue, this paper proposes a nonlinear model predictive control based on data-mechanics hybrid predictive model as a control scheme for path tracking. Firstly, vehicle driving data under different operating conditions were collected as the training dataset for the neural network. The network was trained to fit the error between the prediction model and the vehicle model. Secondly, the trained neural network bias compensation model was validated and integrated into the three-degree-of-freedom vehicle model to obtain the neural network-based compensated model predictive controller. Finally, the designed controller was compared with the uncompensated model predictive controller under different operating conditions. The results showed that the path tracking controller with neural network compensation had better tracking accuracy, reducing tracking errors by 47.01% and 57.55% under constant speed and variable speed conditions, respectively. The effectiveness of the designed controller was verified under high-speed, large curvature conditions.