Abstract:

For the problem of weak adaptability to complex curvature changing conditions in the path tracking of intelligent vehicle， a control method based on RBF neural network compensation of model prediction is proposed. Firstly， the three-degree-of-freedom intelligent vehicle dynamics model is used as the prediction model. Then the state transition error model is obtained by analyzing the linear time-varying equations. The adaptive compensation for error by the RBF neural network is realized to ensure the accuracy of the control and improve the path tracking accuracy. Finally， the complex path curvature changing condition including straight line segment， serpentine segment and double-shift line segment is constructed based on the China Smart Car Competition track. The path tracking performance of the control method in high-speed environment is verified on the semi-real vehicle simulation platform. The results show that the maximum trajectory tracking error is within the range of 0.285 m， and the maximum lateral acceleration is 0.329 9 m/s²， which ensures the accuracy and stability of the path tracking.

Key words: intelligent vehicle, path tracking, complex curvature variation, error compensation