关键词: 车辆工程； 路径跟踪； 横向控制； 轮胎参数估计； 自适应模糊模型预测控制

Abstract: In the context of intelligent vehicle path tracking, the model predictive control (MPC) method with a fixed prediction time domain often suffers from poor local path tracking accuracy and stability. To tackle these issues, an adaptive fuzzy model predictive control (AFMPC) algorithm was designed, which optimized and adjusted the prediction step size of the controller online according to the longitudinal vehicle speed and the curvature of the reference path. Simulink/CarSim were used to conduct a double lane change test at high, medium, and low speeds. The results show that the AFMPC controller achieves overall better path-tracking accuracy than the fixed-parameter MPC. Compared with the optimal fixed-parameter controller under each operating condition, its lateral average error is significantly reduced by 12.25% in the low-speed scenario, remains nearly unchanged (a 0.37% decrease) in the high-speed scenario, and is comparable to that of the optimal fixed-parameter MPC in the medium-speed scenario. Meanwhile, the vehicle yaw rate and centroid sideslip angle maintain a level comparable to those of the fixed-parameter MPC under all three operating conditions. Therefore, adjusting the prediction horizon in real-time according to changes in road and vehicle conditions is beneficial for improving the driving accuracy and stability of intelligent vehicles.

Key words: vehicle engineering; path tracking; lateral control; tire parameter estimation; adaptive fuzzy model predictive control