Given the limited adaptability of conventional single-model model predictive
control (MPC) to the complex driving conditions encountered by intelligent
vehicles on varying road surfaces, as well as the trade-off between modeling
accuracy and computational efficiency, this study proposes an adaptive
multi-model-switching MPC strategy for trajectory tracking control. First, three
prediction models with different levels of complexity are developed, including a
kinematic model, a linear two-degree-of-freedom (2-DOF) dynamic model, and a
nonlinear tire dynamic model. A model-switching mechanism based on vehicle
speed, road curvature, and road-surface friction is then designed to select the
appropriate prediction model according to the prevailing driving conditions. In
addition, a lateral-deviation feedback correction strategy is introduced to
further improve tracking accuracy and smooth the steering response. Finally,
joint simulations using MATLAB/Simulink and CarSim are conducted under four
representative driving scenarios. The results show that, compared with
conventional single-model MPC, the proposed strategy effectively reduces lateral
tracking errors and steering-angle fluctuations while maintaining favorable
computational efficiency and vehicle stability. These results demonstrate the
potential of the proposed method to achieve accurate and robust trajectory
tracking for intelligent vehicles under complex and varying operating
conditions.