Comparison of Model Predictive Control Algorithms Based on Different Reference Models

Model predictive control (MPC) plays a crucial role in advancing intelligent vehicle technologies. Controllers designed based on various vehicle reference models, including kinematic and dynamic models (both linear and nonlinear), often demonstrate significant differences in control performance. This study contributes by comparing three different MPC control methods and proposing a comprehensive evaluation criterion that considers tracking accuracy, stability, and computational efficiency across various MPC designs. Joint simulations using CarSim and MATLAB/Simulink reveal distinct performance characteristics among the MPC variants. Specifically, kinematic MPC (KMPC) exhibits superior performance at low speeds, linear model predictive control (LMPC) performs best at moderate speeds, and nonlinear MPC (NMPC) achieves optimal performance at high speeds. These findings highlight the adaptive nature of MPC strategies to varying vehicle dynamics and operational conditions, emphasizing the importance of selecting the appropriate MPC design based on the speed regime for maximizing control effectiveness in intelligent vehicles.