In traditional vehicle trajectory tracking processes, the dynamic model of the vehicle may not accurately represent complex and nonlinear vehicle behaviors. Moreover, conventional control methods may perform poorly when dealing with system uncertainties and disturbances, facing challenges in real-time computation. To address these issues, this paper proposes an autonomous driving control method based on control-affine feedforward neural network (CAFNN) and fast tube model predictive control (tube-MPC). This method utilizes CAFNN for system dynamic identification, replacing traditional mathematical modeling with data-driven neural network pattern recognition to more accurately describe the vehicle’s nonlinear dynamic characteristics. On this basis, the proposed tube-MPC structure is divided into two parts: nominal MPC and sliding mode control (SMC). The nominal MPC controller associates the MPC problem with a linear complementarity problem (LCP) using a ramp function, enabling rapid computation of the quadratic programming (QP) solution through piecewise affine (PWA) functions; the auxiliary SMC controller employs multi-power sliding mode reaching laws to enhance the system’s robustness against external disturbances and model uncertainties. This control strategy demonstrates high accuracy and stability in vehicle trajectory tracking under complex road conditions, providing strong support for the advancement of autonomous driving technology.
Dai et al. (Sat,) studied this question.