Finite-time robust motion control of intelligent vehicles based on neural approximation and disturbance estimation

This paper addresses the inherent trade off between rapid convergence and high‑precision disturbance rejection in intelligent vehicle motion control under severe nonlinearities, parametric uncertainties, and time varying external perturbations. A composite finite‑time robust control strategy is proposed by synergistically integrating a nonsingular fast terminal sliding mode (NFTSM) surface, radial basis function neural networks (RBFNN), and a finite time disturbance observer (FTDO). The NFTSM guarantees singularity free finite time convergence of the tracking errors. The RBFNN approximates the slowly varying unknown dynamics, while the FTDO compensates online for the neural approximation residual and fast varying disturbances in a cascaded "coarse to fine" manner. Lyapunov-based Extensive co‑simulations with MATLAB/Simulink, conducted under double lane change and sinusoidal path scenarios with simultaneous road friction drops and stochastic lateral wind gusts, demonstrate that the proposed framework significantly outperforms conventional nonsingular terminal sliding mode control and disturbance observer based sliding mode control in terms of tracking accuracy, convergence speed, and chattering suppression.