Physics-Guided Neural Disturbance Observer Based Saturated Barrier-Adaptive Sliding Mode Control for 4WIS/4WID Vehicle Trajectory Tracking

This article proposes a disturbance-aware kinematic control framework specifically designed for four-wheel independent steering and driving (4WIS/4WID) mobile platforms. A physically consistent kinematic error model is established to accurately capture the steering geometry and enforce no-slip constraints. To cope with modeling uncertainties and time-varying disturbances, a hybrid disturbance observer is developed that integrates a low-noise physical proxy with an adaptively updated random-feature neural residual. Furthermore, a saturated barrier-adaptive sliding mode (satBASM) controller is developed to keep the control effort bounded and drive the pose errors ultimately into a prescribed region under actuator saturation. By integrating the observer with the controller, a disturbance-compensated velocity command law is obtained, enhancing the closed-loop robustness against nonlinear coupling effects and external perturbations. Hardware experiments conducted on a full-scale 4WIS/4WID platform demonstrate that the proposed method provides superior disturbance rejection and significantly improves trajectory-tracking accuracy over existing kinematic control schemes.