Robust Predefined-Time Zeroing Neural Network for Trajectory Tracking of 4WS Mobile Robots

Four-wheel steering (4WS) mobile robots possess enhanced maneuverability but involve complex kinematic modeling and control due to their coupled multi-degree-of-freedom dynamics. To reduce complexity, this article introduces an equivalence relation at the kinematic level to transform the 4WS model into a two-wheel steering (2WS) model, simplifying the model without imposing additional dynamics. Furthermore, to overcome the limitations of conventional controllers in tracking speed and robustness, we propose a robust predefined-time zeroing neural network (RPTZNN) controller. The predefined-time mechanism enables the upper bound of convergence time to be explicitly assigned in advance according to performance requirements, thereby guaranteeing time-constrained tracking performance. Specifically, two novel activation functions and corresponding convergence parameters are constructed to derive explicit predefined-time design formulas. Subsequently, a cascade-based control framework is developed to regulate the position and orientation of the robot, forming the RPTZNN controller. Afterward, theoretical analysis proves that the upper bound of convergence time depends solely on the predefined-time parameters and is independent of initial conditions, while robustness against bounded disturbances is preserved. Finally, comparative simulations validate that the proposed method achieves faster and more robust trajectory tracking than the conventional controllers.