Robust tracking control of uncertain dynamic nonholonomic systems using recurrent neural networks

The tracking problem for a class of dynamic nonholonomic systems with uncertainties is considered. First, under the assumption that the dynamics of the nonholonomic systems are exactly known without uncertainties, a simpler model-based controller is proposed by means of cascade design approach, in which the virtual velocity controller is linear, and the actual torque controller is derived by conventional computed-torque law. Then, to deal with uncertainties, a recurrent neural network control system is developed without requiring explicit knowledge of the system dynamics. The closed-loop stability analysis is presented based on a technical lemma developed for nonlinear cascaded systems with vanishing disturbances. Comparing with the existing results, the resulting control system has a simpler structure, and can deal with parametric uncertainties as well as non-parametric uncertainties, yet guarantees asymptotic stability of the tracking error dynamics. Simulation results for a wheeled mobile robot verify the good tracking performance and robustness of the proposed control system.