Adaptive Tracking Control of Nonholonomic Mobile Manipulators Using Recurrent Neural Networks

The trajectory tracking problem is considered for a class of nonholonomic mobile manipulators in the presence of uncertainties and disturbances. First, under the assumption that the kinematic subsystem of mobile manipulator is capable of being transformed into the chained form and the dynamic subsystem of mobile manipulator is exactly known without considering external disturbances, a model-based controller is designed at the torque level using backstepping design technology. However, the model-based control may be inapplicable for practical applications, as the uncertainties and disturbances do exist in the dynamics of mobile manipulators inevitably. Thus, a Recurrent Neural Network (RNN) based control system is developed without requiring explicit knowledge of the system dynamics. The control system comprises a RNN identifier and a compensation controller, in which the RNN is utilized to identify the unknown dynamics on-line, and the compensation controller is presented to compensate the approximation error and external disturbances. The online adaptive laws of the control system are derived in the Lyapunov sense so that the stability of the system can be guaranteed. Finally, simulation results for a wheeled mobile manipulator are provided to show the good tracking performance and robustness of the proposed control method.