Robust Adaptive Neural Control for a Robotic System with Performance Degradation of Actuators

It is useful and effective that actuator torque coefficients applied at each joint are bounded nonlinear time-varying, rather than normally unity all the time. The bounded nonlinear time-varying actuator torque coefficients as well as uncertainties may cause performance degradation of the robot or the actuators in a robotic system. This work presents a design methodology of a robust adaptive neural controller to overcome the performance degradation for an uncertain robotic system with actuator failures. The proposed control scheme uses a GFN (Gaussian function network) approximating a nonlinear dynamic terms. An adaptive control input is added to overcome not only the uncertainties such as the approximation error, parameter variations and external disturbances, but also actuator failures. The proposed controller can improve performance degradation and achieve task completion despite actuator failures and uncertainties. Simulation results are shown to verify the validity of the proposed control scheme