Robust Adaptive Compensation for a Class of Nonlinear Systems

Abstract

A robust adaptive design method is proposed for the on-line compensation of uncertainties, for a class of nonlinear systems. As an extension of previous work, the adaptive part of the control law uses a constructive Gaussian network without any prior training, and the control law provides robustness using a systematically designed sliding mode term. In the design, learning and control bounds are guaranteed by properly constructing the control architecture using the proposed methods. The robust adaptive control strategy, with the proposed design guidelines, has been validated using a hardware example case of a nonlinear robotic linkage system. Experiments have shown that the inclusion of the proposed stable learning and robust terms into the control design, using the proposed constructive methods, results in improved system performance for the example case system.