Smooth switching adaptive model reference control of robots using neural networks

A hybrid controller for the tracking of a model reference of robots is presented. It consists of four parts: a neural network (NN) for resembling the unknown nonlinearities of the robot; an adaptive control for compensating the resembled nonlinearities; a high-gain control which takes over temporarily once the former is approaching singularity; last, a robust control to counteract the degradation due to the approximation errors. Such an approach preserves the advantages of adaptive control scheme while avoids running into singularity at the same time by incorporating the temporary high-gain control. Moreover, the switching mechanism is absolutely smooth and hence does not incur any chattering behavior. Simulation results demonstrating the validity of the proposed design are given in the final.