A neural-network compensator with fuzzy robustification terms for improved design of adaptive control of robot manipulators

A sum radial-basis-function neural-network (NN) compensator with computed-torque control and novel weight-tuning algorithms is proposed to improve tracking performance and to account for structured/unstructured uncertainties of robot manipulators. The proposed weight-tuning algorithms do not require the initial NN weights to be small. The bounds of NN weights are guaranteed to be convergent in the sense of Lyapunov. The effectiveness of the proposed algorithm is demonstrated using a laboratory robot manipulator.