Robust and optimal control for robotic manipulator based on linear-parameter-neural-networks

Abstract

Stability analysis of neural-network-based nonlinear control has presented great difficulties. For a rigid-body robotic manipulator whose nonlinearities are unknown, we employed linear-parameter-neural-networks to approximate on-line the unknown nonlinearities and then succeeded in designing the control law and the adaptive law of neural network weights. It is shown that the proposed control is continuous, guarantees global stability without knowledge of nonlinear dynamics, and ensures a finite upper bound on the attenuation performance index. That is, the proposed control is both robust and optimal. Simulation results show that the controller we proposed exhibits robustness and excellent tracking performance.