Hybrid feedforward-feedback robust adaptive extreme learning control for Euler-Lagrange systems

In this paper, a feedforward-feedback robust adaptive extreme learning control scheme for Eular-Lagrange systems under unknown uncertainties is proposed. System unknown uncertainties can be effectively approximated through the established feedforward extreme learning neural network approxima-tor. Compared with traditional feedback approxiamation, only reference signals are needed as extreme learning approximator inputs rather than reference signals and tracking errors, which means not only the input dimensions can be reduced, but also hidden nodes can be randomly determined, and thereby eventually leading to lower computation consumption so as to facilitate the practical application. Moreover, a H-infinity term is employed to dominate the influence of approximation errors. Simulation studies are provided to demonstrate that the hybrid control scheme is effective.