An Inversion-Based Neuro-Controller for Robot Manipulators

Neuro-controllers that are applicable to a wider range of dynamical systems are the focus of this paper. The proposed controllers are inversion-based wherein the inverse of the dynamical system is modelled using a feedforward neural network with a single hidden layer and trained with the online sequential learning algorithm. We demonstrate that the proposed controller is applicable to both single- and twolink robot manipulators, and is able to handle flexible joints and unmodelled dynamics.