A Neural Network-based Impedance Controller For a Redundantly Actuated Closed-Chain Robot Manipulator

Authors:	Mesbah-Nejad Asghar, The University of Western Ontario, Canada Moallem Mehrdad, The University of Western Ontario, Canada
Topic:	4.3 Robotics
Session:	Robot Control II
Keywords:	Redundant actuation, closed-chain manipulator, neural networks, impedance control.

Abstract

In this paper, a neural network (NN) impedance controller is proposed to control position/force of the tip of a 3 DOF redundantly actuated closed-chain manipulator. The manipulator is in contact with an unknown environment. The structure of the controller is derived using a filtered error approach in which no off-line learning phase is needed. The actuator redundancy is resoled by augmentation of the Jacobian matrix of the manipulator. Simulation results are presented that illustrate strength of the proposed controller in the presence of model uncertainties and external disturbances.