Neuro-fuzzy control of rigid and flexible-joint robotic manipulator

Abstract

Robot control can be decomposed into specification, control algorithm, sensory and torque control levels. The paper deals with the control algorithm level of industrial robots. In this paper, a neural implementation of the fuzzy controller has been proposed. A neuro-fuzzy controller (NFC) is used for the trajectory tracking of four degree-of-freedom rigid-link flexible and rigid joint SCARA-type manipulators. Online fine tuning of NFCs is proposed using the backpropagation algorithm. This paper contains a comparison of NFCs before and after fine tuning. A traditional fuzzy controller is compared with an NFC, using a simulation method, with respect to the trajectory tracking control of a SCARA-type manipulator. The results illustrate the usefulness of NFCs.