FORCE/ POSITION ADAPTIVE CONTROL FOR ROBOT MANIPULATOR USING RADIAL BASIC FUNCTION NEURAL NETWORK

Abstract

This paper presents an adaptive control method using radial basic funtion
neural network (RBFNN) to control the force/position of robot manipulators under
working environment constraint. Compare to the tranditional sliding controller, the
RBFNN controller has the advantage of being able to learn and approximate
unknown nonlinear functions with arbitrary precision regardless of the various
system parameters while the sliding controller need to accurately calculate the
nonlinear functions so the chatering occurs under the affect of the uncertain system
parameters and disturbance. The adaptive controller using RBFNN will update the
online neural network weights so that the output vectors of neural network are
trained online to approximate uncertainty components of the system. Besides, the
control and adaptive law are calculated base on the use of Lyapunov function. The
simulation results of A465 CRS robotics using Matlab Simulink software guarantee
the accuracy and reliability of the position/force end - effector robot manipulators.