Neural network servo control of a robot manipulator joint in real-time

Abstract

Empirical results are presented for a real-time experiment where a feedforward neural network is used to replace an existing proportional-derivative (P-D) servo controller and trained online in real-time using a performance measure to control positioning of a robot manipulator joint. The optimized neural network controller was tested online as both a fixed and adaptive controller, with step and ramp input commands, under different joint loading conditions. The results are compared to the existing P-D controller.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>