Stable visual servoing with neural network compensation

Abstract

We propose a stable 2D visual servoing algorithm for planar robot manipulators. We assume that gravity and friction are unknown and that there exists modeling errors in the vision system. By using a radial basis function neural network, it is shown that these uncertainties can be compensated. We prove that without or with unmodeled dynamics, the 2D visual servoing with neural networks compensation is Lyapunov stable.