Identification of Flexible Robotic Manipulators Using Neural Networks

Abstract

The control of flexible-link robotic manipulators is a challenging problem because of the high degree of nonlinearity possessed by these systems. Equally challenging is the development of dynamic models and/or identification techniques that would allow real-time processing of all the information needed for controlling these manipulators. This article highlights some of the difficulties and problems that arise when using current approaches to control these systems. In this article, we utilize artificial neural networks as an alternative to perform identification of nonlinear dynamic systems, such as flexible manipulators, in real time. A particular type of neural network, radial basis function neural network, is examined along with the orthogonal least-squares learning technique, which linearizes the parameters of the network. The method is applied to the identification of an experimental Single-link flexible manipulator system and the results are shown and discussed.