Neural Network Modeling of a Flexible Manipulator Robot

Abstract

This paper presents an artificial neural networks application for a flexible process modeling. A flexible planar single-link manipulator robot is considered. The dynamic behavior of this process is described using Lagrange equations and finite elements method. The artificial neural networks are all variations on the parallel distributed processing (PDP) idea. The architecture of each network is based on very similar building blocks which perform the processing. Therefore, two feed-forward and recurrent neural networks are developed and trained using back-propagation algorithm to identify the dynamics of the flexible process. Simulation results of the system responses are given and discussed in terms of level of error reduction. Finally, a conclusion encloses the paper. © 2012 IFIP International Federation for Information Processing.