Dynamic model identification of 2-axes PAM robot arm using neural MIMO NARX model

Abstract

In this paper, a novel Forward Dynamic MIMO Neural NARX model is used for simultaneously modeling and identifying both joints of the 2-axes PAM robot arm's forward dynamic model. The contact force variation and highly nonlinear cross effect of both links of the 2-axes PAM robot arm are modeled thoroughly through a Forward Neural MIMO NARX Model-based identification process using experiment input-output training data. The results show that the novel Forward Dynamic Neural MIMO NARX Model trained by Back Propagation learning algorithm yields outstanding performance and perfect accuracy. © 2009 Springer Berlin Heidelberg.