NARMAX modeling of a two-link flexible robot

Abstract

Modeling of a two-link flexible robot as a nonlinear autoregressive, moving average with exogenous input (NARMAX) model is considered in this paper. The advantage of using NARMAX model over functional series representation for complex nonlinear systems is well established due to less number of parameters involved in the former. Next, the realtime identification for the NARMAX model parameters is acquired by using the recursive extended least square (RELS) algorithm. Model validation for a two-link flexible robot in real time is established by operating the robot under variable payload conditions. The experimental results show that the proposed NARMAX method provides better identification of the TLFR dynamics compared to ARMAX model.