Non-parametric modeling of double-link flexible robot manipulator based on NNARX model structure

The double-link flexible robot manipulator (DLFR) is a highly non-linear system. The development of existing linear models involves a lot of assumptions and approximations in order to reduce the complex calculation. Due to that, this paper presents a non-linear Auto Regressive model with eXogenous inputs (NARX) for DLFR and neural network is used to predict the non-linear segment. Thus, NNARX structure is used to develop four separate model of double link flexible manipulator from torque input to hub angle and from torque input to end point accelerations of each link. The performance of NNARX models are validated via One Step Ahead, Mean Squared Error and correlation tests.