Identification techniques for a robotic link

The paper presents the theoretical background and the actual results of both linear and nonlinear identification procedures for the links of an industrial 6-axes articulated robot. The identification results are aimed at developing a position controller for repetitive trajectory displacements by implementing and comparing a linear (ARMAX) and a nonlinear neural network based (NARX) technique. Theoretical aspects and realization of all procedures are presented, that, together with experimental results, state the superiority of the nonlinear approach in most operational points