In this paper, we study the identification of industrial robot dynamic models. Since the models are linear with respect to the parameters, the usual identification technique is based on the Least-Squares method. That requires a careful preprocessing of the data to obtain consistent estimates. In this paper, we carefully detail this process and propose a new procedure based on Kalman filtering and fixed interval smoothing. This new technique is compared to usual one with experimental data considering an industrial robot arm. The obtained results show that the proposed technique is a credible alternative, especially if the system bandwidth is unknown.
展开▼
