An automated instrumental variable method for rigid industrial robot identification

Abstract

Industrial robots must be operated in closed-loop since they are electro-mechanical systems with double integrator behaviour. Their mechanical model, called the Inverse Dynamic Identification Model (IDIM), is based on Newton’s laws and has the advantage of being linear with respect to the parameters. The Instrumental Variable (IDIM-IV) method provides a robust solution to the closed-loop estimation problem. This method relies on a tailor-made prefiltering process in order to estimate accurate parameters. An alternative and automatic way of constructing the observation matrix has been recently introduced. If this methodology provides appropriate estimated parameters, it can fail to estimate the variances of those parameters. In this paper, an identification of the additive noise characteristics is included in the process to obtain correct and lower variances of the IDIM parameters. The evaluation of the new estimation algorithm on a one degree-of-freedom rigid robot shows that it improves statistical efficiency, while minimizing the a priori knowledge required from the practitioner.