A pragmatic and systematic statistical analysis for identification of industrial robots

Abstract

Identification of industrial robots is a prolific topic that has been deeply investigated over the last three decades. The standard method is based on the use of the inverse dynamic model and the least-squares estimation (IDIM-LS method) while robots are operating in closed loop by tracking exciting trajectories. Recently, in order to secure the consistency of the parameters estimates, an instrumental variable (IV) approach, called IDIM-IV method, has been designed and experimentally validated. However, the statistical analysis of estimates was not treated. Surprisingly, this topic is rarely addressed in mechatronics whereas it has been deeply investigated in automatic control. This paper aims at bridging the gap between these two communities by presenting a pragmatic statistical analysis of the IDIM-IV estimates. This analysis consists of a two-step procedure: first, the consistency of the IDIM-IV estimates is validated by the Revised Durbin-Wu-Hausman test, and then the statistical analysis of the IDIM-IV residuals is treated. This two-step approach is experimentally validated on the TX40 robot.