An in vivo experiment to assess the validity of the log linearized hunt-crossley model for contacts of robots with the human abdomen

Abstract

A key issue in Human-Robot physical interaction is the real-time perception of contact impedance by the robot. The Hunt-Crossley (HC) model is a popular model of contact force with soft biological tissues as it enjoys accuracy with low-complexity properties and its parameters are physically sound. Because the original HC model is non-linear, the current best known approach of real-time identification consists in identifying the parameters of a log linearized version of the HC model, by means of a Recursive Least Squares (RLS) algorithm. But, the final model used for exploitation in robot control, is the original non-linear HC model with the previously identified parameters. Hence, this approach may be questionable concerning the modeling accuracy and some authors prefer rejecting the HC model. This paper presents for the first time an in vivo experiment to assess the performances of contact models with the human abdomen. In particular we show here through a statistical analysis, that the log linearized HC model should be considered as a contact model on its own and replace the original non-linear HC model for both identification and exploitation.