Distinguishability and identifiability testing of contact state models

Abstract

An important component of compliant motion control is the estimation of contact states during task execution. This paper addresses two fundamental questions that must be answered when formulating mathematical descriptions of contact states: are the contact states distinguishable from each other? and can the unknown or imprecisely known parameters in these descriptions be identified? An analytical method is presented for evaluating the distinguishability and identifiability questions of a set of contact state models described by non-linear algebraic equations. In contrast to existing, on-line numerical techniques, this approach can be used during the design phase of a contact state estimator to select contact models and robot sensors to ensure feasibility of the estimator. The approach is illustrated using contacts between polygonal and polyhedral objects.