A Shape Description Model by Using Sensor Data From Touch

Abstract

In this paper we consider the problem of recognizing the shape of a 3D object using tactile sensing by a dexterous robot hand. Our approach uses multiple fingers to slide along the surface of the object. From the sensing contact points we extracts a number of 3D points belonging to the surface of the object. The unknown surface Γ of the object is determined by using an “n-ellipsoid” model (Bonnet [1]). The set of parameters that define the surface Γ is determined such that the n-ellipsoid best fits the set of data points.