A tactile sensor with automatic learning capability for industrial parts inspection

Abstract

This paper details a prototype tactile sensing system for automated parts inspection. This is a two-dimensional array sensor which is capable of extracting 3-D information from a contact footprint. A Moire fringe method for reading the gratings of the sensing elements is employed. The effect of any local manufacturing defects in grating formulation are eliminated in measurement by the principle of averaging of a number of grating lines. A learning algorithm is devised to generate unique part codes from raw tactile data. The use of this algorithm for automated part identification is also presented.