Robotic Perception of Material

Abstract

In this paper, we develop a conceptual framework in which acts of manipulation are undertaken for the sake of perceiving material. Within this framework, we disambiguate different materials by actively contacting and probing them, and by sensing the resulting forces, displacements, and sounds. We report experimental results from four separate implementations of this framework using a variety of sensory modalities, including force, vision, and audition. For each implementation, we identify sensor-derived measures that are diagnostic of material properties, and use those measures to categorize objects by their material class. Based on the experimental results, we conclude that the issue of shape-in variance is of critical importance for future work.