PROJECT TITLE :

Benchmarking in Manipulation Research: Using the Yale-CMU-Berkeley Object and Model Set

ABSTRACT:

In this article, we present the Yale-Carnegie Mellon University (CMU)-Berkeley (YCB) object and model set, intended to be used to facilitate benchmarking in robotic manipulation analysis. The objects in the set are designed to hide a wide selection of aspects of the manipulation downside. The set includes objects of way of life with completely different shapes, sizes, textures, weights, and rigidities also some widely used manipulation tests. The associated database provides high-resolution red, inexperienced, blue, plus depth (RGB-D) scans, physical properties, and geometric models of the objects for straightforward incorporation into manipulation and coming up with software platforms. Besides describing the objects and models in the set along with how they were chosen and derived, we give a framework and a number of example task protocols, laying out how the set can be used to quantitatively evaluate a range of manipulation approaches, including designing, learning, mechanical style, management, and several others. A comprehensive literature survey on the existing benchmarks and object data sets is additionally presented, and their scope and limitations are discussed. The YCB set can be freely distributed to analysis groups worldwide at a series of tutorials at robotics conferences. Subsequent sets will be, otherwise, accessible to purchase at a cheap cost. It's our hope that the prepared availability of this set along with the bottom laid in terms of protocol templates can enable the community of manipulation researchers to additional simply compare approaches and continually evolve standardized benchmarking tests and metrics as the field matures.