Peter K. Allen
Department of Computer Science
Columbia University

"Data Driven Grasping Using the Bajcsy Principles:
Human Modeling, Active Sensing, and Lots of Results"

To grasp a novel object, we can index into a database of known 3D models and use precomputed grasp data for those models to suggest a new grasp. We refer to this idea as data-driven grasping.  It begins with our construction of a database consisting of several hands, thousands of objects, and over 230,000 unique grasps of objects.  We then describe a new grasp planner that requires only partial 3D data of an object, such as that acquired by range scans, to find a correct grasp. To achieve this, we introduce a new shape descriptor for partial 3D range data, along with an alignment method that can rigidly register partial 3D models to models that are globally similar but not identical. Our method uses SIFT features of depth images, and encapsulates nearby views of an object in a compact shape descriptor.  More importantly, this planner is based upon the  Bajcsy Principles: Human Modeling, Active Sensing, and lots of results

Wednesday, April 22nd, 2009
8:30 am - 5:30 pm

School of Engineering and Applied Science

University of Pennsylvania

210 South 33rd Street

Berger Auditorium

Skirkanich Hall

GRASP Lab - Levine Hall 4th Floor

For more information regarding our speaker please visit:

http://www1.cs.columbia.edu/~allen/