RI: Small: Grasping the Inaccessible

Carnegie Mellon University, Pittsburgh PA

Investigators

Abstract

Robots can only grasp "accessible" objects, objects with clearance for the robot's gripper or fingers. Unfortunately, objects are often kept tightly packed, for efficient storage or transport, like the silverware in a drawer. Acquiring such an inaccessible object is called "unpacking". Unpacking requires the shifting of an object, or the surrounding clutter, without a conventional grasp. In current practice, unpacking is almost always performed by humans. This project identifies different classes of unpacking problems, and explores solutions for each class. Unpacking by robots will increase the impact of robotics in applications of strategic and economic importance such as manufacturing and logistics. Unpacking problems also arise in the household and in health care institutions. The project's activities include classification based on the object, and the surrounding clutter. The project proceeds by refining a preliminary classification, identifying representative benchmark problems for each class, and identifying the challenges particular to each class or cutting across several classes. The project will explore novel effector designs, controls, modeling, and planning. One key broadly applicable approach, when access to an object's surface is limited, is to use "partial closure" -- insufficient for a stable grasp, but sufficient to shift an object or the surrounding clutter, increasing accessibility of the object. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.