Picking up soft 3D objects with two fingers

Abstract

This paper describes a simple strategy for a robot hand to grasp and lift a deformable 3D object resting on a table. Inspired by the human hand grasping, the strategy employs two rigid hemispherical fingers to first squeeze the object until a secure grip is achieved under contact friction, and then translate vertically upward. During the squeeze, a lift test is repeatedly conducted to determine if the maximum hypothetical liftable weight of the object reaches its real weight. Actual lifting follows when the test is passed. The object's deformation and configuration of contact with the fingers and the plane are tracked using the finite element method (FEM) in an event-driven manner, based on varying contact displacements induced by the finger movements. The gravitational force acting on the object is accounted for. Slips inside the contact regions are determined from solving quadratic systems. Experiment has been conducted to demonstrate the efficiency and accuracy for this sensorless grasping approach.