Task dependent human-like grasping

Abstract

In this study we report our preliminary work on investigating task dependent grasping in humans and realization of this skill on a robotic platform. For human data acquisition, subjects were asked to reach and grasp two objects (a claw-hammer and a pen) positioned in four different orientations. The subjects were primed by specification of the task to be performed with the objects after the grasp. For the hammer, the task could be one of hammering or prying a nail, or transporting the hammer; for the pen the task could be transporting or using the pen for drawing a circle. The human movements were recorded using a motion capture system and analyzed off-line. The results indicate that humans deploy different grasps for objects depending on the task specification, even though the orientation and location of the objects are kept the same. The human data was also used to derive robot trajectories that were used to implement task depended grasping on a 7-DOF robotic arm (Mitsubishi, PA-10) and a 16-DOF robotic hand (Gifu Hand, Dainichi Co. Ltd.).