Human Prehension and Dexterous Robot Hands

Abstract

The human hand is versatile in its interactions with the envi ronment, demonstrating skills that designers of dexterous robot hands would like to emulate. Various postures and features of the human hand combine to provide a great deal of functional ity. In this article, an analysis of human prehensile capability is presented, comparing a symbolic description of prehension to an opposition-space, parameterized framework. By pointing out features of the human hand in prehensile tasks, we hope to offer insights that designers can use for building more versatile robot and prosthetic hands.

On the representation of anthropomorphic robot hands: Shape versus function

“…Since then, this broad view has been further developed into finer categories with the aim of labeling hand-shapes according to their functional abilities. More than 22 such grasp taxonomies are reviewed in [9]. Recently, a comprehensive taxonomy was provided by [13] with the goal of determining the largest set of different grasp types that can be found in the literature.…”

“…Precision and power have been long been accepted as basic functional qualities by a line of researchers motivated to functionally categorize hand shape for varied reasons like prosthetics, anthropology, realistic animation and also robotic grasping and manipulation. This has resulted in a series of increasingly granular grasp taxonomies (see [9], [13] for comprehensive surveys) .…”

“…Thus complex mixtures of power and precision are possible, which in turn can produce a rich set of effects on a grasped object. This raises the question of "how to grasp" which implies determining how to shape the hand into a prehensile posture that enables it to exert the task related forces on the object concerned and also constrains its efficacy as a sensing device able to gather information about the state of the hand-object interaction [9]. …”

Platform portable anthropomorphic grasping with the bielefeld 20-DOF shadow and 9-DOF TUM hand

“…5) of the grasp taxonomy we propose has to be chosen. In contrast to detailed taxonomies, like [12] or [27], all of these grasp types are realizable with most robot hands possessing at least three fingers. With our grasp classification, major hand potentials can be utilized (precision, power, pinch), where we further subdivide precision grasps according to the number of involved fingers (two, all).…”

Robotic Grasping: A Generic Neural Network Architecture

“…In fact, in order to better understand the different parameters involved during grasping, several studies have been conducted to classify human prehension. Their main principle is that prototypical hand shapes are used to perform classes of tasks according to particular hand, environment and task related constraints (Napier, 1956;Iberall et al, 1986;Cutkosky, 1987;Tomovic et al, 1987;Bekey et al, 1993;Iberall, 1997;Saito & Nagata, 1999). These classifications have been intensively used to propose solutions in the robotics field, especially in the frame of artificial intelligence and expert systems (Pook & Ballard, 1993;Kang & Ikeuchi, 1997;).…”