A Quantification of Machine Dexterity Applied to an Assembly Task

Abstract

This article reviews the need for a quantitative measure of dexterity and considers intuitive notions of dexterity based on the scaling laws of biological systems. Task/manipulator interactions are quantified in the human perspective, and machine dexterity is shown to have parallel measures and limitations. A nondimensional definition of dexterity is pro posed based on the information content of task/effector inter actions. It is shown to provide a quantitative measure be tween different robotic systems that have varying degrees of dexterity and interactions between a specific assembly task being performed. Values for the index of difficulty are shown to vary in the work space, and the loci of maximum dexterity that indicate the most favorable task/effector arrangements are determined. The notion of a set of principal axes of dex terity is defined and located in the task/manipulator work space. The peg-in-hole task is used as an example application of this approach.