Grasp quality metrics for robot hands benchmarking

Abstract

This paper addresses the problem of evaluating the grasping capabilities of a robot hand. The purpose of such evaluation is to provide numerical results that enable the comparison of different aspects of robot hand designs which can lead to improve their performance. The methodology proposed in this paper consists of computing an exhaustive amount of grasp configurations for each robot hand over a wide variety of objects on a simulation environment. These grasps are evaluated through a set of six selected grasp metrics. Three hand models are compared to test the suitability of the methodology.