Robotic stiffness control and calibration as applied to human grasping tasks

In this paper, we study stiffness analysis as applied to human grasping. Grasp stiffness has been demonstrated to be useful for modeling and controlling robotic manipulators. The computation of general linear ℜ ^3x3 stiffness matrices for grasping, which can be decomposed into symmetric (conservative) and antisymmetric (nonconservative) components, offers physical insights for stiffness control in robotics as well as human grasping. Methods of stiffness calibration, using least-squares best fits with and without symmetry constraints, are presented and applied to the force and displacement data obtained from grasping tasks to study human grasping behaviors. The results of this study show that a linear relationship between force and displacement is capable of capturing the characteristics of the experimental data of human grasps for which displacements are small (on the order of one to seven mm). Different measures, proposed and developed in the robotics literature, are employed to predict the behavior of human grasps in reacting to externally applied loads.