Effects of multi-joint stiffness on the stable, Cartesian workspace for dexterous manipulation with the anatomically correct testbed hand

Human hands perform amazingly complex dexterous in-hand manipulation tasks stably through modulation of musculotendon stiffness, setting a high standard for the capabilities of robotic manipulators. Applying passive stiffness boundary analysis to quantify human-like stiffness modulation properties on a human-like physical simulation platform, the Anatomically Correct Testbed (ACT) hand, has the potential to further our understanding of the human stiffness parameter modulation that enables this manipulation performance. In this work, a model of human-like index finger endtip stiffness is developed to quantify the passively stable region within a 3D workspace for stable manipulation. The stiffness model shows that biomechanical properties vary throughout the workspace with the greatest stiffness volume at proximal postures enabling efficient opposition of the anatomical thumb for precision grasping tasks. Using biomechanically informed control principles on the ACT hand index finger increased the stable manipulation region and trajectory tracking performance. Finally, the effect of external forces on the stable region was characterized as an inverse linear relationship to the magnitude of the applied force and the direction of the applied force highlighted the observed biomechanical constraints which make the positive ulnar axis ideal for opposing the thumb. The resulting method has the potential for general optimization of tendon-driven system stability on a task basis in 3D space, bringing robotic manipulators closer to human levels of dexterity.