Direct-Drive Hands: Making Robot Hands Transparent and Reactive to Contacts

Abstract

Industrial manipulators and end-effectors are a vital driver of the automation revolution. These robot hands, designed to reject disturbances with stiffness and strength, are inferior to their human counterparts. Human hands are dexterous and nimble effectors capable of a variety of interactions with the environment.

Through this thesis we wish to answer a question: How can we make robot grippers better at interacting with the environment? A key idea essential for supporting a wide range of interactions with the environment is transparency: the efficient transfer of force and motion between the robot and the task.
In search for an answer, we explore the key developments that allow robots to be reactive to contacts. We analyze the space of actuators to determine parameters that affect transparency and define a metric to characterize transparency by the collision reflex response: the impulse transferred to a rigid, fixed object in a collision.

We show that direct-drive is a favorable transmission choice for improving transparency.
Recent advances in motor technology have triggered a resurgence in direct-drive and quasi-direct-drive systems. The MIT Cheetah, Ghost Robotics' Minitaur, and Agility Robotics' Cassie are all examples of successful legged robots that have improved transparency without compromising their torque specifications.

We have developed the DDhand, a transparent direct-drive gripper inspired by the direct-drive robots from the 80's. The DDHand is a 4 degree of freedom, two fingered gripper with a parallel five bar linkage to keep link reflected inertias low. The gripper improves on speed, bandwidth and mechanical simplicity over other comparable grippers.

Transparency in robotic grippers allows for performance gains in contact-rich and dynamic behaviors and new approaches in intrinsic contact sensing. We present the smack-and-snatch grasp: a contact-informed grasping behavior that can grasp an object on a table of unknown height in a rapid arm motion. We show a velocity-based contact localization algorithm that can localize contact locations using only position and velocity sensing. Finally, we demonstrate the viability of the DDHand in industrial applications.