Image: Movement
Control Laboratory/University of Washington
(February 18, 2016) There
are two generalized schools of thought when it comes to robot hand design. You
have robot hands that are simple and straightforward and get the job done, like
two- or three-finger grippers that can reliably do many (if not most) things
well without any fuss. And then you have very complex hands with four fingers
and a thumb that are designed to closely mimic human hands, on the theory that
human hands were intelligently designed by millions of years of evolution, and
we’ve designed all of our stuff around them anyway, so if you want your robot
to be able to do as many things as possible as well as possible you want a hand
that’s as humanlike as possible.
Because of the inherent complexity of a real human hand,
biomimetic anthropomorphic hands inevitably involve lots of compromises to get
them to work properly while maintaining a human-ish form factor. Zhe Xu and
Emanuel Todorov from the University of Washington, in Seattle, have gone crazy
and built the most detailed and kinematically accurate biomimetic
anthropomorphic robotic hand that we’ve ever seen, with the ultimate goal of
replacing human hands completely.
Here’s why it was important for them to design a new kind of
robotic hand, according to Xu:
“The conventional
approach to designing anthropomorphic robotic hands often involves mechanizing
biological parts with hinges, linkages, and gimbals in order to simplify the
seemingly complicated human counterparts. This approach is helpful for
understanding and approximating the kinematics of the human hand in general,
but inevitably introduces undesirable discrepancies between the human and
robotic hands since most of those salient biomechanical features of the human
hand are discarded in the mechanizing process. The inherent mismatch between
mechanisms of these robotic hands and biomechanics of human hands essentially
prevents us from using natural hand motion to directly control them. Thus none
of the existing anthropomorphic robotic hands can achieve the human-level
dexterity yet.”
With their hand, Xu and Todorov decided to start from
scratch, duplicating the human hand mechanically in as accurate a way as
possible. First, they laser scanned a human skeleton hand, and then 3D-printed
artificial bones to match, which allowed them to duplicate the unfixed joint
axes that we have, as Xu explains:
“For example, the
motion of our opposable thumb is relying the complicated shape of the trapezium
bone located at the carpometacarpal (CMC) joint. Due to the irregular shape of
the trapezium bone, the exact locations of CMC joint axes are not fixed. Thus,
none of the existing anthropomorphic robotic hands can restore the natural
thumb motions with conventional mechanical joints that require fixed rotation
axes. We 3D print artificial bones from the laser-scanned model of a cadaver
skeleton hand and connect them with artificial finger joints whose range of
motion, stiffness, and dynamic behaviors are very close to their human
counterparts. Our robotic hand design uniquely preserves the important
biomechanical information of the human hand on anatomical level.”
video >>
read also (cs.washington.edu):
Design of a Highly Biomimetic Anthropomorphic Robotic Hand towards Artificial Limb Regeneration >>
Optimal Control with Learned Local Models: Applicationto Dexterous Manipulation >>
Design of an Anthropomorphic Robotic Finger Systemwith Biomimetic Artificial Joints >>
The UW Hand: A Low-cost, 20-DOF Tendon-driven Hand with Fast and Compliant Actuation >>
A Low-cost and Modular, 20-DOF Anthropomorphic Robotic Hand: Design, Actuation and Modeling >>