Boston Dynamics redesigned humanoid hand

19 views
Skip to first unread message

camp .

unread,
Oct 10, 2026, 2:00:20 AM (yesterday) Oct 10
to HomeBrew Robotics Club

Chris Albertson

unread,
Oct 10, 2026, 3:35:56 PM (13 hours ago) Oct 10
to hbrob...@googlegroups.com
I like that BD had employees do a physical experiment by taping two fingers together for a full day.   This is basically observing nature.  I liked to look at my dog.  I ask, How does she turn? Is the rotation axis through the front legs (shoulders) or the rear legs?

I looked at grasping too.   When do I decide to use more pressure vs. when do I decide to lift from the bottom?  Either method can avoid a slip.   There must be some kind of rule.   Taping the pinky finger was a great way to know if we really use it for daily tasks. 

This isn’t real science, just looking at nature for some clues about how to make our robot's motion better.

Charles Nolan

unread,
Oct 10, 2026, 3:44:13 PM (13 hours ago) Oct 10
to hbrob...@googlegroups.com
G'Day Chris,

Thank you for sharing.

Regards,
Chuck

--
You received this message because you are subscribed to the Google Groups "HomeBrew Robotics Club" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hbrobotics+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/hbrobotics/53D27303-DB1C-4C45-ADFA-D49A12FA9399%40gmail.com.

Karim Virani

unread,
Oct 10, 2026, 11:23:49 PM (5 hours ago) Oct 10
to hbrob...@googlegroups.com
I like the approach BD is taking. but we should always be cautious about expectations. the industry is just now approaching the controllable degrees of freedom we see in the human hand (tho not really if you include the curl of the palm that is almost never attempted), but we are nowhere near approaching the sophisticated sensing in biology. i love this slide that shows the problem from a sensing perspective:

image.png


--
You received this message because you are subscribed to the Google Groups "HomeBrew Robotics Club" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hbrobotics+...@googlegroups.com.

Chris Albertson

unread,
2:38 AM (2 hours ago) 2:38 AM
to hbrob...@googlegroups.com
Just recently in the last couple of days I found a cheap way to sense complex pressures on a gripper.

A pressure sensor with 100 “pixels” per finger is expensive, but Amazon just delivered a suitable camera, cost about $16. 

The robot can self-train to learn how much deformation is needed to grip different kinds of objects.

Here is a not-to-bad video that I did not make, that shows how a "fin-ray" deformable finger can make grip force visible in a camera.   It will be better when it is taken from a wrist camera.   Yes, there are better hands, but my goal is only to get the training environment to work.  The goal is to develop a process.   BTW, is you need a good rob-arm to suport software dev google “SO-101"

Reply all
Reply to author
Forward
0 new messages