Comments on: Dexter: Dynamically balancing biped

By: Jeff Kramer

Jeff Kramer — Mon, 05 Mar 2007 05:52:50 +0000

I’m impressed by the compliance in the joints as well though – they make a good point about the required frame stiffness in the other walkers.

I wonder how the different walkers perform on rougher terrain and with loads?

By: Mr Robotics

Mr Robotics — Thu, 01 Mar 2007 19:39:29 +0000

Asimo moves back, so does qrio. And ZMP won’t get you over a teeter-totter when you cross the fulcrum – which is why many humans screw it up.

By: Head Rotor

Head Rotor — Thu, 01 Mar 2007 08:11:24 +0000

OK. I stand (cough cough) corrected.

But isn’t there a quantitative difference between Zero Moment Point (Asimo, et al.) and Dexter? When you shove Asimo, does he actually move his feet to compensate, or just kind of push back from a static stance? I get the feeling that the “dynamic” behaviors of Qrio, etc. are kind of just add-ons to the basic ZMP — where you keep the center of gravity above the feet at all times. Or maybe it’s just a matter of how you define “dynamic.”

By: Mr Robotics

Mr Robotics — Thu, 01 Mar 2007 01:34:54 +0000

Trevor’s robots are very cool!

But actually, a great many robots walk dynamically – including the referenced Asimo. The original (~4 years ago) was choreographed movement, but I saw the latest version at CES in Jan 2007, and they did a demo of it’s dynamic gait – both push back and running/jogging on uneven surfaces (which gets to the point where both feet at some points are off the ground at the same time.)

Jin Sato did the same thing about 2 years ago with off-the-shelf gyros on his 18″ Pirkus (one of which trundles around my office.)

Sony perfected dynamic gaits with the Qrio (nee SDR-4) about 3 years ago, which was really impressive to watch as it went up and down teeter-totters and over pebbles.

Team Osaka takes it further, with 9 DOF sensing (X,Y,Z;R,P,Y;V,S,A) and constant joint positional feedback.

Which isn’t to say what Trevor did isn’t impressive, but there are a great many humanoids with dynamic gait – so the “first” claims are a bit off.

-Mister Robotics, who spends WAY too much time studying gait theory, when he should be working. Wait… that is work. Woo hoo!