6 ms·

They should really include a video, because without it this article isn't particularly helpful. https://youtu.be/wwH40rYJt9g https://youtu.be/wwH40rYJt9g

sethammons 5y ago
that is really, really cool. Leveraging spring-like "tendons" to do the leg lifting and resulting in 1/4 the power usage of "normal" robot legs modeled after humans. Great visualizations.
- Animats 5y ago
  That's been done before. It's just new for Pratt, who previously used series elastic actuators.[1] Those don't do energy recovery; they fake it using a jackscrew and a stiff spring. Running humans have about 70% elastic energy recovery. Cheetahs have around 90%. What you want in a leg muscle is a spring whose spring constant, damping constant, and neutral point you can control. That can be done pneumatically (you need compressed air), hydraulically (you need hydraulic accumulators), mechanically (two springs operated by separate actuators), or electrically (a fast back-driveable motor drive that generates power during compression). All those options have been tried. There's no really good way to make a muscle-like actuator with power to weight ratios comparable to biology. Systems with a fixed spring work, but under a narrower range of conditions. [1] http://www.ai.mit.edu/projects/leglab/robots/Spring_Flamingo/misc_info/faq.html http://www.ai.mit.edu/projects/leglab/robots/Spring_Flamingo...
  - garbagetime 4y ago
    Hello, do you have a source for running humans elastic energy recovery being 70%?
darkteflon 5y ago
BattleTech nailed this decades ago. Watching this took me right back to my childhood. Edit: AT-ST Walker, too.
- duskwuff 5y ago
  It's interesting that most fictional depictions of legged robots (outside of explicitly humanoid robots) use digitigrade legs. Did the artists model their art after animals which happened to have these legs, or did they subconsciously realize that this would be more effective?
  - marcosdumay 5y ago
    Isn't it obvious? After you see an ostrich running, there is little doubt about its efficiency. Now, actually capturing that efficiency in a real system takes some really hard engineering work. Fictional robots do not need to conserve energy, compensate for impacts, etc.
  - beaconstudios 4y ago
    Now you have me imagining mechs with human legs and it's making me uncomfortable. A bit like those "birds with human arms" photo edits.
- girvo 5y ago
  It really does look exactly like a BattleMech from one of the MechWarrior video games doesn't it! Thats fascinating. Reminds me most of the Locust: https://www.sarna.net/wiki/Locust https://www.sarna.net/wiki/Locust https://sites.google.com/site/fineartofmechcombat/_/rsrc/1468738697145/home/general-guide/locust/xctl.jpg https://sites.google.com/site/fineartofmechcombat/_/rsrc/146...
  - hef19898 5y ago
    Wouldn't a Spider be the better scout 'Mech? It has jump jets!
    - girvo 4y ago
      Oh for sure! It just reminded me of exactly how the Locust’s legs are animated in MW5
- 5y ago
ahaferburg 5y ago
Looking at the video I can't imagine this thing achieving balance while walking. It would fall to the side immediately if you removed the guide that held it upright. This is not a walking robot yet, it's a leg design. The movement looks plausible from the side, but not from the front. It does look promising, but putting "achieves energy-efficient gait" in the title of their publication seems a bit disingenious. Putting out pictures of the thing standing - but not moving - in nature enviroments also doesn't help. Balance is the hard part. Has anybody been able to find a download link for the actual publication?
- loonster 4y ago
  I do wonder if it can walk backwards or handle uneven terrain. The bird likely has a muscle to extend the foot when needed, the bot does not.
- regularfry 4y ago
  Well yes, and they say why in the article. It doesn't yet have the degree of freedom in its hips to balance properly. That's not especially hard though - a pivot around the main spring axis would do the trick, and the dynamics of that part are well-understood at this point.

sethammons 5y ago

that is really, really cool. Leveraging spring-like "tendons" to do the leg lifting and resulting in 1/4 the power usage of "normal" robot legs modeled after humans. Great visualizations.

Animats 5y ago
That's been done before. It's just new for Pratt, who previously used series elastic actuators.[1] Those don't do energy recovery; they fake it using a jackscrew and a stiff spring. Running humans have about 70% elastic energy recovery. Cheetahs have around 90%. What you want in a leg muscle is a spring whose spring constant, damping constant, and neutral point you can control. That can be done pneumatically (you need compressed air), hydraulically (you need hydraulic accumulators), mechanically (two springs operated by separate actuators), or electrically (a fast back-driveable motor drive that generates power during compression). All those options have been tried. There's no really good way to make a muscle-like actuator with power to weight ratios comparable to biology. Systems with a fixed spring work, but under a narrower range of conditions. [1] http://www.ai.mit.edu/projects/leglab/robots/Spring_Flamingo/misc_info/faq.html http://www.ai.mit.edu/projects/leglab/robots/Spring_Flamingo...
- garbagetime 4y ago
  Hello, do you have a source for running humans elastic energy recovery being 70%?

darkteflon 5y ago

BattleTech nailed this decades ago. Watching this took me right back to my childhood. Edit: AT-ST Walker, too.

duskwuff 5y ago
It's interesting that most fictional depictions of legged robots (outside of explicitly humanoid robots) use digitigrade legs. Did the artists model their art after animals which happened to have these legs, or did they subconsciously realize that this would be more effective?
- marcosdumay 5y ago
  Isn't it obvious? After you see an ostrich running, there is little doubt about its efficiency. Now, actually capturing that efficiency in a real system takes some really hard engineering work. Fictional robots do not need to conserve energy, compensate for impacts, etc.
- beaconstudios 4y ago
  Now you have me imagining mechs with human legs and it's making me uncomfortable. A bit like those "birds with human arms" photo edits.
girvo 5y ago
It really does look exactly like a BattleMech from one of the MechWarrior video games doesn't it! Thats fascinating. Reminds me most of the Locust: https://www.sarna.net/wiki/Locust https://www.sarna.net/wiki/Locust https://sites.google.com/site/fineartofmechcombat/_/rsrc/1468738697145/home/general-guide/locust/xctl.jpg https://sites.google.com/site/fineartofmechcombat/_/rsrc/146...
- hef19898 5y ago
  Wouldn't a Spider be the better scout 'Mech? It has jump jets!
  - girvo 4y ago
    Oh for sure! It just reminded me of exactly how the Locust’s legs are animated in MW5
5y ago

ahaferburg 5y ago

Looking at the video I can't imagine this thing achieving balance while walking. It would fall to the side immediately if you removed the guide that held it upright. This is not a walking robot yet, it's a leg design. The movement looks plausible from the side, but not from the front. It does look promising, but putting "achieves energy-efficient gait" in the title of their publication seems a bit disingenious. Putting out pictures of the thing standing - but not moving - in nature enviroments also doesn't help. Balance is the hard part. Has anybody been able to find a download link for the actual publication?

loonster 4y ago
I do wonder if it can walk backwards or handle uneven terrain. The bird likely has a muscle to extend the foot when needed, the bot does not.
regularfry 4y ago
Well yes, and they say why in the article. It doesn't yet have the degree of freedom in its hips to balance properly. That's not especially hard though - a pivot around the main spring axis would do the trick, and the dynamics of that part are well-understood at this point.