MegaButts 3 years ago

They should really include a video, because without it this article isn't particularly helpful.

https://youtu.be/wwH40rYJt9g

  • darkteflon 3 years ago

    BattleTech nailed this decades ago. Watching this took me right back to my childhood.

    Edit: AT-ST Walker, too.

    • duskwuff 3 years 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 3 years 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 3 years 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.

    • hef19898 3 years ago

      Another thing that we are going to see in combat vehicles in the next decades, IMHO, is something like OmniMechs. Meaning that future weapon systems, e.g. fighters or tanks (not the upcoming generation but those after that, F-35's successor' successors) will feature LRUs (line-replaceable units) that carry different mission capabilities. That way, a fighter can land as a ground support fighter or drone controller and lift of as air superiority fighter by swapping a hand full of modules out.

      • dumpsterdiver 3 years ago

        It certainly sounds terrifying having an adversary that can pivot into superiority based on the situation, but I wonder how practical it would be when faced with an opponent who simply brought the right amount of non-modular vehicles in the first place?

        If you have to pivot to gain air superiority, there's a really good chance that you've already lost.

        • dwighttk 3 years ago

          I feel like in theory you’d be able to adjust the mix across the battlefield throughout the battle and it would be less about one plane’s ability to pivot.

          And in theory, theory is identical to practice ;)

          • dumpsterdiver 3 years ago

            Everyone has a plan until they get punched in the mouth.

            - Mike Tyson

      • kingkawn 3 years ago

        Sounds super expensive and nothing a javelin can’t take care of

        • hef19898 3 years ago

          Like bazookas took care of Panzer IVs? I like how everyone just heard about Javelins and now thinks they rendered all armored vehicles obsolete over night.

          EDIT: You now what takes care of a Javelin, or rather the guy using it? A bat or a knife.

          • kingkawn 3 years ago

            From everything we’ve seen in warfare over the last few decades javelins and other shoulder mounted anti-tank weapons do render heavy armor highly vulnerable. New armor solutions that are even more expensive than existing ones but no more resilient against these weapons is throwing money down the drain. Sorry it fractures cherished anime fantasies.

            • hef19898 3 years ago

              And there I was convinced the best anti-tank weapon was a sharp Katana... Combined arms are a thing, and tanks were never well suited for urban environments. And vulnerable is the opposite of obsolete, everything and everyone is vulnerable on a battlefield.

              The race between armor and anti-armor is as old as armor, it is what gave us Javelins in the first place. I don't see any reason why that is going end, ever.

              • kingkawn 3 years ago

                Which is why nobody is going to produce something that is shiny and impressive but offers no added strategic advantage against existing counter-measures in this old struggle.

                • hef19898 3 years ago

                  France/Germany, the US, South Korea / Turkey are launching new MBT programs at the moment, whether or not you believe the entry into service dates of 2035 for the Franco-German or US programs or not.

                  Not to forget that the existing models are constantly being upgraded.

                  That being said, shiny and impressive is usually enough to get expensive and job creating military programs of the ground, isn't it?

  • ahaferburg 3 years 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?

    • regularfry 3 years 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.

    • loonster 3 years 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.

  • sethammons 3 years 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 3 years 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...

      • garbagetime 3 years ago

        Hello, do you have a source for running humans elastic energy recovery being 70%?

sandworm101 3 years ago

Because birds have been walking on two legs for several hundred million years, back to long before they learned to fly. Humans... not so much. It is no suprise that a bipedal prey animal would be a better runner than an ape that started standing up yesturday. We are basically the only animal that walks the way we do. The bird bodyform was already the proven betatest winner long before robots were a thing.

  • DoingIsLearning 3 years ago

    I don't have the evidence at hand but I am very sure from previous reading that our gait is still more energy efficient than other birds/animals.

    It is not about years of evolution it is about purpose.

    The downside is that our gait is basically a process of controlled falling, this need a really complex proprioceptive and neural control. The advantage is that we can travel _extremely_ long distances with minimal caloric intake when compared with other animals.

    So the claim from these researchers is that this system is efficient without the need for complex control systems (like in a human model). In practice a bird locomotor system is easier to control than a human one, it does not make the bird system more efficient because birds have been on earth longer.

    • kayodelycaon 3 years ago

      Our gait isn't the reason humans can travel long-distance. It's that humans have more effective cooling and don't overheat in high temperatures.

      In cold climates, gray wolves have advantage in ultramarathon distances, provided they don't overheat. They can maintain 5~6mph on rough surfaces. Humans maintain 4~5 mph on prepared surfaces. (Wolves are lighter and their four clawed feet have significantly better traction than our two feet.)

      Any surprise we domesticated dogs? They can keep up with us. :)

      • DoingIsLearning 3 years ago

        Speed/traction have little to do with energy efficiency. The cooling adaptions we have allows to outperform most animals in long distance _running_ which is how the whole concept of 'persistence hunting' is based on.

        What I was pointing to is energy-efficiency in _walking_ long distances (we migrated long distances by walking not running).

        The bi-pedal flat foot heel-strike and roll gait that we have is the most energy efficient gait, the dogs in your example can keep up with us but still spend more energy per mile per kilo walking on four legs.

        The downside is that it is actually pretty hard to keep your center of mass on top of such a small Base of Support between your feet.

      • sandworm101 3 years ago

        Dogs have more than four feet. For purposes of traction, they have at least twenty. Each paw has five pads that contact the ground independent of one another. Add in their nails, and they have 36 articulated points of contact. Our toes are no longer load-bearing. Whereas we walk with the bulk of our weight on maybe four pads (heel+ball on each foot) dogs walk on an octopus-like network of structures. They will always have better traction.

    • adrian_b 3 years ago

      This fact about the energy efficiency of human locomotion is frequently mentioned.

      However, what is not mentioned is that this high efficiency is attained only when humans walk or when they run slowly, much more slowly than almost any other animal of a similar size.

      This ability would not have been of any use, had it not developed simultaneously with the ability of throwing sticks and stones much faster than any prey animal could accelerate.

      High endurance, even at slow speeds, was perfect for following the track of an animal wounded by a throwing weapon, until it became exhausted and it could be caught easily.

      Otherwise, walking randomly across the plains, with the hope of encountering by chance an animal which was exhausted enough to not be able to escape from your very slow, but efficient, running, would have been a losing strategy.

      • sandworm101 3 years ago

        Humans also had one advantage: sweating. We are almost the only animal that can perspire across the bulk of our skin. So we can keep walking/running in the heat of the day when our prey and our predators cannot. Given enough time, in hot weather, humans can harass and run down uninjured deer and even horses. Our method of walking may not be efficient, but our ability to handle heat means we can keep moving long after everyone else has to hide in the shade.

        Just look at anyone walking a dog on a summer day. The dog is probably panting. The human is not. In a distance race, the human could easily outrun the dog.

        • adrian_b 3 years ago

          I agree with this.

          There was not a single new ability which ensured that humans became the most dangerous hunting animals, but an ensemble of correlated abilities, and improved sweating was certainly one of them.

          Regarding the dogs, that is why their relatives are seldom important predators in warm climates.

          Even the best adapted to warmth among canids, like the African hunting dogs, have to remain much smaller than the large wolves of cold climates, for adequate cooling.

          When you compare humans with African predators, humans still have better cooling, but the difference is not so large as when you compare them with domesticated wolves, a.k.a. dogs.

      • DoingIsLearning 3 years ago

        Wholly agree, which is why evolution wise we 'waste' a lot of resources on a big brain that is a huge energy sink.

        However my point was about GP's claim that birds gait was more efficient because they have been around for longer, which I disagree with.

  • ramblerman 3 years ago

    It's an interesting thought. I guess that also implies that the way our hindlegs bend is optimized for quad-pedal movement.

awinter-py 3 years ago

bipedal bird to beat is the ostrich

anatomy of the intermediate palaeotis weigelti shows that if we're serious about this biped project, we need to move the pelvis around, fuse the shoulderblades, and reduce the # of toes + bones in the foot. (weigelti has 3, modern ostriches have 2)

unclear if we'll be able to climb trees or throw things at that point -- might be able to use the neck as a sort of trebuchet

https://pterosaurheresies.wordpress.com/2018/01/15/palaeotis...

https://palaeovertebrata.com/Articles/sendFile/270/published...

jpm_sd 3 years ago

Agility Robotics has been building similar robots for about 5 years now. They have recently commercialized a full humanoid for warehouse applications.

https://www.agilityrobotics.com/news

  • d--b 3 years ago

    That's weird, warehouses have flat floors, so robots on wheels would be much simpler.

    • emteycz 3 years ago

      Only until something falls down on the floor or the robot needs to use stairs etc. Loading into smaller vehicles is also not done with wheels but with hands and feet.

  • prox 3 years ago

    Why do these big standing robots usually have tiny tiny feet? It seems counter intuitive.

    • smegsicle 3 years ago

      maybe a smaller contact point means a simpler model that doesn't have to take the orientation of a larger foot into account?

proc0 3 years ago

Hmm, so birds have an AI research department now? Title could be fixed by adding "anatomy", so "bird anatomy makes better bipedal bots".

Anyway, it seems reverse knee legs aren't great for upright bipedal walking, especially if you want hands to do something else. Bipedal animals with reverse knees have either wings or very small hands (kangaroo), and I would guess it's because the center of gravity with reverse knees has to be relatively low. Humans evolved to walk upright for purposes of using hands at the same time.

  • aimor 3 years ago

    "Reverse knees" is a funny term for ankles.

    Penguins are another unique animal, they have very low ankles (and high knees, hence the waddling) and also seem to use their hands more than other birds.

    https://images.fineartamerica.com/images/artworkimages/mediu...

    • iggldiggl 3 years ago

      And kangaroos? Okay, so their knees are possibly somewhat more inconspicuous, but looking at a kangaroo skeleton, what to my eyes looks like their ankles also actually is their ankles.

      Whereas with birds the confusion comes in because

      a) they're walking on their toes, so what to us looks like an ankle is in fact a toe joint, and then moving further up what we think looks like a "reverse knee" is in fact the actual ankle, and

      b) their true knees are usually hidden somewhere beneath their feathers

    • usrusr 3 years ago

      And even we humans switch to "heel up" for high intensity bipedalism. Are we are secret members of the reverse knees club?

  • olodus 3 years ago

    All this time we were too focused on automation stealing our jobs that we didn't realize the birds were stealing our automation-creation jobs!

GauntletWizard 3 years ago

How will this scale to quadrupeds? It seems novel, but birds are lightweight and topheavy, with significant active balance redistributing their weight front-to-back. Look at how much head movement is involved in a bird walking.

For light loads, this may work very well, but most robotic walkers are being developed as pack animals and I think that's the clear niche for bipedal robots.

  • enasterosophes 3 years ago

    You seem to presume that this specific technology should scale to quadrupeds.

    Other people are already working on quadrupeds. You may have heard of Boston Dynamics.

    Not everything needs to do the same thing as everything else.

  • tuatoru 3 years ago

    Have you heard of ostrich races and emu races? Birds can carry reasonably heavy loads. Quickly.

    Edit: much more energy-efficient than a drone delivery service, and probably not that different in speed for medium distances.

    The difficult problem, as with any mechanical replica of a biological mechanism (tensegrity energy recovery locomotion in this case), will be long term reliability.

    • regularfry 3 years ago

      And safety for bystanders.

  • makeitdouble 3 years ago

    Wouldn't quadrupeds be a separate problem space with pretty different solutions ?

    In particular, most quadrupeds have a strategy to keep two legs touching as much as possible, which fundamentally changes the balance game.

  • boomlinde 3 years ago

    How will a boat propeller scale to an airplane jet turbine?

luismedel 3 years ago

We're nearer to get an ED-209 in our streets.

ncmncm 3 years ago

TIL birds make robots. :)

I always thought making robots with legs like humans was a dumb idea. Birds clearly have the right idea.

  • autokad 3 years ago

    we are built and optimized for our respective uses cases.

    eventually after we figure enough things out, it wont make sense to use any animal as a template. we wont need to take in any of the drawbacks the animal design has because our robot doesnt have to worry about the things animals have to worry about

    • ncmncm 3 years ago

      We really are not optimal in any particular way. We started out as monkeys, and have done the best we could from a bad starting point. Turning the spine vertical really did not work out well at all, and knees are unfortunate at best.

      Probably worse than all that is how we go all to hell when we don't get exercise. Cats can sleep 18 hours a day without ever getting out of shape. Bears can can get super fat and then sleep for months, and be fine. A lot of birds can go a week without drinking or eating, working out the whole time with no sleep or even rest breaks.

      • lloeki 3 years ago

        > We really are not optimal in any particular way.

        Human running sounds pretty close to optimal to me, only not on the usually expected dimension: https://youtu.be/826HMLoiE_o

      • gherkinnn 3 years ago

        You just watch cats get fat once they hit their post-industrialisation phase.

    • Retric 3 years ago

      Birds had a lot longer to optimize for bipedal walking than humans did. If we had stayed bipedal without developing advanced technology it’s likely we would have gotten a lot more efficient at it over 10’s of millions of years.

    • marcosdumay 3 years ago

      Hum... We are a bit optimized for flexibility, and a lot the way we are by accident. As many people already said in other threads, we are probably the least optimized walking animal of all, because our body is recent and walking efficiency was never as relevant as other characteristics (like flexibility).

      The birds body, by their turn, appeared in a time closer to the first animals than to today, took over the lands of the world in large part because of their ability to run, and kept them for more time than any other kind of animal.

      There are some freedom in robot bodies that animals don't have. So it's not immediately obvious that an animal inspired body is perfect. But birds have had hundreds of millions of years to perfect their shape, it's not clear that we can do any better either.

  • rbanffy 3 years ago

    That’s precisely how I read he title. A shame. It would be a far more amusing world.