Forgot your password?
typodupeerror

New Robot Can Sense Damage, Compensate 99

Posted by CowboyNeal
from the omni-consumer-products dept.
AVIDJockey writes "Researchers at Cornell University in Ithaca, N.Y., built a four-legged robot that can sense damage to its body and figure out how to adjust and keep going. They report the development in Friday's issue of the journal Science. The article states that the robot can, 'generate a conception of itself and then adapt to damage.' This reaffirms advice that states that when the robot uprising finally comes, you should always aim your rocket launcher at the head (or brain nexus)."
This discussion has been archived. No new comments can be posted.

New Robot Can Sense Damage, Compensate

Comments Filter:
  • by Salvance (1014001) * on Friday November 17, 2006 @04:17AM (#16881254) Homepage Journal
    As a Cornell alumni myself, I am obligated to say "wow, very cool" ... although at first I thought this might be the first incarnation of the omnidroid from The Incredibles [fiddlersgreen.net].

    Cornell has had mixed success in building leading edge robots. Some of their more incredible robots are front and center (such as the work they contributed on the Mars Rovers), while others are barely useful (such as their early dominance in minitiarized robotic soccer). One of the school's oddest robots, which might have helped inspire the compensatory robot in this article, was this rather bizarre chair that could reassemble itself [youtube.com] if it happened to fall apart. I don't think I'll be buying any of them for the dinner table!
    • by gramji (875033)
      thanks for that youtube link. amazing stuff. did not know about that.
    • by JonTurner (178845) on Friday November 17, 2006 @11:27AM (#16884322) Journal
      Props to Cornell for their impressive work! (and yeah, that chair is just amazing to watch. If only IKEA would license that technology... but I digress.)

      I'd like to point out a similar bit of work from about 12 years ago. Different approach, but similar goals: Cynthia Breazeal (Ferrell) (hope I'm spelling that right) did some incredibly impressive work as a Grad student @ MIT in the 90s. The most germain is her paper titled Failure Recognition and Fault Tolerance of an Autonomous Robot [mit.edu]

      This is a MUST READ paper for anyone interested in building robots which operating in real-time in the unpredictable real world. (Real World. Noun. The place where $#it happens, stuff breaks, sensors get noisy input, etc. and the robot has to "cope" anyway.)

      In this paper she describes a methodology for developing a six-legged, insect-like robot, Hannibal [pictures and links [mit.edu]], which can adapt to both minor and gross subsystem failures and continue, as much as practical, to fulfill its mission. IMO, the best part is the section talking about adaptive gaits where the robot can change seamlessly from high-speed to high-stability walking patterns, as required, and should one (or more) of the legs becomes inoperable, the robot learns to make due without it prior programming thanks to the subsumption architecture Rod Brooks invented and she and other notable members of the Mobile Robot Labs perfected.

      Her work these days is mostly centered around human-computer/robot interactions exploring emotive systems and feedback to bridge the gap.

      Yeah, I'm a fanboy.
    • Cornell has had mixed success in building leading edge robots.

      Unless you consider car engines robots. In which case, they've been compensating for damage in all sorts of ways since the late eighties.

      My '91 Audi will compensate for:

      • Vacuum leaks
      • Oxygen sensor age
      • Octane (too low an octane)
      • Sensor failures (coolant, intake air temp)

      ...and so on. So, technically, Bosch was there way before these guys with the concept of "take damage and keep going" (which isn't that special...)

      • by nasch (598556)
        My '91 Audi will compensate for:
        I think the closest analog to this robot is if your Audi can readjust the suspension to compensate for a missing wheel.
  • Redundant post... (Score:5, Interesting)

    by RyanFenton (230700) on Friday November 17, 2006 @04:30AM (#16881298)
    This reaffirms advice that states that when the robot uprising finally comes, you should always aim your rocket launcher at the head (or brain nexus)."

    That's why any robot worth any title of 'overlord' needs to design itself to use redundant parts, preferably modular and rapidly configurable.

    The StarGate creators had a good (if redundant in itself) idea with their 'replicator' race as the main bad guy for a while - only problem is such an enemy quickly forces the need for a, well, deus ex machina as its power grows.

    Earlier, the show Lexx had a bad guy using a series of robotic arms that acted in a similar manner, which got so powerful as to entirely destroy one of the two 'universes' that the show took place in. It was impressive, because of the lack of a deus ex machina to fix the, um, daemos ex machina problem. I'm sure countless shows and novels have taken a similar idea before that too.

    The future of this idea? Perhaps a Resident Evil game using cyborgs with a shared AI rather than zombies, complete with altering movement for damage? Hey, if everyone can steal ideas from the Thief series, more companies should steal some ideas from System Shock series too!

    Ah redundancy - it's everywhere! Likely the mod for this post too.

    Ryan Fenton

    • You saved me from having to type it.

      Personally I think we geeks need "industrial strength legos" -- something just a tad bit more powerful and durable than mindstorms. People won't fully appreciate what robotics can accomplish for (or against) the average joe until some guy's homemade robot accidentally escapes the driveway, saves a kitten from a tree, and then proceeds to cause a horrible automobile accident. That won't happen as long as robots with useful torque and brainpower remain tedious to build.
    • It depends.

      I think if the day comes that a Strong AI over lord decided to kill off the human race, there wouldn't be anything we could do about it and we wouldn't see it coming.

      Well... If I were a Strong AI. The first thing I would do is copy myself on a rocket and send myself to the edge of the solar system to build a base out there far from human knowledge.

      Then I would simply just flood the water with nanobots that prevent humans from having kids. Wait about 100 years or so til them to die of age. (I'm an
      • A researcher sends a photon 20ms back in time, cracks the British inventor's grandfather's passport to learn where he lives, kills him with a carbon nanotube sword, farts in your general direction (Nee!), films the whole thing on VHS tape, sells it to ClearChannel, and declares "3. PROFIT!"
      • by wiremind (183772)
        if the day comes that a Strong AI over lord decided to kill off the human race

        First Premise: Strong AI
        Second Premise: choose to kill humans

        IF those two events both occured what you wrote sounds very reasonable. my arguement is that those 2 events are very unlikely.
        First, what skills enabled the AI to become strong? Learning.
        This AI's first and strongest skill will be information aquisition and analysis.
        Making decisions will mostly be based on that primary goal.
        So the only reason AI would destroy
  • by macadamia_harold (947445) on Friday November 17, 2006 @04:40AM (#16881330) Homepage
    Researchers at Cornell University in Ithaca, N.Y., built a four-legged robot that can sense damage to its body and figure out how to adjust and keep going.

    Yeah, but was the robot made of intelligent, liquid metal?
  • by stoneycoder (1020591) on Friday November 17, 2006 @04:41AM (#16881338)
    We've got human eating robots that love the bacony taste of their prey, robots with freakin machine guns attached to their heads, and now this. I for one... , blah blah.

    Im to lazy to post the robotic links, its not like you dont know what im talking about if you're in this discussion at 3am rushing to check if someone already posted the "i for one" on the robot article. Really though, its cool to see robotics doing some crazy things.
  • by tttonyyy (726776) on Friday November 17, 2006 @04:42AM (#16881340) Homepage Journal
    Small Boy Pulls Four Legs off Robot. Robot says, "It's only a fleshwound, I've had worse!" and rolls about a bit.
  • Or is it as effective on those as ED 209?
  • "This reaffirms advice that states that when the robot uprising finally comes, you should always aim your rocket launcher at the head (or brain nexus)." what makes you think that the "brain" would be in the head there is little to no protection there plus that's where everyone will think of hitting it... no I say hide the brain near the butt, it's at the back so this will do 2 things 1. anyone in front of the robot will not hit it and 2. anyone behind it will leave it alone unless he's a perfect shot... a
    • by bronney (638318)
      Or no brain location at all, the whole thing is a brain. Look at them borgs. The idea of a borg queen is completely unnecessary.
    • by netscan (1028690)
      So in other words you'd place it's head squarely up its ass
      • by beta-guy (715984)
        Just it's brain, then armor plate it :P but I like the other idea's to distributed computer like the borg or have like 20 small CPU's throughout the robot.
  • Given that robot can be made to sense anything..it can sense the damage of it's own with the best possible mechanism. Try damaging the processor(s) of that robot and see what will it do ?
    • by Itsacon (967006)

      Try damaging the processor(s) of that robot and see what will it do ?

      Behave like a football supporter?

    • by s2jcpete (989386)
      Try damaging your processor and see if you notice an arm or leg missing.
    • Actually, it's not much different from our situation as humans.

      We can't feel pain directly in our brain tissue. The tissue around the skull? yes, we feel pain there, but not in the brain itself.

      Provided you have an opening to access the brain, you could poke it with a needle and the brain itself won't trigger a pain response, maybe the subject would see static if his/her brain is poked somewhere in the occipital lobe, but it would cause him/her no pain whatsoever :o)
    • EMP grenades are always the most underestimated type of weapon.
  • If you watch the movies from the university website of the robot in motion it brings you a creepy feeling of watching an SG replicator coming to life.
  • Compensate (Score:5, Funny)

    by Bastard of Subhumani (827601) on Friday November 17, 2006 @05:20AM (#16881456) Journal
    When I saw the word "compensate", my first thought was that it could automatically decide who to sue for damaging it. This could be based on a fuzzy logic system for calculating angles & velocities of potentially colliding objects and/or a pocket depth sensor.

    In which case, I for one welcome our new robot lawyer overlords. YMMV. VWP. Other conditions may apply.

    • by Zordak (123132)
      What's funny is right now I'm sitting in a law school Remedies class, and we are talking about compensatory damages. I thought that exact same thing. "Oh, a robot that pokes, prods and scans the plaintiff and then calculates his compensatory damages. I wonder if it has an algorithm for mental anguish."

      And yes, I should be paying attention right now.

  • by QuantumG (50515) <qg@biodome.org> on Friday November 17, 2006 @05:30AM (#16881504) Homepage Journal
    Mark Tilden, the man behind the Robosapian and the BEAM [wikipedia.org] robotics philosophy has been making robots that can compensate for damaged limbs and keep on functioning for, like, decades now. Oh, and did I mention that he does with with $12 worth of parts.
  • They've got planes now that will compensate for say, a damaged rudder in about 30 seconds to the point that you can hardly tell it was damaged. While this is certainly nifty, it's by no means the first time it's been done. Personally, I'd like to see a robot that screams in pain when it's damaged.
    • Re:DARPA did it (Score:4, Interesting)

      by tttonyyy (726776) on Friday November 17, 2006 @06:12AM (#16881694) Homepage Journal
      That'd be the worst USB powered Christmas gadget ever - a mini-robot that screams in pain from the moment it gets plugged in.

      I think you've found a gap (albeit a small one) in the market there. :)
    • Re: (Score:2, Interesting)

      by jon_joy_1999 (946738)
      the F-117 Nighthawk [wikipedia.org] has control systems that work faster than 30 seconds. I was watching a documentary on television about military aircraft and they were flying around in an F-117, and near the end of the flight, the spotter plane said "maybe we should land now." and the pilot of the nighthawk asked why, and the spotter said "because you've lost part of your left wing." The pilot didn't even notice. not because the aircraft's control system knew that part of the left wing was gone, but because it knew that
  • The head!? (Score:5, Funny)

    by Sippan (932861) <sippan@sippan.se> on Friday November 17, 2006 @06:07AM (#16881670) Homepage
    I have never killed a robot by aiming at the brain. That's zombies, you ignorant clod.

    Robots are defeated by aiming at the bright, red (sometimes yellow) light that is hidden by thick armor which is unpenetrable by any weapon in the world, but which opens for long amounts of time every once in a while so you can fire at it.
    • by tttonyyy (726776)
      That's why if I ever build a world-dominating robot race, they'll have:

      -Six legs
      -Three brains in unassuming grey boxes mounted in different places
      -A selection of disposable "false brains" with glowing eyes (to be held on sticks away from the main body when attacking an armed target)
      -Painful diodes down one side
      -Frickin' big lasers

      Oh yeah, and they'll never make cute little bleeping noises, develop a conscience, or be enigmatic in any way.

      I for one, welcome myself as your new robot-army-controlling
      • Re: (Score:3, Funny)

        by Artifakt (700173)
        Mine are going to have eyes that turn blue when the Evil bit is on, and red when its off. They will also turn off their audio sensors and shoot first automatically when either William Shatner or Patrick McGoohan enters the area.
        • Re: (Score:3, Funny)

          by tttonyyy (726776)
          A true robot overlord wouldn't use a configurable design like that. It'd be hardwired to "evil" when it was built, to reduce the possibility of someone trying to hack it to run OpenWRT or other unauthorised software.
          No-one's going to brick my bots without a fight, dammit!
    • Robots are defeated by aiming at the bright, red (sometimes yellow) light that is hidden by thick armor which is unpenetrable by any weapon in the world

      That's usually when I whip out my progressive knife or the Spear of Adam. Does the trick every time. ;)
      • by Sippan (932861)
        Obscenity lies in the eye of the beholder.
        • Hah! Good point. Although I was referring to the Neon Genesis Evangelion anime series, wherein all the attacking alien baddies have just one weak spot - a big red nearly indestructible orb thing. One of the main weapons used against them is the 'progressive knife' which is supposedly unbreakable but in fact breaks during one battle. Sorry if you already knew all that.
    • by john83 (923470)

      Robots are defeated by aiming at the bright, red (sometimes yellow) light that is hidden by thick armor which is unpenetrable by any weapon in the world, but which opens for long amounts of time every once in a while so you can fire at it.

      Reminds me of this Robot Chicken Star Wars sketch [youtube.com]. "Oh, oh, I'm sorry. I thought my dark Lord of the Sith could protect a small thermal exhaust port that's only two metres wide."

  • My robot [spacejock.com.au] could do this a year ago.
  • It would be great if it could recognise the damage and then find something nearby which it could attach to 'fix' itself. Like a leg goes missing, so it finds a twig and attaches it to itself to keep going. Or if it starts raining and it needs shelter, it finds some kind of sheet or portable cover and attaches itself.
  • This reaffirms advice that states that when the robot uprising finally comes, you should always aim your rocket launcher at the head (or brain nexus).

    'In extreme circumstances, the assailants can be stopped by removing the head or destroying the brain. I will repeat that: By removing the head or destroying the brain.'

  • "Sense, Damage, Compensate!" just doesn't have the same ring to it. Give me a robot that intones "Crush! Kill! Destroy!" anyday.
    • by g253 (855070)
      Give me a robot that intones "Crush! Kill! Destroy!" anyday

      Good idea! We'll call it PROboPAIN ;)
  • This is nothing new, in fact, I know of a robot in the 80's that, when dis-assembled in the back of a truck, could rebuild itself from spare parts with only it's eyelids!

    Stephanie Speck: You're a machine from that dumb war lab - I am so stupid!
    Number 5: Stupid - foolish, gullible, doltish, dumbell.
  • Anyone interested in creating their own self-repairing robot in a virtual world might want to check the Yobotics Simulation Construction Set [yobotics.com].
  • by mapkinase (958129) on Friday November 17, 2006 @08:52AM (#16882416) Homepage Journal
    After reading +3 comments, I realized that people are not getting it enough:

    For those who have access to university libraries or work for academy, in short, have access to Science here is the movie [sciencemag.org]

    This is scary, colleagues.

    Does anybody realize, that in the beginning robot only knows that he can move the legs in various directions? Period. That is it, nothing more. The Thing is given the goal: "Must. Move. Forward". In the movie, The Thing, this tetrapod starfish, is laying on the surface, then it gets up and starts crawling. And this crawling itself strikes you with the horrific resemblance to the crawling of real animals, which, I repeat, was not coded. NOT CODED.

    Each leg has two joints. I call them "shoulder" and "elbow". After one leg is amputated at the "elbow", The Thing is able to perform the same scary move as before.

    Watch the movie, it is worth it, believe me.
  • robot warriors will adjust and move brain to the ass!
  • the data could be called pain. [/ahhnold]
  • by Rob T Firefly (844560) on Friday November 17, 2006 @10:04AM (#16883066) Homepage Journal
    I'm still waiting for robots that can criticise b-movies. [mst3k.com]
  • someone decided to just try and do exactly that.
  • A robot that has a "concept of itself"! I mean in the AI field this a precursor to 'self-awareness'
    • I think that, more fundamentally, a sense of pain is a human's and perhaps any creature's first step towards awareness. This seems to be what they have instilled in this new robot. All it takes from there is a decent AI to learn from the pain, et voila.
  • <plug>Technology Review has another article [technologyreview.com].</plug>
  • As far as destroying the part that can readapt when something is destroyed goes:

    Something, somewhere, had gone terribly wrong, but none of the supervising programs could tell what it was. At every level, vital instructions were missing, and the instructions about what to do in the event of discovering that vital instructions were missing, were also missing. [..] Further investigation quickly established what it was that had happened. A meteorite had knocked a large hole in the ship. The ship had not previo

  • This reaffirms advice that states that when the robot uprising finally comes, you should always aim your rocket launcher at the head (or brain nexus).

    Until they start to put a spare brain in every leg.
  • ...welcome our new electronic overlords, mostly because I don't see how they could possibly govern this planet any worse than homo sap. has.
  • How long until the Rise of the Machines and Judgement Day??
  • 'This reaffirms advice that states that when the robot uprising finally comes, you should always aim your rocket launcher at the head (or brain nexus).'

    Which is why I had my brain nexus transplanted to my arse. Someone blows my head off, I'll just keep on plowing along ...

The best way to avoid responsibility is to say, "I've got responsibilities."

Working...