Follow Slashdot stories on Twitter

 


Forgot your password?
Close
typodupeerror
×
Technology

The Largest Unpiloted Legged Robot Yet 98

Posted by timothy from the my-gawd-*look*-at-that! dept.
An unnamed correspondent writes: "Ever wanted your own dinosaur? Well slap some skin on this baby and you could." This beast looks like a steel elephant, features unusual motor-less joints, and takes a 700Mhz CPU to control each leg.
This discussion has been archived. No new comments can be posted.

The Largest Unpiloted Legged Robot Yet More

The Largest Unpiloted Legged Robot Yet

Comments Filter:

  • Something I'd like to see (Score:5)

    by Aggrazel ( 13616 ) <aggrazel@gmail.com> on Monday March 05, 2001 @08:18AM (#383896) Journal
    I would pay big bucks to see this thing attack the hosts of Comedy Central's Battlebots.

    I think the british version is much better...

  • Will it be road tested? [robotwars.co.uk]

    What I love about that show is how robots built from elastic bands, cardboard and milk bottle tops manage to beat the big guys.

  • Now I can get my own personal giant ED-209 guard robot within the next few years! Now no one will fuck with me again!

    I haven't been this excited since I saw Kenny on South Park wearing an ED-209 halloween costume...

  • 4x4 SUV 's will be replaced by AT AT's

  • Imagine working on this beast, even a single mistake would mean a catastrophic fall, probably setting you back months.

    So in addition to the increased mass/power supply issues of making a large robot, you can't make mistakes, which compounds the design hassles even further.

  • how many walking creatures do you know w/ motors on their joints?

    i certainly don't know of many
  • this thing is more than 11,000 pounds. thats 5,500 on each foot. if you see the pics, look at the size of each foot. I don't know if it can just walk around with out breaking up the ground, thier going to need to take some weight of before it just walks around a park


    ________

  • May they never ride unicycles. Think, first they're riding unicycles, then, their mimes!! Real mimes are bad enough, but now robotic onces that can go years (hours if it runs Windows) of miming without stopping...

    Perhaps that is what is meant by the apocolypse.

    :

  • Looks like they wasted all of their 700MHz processors in the legs and had to settle for a 25MHz Math Co-Processor lacking i80486.

  • Re:The real question is... (Score:4)

    by Arethan ( 223197 ) on Monday March 05, 2001 @08:25AM (#383905) Journal
    I've never been very impressed with Battlebots. I'd rather see autonomous robots fight. You don't need image recognition to do it either. Stick a transponder on each robot for positioning. Robot creators are allowed to do anything they want to prevent the transponder from working, but they are not allowed to physically shut it off. (It could run on its own power source quite easily.) However, ecm bursts for up to 10 seconds could be allowed. Trouble is, each transponder runs on a different frequency, so the ecm would have to be programmable. hehe.

    Then, you throw away the lame remote controls, and focus on some good AI routines instead.

    The whole idea would increase the overall cost and time to build for each robot, but it would be so much more interesting!
  • ... are suddenly not near as cool.

    Soon as I find a toy I like, I go get it and something better comes out. I remember growing up and getting dicked on toys, now they have all these fancy toys that change and mutate.

  • When the thing walks, the entire body trembles because of the added stress on the other legs. This causes the entire frame to flex a few inches. I'd be guessing this is yet another reason why this is one of the only (correct me if I'm wrong) five and a half ton robots.

  • Sniff the wireless commands....AND.... (Score:3)

    by ellingtp ( 198719 ) on Monday March 05, 2001 @08:28AM (#383908)
    Imagine the thrill of hacking this thing. Issue your own commands, like, STOMP, CHARGE, SPIN IN CIRCLES TILL YOU PUKE, etc....hope that wireless command is encrypted.
  • ok here is it, cut and past verbatem from the website

    Like a turn-of-the-century hunter returning from safari, Frank Mezzatesta stands next to a huge wooden crate he gleefully says contains something wild, a beast never before known to man. And he can hardly wait to show it off. With the click of a mouse, the front of the crate crashes to the floor, revealing an enormous metal monster 13 feet tall, 18 feet long, and weighing 11,000 pounds-- the basic statistics of a loaded delivery truck. The creature seems to hesitate for a moment, then moves forward. With deliberate but surprisingly lithe steps, it strides across the floor, shifting its weight with the grace of a cat as it lifts each foot in turn. Nearing a small group of people, it leans toward them, then sways from side to side as if trying to decide whether to charge them, eat them, or ignore them. "At this point in a previous demonstration, one woman got up and ran," says Mezzatesta. The beast then takes a few steps backward, turns slightly, and begins to dance. Swiveling left and right as much as 7 degrees, it keeps its feet in place, making large undulations and 15-inch-deep knee bends. Before it is sent back to its box, some brave observers amble over for a better look.

    To understand what Dino's inventors are attempting, imagine this behemoth covered with a shell that makes it look like a dinosaur. Then imagine it roaming freely on its own-- the world's first truly autonomous robot. Photo by Jan Staller

    Mezzatesta is no big-game hunter; he's an engineer, and his beast is a robot dubbed Dino. It is the largest robot ever built that has legs and doesn't have a human inside. It contains its own power and moves autonomously after receiving basic instructions like "move forward." Ultimately, a version of Dino may be covered with a skin to make it look more like a triceratops. If challenging problems are solved, it could be let loose in theme parks to roam on its own. A machine that knows where it is, can make its own decisions, and can move around as easily as a living animal has long been the Holy Grail of roboticists. No one has yet been able to achieve this feat, even with small, wheeled robots.

    Birthed by Walt Disney Imagineering Research and Development, Dino was dreamed up by Danny Hillis, the man who invented massively parallel supercomputers in the 1980s. "I always wanted to build a robot dinosaur," says Hillis, who, as a Disney Fellow, ran the Dino project from 1998 until 2000, when he cofounded Applied Minds, Inc. His team of engineers and scientists was recruited from universities like the Massachusetts Institute of Technology and the University of California at Los Angeles. They went to work in a walled-off area of a large warehouse near the airport in Burbank. To one side of the robot sits a bank of computers where they created much of Dino's software and one console that wirelessly sends the robot simple commands like "walk backward."

    Beyond those commands, Dino carries everything aboard needed to control itself-- power from a bank of 55 sealed lead-acid batteries, three electric motors to move each leg, a Pentium 700 megahertz processor for each leg, and a central computer that receives commands, loads appropriate software from the onboard memory, and coordinates each leg's response. A gyroscope tells the robot how much it is leaning, and lasers at each ankle measure the distance to the ground to help calculate how a step should be taken. Sensors tell Dino how far each motor has rotated and the distance the robot has actually moved. The robot constantly compares feedback readings to make sure the multiple measurements make sense. It also monitors motors for current spikes or high temperatures to determine whether too much force is being applied, and it tracks velocity and acceleration limits.

    Dino is impressive, but it is still a work in progress. Hillis originally wanted to power the robot with a Corvette V-8 engine that pressurized oil-filled hydraulic actuators. A test of that setup turned out to be far too noisy and cumbersome. Now smooth, quiet electric motors and batteries have replaced gasoline engines and hoses.

    Damn lameness filter. look at my next post


    ________

  • You can have your own force of Dinobots(well, Grimlock at least)
  • New Fox shows like "when dinosaurs attack, part 5!!!"

    Or people overclocking their 700 Mhz legs, to make them run faster...

    ---

  • Creating a full-scale Dino from simulations was an engineering challenge. "If you just scaled up a grasshopper to the size of an elephant, it would crush itself," says Hillis, because increases in scale create geometrically larger increases in weight. Make anything twice as big and it gets eight times heavier, so all its supports have to be thicker-- which makes it even heavier. That extra mass, in turn, shoots up the forces of momentum and therefore the power necessary to get the robot moving and to stop it once it is moving. "The heavier and larger you scale an object, the more dynamic it becomes," says Gill Pratt, head of the Leg Lab group at MIT's Artificial Intelligence Laboratory, which is building its own, but much smaller, walking dinosaur (see "Walking on Two Legs," below). The more dynamic the robot, the more difficult it is to keep balanced when walking.

    Small robots with legs are also easier to build because they can fall over without causing much damage-- they don't have as far to fall or as much mass. "We'd be in a whole lot of trouble if this thing fell over," says Akhil Madhani, a mechanical engineer on the project. Madhani and other designers kept Dino's legs light by putting all the motors in the shoulders and then using a series of aluminum linkages and steel ball screws to transfer power through the knees to the ankles. Still, when Dino raises a foot, its body flexes about two inches, which sends vibrations through the whole chassis. "If you have ten thousand pounds vibrating a few inches, the forces are dramatic, maybe a thousand pounds back and forth," says Alexis Wieland, an applied mathematician who worked on the robot's software. "So all the walks are smoother than you'd really think is necessary, because any jarring of the body can produce enormous forces." Worse, when Dino lifts a foot, frame-flexing changes the distance between the three feet still on the ground, trying to pull them apart. By monitoring its motor currents, which shoot up due to the increased forces, Dino can compensate. "It's another level of intelligence that says: 'I'm exactly where I want to be, but, gosh, I'm fighting myself. Let me just move a few millimeters here and, oh, there it goes; the currents are lower,' " says Mezzatesta.

    When Dino's sensors don't agree with its software or one another, it stops. "It is truly autonomous; there's no preplanned trajectory," says Wieland. Without any human input, Dino can shift its weight and move its feet until the motor sensors tell it that it has reached its original starting stance. Still, the team has had to learn to trust Dino. "When you're watching something generate trajectories on the fly and you don't know up front what it's going to do and the whole thing could fall over if it does the wrong thing, it's very nice to see it do the right thing," says Madhani. "But it gets your heart rate up."

    OK taco you need to explain this lameness filter thing, it keeps cutting me off, ican't see why..... read the next post for more


    ________

  • The sooner I can "rearrange" my boss's Mercedes with a giant robot the better...
    <Insert evil villain laughter here>

    Gekigangar Walk!

    -----

  • Right now Dino should be able to handle uneven ground or walk up hills, and engineers plan to take it outside for a test run soon. Dino doesn't yet have sensors to tell it where it is on the planet or how to navigate around obstacles like people. All of that would be necessary before it could run around on its own. "But I wouldn't be that concerned about it, because much of that has been developed for wheeled mobile robots," says Martin Buehler, who's built small, running quadruped machines at the Ambulatory Robotics Lab of McGill University in Montreal. He is impressed by the work done on Dino. "The hard problem with all these robots is just to get the basic mechanics, dynamics, and control right."

    Dino has laser gyroscopes that give it a balancing mechanism "sort of like an inner ear," Hillis says. "But right now it's not smart enough to take advantage of that. Once it starts being able to do that, feels itself start to trip and catches itself, I think it will start looking more and more natural. But that's a big software job and nobody's ever done it before, so it's going to take a long time. I think eventually you'll have lots of things like this walking around. Some of them will look like robots, and some of them will look like dragons, and some of them will look like big animals like rhinoceroses or woolly mammoths or imaginary animals, all kinds of things." Buehler notes that legged robots like Dino also would be able to perform extraordinary services like fire fighting, containing nuclear and chemical hazards, defusing bombs, searching for land mines, even exploring outer space. The key to all these activities, Buehler points out, is the superior mobility of legs that work as well as Dino's.

    Before the gigantic robot can go anywhere, though, it must be able to make its own decisions about where it should put its foot next without stepping on someone. Eric Haseltine, who heads research and development for Disney, says his team of Imagineers is already working on such artificial-intelligence technology for virtual-reality beings, and those programs might be reusable in actual machines like Dino. "This is a test bed that puts us on a road map toward intelligent, self-directed characters," he says. "We want them to be able to move around, react, learn, and behave on their own."

    Autonomous dinosaurs roaming at will are likely to amaze and amuse us, but they will still be machines run by computers, and thus unlikely to operate perfectly. So as independent as they may become, Haseltine thinks, there will always have to be someone nearby watching-- with a finger on a kill switch.


    ________

  • I'll have a Marauder, please. The AC-5 and dual PPCs will come in handy during the evening commute.

    Hey, Buddy, will you drive better with your cell phone at 5,000F?

    On second thought, maybe something with jump-jets. Just no Morton Thiocol parts ...

    --

  • From Merriam-Webster Online [m-w.com]:
    Main Entry: motor
    Pronunciation: 'mO-t&r
    Function: noun
    Etymology: Latin, from movEre to move
    1 : one that imparts motion; specifically : PRIME MOVER
    2 : any of various power units that develop energy or impart motion: as a : a small compact engine b : INTERNAL COMBUSTION ENGINE; especially : a gasoline engine c : a rotating machine that transforms electrical energy into mechanical energy

    Definition 1 implies that almost all living-creature joints come with motors - they're called muscles. Sadly, the site is /.-ed, so who knows what exactly is being said?

  • Since it seems like Disney has a commercial interest in this project, for them it is far from a criminal waste of resources. This is going to be the next "Pirates of the Caribbean", literally.

    I am sure Disney would love to have a section of Disneyland with characters from its movie "dinosaur" walking around and acting like 'real' dinosaurs. Since they have a history of pulling off (almost)cool rides, I am sure this one will be a winner too. Or, at least a money-maker.

    Sooo, maybe you should buy some disney stock and donate your profit to charity.
  • for some reason i think that it would exceed the weight limits.
  • Walking on Two Legs Not all ancient reptiles were giants. Troödon, for example, was a late Cretaceous dinosaur that grew up to six feet long and four feet tall. Over the past four years, Peter Dilworth, a robotics researcher at MIT's Leg Lab, has built a 10-pound version of that dinosaur called Troody, the first two-legged dinosaur robot. Dilworth consulted with paleontologist Gregory S. Paul to ensure that the dimensions are as true to the real thing as possible, based on fossils. Troody has 16 electric motors distributed among its hips, knees, ankles, feet, and tail. There are two sensors on each motor to feed joint angle and force readings to an onboard computer and a gyroscope to sense which way is up and in which direction the robot is moving. When Dilworth turns the robot on, it starts from a crouch like that of a nesting bird, then rises and rocks slightly before getting its balance. Although Dilworth gives Troody commands like "move forward" from a joystick, the robot plans its own movements. "Seven hundred times a second, an onboard computer reads all the sensors, does a bunch of math to figure out where all its joints are positioned, where its mass is, all its physical properties," says Dilworth. "Then a control algorithm decides whether it should be swinging a leg or how it should be moving." Dilworth hopes to eventually have Troodys running around museums so that visitors can get a more realistic sense of what dinosaurs may have been like. -- F.S.


    ________

  • Given such a robot the potential for abuse of the technology seems rather high. What if your next door neighbour purchases it to use it as an e-Bouncer that can just walk into your apartment and knock you out? What if it was used to damage properties destroy lives and wreck general havoc? I am not saying it is all wrong but surely some control over the modern technology is a must at this point. Now that AI is advancing in leaps and bounds soon these things will become self aware and the doomsday scenarios that were the domain of SF books can well become a reality. Research on such robots should be controlled by independent international organizations to prevent any potential abuse of the technology.
  • Took a while for me to get to the artical, seems to be slashdoted. anyway here is the home page for the project, it has some quicktimes's, pictures and other stuff.
    [mit.edu]
    http://www.ai.mit.edu/projects/leglab/robots/rob ot s.html

    -Jon

    Streamripper [sourceforge.net]

  • Mirror (Score:5)

    by mberman ( 93546 ) <mbermanNO@SPAMearthling.net> on Monday March 05, 2001 @08:49AM (#383922) Homepage
    This one seemed to get /.ed pretty damn fast, so here's a mirror [cmu.edu].
  • 1 - jurassic park
    2 - westworld
    3 - itchy and scratchy land (where nothing can posibli go wrong)
  • If you like cool bots, check out the book Robo Sapiens. It has about fifty 2-3 page interviews with roboticists and tons of nice photos. It's mostly aimed at laymen, so don't get mad if it's not techie enough for you.

    Check out the Amazon Link at Robo sapiens: Evolution of a new species. [amazon.com]

  • yes, but i think the /. article was refering to 2. a, b, or c

    *shrug*

    i'm sure it was prolly just the submitter, not the article being bizarre...

    but yes under 1. muscles are motors, and of course anything that moves would have a motor, as something imparts motion on it, through physical contact, gravitation, magnetisim or some such thing...

  • I agree with your point of view, for the most part, however *this* particular site is hosted by discovery.com

    okay, so it IS a cable channel, but they should *still* be able to handle a bit of a heavy load eh?
  • If you look at the link this is to www.discover.com. If a site that big and advertised is running on a partial T1 they needed a whake up call in any case. Also yes it is pretty much true that the site linked to in this case *is* making money off of ads. And oh yes they would be more than happy to sue /. over mirroring their stuff. You should pick your stories better for posting this stuff.
  • ...I just rented Patlabor 1 yesterday...

    ------------------
  • If it has a pilot it is not a robot.

    Whether the pilot is on board or 1e6km away. As the long as the maching is directed by a human it is not a robot.

    I hate to see these "news" items about robots that really mean remote controlled devices. Lets keep the terms straight and not get led on by the popular notion that anything remotely high tech is a robot.

    Battlebots are not robots. They are funky remote controlled cars playing a mean version of the smash up derby.

    With power steering, abs brakes, and ignition control you could probably call most modern car fly by wire systems and describe them as robots using the "great" distinctions of the unwashed public.

  • Re:watch out (Score:5)

    by GypC ( 7592 ) on Monday March 05, 2001 @09:05AM (#383930) Homepage Journal

    Elephants manage just fine on grass, and they often weigh over 10,000 pounds. In fact, they are downright stealthy on turf. Ever been snuck up on by an elephant? Talk about a good scare, I thought hair on the back of my neck would never lie back down.

    I do not find in orthodox Christianity one redeeming feature.

  • that something that could grow to take over small cities and such was created by Disney? I am not really jumping on the conspicary boat, as much as pointing out, there is already a Disney Police force and now they have this....
  • I thought NASA built a six-legged version of this bot first, it just wasn't very self directed. (Required a human to operate it). And it sounds like the really big challenges are still how to get the thing to move mechanically, not intelligently. After all, the article does mention that small wheeled robots have already made big strides in 'intelligent' movement processes. Please enlighten me on where I can find info again on the NASA project, as I thought their walker was much more functional on a mechanical level than this one.
  • Unfortunately, it'd be rather boring: the state of the art in robot AI still takes a few minutes to recoginize where it is in a room, much less identify/track another hostile robot (particularly if it's moving at high speed). Of course, you still have the problem of disabling the other robot, which even human drivers seem to suck at.

    Kurdt
  • If I wanted to go destroy my neighbor's place, which of the following would be easier?

    a) build or purchase a 13'x18'x8' robot worth more than i'll ever have in my entire life, whose stability is questionable and who weighs over 5.5 tons (not sure if my freight elevator is approved for that much...)

    or...

    b) go out to my local sporting goods store, buy a baseball bat, and do it the old-fashioned way

    why does it being called a "robot" necessarily make it so scary? humans are quite good at messing each other up already, and i really don't think 1.1 klbs of aluminum that moves by itself (sorta) is going to make that all that much easier. remember: a shotgun will always be cheaper than a Killer Robot of Death.

  • What the hell is it doing with a 700MHz processor in EACH leg? The computations needed could easily be done with a 486-66.
    Must...restrain...flame....

    Maybe for just the movement computations. But I would imagine the added computing power is useful in analyzing the input coming from the multiple sensors used to determine height off the ground and all that other non-sense that was mentioned in the article (You did read the article didn't you? Oh wait, maybe it was slashdotted, so I'll give you a break). There's a lot more going on here than just a simple wind-up walk-along toy does. A lot more.

  • Ok, you need a 700mhz processor to control the sensors for balance and where the ground is etc etc...

    So if you wish to run, you'd need to process quicker. Would that mean that the faster you want to go the faster the chips?

    Interesting how our mind can't compute numbers so quick but we can already run :)

    ---
  • This probably seems too obvious. but could you imagine Disney filming a movie with these things??

    Jurassic Park: XXVII

    Well,the movies could be filmed in real time without having to wait for a render farm to create your effect!

    Come to think of it, the technology could have a couple practical uses -- mobile autonomous camera control, rough-terrain heavy equipment transport, automatic bug squisher... OK, I'm stretching, but you get the idea.

    In addition, if Disney ever decides to license or share this technology, imagine what such a robot could be used for... Earthquake rescue, anybody? Terrorist negotiations? Unmanned ice-cream vendor? (it'll be a real hit with the kiddies at the park!)

  • I wasn't able to get back on to the page to check, but I don't remember a great level of detail on the computations that are being done. Non-linear control problems associated with robotics aren't generally thought of as the easiest thing in the world. There's a lot of math behind the stabilization of non-linear systems. It's possible that they are using a fairly computationally intensive method. This would not be like slapping together a PID controller and tuning the parameters until it can walk without falling over.

    Whether they really need 700MHz processors in each leg, I can't say. But it's not a possibility you could dismiss without knowing a fair bit about non-linear control methods and what method was used in particular in this case. In any event these processors are going to be doing a hell of a lot more than "telling" the legs to move up and down regularly.

  • Finally, i can replace the "Imperial Walker" toy that mom threw out when i was ten....
  • I was thinking of cops on horses / cops on an 11,000 lb dinosaur like machine...
    But robocop good too. . .

    I have a shotgun, a shovel and 30 acres behind the barn.

  • If you put 3 transmitters in the arena and 1 on each robot, they could very easily determine their location and find the other bot.

    A really good AI should be better at disabling opponents, because it can much better coordinate the use of several degrees of freedom. The reason human drivers suck is that they are trying to remotely manipulate their bots with an incredibly low bandwidth connection (human fingers). Robots designed to have microsecond reflexes would rock.

    I do agree, at least at first, an autonomous divison would be much more fun for the participants, but much less fun for the spectators. But I think it is possible.
  • Make that 2750 (4 legs).

    The weight isn't that bad if it has big feet (and it does). Those feet are probably bigger than the contact patches of a truck's tires, so it can go anywhere a truck could.

  • Look at your leg. (Score:4)

    by Russ Nelson ( 33911 ) <slashdot@russnelson.com> on Monday March 05, 2001 @09:26AM (#383943) Homepage
    Look at your leg. See that bulgey thing below your knee, in the back? That's the linear motor that powers your ankle. The dinosaur robot has *no* motors in its legs. None.
    -russ
  • 5x700Mhz (4 legs, one main). It seems that the processors are there for "future" expansion.

    I'm sure that they designed it to be a little "overkill" too. Shit, this thing weighs 11,000 pounds, is mostly all custom made parts, etc... Another $500 is bullshit for this project.

    Oh. I'm assuming height off ground is provided by a laser system that does its own calculations.

    So I'm somewhat redeemed, if not entirely clear.

    Like I said, its monday. I need coffee.
    and wow, it is dead.
    fuck.
    we just killed discover.com, even the front page is dead.


    I have a shotgun, a shovel and 30 acres behind the barn.

  • 5x700Mhz (4 legs, one main). It seems that the processors are there for "future" expansion.

    I'm sure that they designed it to be a little "overkill" too. Shit, this thing weighs 11,000 pounds, is mostly all custom made parts, probably cost a shitload for the metal alone. etc...
    Another $500 is bullshit for this project.


    I have a shotgun, a shovel and 30 acres behind the barn.

  • 700Mhz per leg? That tells me a) they forgot to optimize their code or b) it's just really bloated. I just can't imagine it needs that much processing power to make those calculations. I hope this is a case of, "throw a lot of hardware at the problem just in case."

  • Re:A 700 MHz CPU for each leg... (Score:5)

    by Rupert ( 28001 ) on Monday March 05, 2001 @09:40AM (#383947) Homepage Journal
    Funny, but also sad. I'm sure this is a fine piece of engineering, but 5.1GB of RAM? Is it running Whistler^H^H^H^H^H^H^H^HWindows XP in every shoulder?

    With the 4x700MHz PIIIs, that's significantly more processing power than your average dinosaur brain, and they walked a lot better than this beast.

    Time to give up on this problem, chaps. AI researchers used to work on chess, because they thought they could never brute force the problem. Now they've done the same to walking. Time to move on to face recognition, or something else.

    --
  • The "Mindrover" game provides a 3D simulation of the game you'd like. You build your robots and program them, then test them against others. Robots can be distributed to other competitors either closed(binary only) or open sourced(including the logic). Game is available at: http://www.cognitoy.com/mindrover/mindrover.htm

    LetterJ
    Head Geek
  • Remove any spaces that /. inserts

    freenet:CHK@L~ijwvGY4muOIqcqPb2HSBlhxEYOAwE,Ol4HRW 4g1y0-oFhYMK0AKA
    freenet:KSK@www.discover.com/mar_01/featrobots.h tm l

  • Re:Sniff the wireless commands....AND.... (Score:5)

    by jheinen ( 82399 ) on Monday March 05, 2001 @09:56AM (#383950) Homepage
    Since the "Lone Gunmen" have shown us that it is possible to hack into the primary flight controls of an airliner in flight, hacking this bad boy should be no problem. What this thing REALLY needs is a couple of Octium IVs (tm).

    -Vercingetorix
  • I want to see two of these beasts mating. Now that would be a robotic scene you wouldn't forget in a hurry...
    --
  • How's the British version different?
  • I'm sure they would tether it to some supporting rig during testing :p

  • While this thing is nifty, it's still no match for the real mech that Mechanized Propulsion Systems is putting together

    Mechanized Propulsion Systems [mechaps.com] say they're building a real mech, full size and human piloted!!!! Check them out :-)

  • I think the phrase "potential for abuse" generally signals a troll or wry humor in any tech article here.
    ___
  • 5x700Mhz (4 legs, one main). It seems that the processors are there for "future" expansion.

    Plus, if the leg movement is based on corrective movement, a faster processor will minimize the vibration caused by minute changes in position.
    ___

  • Seriously, all humor aside, I think you'd have a hard time getting enough contestants to justify it. The only thing I can think of would be getting a number of colleges to put together teams or something. However, a number of them already compete in similar but more serious events, such as the "International Aerial Robotics Competition"

    Well I think you said it... you'd probably get plenty of entries, but maybe more from colleges and companies than individuals.. or perhaps not..I'd certainly never build an r/c bot (dull), but an autonomous one would be more fun!

    As well as the aerial robotics thing (a bit too hard, really!), there's also the autonomous robotics soccer competition - teams of opposing robots. Not only do these things track each other and the ball in real time, but they can plan intercept paths to hit the moving ball, and even manage to plot intercept paths to pass the ball to each other! Rather than having a dumb selfish "each man for himself algorithm", the better ones at least use team play. I remember reading about one team who's genetic algorithm optimized team algorithm resulted in player roles and positioning very similar to those in real life soccer!!

    There's a web site for the competitions - should be easy to find. AAAI magazine also covers the competitions.
  • Well, first of all its not just about whose bot can beat up the other guys. While that's slightly entertaining, the British version has obstacle courses and races that the bots have to get through in order to get to the deathmatch mode. Means you have to design your bot to not only be battle capable, but also have the ability to maneuver well.

    Plus Robot Wars, (what its called) the few episodes I have seen were hosted by Craig Charles ("Lister" of "Red Dwarf") who is a lot funnier than those idiots on Comedy Central. IMO.
  • Actually walking and balancing is harder than face recognition (which is old technology now anyway).

    If you want to see better walkers then hed over to the MIT web site and visit their "leg lab". This is where walking/hooping/etc robots originated. They even have one with one leg that hops up and down like on a pogo stick!

  • A number of things
    1) house robots. these things are cool, they try to get either player

    2) THe deathtraps are much cooler....can you say flamming pit of fire.

    3) better robots...dont know why, they are just better

    4) better announcers....hey they just suck less
  • Is this reviewed in Tom's Hardware Guide? :)

    .kb
  • It does sound more interesting.

    Do they show the British version on American TV?

    They had Adam Corrola (and his man show sidekick) as guests on Battlebots, and they were pretty funny, but I agree the hosts suck.

  • Ach, you maroon! Don't theorize about stuff you know nothing about. To say that dinos had significantly less processing power than 4x700MHz PIIIs is silly. It's comparing apples and oranges, and the apples evolved over millions of years and employed genetic algorithms to perfect the most simple seeming actions to an art. The oranges are layers of copper and semiconductor! Research into robotic motion is difficult. Simulators are helpful, but it's tough to model what we can realize - it's much easier to model stuff that just "works". The real-world engineering and design that's required to approximate something as simple as a cockroach is still WAY beyond our abilities as engineers and mechatronics people. Building stuff is a good way to go. You bring the REAL engineering challenges to the top of the pile. Linear actuators. Power dissipation. Balancing problems. Data fusion. Reactive neural networks. Shit man, you're dismissing a whole lot more than a model there. Dork.
  • Try http://www.robocup.org

    Or...
    http://www.mae.cornell.edu/robocup/RoboCup.html
    ...for the Robocup Champions of the Solar System from Cornell University. Go Big Red!

    These robots are super-cool!

    The Cornell link has a bunch of other autonomous vehicle info too.
  • ...in under 22 minutes, if I remember the episode correctly.
  • If challenging problems are solved, it could be let loose in theme parks to roam on its own.

    If it is supposed to give people the right of way (probably a good idea), how will one of these things get anywhere in a Disney theme park? Those places are generally packed shoulder to shoulder.
    ___

  • Yeah, they show it on some PBS, like a lot of other British shows. :) Thats how I've seen it anyway.
  • Anyone, given budget and time, can throw enough hardware at a problem (making a dinosaur walk, beating the world chess champion) to hide the fact that they haven't solved the problem. That was my point.

    --
  • It used to be on PBS, but it moved out of their lineup last year. However, they are planning a US tournament. Check out the website http://www.robotwars.co.uk/ [robotwars.co.uk]
  • > How's the British version different?

    Apparently, in the British version, the robot makers actually practice with their robots for more than 15 minutes before the show.
  • Offtopic? What an intelligent moderator. THIS post is offtopic.
  • > Those feet are probably bigger than the contact
    > patches of a truck's tires, so it can go
    > anywhere a truck could.

    I.e. cement roads, tar roads, and roads of hard-packed, dry dirt, on all of which it can carefully step over the boulders all over such roads.
  • Actually, that is significantly less than a dino brain. If you were to assume that a dino brain was capable of processing 1/100 of 1% of what we humans are, and this is reasonable... Humans have on the order of 10^15 dendritic connections in their brains, a p3 has somewhere in the 10^7 range, connections. 10^15 * .0001 = 10^11 or 4 orders of magnitude higher than a P3. I have met plenty of people who I suspect were not 10,000 times smarter than a dino.
  • That's Jimmy Kimmel, formerly the sidekick on "Win Ben Stein's Money".
  • Ya, i think someone really hates me out there. I seem to be targeted like that.

    -Jon

    Streamripper [sourceforge.net]

  • Damit, moderators when I posted this I was the only one to provide a way for everyone to read the story, note the number of the post. If fact this was at one time moderated +5 (informative). when you moderate take note of the order. Anyway i'm back to a karma of 42, incert H2G2 referance here


    ________

  • Hmmmm. 11,000 divided by 4 is 5500? My calculator suggests it's closer to 2750 per foot.

  • Re:A 700 MHz CPU for each leg... (Score:5)

    by Animats ( 122034 ) on Monday March 05, 2001 @12:24PM (#383978) Homepage
    What makes you think biology doesn't brute-force the problem? Over half the human brain, and more than half of the lower mammal brains, is devoted to balance and coordination. It's a reasonably compute-intensive problem.

    The first big insight on the problem was when Raibert figured out that balance is more important than gait. Locomotion researchers had been obsessing on gait all the way back to Muybridge, and never understood gaits beyond the walk. That's why Raibert did the one-legged hopper, which forced him to focus on balance. This provided the insight that cracked running. The basic concept is that in stance, the goal is to level the body, and in flight, the goal is to land with the foot at the "zero point" landing point which will maintain the current speed and direction. Displacing the landing point slightly from the zero point results in a turn or speed change, and that's how you steer. Very neat.

    My big insight on this is that traction control is more important than balance. I figured out (and, of course, patented [delphion.com]) how to do anti-slip control for legs. This is necessary to run on hills. [animats.com] One interesting result was that it finally became clear why legs have three joints, considering that two are sufficient to place the foot anywhere. The third (ankle on human, hock on the quadrupeds) joint gives the ability to control the direction of the contact force, which is a big win on non-flat surfaces. This is most true for animals like horses, which have hind legs with three sections of about equal length, but it's true for humans, too. Try climbing in rigid ski boots that lock the ankle joint.

    Lots of people have built walkers. It's building a runner that makes it serious.

  • kewl now I can have my own Veritech fighter, Valkyrie class complete with Gerwalk and Battloid formations.. go tech people ! :)
  • As opposed to a flaming pit of ICE, which just results in a lot of water.

    Which would work fairly well, considering their robots.
    ----
  • We can run through trial and error... there are 'monkey' robots that can learn to swing through trian and error. Falling over would ruin this behemoth, though, so it has to know how on the first try.
  • *cough* TROLL *cough*
  • This [mit.edu] is pretty neat. Found the link at the bottom of the main story. This bipedal robot is actually walking, shifting its weight and such. Quite impressive, it's attached to 2 cables but its easy to see they're not holding it up or helping it in any way as both have quite a bit of slack. Neat stuff.

    "// this is the most hacked, evil, bastardized thing I've ever seen. kjb"

  • What about guns and cars? Both of those, realisticly are more dangerous.
    But the human race hasn't been wiped out by either of those invetions.

    I think that you should be more worried about gun control and car laws at the moment.

    I don't see how someone would be alowed to just go around with one of thses things and knock people out. Someone could do the same thing with a baseball bat. Just becasue it's a new peice of technology. Dosn't mean the laws change.

    Although. you do have a good point about the future. when AI starts to get more of the I in it.

  • With the 4x700MHz PIIIs, that's significantly more processing power than your average dinosaur brain, and they walked a lot better than this beast.

    Hardly! I'm not even a biologist. But I know that probably isn't true, considering how much we don't know about our own brains for starters.

    Just becasue a dinosaur can't do maths aswell as, or as fast as, a computer, it dosn't mean it isn't as powerfull. Don't foget that a dino, or human brain has to control far more muscles than that robot. than there's the fact that it dosn't have to worry about a complex digestive system. Blood and oxgen curcualtion, high res vision, millions of touch, temperature sensors all over the body. And of course the biggy... that they can't learn in the same way as we, or dino's do/did.

    Don't forget, brains work in an entirely differnt way to a computer chip. A computer chip can only do simple things fast (0's and 1's). And that's about it.

  • A beowulf cluster of these?

    But seriously, that's essentially what the bot's control system sounds like...

    So when will we see something like this on Battlebots?
  • At last, someone who agrees with me! Geez I hate that when people refer to things like Battlebots as "robots". I'd actually like to build a Battlebot one day. But it will be fully autonomous (probably with a remote control to help "coach" it). That's the only way I could go on that show and retain any of my dignity.. What do you think? I say we start entering autonomous Battlebots into the competition. Give these pansies a run for their money. Sure, the RCs will win at first. Heck they'll probably kick our butts at first. But once we get the hang of it... a human simply can't compete. (Plus we wouldn't have to deal with the shame of going on national TV and calling our remote controlled car a "robot".)

  • Couple problems with that conspiracy theory...

    1) The cost to build one of these things would be far greater than the cost of the largest rendering farms.

    2) 3D render speeds seem to be doubling at nearly _twice_ the rate of Moore's Law [intel.com]. A frame that takes an hour to render today, will take only 15 minutes to render in 18-24 months.

    Have you seen some of the GeForce3 footage [nvidia.com]??

    It won't be long before movies like Jurassic Park will be rendered in real time at studios... and not long after that, in our own homes. But by then studios will probably have replaced most actors (and sets) with inexpensive, high quality CGI dopplegangers, thus increasing total production time again to several years.

  • Mechanized Propulsion Systems has done nothing more than tinker with legos, and claim their first prototype will be an 8m tall giant walking heap of metal constructed mostly from scrap I-beams. It looks to be either an elaborate joke (which they deny in their message board) or someone's very half-baked idea to scam the feds for research funding (which they state as a goal on their front page).

    The closest thing we have to mecha in our future will probably be a decendant of Asimo [honda.com].

  • No reason to worry about your neighbour - sending his robot to attack you would be just as illegal as shooting you.
    Governments, however, will be interested in this technology -hello BattleMechs!!
  • The reason human drivers suck is that they are trying to remotely manipulate their bots with an incredibly low bandwidth connection (human fingers).

    Look at a good pianist. I don't think the bandwidth limitation is with human hands and fingers. Radio remote controls as used in battlebots are legacy systems, designed to control model cars and kludged to control model airplanes. These applications don't involve combat and really just don't require that much bandwidth.

    I also think that microsecond reflexes are probably overkill; useful reflex time is limited by the inherent acceleration and deceleration times of the robot's parts. Even cats and mongeese get by with millisecond reflexes.

    I'd like to see somebody design a battlebot where they focus on a high-bandwidth control system rather than a bad-ass weapon. (Most of the weapons end up looking pretty lame anyway.) Video cameras are cheap these days, so no reason the operator can't where a headset that gives him a robot's-eye vantagepoint. There are analog joysticks and 6DOF controllers. Bitstreams from multiple controllers could easily [unc.edu] be shipped over a radio channel (though it probably makes sense to keep the video stream separate [x10.com].

  • AI is not advancing by leaps and bounds. In fact, I don't think there's been a really significant development in AI in years -- everyone is just working on getting their current systems doing what they want.

    People have this impression that because chips are getting faster, and more software is in circulation, the practice must be advancing at a tremendous rate. It's ain't. Advances in software are few and far between, and take decades to become widely accepted (think OO, or functional programming). The quality of software hasn't improved much, and may even have dropped some. Complexity is up, but it's mostly the result of bloat on existing systems and not the gobs of new functionality. Even hardware isn't moving that quickly if look at architectures; how long have the basic IA32, Sparc, MIPS, etc, designs been around?

    AI is, if anything, worse off. It's both a software field and a science with close ties to psychology. It's stuck with both to sad state of engineering, and the slow, evolutionary methods of science. The appearance that AI is quickly advancing is mostly the result of poor reporting and confusion between the actual state and predicted state of the art. Every time some looney writes a book about nano-AI's taking over the world, that story gets run on /. and a billion other places and people come away with the impression that we're on the verge of having little AI's. We aren't. We aren't even on the verge of big AI's. Assuming we can actually build one -- and that's still a huge if -- it'll be decades, maybe centuries, and it's will probably be far more alien than anyone gives AI credit for.

  • I am not a robotics researcher, so this is based on intuition and what I've read and seen.

    I think that one of the reasons that animals 'have it so easy' is that we've got so many more sensors, as well as the 'circuitry' to put all of that sensory data together. And I don't mean the strict inner-ear stuff, either. I mean the somatosensory (joint position and such) senses. How many robots 'know' where all of their limbs are as easily as I can know exactly where my hand is with my eyes closed? Furthermore, humans (and presumably other animals) have the benefit of learning exactly what works for their body, motion-wise. If my brain was suddenly put into the body of someone differently sized than mine, I'd probably be unable to walk for a while. I'm guessing that most robots don't have the benefit of years of experience learning how to move optimally. This kind of thing is why it needs a 700Mhz processor in each leg.

    And another thing is that, however it gets that way, the brain tissue that controls movement and the coordination is optimized, both to begin with and then fine-tuned over time. If robotics researchers could make the equivalent of a DSP for balance and coordination, I'm sure the specs would appear much more 'reasonable.' After all, a quad Pentium III system running, say, a chess program will tend to lose to a world-class human chess player, and chess is an inherently mathematical problem! Walking is so much more...well, certainly not mathematical, really.

    Anyway, those are just some of my idle ramblings on the subject. Maybe that helps, I dunno.

    (One thing I question is your notion of more than half of the human brain being devoted to coordination and balance...doing some searching reveals that, from the brain diagrams I looked at, roughly a tenth of the frontal lobe is given over to 'movement.' Also, some parts of the basal ganglia are involved with inhibitory balance, reflexes, and the initiation of voluntary movement. Coordination, by which I mean the coordination of multiple sensory inputs and movements, is the domain of the right hemisphere, but I imagine that's not all of what it does. Also, that definition of coordination seems substantially separate from simple balance. That is, it's gymnastics and playing the piano, not walking while whistling. It may take a large chunk of the human brain, but it's so much more than what robots can do now that I imagine that it would only take a small portion of it to replicate simple movement and things like rubbing your stomach while patting your head (which is a silly example, I know).

Slashdot Top Deals

Children begin by loving their parents. After a time they judge them. Rarely, if ever, do they forgive them. - Oscar Wilde

Close