Robots and walking

Discussion in 'Intelligence & Machines' started by wet1, Aug 22, 2001.

Thread Status:
Not open for further replies.
  1. wet1 Wanderer Registered Senior Member

    Messages:
    8,616
    One of the truely difficult tasks to get a robot to do is walk. Multilegged robots have a better track record when it comes to mobility. But two legged creations have a more difficult time of it due to balance, texture of the ground, and it inherent necessity of partially falling forwad to move. Here is a news article from USA TODAY on a few successes.

    Dinosaurs walk again, thanks to technology
    By Kathleen Fackelmann, USA TODAY


    Please Register or Log in to view the hidden image!



    http://www.usatoday.com/life/cyber/_photos/dinosaur-robot-inside.jpg

    It took Peter Dilworth five years to create Troody, a robotic dinosaur.
    But putting Troody together, bolt by bolt, was the easy part. Dilworth's goal was to give Troody the gift of motion. To do that, he had to give the robot a brain — a computer that would tell Troody's legs to move forward despite a changing terrain. In short, Dilworth was trying to teach Troody to walk.
    On Oct. 20, Dilworth achieved his dream. It was 3 a.m. at the Massachusetts Institute of Technology Leg Laboratory in Cambridge. Dilworth had just gotten Troody to balance on one leg. He didn't think the robot was ready to walk yet, but a last-minute hunch made him send a signal to the robot's computer.
    To Dilworth's amazement, Troody started to walk.
    "It was a slow Frankensteinian kind of walk," he says. "But it made it all the way across the desk."
    That walk put Troody in the vanguard of a small group of elite robots that can walk, run, hop and even perform tucked somersaults. As it turns out, getting robots to walk on two legs is a surprisingly difficult task.
    "Troody is the first walking 3-dimensional, two-legged dinosaur ever," Dilworth says. He's understandably proud of the robot, which at 18 inches tall and 4 feet long was built to look like a young Troodon, a vicious meat-eating dinosaur that prowled the Earth during the late Cretaceous period, about 75 million years ago.
    Far-reaching capabilities
    Troody is the prototype for a bigger, sturdier robot that Dilworth hopes to build. Such robots would give museum visitors a better idea of how Troodon dinosaurs actually moved in real life, Dilworth says.
    But Troody's impact goes far beyond the rarified world of museum exhibits. The technology used to make Troody will one day help scientists build smart wheelchairs, robotic legs or other devices that will help disabled people walk again, says Gill Pratt, director of the MIT Leg Lab.
    Tiny but ever more powerful computers will allow robotic researchers to begin fashioning such futuristic devices — possibly within the next decade, says Tucker Balch, an artificial-intelligence researcher at Carnegie Mellon University in Pittsburgh.
    But to realize that dream, researchers must first figure out how to make robots that walk just like their biological counterparts. That's where Troody comes in.
    About five years ago, Dilworth began to build Troody with an assortment of odd-shaped aluminum rods, springs, screws and other parts made to order by the MIT machine shop.
    The robot weighs about 10 pounds. Its head and neck are made of foam and painted to look like a real Troodon, which was about 6 feet long when fully grown.
    "It looks like a big chicken," Dilworth says.
    Real Troodon dinosaurs looked something like big birds, but they were no chickens when it came to hunting prey, says dinosaur expert Michael Brett-Surman of the Smithsonian's National Museum of Natural History in Washington, D.C. Troodon dinosaurs would literally run down and rip apart their prey, usually small mammals or lizards, he says.
    Taking dino baby steps
    It took Dilworth two years to put together the aluminum bolts and springs that make up Troody's body. To give Troody the brainpower to walk, Dilworth put a small computer in the robot's chest. That computer is attached by wires to 32 sensors in the joints of Troody's feet and legs. The wires and sensors, like nerve cells in the human body, constantly take note of the hardness or softness of the ground, and relay that information to the computer.
    The computer also gets information from several other sensors in the chest that tell the computer whether Troody is tilting to the side or has started to fall.
    When Dilworth works a joystick to signal the robot to move forward, Troody's computer reads that information and sends a signal to 16 electric motors that power Troody's legs and feet.
    Each time Troody puts down a foot, the computer adjusts for a variety of factors and tells Troody to take the next step. Most of the time, that means Troody can walk without falling.
    But Troody does fall — and to prevent damage, the wire connecting the joystick to the robot doubles as a safety line, Dilworth says.
    Getting Troody to walk sounds easy, but it took Dilworth one year to get Troody just to stand up. After another year, he got Troody to step side to side. Three years later, he got Troody to balance on one foot. In the end, five years had passed before the fateful October night when Troody took her first steps.
    "I was elated," Dilworth says. It may not sound like much. After all, a 1-year-old human learns to walk, usually without any help at all. Robotic researchers have long been able to build robots that walk on four legs. But it's much harder to create a machine that can balance and walk about on two legs, Balch says. In fact, Troody is one of the first two-legged robots in the world to achieve the goal of walking, he says.
    M2 is all legs
    In 1997 Japanese scientists at Honda were the first in the world to create a walking two-legged humanoid robot. More recently, Sony scientists did the same. Japanese researchers have been the leaders in building humanlike robots that walk, Balch says.
    MIT researcher Daniel Paluska hopes that his M2 robot will change that.
    Paluska's robot is a metal torso on a pair of legs, with no head — at least not yet. Just like Troody, this robot has a computer that reads information from the feet and legs and adjusts each step.
    M2 can step side to side, but it has trouble walking forward. Paluska hopes to get M2 to walk and turn reliably well. To do that, he and the other MIT researchers are racing to unravel the complex mechanism that allows humans to walk, and turn, and run without really thinking about it.
    If scientists can build a machine that can walk the way a human walks, they might be able to design devices that would replace braces or a wheelchair, Pratt says. For example, he envisions a future in which a disabled person slips on a pair of robotic pants that read sensor information and then give signals to the leg and feet to walk forward.
    As for Troody, the robot is now on display at the Boston Museum of Science. Such exhibits will help educate the public about Troodon dinosaurs, which scientists consider the smartest of all dinosaurs, says dino expert Don Lessem, president of Dinosaur Productions, a consulting firm in Newton, Mass.
    Troodon dinosaurs had a brain that scientists considered massive for the time, Lessem says. A bigger brain probably allowed Troodon dinosaurs to hunt in packs, a relatively complex task that allowed them to attack larger animals, he says.
    Right now, Troody can't solve problems or think or strategize. Dilworth wasn't even trying to give Troody such abilities. It was hard enough to get the robot to walk, he says. Yet that's clearly where the field of artificial intelligence is headed, Balch says. Could scientists one day make a robotic dinosaur with a better brain, one that could plot and carry out an attack on a mammal?
    Absolutely, Dilworth says.
    Stay tuned for Jurassic Park IV.
     
  2. Google AdSense Guest Advertisement



    to hide all adverts.
  3. Cris In search of Immortality Valued Senior Member

    Messages:
    9,199
  4. Google AdSense Guest Advertisement



    to hide all adverts.
  5. Anon1268 Registered Member

    Messages:
    9
    Robotic Walking

    I agree, a nice article, and a very fine accomplishment.

    However, I have always wondered *why* robots need to 'walk'. i can understand it from the research perspective; i.e. we want to understand how walking *works* and marvel at the complexity of a bipedal system, but truly - could not robots achieve motion in other, more machine-efficient ways?

    A robot is, after all, a machine - usually specifically designed for a task. The issue of tough terrain no longer *is* an issue if a robot is only designed for a specific environment... and multi-purpose 'travel apparatus' could follow from this; hovering is an almost perfect solution, if it were only more energy efficient.

    Im not trying to bash the article or the robot, i think its fantastic - and i know this is my first post etc, but it does seem rather blinkered to me to be trying to get robots to 'emulate' human physiology and behaviour. I guess this makes the technology more 'approachable' to the human mind...

    I don't know, I just see the 'robots of the future' being machine-like, because thats what they ARE. I agree with the premise that we have a LOT to learn from nature and our world, but I do not think that this should mean we don't respect our own thoughts and designs.. it is very possible that we have evolved to be *more* intelligent than nature, and i think this is often scorned or frowned upon.

    Sorry to rant in my first post... hi everyone! Please don't flame me to death

    Please Register or Log in to view the hidden image!

     
  6. Google AdSense Guest Advertisement



    to hide all adverts.
  7. Cris In search of Immortality Valued Senior Member

    Messages:
    9,199
    Anon1268,

    Firstly, welcome to sciforums, and ‘flaming’ isn’t a typical activity in these forums, so don’t worry too much.

    I guess humanoid style robots comes from expectations set by sci-fi novels, but I think you are correct there are more efficient and more stable methods of locomotion that could be built into robots. Achieving balance on two legs has been a major hurdle for these robot makers, yet why not use three legs in a tripod fashion, balance almost becomes unnecessary in that case. And then look at the balance and speed that 4 legs can provide as shown by most animals. And all this of course assumes that legs are needed at all.

    But I think the agenda of these robot builders is to produce general purpose robots that can co-exist with humans in human environments, and that generally means having similar dimensions and at least equivalent capabilities as humans.

    Cris
     
  8. wet1 Wanderer Registered Senior Member

    Messages:
    8,616
    Anon1268, I would echo Cris' welcome to Sciforums. Flaming is a rare thing here, usually found with the newcomers but occassionaly heads bump with the other members as each strives to make his point. But on the whole it is not a worry.

    Walking, ie. bipedal motion, is something we have been groomed to expect from scifi and Hollywood. And as far as effiecency goes it is probably way down on the list. But for reacting with humans it is best to generally resemble human characteristics or average Joe will rebuke any general contact. Futher to react with humans indirectly means being with humans and needing to go where they go. Which leads us back to bipedal motion. Ever notice how long it has taken us to equip public buildings for handicap access?
     
  9. Anon1268 Registered Member

    Messages:
    9
    I agree entirely with the sentiment that 'the average joe' will be suspicious of anything which isn't humanoid, or at least 'animal-oid' (? is that a word ?

    Please Register or Log in to view the hidden image!

    ) - but I still can't see that its a good thing, and as usual - as you both point out, it seems that it is partially the fault of hollywood and the Media that we 'expect' humanoid robots. I think that humanoid robots will *have* to be a goal for this reason alone, but I also think that it is a shame that the other options are almost not given a chance. A 'gyroscopic levitation' equipped robot (I don't know much about this technology, I just read about it somewhere when reading about Kamens (sp?) 'Ginger' - could be SF for all i know!) could be small, light, and multi functional depending on requirements (i.e. it could fetch you a beer, if domestic; it could fly down potholes for recon if military), and it could get everywhere that humans can get *and* places where they couldnt.. surely *this* is what we should be aiming at.

    I don't know, im not going to get tied up on this walking issue, as it is part of a bigger thing, which is luddism; which i suspect belongs in a different forum. It bothers me to think that there are advances which we have for all intents and purposes *already made* which are being 'put on hold' by trying to fit them round the very skewed perspective that our 'society' imposes upon us. AHH! I'll tell myself to shut up.

    Thanks for the welcome... hope I can contribute something!
     
  10. wet1 Wanderer Registered Senior Member

    Messages:
    8,616
    I suspect that with these two posts you already have started on the way to just that.... Contributing.
     
  11. Cris In search of Immortality Valued Senior Member

    Messages:
    9,199
    Ginger and IT

    Anon,

    Ha ha, Ginger, or better known as IT, has not yet been revealed. Kamens is a recognized scientist and has invented a number of things, like medical stints, and something else that is well known of which I can’t remember at the moment. He is currently demonstrating a 6 wheel wheel chair that can climb stairs. But for IT he just says that we have a good product but it is not as fantastic as the rumors imply. The best guess, based on a view of a recent patent submitted by Kamens is that IT is some form of small, perhaps 2 wheel personal vehicle.

    Cris
     
  12. Cris In search of Immortality Valued Senior Member

    Messages:
    9,199
    Anon,

    As for humanoid robots. I agree, that once our imagination is let loose with adequate AI software with adequate computing power, then the variations for robot forms will be unimaginable. Specialist in many areas, be able to go places that humans cannot, or with power way beyond human strength. Flying, hovering, small, large, multi-limbed (e.g. greater than 4), multi-eyed, etc, etc,. Why limit ourselves?

    The future looks like being really fascinating.

    Cris
     
  13. wet1 Wanderer Registered Senior Member

    Messages:
    8,616
    Prehaps you had in mind something like this;

    From Newscientist:
    A droid for all seasons
    No job is too unusual for a breed of robots that reinvent themselves
    IMAGINE AN AUTOMATON that can design itself, assemble itself and even kill itself. No, it's not the liquid metal robot from Terminator 2-but this droid can certainly build itself to perform a particular task, melt itself down and recycle itself, say researchers in Massachusetts.

    The scientists have developed what they call a polymorphic robot-a machine that can change its shape to suit the job in hand. Shape-shifting robots could be used as planetary explorers, or for search-and-rescue missions, changing their shape to meet each new challenge and adapting to strange and unpredictable environments.

    The researchers have produced a simple thermoplastic-framed robot, says Hod Lipson, who developed it with Jordan Pollack at Brandeis University, near Boston. "All the robot has to do is find a way to move," he says. The prototype is very basic and has no sensors, so it is unaware of the world, though sensors could be added at a later design stage.

    The idea is that a task will be set for the robot, such as: "Figure out how to move using only one leg and one motor." A computer will then attempt to design a body that will help it to meet this challenge most efficiently. At present, the robot's body is built using the "rapid prototyping" technology common in the car industry, which can produce complex three-dimensional structures very quickly. A device called a 3D printer uses a nozzle to build up progressive layers of thermoplastic, slowly creating the required structure.

    Although 3D printers Polymorphic robots that reinvent themselves are large and cumbersome, says Lipson, much smaller ones could one day be built into a robot, allowing it to change parts of its body, for example, to reshape an arm to produce a new tool for a novel situation. Mark Yim of the Xerox Palo Alto Research Center (PARC) in California says this is one area in which polymorphic robots could be most useful. There's no point in taking an entire toolkit into space, he says, when you don't know which tools you'll need: a single robot arm can be shaped to do the job of all of them.

    It is also conceivable, says Lipson, that the 3D printing technology will allow several materials to be printed, including conductive, nonconductive and even semiconductive materials. "Wires, motors and logic circuits, as well as structure, could be printed in one pass without the need for assembly," Lipson predicts.

    With each new task, the look of a polymorphic robot is impossible to predict, because each design is "evolved" using a genetic algorithm. The physical structure, and the neural network that will be the brains of the proposed robot, are treated like genetic information that can be combined and mutated in simulation to produce entirely new designs. The "fitness" of these offspring is then evaluated and the best are "bred" to produce more offspring. This process is repeated many times until the design has evolved to do the best job.

    To keep things simple, Lipson allowed the algorithm only basic components with which to design the prototype robot: straight plastic bars of varying lengths and electric motors that can extend or shrink the length of a bar. Joints are all ball-and-socket designs, as these are easily created by a 3D printer. From these basic parameters a host of complex-sometimes lifelike-structures have been evolved.

    Some versions push themselves along on one leg, while others produce a hinge-like motion and crawl about like a fish out of water. Yet another moves sideways like a crab. "The robot is ready to move when it comes out of the printer," says Lipson. Its motor, however, must be inserted by a person. But the aim is to make the robots totally independent, much like the vengeful shape-shifter in Terminator 2.

    When the robot has performed its task, it offers itself up to be melted down, so its thermoplastic components can be recycled into another useful droid by the 3D printer.

    The idea of building and melting down robots is novel, says Yim, who makes modular robots that reshape themselves by fitting smaller robots together. "I've never seen anything like it." But he warns that to make truly useful robots, stronger plastics and more materials are needed.
     
  14. Cris In search of Immortality Valued Senior Member

    Messages:
    9,199
    wet1,

    Awright, that's what I mean by imagination. Neat stuff.

    Cris
     
  15. Anon1268 Registered Member

    Messages:
    9
    Yeah, thats what i'm talking about...

    cut out the human factor, which by inference cuts out the luddite factor.

    I remember reading that article and seeing some pictures on the web. Pretty good ideas.. why aren't we doing more of it?

    i say 'Wind 'em up, and let em go'.. and if they turn round and attempt to destroy us, thats what happens.
     
Thread Status:
Not open for further replies.

Share This Page