Slashdot Mirror

← Back to Stories (view on slashdot.org)

Robosapien: Latest Toy Robot From Mark Tilden

Posted by ryuzaki0 on from the soldering-skills-essential dept.

Onnimikki writes "Mark Tilden has been building really cool BEAM robots for a long time. Now, he's come up with RoboSapien, a toy that no self-respecting geek can go without. Videos of the RoboSapien at the 2004 New York City Toy Fair have been made available by Solarbotics. Mark offers some really good explanations about what makes them work."

13 of 181 comments (clear)

  1. Popular science quote by t0qer · · Score: 2, Insightful

    At the bottom of the page they had 4 links, the one to popular science had this to say.

    the 14-inch-tall RoboSapien, which will retail for about $80 when it hits stores later this year, uses analog transistors to react to signals from the world around it.

    How is this different from the aibo?

    1. Re:Popular science quote by kertong · · Score: 5, Insightful

      it isn't $600. :)

  2. Mirrors? by chris-johnson · · Score: 1, Insightful

    Can anybody who manages to get to the page make a mirror, maybe even .torrent's for the videos? It's been maybe 5 minutes and hasn't loaded for me yet.

    --

    <wik>/bin/finger that girl in the back row of machines.
  3. This is not especially interesting by Anonymous Coward · · Score: 5, Insightful

    Pray tell, why is this an especially interesting development? This toy is basically a remote-controlled device. Far from a "robot", like AIBO or QRIO which actually have autonomous capabilities and can decide to do things on their own.

    I move to strike the word "robot" from any device that is not autonomous in some fashion... :)

    1. Re:This is not especially interesting by EtherealSys · · Score: 4, Insightful

      In my opinion, this is an exciting article, not because of the robot's capabilities, but because of its price tag. This is bringing some pretty sophisticated robotics to a completely different market than the AIBO or QRIO. If these motor skills can really be done at such a price, there's no reason why we shouldnt see pretty dramatic drops in the prices of the more expensive tech toys.

      --

    2. Re:This is not especially interesting by kfg · · Score: 2, Insightful

      Is a programable industrial robot a robot?

      KFG

    3. Re:This is not especially interesting by SWTP_OS9 · · Score: 2, Insightful

      Actualy human are basicaly an analog unit. Good concept of using analog which in motor control works better than trying to do it digitaly.

      If this thing runs! Then pitty the poor cat population! { Evil laughter goes here! }

    4. Re:This is not especially interesting by kfg · · Score: 2, Insightful

      Well, it's an interesting opinion, although it differs from the dictionary definition. Nor am I entirely certain your conclusions about the toy robot are true. I don't see why adding sensors and a microcontroller to a fully functional mechanical man wouldn't create an autonomous mechanical man, or the difference between this mechanical man's controller being remote from himself and the industrial robot's controller being remote from itself.

      I could add such sensors and microcontrolers to my R/C car easily enough and your definition seems to rely on being able to add things to a machine to make it a robot.

      You'll also have to hold my hand through the micro scale control of movements issue. I'm at a loss as the to relevance of that.

      I would hold that my analog thermostat, which has a sensor and reacts autonomously to enviromental factors to perform a useful mechcanical funtion is a robot, although lacking a microcontoller, or any electronic way of modifying its behaviour.

      I take it you would disagree with me, and also hold that BEAM robots are not robots if they lack a microcontroler?

      KFG

    5. Re:This is not especially interesting by randyest · · Score: 2, Insightful

      An analog motor controller is usually a closed-loop feedback control system consisting or resistors, capacitors, inductors, and amplifying transistors. Which is not really the forte of ASICs. I'm not sure why you're under the impression there's an ASIC in the Robosapien -- I didn't see that anywhere, and I read every link I could find. I guess there could be, but that'd be overkill -- a simple PIC running a very simple state machine could handle this:

      Using the ergonomic remote control, you can command Robosapien(TM) to perform up to 67 pre-programmed functions including pick-up, throw, high-five, whistle, dance and three different karate moves.

      Robosapien(TM) is fully programmable. He can perform a programmed chain of commands in any combination of moves that you select. For example, you can create your own dance sequence or program him to walk straight, turn left and give your buddy a high five.

      The cool thing about this isn't the complexity of the system, or it's ability to be as autonomous as even a Roomba, rather its lifelike motion accomplished by very fast, very cheap (but not at all flexible or extensible) analog closed-feedback loop control circuits.

      Analog circuits, BTW, are much harder and more expensive to design than digital logic. Most cheap PICs and simple controllers include no such analog circuits (just A/D and/or D/A if you spend a bit more). Certainly not the highly custom and tuned ones needed here. ASICs also cost more for custom analog cores. Your ASIC vendor may give you a stock, commonly-used analog core for free (PLL, DLL, D/A, A/D, etc.) but you'll pay for custom layout, at least one test chip, and lots more per part to design your own RLC/amplfier circuit into an ASIC than just using the stock cellbase or gatearray digital logic.

      And, without being even more expensive, the 1% accuracy you quote for passive component variations on cmos (the cheap) process is way low. More like 10-500%, especially for the relatively large (~kOhm, mH, and mF) R, L, and C values needed for control systems. It's easier to make really small R, L, and C's on cmos, but the accuracy is poor unless you pay a lot more than $80. Moreover, the amplifiers you need are too strong for cheap cmos processors -- why buy enough die for tens of thousands of tiny weak transisors optimized for on/off operation (which is the smallest feasible die to make these days) when you only need a few strong ones configured to amplify? Discrete would be cheaper and more precise/tunable in this case.

      --
      everything in moderation
  4. Re:World domination robot by sydlexic · · Score: 1, Insightful

    We would of, if Halliburton made 'em.

    there's more truth to that statement than you can possibly know.

  5. Grand & Tilden by Usquebaugh · · Score: 3, Insightful

    I'm sort of an amateur AI/AL person, unlike the MIT clowns I admit to it :-)

    There is a great deal in common between this and the game/work of Steve Grand. Steve has started to work with robotics and I think this a mistake. He could have taken his software to the next level.

    Both Grand and Tilden feel that you can create life with very simple processes. You do not need to spell about how something is to behave but what something is. This is a fundamental change from the traditional AI/AL approach.

    The exciting thing is that the approach of using simple processes is paying dividends. Where Grand might explain conciousness, Tilden can explain physiology.

    Where is computing going in the future, take a look at the work of these two gentlemen and see for your self.

  6. This is _VERY_ interesting.... by Usquebaugh · · Score: 5, Insightful

    The real world can be analog and digital at the same time. It's red not blue. it's a sort of pinkish red. Is time discreet or continuos? What do you mean you do not know!

    Fixed, what is fixed? There are a lot of fixed values in the human body. In fact most of the body is based on very fixed processes. Feed back, is a very fixed response. The complexity comes with the sheer number of feedback systems working in parrallel. We cannot model this complexity with a pre-programmed system, but it may be possible to simulate the feedback and then set those loose to model the system.

    Have you _EVER_ worked with a digital robot, adding a new senosr is not easy? Adding a new response is not easy. In fact this is one of the main stumbling blocks of digital robots. Everytime you add a new sensor you have to explicity program for it. That means the robot is limited by the imagination/time of the designer.

    In response to your last paragraph, take a look at beam robots. See how they can do tasks with a few components that complex digital robots cannot. See how they deal with component failures. Think about how this ties back to nature. See that tieing into a feedback circuit is easy, but ultimately unpredictable.

    This whole area is opening up after 50+ years of going in the wrong direction and achieving only predictable systems. AI/AL is embracing simple systems that combine automatically to implement complexity.

    Read Stephen Wolfram, Steve Grand and Mark Tilden. All three are showing that unpredictable complexity can be modeled by designing simple feedback systems and then letting them interfere with each other. Chaos theory is the underlying mathamatics.

    To cast aside this arena as just a cheap toy is to be blind to the sheer scope of the undertaking.

    Orville, Wilbour put down that paper plane it's just a toy.

    1. Re:This is _VERY_ interesting.... by SnowZero · · Score: 2, Insightful

      In response to your last paragraph, take a look at beam robots. See how they can do tasks with a few components that complex digital robots cannot.

      ... and see all the things complex digital robots can do that beam robots cannot. I'll stop working on traditional robots when someone wins with a beam robot in RoboCup. I'm not holding my breath.

      It's kind of like saying launching satellites is trivial because you can build cheap and simple model rockets. Or this: O(n^2) algoritms are usually a lot simpler conceptually than the O(n log n) ones, but the simple ones don't scale.