Become a fan of Slashdot on Facebook


Forgot your password?
News Technology

Artificial Retinas Can Balance a Pencil On Its End 165

mikejuk writes "A team of researchers has built a neural information system that is good enough and fast enough to balance a pencil in real time. If you think it's an easy task, try it! The Institute of Neuroinformatics, ETH / University Zurich have used what look like video cameras to do the job but in fact they are analog silicon retinas. They work so fast that even with fairly basic hardware they can balance a pencil."

This discussion has been archived. No new comments can be posted.

Artificial Retinas Can Balance a Pencil On Its End

Comments Filter:
  • This is impressive bit of tech. Robotic vision has historically been a tough field.
    Anyone knowledgeable enough on the subject to speculate on the implications or interesting uses of this technology?

    • I don't think this capability is new. There is a fairly old video of a robotic pogo stick bouncing around a lab somewhere. That must be more difficult to build. The descent engine of the Apollo LM used vectored thrust to damp out power used by the attitude control thrusters. I doubt there was much computing power behind it but it balanced the entire LM on top of one engine all the way to the surface. That was using mid-1960s technology.

      • Re:Amazing (Score:4, Insightful)

        by icebraining ( 1313345 ) on Sunday January 23, 2011 @07:02PM (#34976504) Homepage

        That's not really the point, is it? None of those were accomplished using machine vision.

        • But the tricky part of the machine vision problem seems to be the sensor design. Doing this with straightforward machine vision doesn't seem hard either. The background is static and the stick is moving. Plenty of industrial systems use vision to work out which direction a stick is pointing.

          • But the tricky part of the machine vision problem seems to be the sensor design. Doing this with straightforward machine vision doesn't seem hard either. The background is static and the stick is moving. Plenty of industrial systems use vision to work out which direction a stick is pointing.

            Yet the video is titled "Pencil Balancer on Changing Background" (emphasis mine).

          • The background is static and the stick is moving.

            It's not recognizing the pencil that's impressive, it's the response time - the super-low latency required to balance the pencil in real time. I'm not certain it's unprecedented, but it's definitely not something you could accomplish with a webcam.

      • The "Pencil Balancer" video was posted on youtube back in 2008, so yes this is not new
    • Re:Amazing (Score:5, Interesting)

      by pitchpipe ( 708843 ) on Sunday January 23, 2011 @06:54PM (#34976456)

      This is impressive bit of tech.

      This [] really impressed me! No artificial vision involved, but awesome nonetheless. Explanation []

    • No, but if you take this, the flying little drone that can fly through moving hoops, the little drones that built a structure, the advances in "thinking AI" the cameras and processing system for the prison cameras that can read moods and draw conclusions from body language and associations of different people, what we have is the beginning of the end of humans. I for one welcome our soon to be arriving self-aware mechanical overlords.
    • by mcgrew ( 92797 ) *

      As someone who's undergone a vitrectomy [] because of a detached retina, I was disappointed in TFA. Since I underwent the surgery I'm scared shitless of it happening again.

      But TFA had little to say about the artificial retinas or the hardware or software that it used to balance. Are these "artificial retinas" for robots or for people? If they're for robots, how can they be called retinas? Can someone explain this to me, or point me to a less shitty FA? TFA didn't really say anything that wasn't in TFS.

      Note --

  • Video Date: (Score:5, Informative)

    by Shikaku ( 1129753 ) on Sunday January 23, 2011 @06:30PM (#34976282)

    September 26, 2008

  • ... that balancing such an object requires the use of several fancy algorithms:

    This is proof that, just exactly as I asserted, all you need is relatively simple feedback as long as it's fast enough.
    • by artor3 ( 1344997 )

      It does require "fancy" algorithms. It's just that those algorithms are well known. See: inverted pendulums [].

      • by Jane Q. Public ( 1010737 ) on Sunday January 23, 2011 @06:51PM (#34976442)
        Wrong. Read TFA, and if necessary read their paper, and try again. They used a relatively simple feedback mechanism and simple algebra, not Lagrange equations.
      • Pardon me. If you want to be really precise, they used trigonometry, not just algebra. Even so, the equations are simple derivations of angular position.
        • Well, it's all control theory []. If all you want to do is something stupid and trivial like balancing an inverted pendulum, then the math isn't too hard and the algorithm is comparatively simple. If you want to do anything more complex, then you have to start using more complex math.

          It's not so hard to turn the 'inverted pendulum' into a more complex case where simple trigonometry and algebra doesn't work: mount your pendulum on a turntable.

          • But that's separate from the point I was making, which is that the inverted pendulum problem, without further complications, does not require advanced math to solve if your feedback/control loop is fast enough.
            • by SomeKDEUser ( 1243392 ) on Monday January 24, 2011 @09:46AM (#34980322)

              the PID algorithm is four lines of code. the RST also. But to prove the properties of either, you must understand Z-transforms (which really are Laplace transforms for sampled functions).

              You can tune your PID using Ziegler-Nicholls, and that requires absolutely no knowledge of maths. To tune it optimally, you need a very good physical model, and pretty involved maths.

              So I don't know what the other guy's argument were, but you might have been both perfectly right.

              • My point was that complex processing is not necessary. You, on the other hand, are arguing that it's not difficult to implement complex processing. And that may be true, but it's irrelevant to the point I was making.

                Yes, a PID Controller [], which they did not use by the way, is complex, in that it performs integration and several other functions. I don't know what programming language you are proposing, but I don't know of many un-embellished languages that will do integration as well as the PID's other fu
      • by by (1706743) ( 1706744 ) on Sunday January 23, 2011 @07:02PM (#34976502)
        If you have fast enough feedback, then the displacements (and hence, angles) involved in the equations of motion for the inverted pendulum are really small, and hence the transcendentals involved can probably be approximated by the small angle approximation -- and then the "algorithms" (solutions to the equations of motion) are pretty simple I think.

        Also, I'm not sure that standing a pencil on end is the same thing as an inverted pendulum, because the bottom end isn't secured for the pencil (correct?). For an inverted pendulum, if you oscillate the base fast enough the pendulum will remain upright (see the wikipedia article you linked) -- so it's pretty trivial to stand a pencil on end in this fashion, I think (just attach it to something that oscillates -- a speaker will probably do).
    • by pz ( 113803 ) on Sunday January 23, 2011 @10:17PM (#34977590) Journal

      ... that balancing such an object requires the use of several fancy algorithms:

      This is proof that, just exactly as I asserted, all you need is relatively simple feedback as long as it's fast enough.

      Yep, that was me. I guess I should go back to my MIT professors and let them all know that they're full of hooey. I've sure been shown up by Jane Q. Public!

      Or, on the other hand, I could look at the video these fellows provided. Doing that, I might notice that the system is barely stable, very noisy, does not deal with perturbation very well, and accumulates error. I could then read the paper and see, under the section called "VI Control System" it explicitly states that they are using a PD system (proportional and derivative), as described in the system of two differential equations. Then I could read the sentence, "Our system normally balances an object for several minutes before losing it..." which would probably be because they don't have an I term to worry about accumulating error. Lack of an I term makes the system drift, and you can see in the video that it nearly hits the edge of the actuator workspace a few times. Striking the limit of motion would be a catastrophic change in actuator impedance and cause the pencil to be dropped. The fact that they had to include a D term means that there is more than just straight (linear) feedback. But, hey, I guess those MIT professors didn't actually know what they were talking about when they taught us 18.03 Differential Equations. Either that or Ms. Public can't read papers very well, and doesn't recognize a differential equation when she sees one.

      Again, I'll state, Ms. Public, please stay away from designing any systems that are critical to support or protection of human life. You have now repeatedly demonstrated your incompetence to do so in a public forum.

      • Unless you qualify a PID control system as a relatively simple feedback system, particularly when applied to a linearized system (which is what appears to be the case -- extremely fast observation and actuation make it easier to make a small angle assumption). PID controllers are something you can learn in an undergrad control class. While a complete understanding of their behavior and the art of designing one for a given system can be very challenging, the concepts are straightforward. I'd consider a no

      • PID control systems *are* simple. I could build the controller one out of the junk sitting in my toolshed.

        Hell, you can trial-and-error one into working in about half an hour, once you've got the test rig set up. It's not like there's that many variables to control.

  • by Culture20 ( 968837 ) on Sunday January 23, 2011 @06:32PM (#34976296)
    My friends and I used to do that back in middle school to pass the time. Not build artificial retinas; balance pencils on our fingers (on eraser or tip).
    • Yes, I have a bit of a problem with one of the linked articles' assertion that it is impossible for a human being. I know people who can do it reliably, and I have done it myself, albeit only for a short time.
  • Just to get one thing straight: A robot balancing a pencil is not a breakthrough. Similar tasks are standard textbook material, often implemented using fuzzy logic.

    But the way they have done it may or may not be cool. Hard to tell.

    • This is a standard class project at MIT, except that they use a broomstick, not a pencil. MIT also teaches that it takes fancy algorithms to properly do the job. The difference here is that they are using a pretty simple feedback mechanism, without fancy algorithms. It's just very fast.
      • I totally agree that the control aspects of this aren't too hard, but I wanted to clarify a few things:

        1 - The natural frequency of the pendulum goes up as 1/sqrt(length), so a pencil is more difficult than a broomstick.

        2 - In the controls community, fuzzy logic would not be employed to do this. Rather, one would probably use an "energy-pumping" swingup controller to get the pendulum near the critical point, and then a local, linear (maybe with feedback linearization) controller to stabilize the critical p

        • Re:Easy task (Score:4, Informative)

          by Jane Q. Public ( 1010737 ) on Sunday January 23, 2011 @10:22PM (#34977612)
          That is precisely my point. To clarify my own statement a little: the last time I was discussing this topic here on /., someone (I don't remember who) was trying to tell me that it was not possible to do this with a relatively simple feedback mechanism, no matter how fast, and that in fact it was necessary to use Lagrange equations [] as linked to there, or similar, to solve the problem.

          My argument was that using advanced math was not necessary, as long as your feedback and control loop was fast enough. This experiment seems to bear out my side of the argument, since according to their paper they did not use anything beyond what might be considered middle-school math in their solution.
  • 3D vision (Score:2, Insightful)

    by Anonymous Coward

    Unless I'm mistaken, from looking at the picture the camera's "eyes" are placed orthogonally, instead of side by side like a human's. That's an advantage, since we know the machine then has real 3D position info, as opposed to a human's stereoscopic 3D vision. Try it yourself: when you balance a pencil, do you fail more often sideways or towards and away from yourself?

    This is an impressive bit of controls engineering, but let's not compare apples to oranges: the machine is designed for this task, and the hu

  • by Paul Fernhout ( 109597 ) on Sunday January 23, 2011 @06:37PM (#34976334) Homepage

    Similar: []

    It uses high speed visual servoing to dribble a ping-pong ball and to toss and catch a cell phone.

    Ironcially, I am listening to President Obama's speech as I write this, and his advisors (and speech) seem clueless about the changing nature of economics given robotics and other automation, AI, better design, and voluntary social networks (even as I think he means well and it is good for the US that he his helping create some jobs by increasing some exports): []
    Pres. Obama can talk all he wants about "winning a global competition", but the average human worker anywhere is not going to win a competition with advanced robots... Humans need to learn to "cooperate", not "compete".

    Economic solutions (my comments): []

    From a comment I posted yesterday in relation to an (purported) demo of a cold fusion device: []

    In brief, a combination of robotics (and other automation, all made possible by cheaper computing), better design (whether from cold fusion devices or thin-film solar panels), and voluntary social networks (especially with volunteers cooperating through the internet on free and open source digital public works), are decreasing the value of most paid human labor by the law of supply and demand. Cheaper energy will only accelerate this trend, since often you can substitute energy for labor and thought.

    At the same time, demand for goods and services is limited for a variety of reasons. These reasons include some classical ones, like a cyclical credit crunch or a concentration of wealth (with that concentration aided by automation, intellectual monopolies, and the rich getting richer and buying up more and more resources like land for rent seeking). The reasons also including some heterodox alternative economics ones, like people moving up Maslow's Hierarchy of Needs as they get a lot of "stuff" and move on to other pursuits than materialism (including spiritual aspirations, self-actualization, and social connections in communities), and as people embrace a growing environmental consciousness of "Reduce, Reuse, Recycle" to protect the biosphere.

    In general, mainstream economists ignore these issues or have very unexamined beliefs about them. Imaginative innovation, like economist Julian Simon talks about in "The Ultimate Resource", makes possible many wonderful potentialities if we think them through. Please don't let your inventiveness or cold fusion get blamed for any issues caused by unimaginative scarcity-based economic models held onto with almost a religious fervor by so many (see "The Market as God" by theologian Harvey Cox in the Atlantic). Mainstream economist have long used such scarcity-based models to apologize for an overly hierarchical social order that we probably did not even need in the past -- search on "The Mythology of Wealth". Still, some degree of centralization can be a good thing; see Manuel De Landa on "meshworks and hierarchies", and how they keep turning into each other and how all real systems are mixtures of both. So, we need to think and experiment regarding ways to allow our 21st century society to function in a healthy way given all the 21st century technology people like yourself are busy creating in all sorts of areas.

    A New York Times article called: "They Did Their Homework (800 Years of

  • Don't they already do this sort of thing using broomsticks and artificial neural networks? Its basically the same thing isn't it?

    • by Ksevio ( 865461 )

      As TFA mentions, the larger the object, the slower the reaction time is needed.

      So balancing a pencil is more impressive than a broomstick because it requires quicker reactions.

    • Try balancing a broomstick on one finger, now try the same thing with a pencil.
  • But... (Score:5, Funny)

    by 93 Escort Wagon ( 326346 ) on Sunday January 23, 2011 @06:45PM (#34976390)

    Can those fancy algorithms make a pencil disappear?

  • by MichaelSmith ( 789609 ) on Sunday January 23, 2011 @06:48PM (#34976422) Homepage Journal

    It never seems to be able to damp down the movement. It should be able to reduce amplitude to less then a centimetre or so.

  • by drewm1980 ( 902779 ) on Sunday January 23, 2011 @07:03PM (#34976516)

    I have seen this demo in person and chatted at length with its creator. It uses a custom sensor chip that does some analog temporal filtering and thresholding of light intensity at each pixel, sending events when the threshold is crossed. The intent of the authors seems to be to mimic the human visual system in silicon, even if it makes no engineering sense whatsoever. The demo was extremely sensitive to fluorescent lighting; the author had to run out and buy an incandescent desk lamp to get it to work at all. The event-based image representation makes it incompatible with everything that has been learned in computer vision over the last decade.

    • by Anonymous Coward on Sunday January 23, 2011 @07:17PM (#34976584)

      In case anyone misinterprets your comment.. The fact that it is incompatible with the last decade of computer vision doesn't make it wrong, nor does it make the previous decade of research in computer vision wrong. As you wrote, different philosophies behind the solutions.

    • > The event-based image representation makes it incompatible with everything that has been learned in computer vision over the last decade

      I think you are grossly exaggerating. The so called "events" are simply difference images, and they have long been used for representation of motion where accurate representation is not needed and computational power is scarce.
  • but the eyes are at right angles to each other and so far apart :)

  • Ow! (Score:2, Funny)

    by Greyfox ( 87712 )
    I tried balancing a pencil on my eyeball, and now I need a new retina. Perhaps these guys can sell me one...
  • by lennier ( 44736 ) on Sunday January 23, 2011 @08:39PM (#34976972) Homepage

    Stop playing with that pencil. It's all fun and games until someone loses a silicon retina.

"Yeah, but you're taking the universe out of context."