Brainless bots exhibit swarming

Harvard University researchers show how simple, brainless "bristle-bots" (like those you can make yourself or purchase for $6 as a "Hexbug Nano") exhibit swarming behavior when contained in a small area. According to the scientists, when this kind of behavior is seen in the natural world, among termites for example, it's "linked with insect cognition and social interactions. Our study shows how the behavioral repertoire of these physically interacting automatons controlled by one parameter translates into the mechanical intelligence of swarms." "Swarming, swirling and stasis in sequestered bristle-bots" (PDF)


  1. If you enjoyed that, check out this video of almost-as-dumb robots doing some useful work. This would work with the bristle bots too. 

    1. Now THAT is some sweet emergence.  About three orders of magnitude sweeter than the original b-bots-in-a-box video.

  2. The bots are confined.  If there were not confined, as is the case with real swarming creatures, the bots would spread out randomly and not exhibit any type of swarming.  I honestly don’t think anything interesting or useful was gathered from this experiment..

    1. So you’ve tested it then? 

      It’s an interesting demonstration of a previously unknown phenomena. I’m sure there’s some mathematician out there who’s seen this and is already trying to come up with a formula or algorithm to describe the  behavior. If that ends up happening, then the experiment was certainly useful.

      1. I’m not sure there is any ‘behaviour’ to document. They are sticking together in groups because when a few of them get together they are confined in space by each other. There’s no group intelligence emerging here, as with animals, and as implied by the video.

      2. It’s not previously unknown, or at least it’s very similar indeed to several things that have been done before, and the differences are not very significant. Compare with the video I posted above, for example, which would work exactly the same with these little robots and is probably a more surprising effect. 

        The Harvard video uses images of flocks to motivate the work, but this is not flocking behaviour as Dan correctly points out below, along with pauxide above. The authors must know this, and I think it’s rather dodgy to draw the comparison. They could have used images of animals all packed into a circle which would be more analogous, but they chose not to.  It’s also rather surprising where they chose to publish the results, since there is a journal of Swarm Intelligence which would be the natural home for this stuff, edited by people who know the field very well. 

        The work is fine, don’t get me wrong. The analysis in the paper is very good. But the presentation and venue are…. odd.

        (credentials: I’m the guy in the box-pushing video I linked above).

        1. Is your box-pushing demo qualitatively different than random vibration of disks/spheres confined on a plane and their clustering with near hexagonal close-packing?

    2. Pauxicide… come on, dude! Those automata are “…sens[ing] the environment
      non-locally via the eff ects of confi nement and substrate topography.”  If you persist in making common-sense observations and writing in clear English, you will go nowhere in the world of contemporary science.  Nowhere!

  3. I’m coffeeless and therefore inarticulate, but there’s something about making a link between mechanical vibration-induced patterns and biological behavior that seems a little hand-wavy to me.

    1. So, you’re saying that biological behavior occurs without motion?

      Or are you asserting some sort of intelligence is behind it, so it is qualitatively different?

      Sounds a little hand-wavy to me. We’ve been discovering for years that so many complex biological behaviors are also exhibited by mechanical systems. Which makes us ask how much intelligent behavior is actually behind those “complex” behaviors, and how much they are emergent patterns.

      As someone pointed out above — everything is constrained. Birds are constrained by topography, wind currents, weather predator avoidance, food-seeking, etc. Flocking may be nothing more than a combination of these behaviors, or it might be something more. For a long time, we all thought it was obviously something more, a gray-matter, hand-wavy brain making some sort of bird-brain decisions.

      But it might not be.

      Same thing for us. Laughing at cat videos could be a sign of high intelligence, or it could be an emergent behavior from a complex, but mechanically and chemically-determined system.

      OR it could be a sign that something can do an awful lot of complex things, and not require awareness or intelligence. And might even be better at it than we are.


      EDIT: Chris Palmer, below, sums it up much better than I did.

  4. It’s the part where they start to exhibit blasphemous worship of Zcerneboch that really creeps me out.

  5. I think anyone who ever tried to play one of those vibrating football games that were popular in the early ’70s has seen this “swarming” behavior before.

  6. the other problem with this theory (besides it only works when confined) is that the swarms you see in nature mostly DON’T bang into each other randomly as a method of regrouping (maybe mites do, but not birds nor mammals)

    1. Actually, if you’ve ever coded any swarming behavior programs, like boids, one of the “senses” the boids have is proximity to one another. Basically, they try to stay close to each other (usually with a forward vision – “go toward others that you see”), but not too close (“If you get too close, turn away). With those two senses, combined with some random walk algorithms, they behave remarkably like fish or birds. 

      In this experiment, the constrained environment is substituted for the “stay close together” motivation and the physical touching is substituted for the “don’t get too close” motivation. So while this result may not seem earth shattering, it is an interesting physical model of complex group behaviors in extremely “dumb” individual agents.

  7. I thought it was very interesting, especially as someone who used to work with genetic algorithms to get interesting life-like behaviors out of  tiny robots (Kheperas), both real and simulated.

    However, did they need to make the video so damn serious? The piano music and slow talk-over? It’s an interesting thing — make it interesting! If you didn’t listen to the content, you’d assume it was some heart-wrenching documentary about beggar children selling their feet for food.

  8. I was disappointed at the end that they hadn’t modelled the behaviour of Daily Mail readers and drawn the parallels between brainless minds and brainless bots.

  9. To me, what’s interesting isn’t the supposed intelligence of the mechanical swarm, but the mechanical nature of organic clusters. It surprised me they were quoting Turing, I mean c’mon.. obviously the bristle bots have no collective hive mind.   But the inverse is an eye-opener.. that natural swarms of fish/birds/people aren’t so much a hyper-aware negotation of personal space, but a more mechanical function of animals-as-pinballs. That’s interesting.

    As for the confinement argument, yeah that’s valid, but organic clusters experience all kinds of natural barriers and influences. Flocks of birds/fish/cars might be influenced by any number of air/water currents, traffic controls, danger avoidance, food or sex drive, you name it. We’re animals after all, and not free radicals drifting in a void.

    So one could say we’re teleologically driven individuals, with like needs in groups, affected by similar motives and constraints, resulting in movement patterns which resemble sensory interconnectedness, but might be better described as mechanical boundary response. That’s what I say, anyways.

  10. A more general look at why swarming is worth studying here: 

  11. I’d be interested to see the same experiment with the same parameters performed in differently-shaped confinement spaces.  Does the so-called “swarming” persist, or is it a function of the circular confinement?  Does the “critical number” change? If “spinners” and “walkers” both end up rotating around the edges, what if you combine the two populations?  It seems this is an incompletely-explored idea presented as a conclusion.  It’s not a boring idea, but I think the investigators stopped as soon as they had something that looked vaguely reportable.

Comments are closed.