Dog Pod Grid One Step Closer To Reality

What’s better than one amazingly acrobatic quadcopter? How about a swarm of acrobatic micro-quadcopters? It’s not a rhetorical question, but an experimental reality. A team at the University of Pennsylvania are showing off their latest round of hovering robots which can move in formation and alter their orientation as a swarm.

You may remember us salivating over the unbelievable stunts the team pulled off with a singleĀ ‘copter back in 2010. That device needed a sophisticated camera installation to give provide feedback, and this uses the same framework. But we don’t that detracts from the achievement; it’s simply a future hurdle for the project.

The video after the break shows some of the stunts the slew of whirring devices are capable of. Watching them move as a grid, and even landing simultaneously, we can’t help but think of the Dog Pod Grid from Neal Stephenson’s book The Diamond Age. It was used as a protection system, keeping unwanted flying intruders out. Doesn’t sound so far-fetched any more, does it?

38 thoughts on “Dog Pod Grid One Step Closer To Reality

  1. IIRC they are using feedback from video cameras for command and control of the quads, hence the use of white sheets everywhere.

    Now, if it were done using gps/magnetometers/accelerometers only with no video feedback, THEN I’d be concerned..


      1. care to elaborate ? Seems like a big step to me.

        Going from a central sensor system that knows the absolute position of all craft, to individual positions from inertial measurements and without a static external frame of reference.

        Nevertheless, very nice results.

      2. Moving to internally controlled and sensed is a massive step….

        They are using the Vicon Vision System….50k$ a camera which is capable of tracking the silver ball positions in infrared around error of millimeters and at 500hz feedback……

        Its interesting, and a good way to demo the formation and swarm dynamics, but outside that room, this has yet to be achieved… Although people are getting closer….

      3. It is a HUGE step to get IMU’s working to that level of precision. While im sure we will get there one day – we are not there yet.

        If the IMU is off by even 1deg the platform will start drifting in the wrong direction, and it’s that drift that is hard to detect.

      4. As far as I’m aware… most of the hardware out there today is very limited in the range of forces it can account for in small packages. You can certainly string many many units together but in a device so small, power budgets, space, and weight quickly become issues. I’ve got no doubts that there is hardware coming/available that’s capable of what needs to be done, but it’s certainly not trivial, and will require at least some advances in what hardware is currently available.

    1. I am a bit concerned; internal navigational reference is a tiny step, attaching IR cameras and tranquillizer darts is another tiny step.#OccupyDirtNap#
      Morally ambiguous potential aside, I think this is pretty neat: a swarm of these fetching your cappuccino and a copy of the Times.

  2. How come the army is not using these as battlefield recon already? These have all the qualities one may need: small, easily portable and deployable, able to carry an imager, and maneuverable enough to negotiate urban terrain.
    Each soldier may carry several, each with a small CMOS sensor (compared to those found in mobile phones, or even as cheap as those in webcams), that feeds back to a remote station, or even the heads-up display of its operator, and a swarm of these may even be able to use interferometry to simulate a larger aperture camera for increased resolution.

    Advances like these just keep bringing sci-fi and video games closer and closer to reality…

    1. See the comment above. The focus of this research is on nailing down swarm behavior. What they’re not trying to do yet (and, it’s obviously very hard), is find a robust way to track relative and absolute positions of all the agents that would work on a battlefield. Currently, they use high-speed video tracking to do that.

      1. Yes, I understand that. I meant that even without the swarming behavior, with a simple inertial measurement unit to track only movement, possibly with a GPS added, a single drone would be a valuable asset to any fighting team, to scout out enemy positions.

        Swarming and the possibility of a very large array camera would just be an added bonus.

    2. Having read some 1st person accounts of the use of throwable drones in the middle east, it seems that insurgents are wary of the sound of small engines now and are able to effectively conceal themselves or simply place themselves outside the US ROE.

      The trick would be to make the drones difficult to spot either from noise ( the current vector ) or visual inspection ( less of a concern but a concern worth noting ).

      1. Just because the position themselves as a swarm doesn’t mean they have to be flying to be observing.

        Imagine a hundred small drones entering a town and in a matter of minutes positioning themselves for optimum overlapping sensor coverage, then landing/attaching on various surfaces. Just a few minutes to get complete surveillance, and then no more motor noise (until you want to recall them.)

      2. I was thinking more along the lines of tactical recon: following behind the drone close enough that the enemies won’t have a chance of concealing themselves and changing positions. For example, throwing one in window, where it does a quick sweep of the room for booby traps or hiding enemies, or piloting one up a stairwell, where it can move fast enough to leave very little room for reaction before the guns follow up behind it.

        Swarms of them could be used out in open streets, where they could form a very high resolution composite eye over the rooftops or streets to identify targets visually. They need not be quiet in that case, as the ambient noise can mask their rotor noise and the swarm could be dispersed and reformed on command to avoid visual detection.
        Granted, automatic dispersal and reforming would require some significant advancements in the swarm logic, but I believe it to be doable inside five years (believe with my layman’s knowledge, that is…).

  3. I agree that once this does not require the video feedback but rather uses internal ‘meshed’ sensors to work together… that will be amazing.

    Still. Very awesome development.

  4. hmm…perhaps they need to take a cue from nature on this one.
    in this case they use a central authority to give directions, and the swarm follows the directions blindly (literally they can’t see anything)

    in normal swarms what we tend to see is a bottom up method, they try to stay near others of their kind but not so close that they hit. then you paint a beacon that they want and they want to go towards that beacon (this could be a gps or virtual beacon, that is to say a beacon which registers as x meters from the swarm at all times and in some orthogonal direction(which may be derived from azimuth))
    but of course such a thing would give swarming but not a whole lot else, any level of higher cooperation would be quite difficult. but i think its all a matter of getting the rule right.
    for instance you want the swarm to pick up a box, each one has a small suction cup or some mechanism to manipulate the box they all just follow the basic rules (dont run into each other) and the new rule (grab box and pull up) then the new instruction can be hold the box in place, and then bring the box to me, at which point they each try to go to you while holding the box.
    but then again im doing a lot of handwaving too.

    either way this is quite impressive, i look forward to seeing more of what they do.

  5. Am I the only one who kept thinking, with that sound, they would organize a mechanical swarm, find me, and strip me of my flesh somehow?

    Imagine 1000 of these together. Working in unison. All with small red eyes. Frightening.

  6. I think with anything swarm-like the loss of a couple drones should be acceptable and expected. Getting the swarm usable, and to a point where you don’t lose the majority by the end of whatever mission they have should be the goal, and not necessarily to get perfect accurate swarming behaviour right off the bat.

    As far as this post is concerned, really cool project; though it doesn’t seem like they’re tackling the harder problems, heck, I’m not entirely certain what they’ve accomplished is applicable to a decentralized control system.

    Really this just makes me wonder what they’re trying to accomplish.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.