Forgot your password?
typodupeerror
Technology

AirBurr UAV Navigates By Crashing Into Things 74

Posted by samzenpus
from the if-at-first-you-don't-succeed dept.
Zothecula writes "If you've ever watched a fly trying to find its way around a house, you might have noticed that it didn't take a particularly graceful approach – it probably bounced off a lot of windows and walls, until by process of elimination, it found a route that was clear. Well, researchers at Switzerland's EPFL Laboratory of Intelligent Systems are taking that same approach with the latest version of their autonomous AirBurr UAV – it's built to run into things, in order to map and navigate its environment."
This discussion has been archived. No new comments can be posted.

AirBurr UAV Navigates By Crashing Into Things

Comments Filter:
  • by Bob the Super Hamste (1152367) on Monday March 04, 2013 @01:55PM (#43070123) Homepage
    That sounds similar to the approach that some took in my robotics class in college except those robots drove around on wheels and didn't fly. There hopefully is more brains in these things if they are mapping out their environment by doing so but the 64k we had to work with even allowed some some rudimentary mapping ability.
  • Drunk (Score:5, Funny)

    by silanea (1241518) on Monday March 04, 2013 @01:56PM (#43070129)
    So essentially it mimics a drunk person? I have a suspicion I know how the idea for this research project first came up.
    • These days, that's how my Mom drives. Wish I could use the joke tag.

      I wonder if I should call her up and tell she's being replaced by robots.

  • Idiotic approach (Score:4, Insightful)

    by bluefoxlucid (723572) on Monday March 04, 2013 @01:57PM (#43070149) Journal
    Current approaches bounce radio, sound, and light off obstructions, using radar/sonar/laser mapping. This new approach bounces the physical object off obstructions, for the purpose of...? Being more easily detected? Making even more noise? Causing itself and everything around it more damage?
    • Re:Idiotic approach (Score:4, Interesting)

      by BasilBrush (643681) on Monday March 04, 2013 @02:15PM (#43070373)

      I can think of a couple of reasons.

      1) For the purpose of saving the weight of radar/sonar/laser devices. It's a small flying device. Weight matters.

      2) For the purpose of saving the cost of radar/sonar/laser devices.

      • But it's essentially blind if it's not crashing into things. It could at least employ visual imaging analysis if it has a camera. What are we going to do? Deploy UAV dragonflies that buzz around peoples' heads, but are mainly inoffensive? What the hell is the use case for something like this that's neither invisible nor able to avoid obstacles?
        • by arielCo (995647)

          Mapping closed spaces, perhaps recording sounds or sniffing chemicals and reporting back, as suggested in TFA. Maybe we've read too much about military/police applications. Make it smaller and you won't worry about knocking anything important, and it'll be able to slip through smaller openings.

          • by X0563511 (793323)

            Provided it can find them before the battery runs out...

            • by arielCo (995647)

              The saved weight should help a bit. Have you seen quadcopter videos in YouTube? Also, I figure that it might have path-integration capabilities to know when the charge is just enough to go home ("bingo fuel").

              • by X0563511 (793323)

                But how does it find home if it doesn't have traditional sensors?

                • by arielCo (995647)

                  Accelerometers. It mapped the place as it went about, right? There may be a few more bumps on the way back due to linearity/quantization errors.

                  • by X0563511 (793323)

                    You wouldn't have anywhere near the accuracy you'd need - especially after one or more collisions. Even a full-blown laser-gyro INS drifts.

                    • by arielCo (995647)

                      I meant, whatever sensors it used to map the place as a primary mission. Namely:

                      By analyzing the position and force of its collisions, the AirBurr is able to gradually map out its surroundings, establishing where the various boundaries lie

                      Of course, I expect it to bump all the way back, but knowing left from right would be enough to trace back the maze. Worst case, it could radio its rough position as it goes, so you know where it ended up. Here's the project's page including a long exposure photo showing the bumpy path; I have to wait until I get home to watch the video.

      • by X0563511 (793323)

        I wonder if the weight saved by forgoing sensors isn't replaced by the weight necessary for power. Given the extra flight time needed to exercise this rather inefficient navigation method, you'll need more power on-board to do anything useful with it.

        • Not to mention the weight saved by "armor" for the lack of a better term.

          If this thing is going to be constantly bouncing into things... by DESIGN... then it's going to have to be able to take a heavier beating than something that will try to avoid bouncing into things all-together.

    • by ByOhTek (1181381)

      To quote Jeremy Clarkson: "What could possibly go wrong?"

    • by flygeek (460427)

      I don't think it's that crazy. I did some robot navigation work for a thesis project, years ago, when compute power was abysmal and sensor capability was very limited as well. The lesson learned was that navigation and mapping is relatively easy regardless of the data source (we did nav and mapping with an original 128KB Macintosh), but spatial sensor processing is hard and unreliable (nothing that we had available could keep up, and the raw sensor data sucked too).

      The key is reliability; you can certainl

  • it's built to run into things, in order to map and navigate its environment.

    Hey that's neat... Question:

    What happens when it bumps into a weakened structural support, one that just happens to barely be holding the building up?

    I assume the AirBurr is cheap to replace?

    • by arielCo (995647)

      1. Make it light enough. At any rate, it doesn't *crash* so much as "bump its feelers" in traditional robotics lab fashion.
      2. Well, there's no [stereo] camera[s], and just enough storage for the flight log.

  • Sounds like an air-Roomba. Wish my Roomba didn't die. That thing was fun as hell to watch.
    • by MrEricSir (398214)

      Unless you dropped an anvil on it or something you can most likely fix it. There's all kinds of diagnostics built right in to every Roomba, they're easy to open with a standard screwdriver, and there's dozens of repair guides and forums on the internets.

      Besides, if you don't fix it, you automatically forfeit the right to complain about Apple and/or Nintendo's unservicable products.

  • by cellocgw (617879) <<moc.liamg> <ta> <wgcollec>> on Monday March 04, 2013 @02:04PM (#43070221) Journal

    Why don't they just graft tadpole eyes onto its butt?

  • I think it's just a bad programmer trying to close a bug as 'Won't Fix'.

  • by Skiron (735617) on Monday March 04, 2013 @02:12PM (#43070319) Homepage

    This is how the roomba vacuum cleaners navigate, isn't it? What's new?

    • by femtobyte (710429)

      Probably the aerodynamic/mechanical/control-systems design so you can do this while *flying* without being knocked out of the air. That's trickier than keeping a big, heavy hockey puck stable on the floor when it bumbles into walls.

    • by femtobyte (710429)

      Also, the mapping/modeling routine may be quite a bit more sophisticated than a Roomba. The Roomba has a fairly limited "short term memory," and relies heavily on tuned heuristics for how to mix methodical motions (following a wall or spiraling to cover an open area) with enough randomness to efficiently, fairly uniformly cover all areas. No overarching "floor map" is calculated/stored, just info about short-term events. The UAV might actually generate/save a more complete map, so it can retrace its steps a

  • by Shaiku (1045292)

    The flies in my part of California don't bounce off the walls. WTF? Sure, they fly back and forth and in circles, but bouncing off the walls as a form of navigation? I have my doubts as to whether this was truly nature-inspired.

    • Well, I think that's what wasps do when I approach them, but damn, what a pain to be mapped!
    • by azav (469988)

      I saw this in part of San Francisco. There would be these hovering flies (not houseflies) that would circle and hover while darting back and forth over small distances.

      Completely different method of navigation.

  • that won't be able to tolerate being hit. Like a child. Or drying concrete. Or another UAV.

  • by azav (469988)

    that's what the whiskers on cats, rats and dogs are for. They are sensors for what they are about to bump into.

  • OK... so who saw the picture in the article and thought the UAV was responsible for the damage?
    • by TedRiot (899157)
      Replied already to an earlier comment, but my first thought really was "that's what things look like after it has done its navigating"
  • My memory is a little bit hazy, but IIRC from topology: as the number of dimensions increase, the probability of returning to your point of origin in a random walk goes down (assuming you're traversing an infinite space with a possible infinite number of steps). Perhaps I'm mis-reading TFA, or perhaps there's not enough information posted, but assuming these autonomous UAV's utilize a random walk to map it's environment, how can one guarantee it can effectively map a 3 dimensional space?
  • by EvilSS (557649) on Monday March 04, 2013 @02:57PM (#43070827)
    It's going to ding my car, damage my walls, break a window, and knock over a lamp before it assassinates me in my home now?
  • Put short range radar in the thing and have it stop just short of the obstacle. Less damage to the environment, and less strain on the device.
  • Funrobot is MSI's somewhat Heath Robinson robotics brand. They've got high end ones with ultrasonic sensors

    This is the M800

    http://www.youtube.com/watch?v=RzV76Zjru5A [youtube.com]

    It works pretty well and can find the docking station to recharge. It is somewhat expensive (US$400 - above most people's impulse buy threshold)

    The R500 has bump sensors

    http://www.youtube.com/watch?v=6lFwhqcLnzo [youtube.com]

    It's cheap (about US$120) but also very irritating. Bump sensors make a lot of noise and it will also get stuck on cables, curtains and

  • So the top of the article has a picture of a building that looked like someone'd been going at the walls with a sledgehammer for a while. Anyone else see this and think "Man, they gotta make that thing a little lighter..."
  • It would be insane to actually release a bunch of these into a space to find people or something like that. But on the other hand, as a research project, it is beyond awesome. Also, it would be really nice to see the majority of VTOL UAVs designed such that they "can't" (for some reasonable value most likely more properly described as "probably won't") chop stuff up with their props, like your face. And perhaps, to have them not kill you if they land on you from very high up. So there's lessons to be learne

  • Failed analogy (Score:5, Informative)

    by arisvega (1414195) on Monday March 04, 2013 @06:03PM (#43073081)

    If you have TRULY observed a fly trying to find its way around a house, you might have noticed that it in fact takes a very GRACEFUL approach: it never bumps to anything but almost completely transparent objects (as do many birds), and its true grace can be readily observed through 1500 fps videos.

    It is one of the animals with the highest flight maneuverability, as two of its wings have evolved to counterweights: not only it can hover and take-off backwards, but it can land upside-down, and does so very skillfully. See youtube and BBC documentaries for further edification.

Aren't you glad you're not getting all the government you pay for now?

Working...