Become a fan of Slashdot on Facebook


Forgot your password?
Hardware Hacking Robotics Build Technology

Quadrocopters Throwing and Catching an Inverted Pendulum 103

derGoldstein writes "We've seen some very impressive aerobatics performed by quadrocopters before, but this is getting ridiculous. Robohub points to the latest advancement from the Flying Machine Arena, which developed algorithms that allow quadrocopters to juggle an inverted pendulum. One of the researchers working on it said, 'We started off with some back-of-the-envelope calculations, wondering whether it would even be physically possible to throw and catch a pendulum. This told us that achieving this maneuver would really push the dynamic capabilities of the system. As it turned out, it is probably the most challenging task we've had our quadrocopters do. With significantly less than one second to measure the pendulum flight and get the catching vehicle in place, it's the combination of mathematical models with real-time trajectory generation, optimal control, and learning from previous iterations that allowed us to implement this.'"

This discussion has been archived.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

Quadrocopters Throwing and Catching an Inverted Pendulum

Comments Filter:
  • It (Score:5, Funny)

    by JustOK ( 667959 ) on Friday February 22, 2013 @11:59AM (#42980215) Journal

    It don't mean a thing if they don't compute that swing.

  • Science. Making ideas possible.
  • Terminators (Score:5, Funny)

    by ArcadeMan ( 2766669 ) on Friday February 22, 2013 @12:02PM (#42980269)

    It is now clearly obvious: in the future, the weapon of choice to fight robots will be an aluminium baseball bat.

    • Re: (Score:2, Funny)

      by Anonymous Coward

      Until they take it from you and balance it on their heads while giving you a robotic pimp slap.

    • How feasible would it be to encase these in donut or hemispherical shells, with grilles for airflow? The number of props is an obvious weakness, can we cover them up entirely?

      • by plover ( 150551 )

        The AR Drone (by Parrot) comes with two shells, one that's a simple body cover, and the other includes integrated rings around the propellers (called the "indoor" shell.) The idea is you determine the level of safety you require in the situation you're in, with safety either applying to the environment, to the plane, or both. The tradeoff is weight, which translates to reduced flight time.

        In this case they're using a room with fabric drapes to absorb the impact of a stray rotor so they can maximize flight

        • I thought the white drapes were so that the drones could easily track the stick's position?

          • by plover ( 150551 )

            That could be. I've seen videos of them flying in rooms lined with cotton netting, too, though, where the builders were obviously trying to minimize damage to the copters. But that might just be old footage. The autonomous drones have gotten a lot more trustworthy and reliable as of late, so maybe they don't worry as much about crashing anymore.

            And flying in an isolated room, they don't seem too worried about incidental damage. The video of the drone assembling the brick tower in a crowded art gallery,

          • by tibit ( 1762298 )

            The drones don't track anything. I don't even know if they have an IMU. All position data is acquired using a video motion capture system. I'd think it may be Vicon with Tracker software.

        • by tibit ( 1762298 )

          They were flying these things apparently over people's heads in a public demo of putting together a brick "building". The room was full of spectators. It almost looked like a live art exhibit.

  • A new way to throw a grenade through a tiny window on the 6th floor on the cheap.

  • ...why do they have to call it an inverted pendulum instead of a stick?
  • These are a blast. []
  • by Anonymous Coward

    I'm impressed.

  • rehersed/pre programmed or not that is cool.

  • by mill3d ( 1647417 ) on Friday February 22, 2013 @01:02PM (#42981125)

    This looks like it was achieved using motion capture equipment, as seen by the usage of retro-reflective balls (see them shine at 1:44). If that's the case, it would imply that the computing was done remotely. Motion capture gear typically works at 120 frames per second, which would give the system enough time resolution to figure this out on the fly with a decent PC on the back-end.

    It'll be a while before the quadrucopters can do this in their own but the program figuring out the catching move is impressive nonetheless.

    • by Anonymous Coward

      It'll be a while before the quadrucopters can do this in their own but the program figuring out the catching move is impressive nonetheless.

      Not really. The benefit you get by transferring the data to an offside server is that it can perform pretty much any calculation. Once you have tried out the algorithm you actually want to do it shouldn't be that hard to build dedicated hardware in that size.

      • by mill3d ( 1647417 )

        Measuring the pendulum's motion in 3D implies having at least stereoscopic vision. While I agree that specific hardware can be made to compute the proper flight once the position and velocity vectors of the pendulum are determined, the quadrucopter would still need to carry the gear required to acquire the pendulum.

        As far as I can tell, that would at least require a 360 radar for object position detection and a couple of high speed, high resolution cameras mounted on 360 degree capable, fast moving track

        • Or, process images from a swarm of individual quadrucopters into a 3D model of the space they inhabit. With enough members in the swarm, you don't need a full, fancy 360 camera setup on each one.

    • by tibit ( 1762298 )

      The way they do it is state of the art at the moment. You pretty much don't want anything extraneous on the quadcopter. If their motion capture is 120Hz as you say, they'd be having an IMU on the quadcopter. If they can do 1000Hz, like they well should if they can afford it, they probably don't need an IMU -- just a completely receive-only quadcopter.

      The quadcopters can't do any of that on their own at the moment because they'd need to know their position in space relative to each other and to the stick. Yo

    • Well, you know what they say: "Nothing is enough for whom enough is too little"

  • Can it be done with a tricopter? Or a bicopter with propellers that pivot?

    • by udippel ( 562132 )

      Interesting question, though as far as I understand that's not what the project was about. To my understanding, the positions and movements are continuously calculated off-site and then, irrespective of the flying objects, any flying object steered, maneuvered and hoovered about to attain the calculated position and flight path. And then maintain constantly what is fed into the Xcopter to balance the rod (pendulum).
      I guess that the quadrocopters were just the easiest and fastest to be maneuvered around, lik

  • That is pretty impressive but it also serves to reinforce the fact that without the über power source, it's just a nifty demo. This is what I keep hearing from SWAT teams who either spend a ton of money (e.g. > $25k) buying one or are looking at getting one and then they discover that they can't put a camera up in the air for hours at a time without landing to change batteries.

    And did anyone else read "Pole Acrobatics" and have a totally different expectation? ;-)

    • by athmanb ( 100367 )

      With how smart these choppers are getting this shouldn't even be a problem. Just buy three of them for any two jobs and have them find their own schedule to "sleep" when they need to so that there's always at least the required number up in the air.

      • It's hard enough for a cash-strapped department to find $25,000 for one. You'd need at least two and enough batteries in various states of being recharged in order to keep eyes on all the time. I know plenty of SWAT guys that would rather spend the money equipping their entire team with thermal night vision.

      • by tibit ( 1762298 )

        Those quadcopters are dumb slave devices. They have no built-in intelligence of any sort.

  • ... I imagine stupid people tricks will remain more popular than smart robot ones.
  • Is the actual balancing performed based solely on the forces applied to the sending/receiving actors? How necessary are the data markers on the stick? I'm wondering that if the two actors knew where each other were in space, and based on the balancing the sending actor has to perform couldn't the receiving actor make a guess where it needs to be in any case? Could I (as an actor in this scenario) make an educated guess as to the length of the stick based on the corrections I have to perform in order to kee
  • They've now got the mopping algorithm locked up; enough with the kitchen use cases get back to the bedroom guys.

  • A puff of smoke at 1:33; I think someone lost a bit of propeller to the pendulum.
  • And I still can't even get my quadcopter for a normal spin without crashing it into the nearest wall or tree!
    How the heck can I get mine to hover at least somewhat in one place? I know these use the cameras for real precise tracking, but mine is the exact opposite it seems. Grrrr

  • Amusement park rides of the future are going to be awesome!

  • As amazing as this is (and it IS friggin' amazing), remember that the human brain does this throw, catch, reposition, recalculate, and respond stuff effortlessly. Just play catch with a ball and your 5 year old child. Animals with pea-sized brains do it great, too! -- Josh
    • by tibit ( 1762298 )

      Those quadcopters are a bit faster than humans arms are, and things are moving towards getting everything even faster still.

  • Very good. Now let's do it with flaming chainsaws.

Can anyone remember when the times were not hard, and money not scarce?