Robohub: Throwing and Catching an Inverted Pendulum with Quadrocopters
February 21, 2013 3:55 PM   Subscribe

Throwing and Catching an Inverted Pendulum with Quadrocopters. From Robohub, a website about all things robotic.
posted by milquetoast (40 comments total) 19 users marked this as a favorite

 
Wow!

I'd like to see Zoe catch and balance a pole on her nose.
posted by notyou at 4:06 PM on February 21, 2013 [2 favorites]


Yeah, super cool, I've seen this around (pitched it to write up at the news site I work for but they didn't bite).

Note that these copters are being assisted by high-speed cameras that monitor angles and such, so not quite ready for the outside world yet. But the military would love one of these to go toss grenades into 4th-story windows, don't you think?
posted by BlackLeotardFront at 4:10 PM on February 21, 2013 [1 favorite]


Wow, that is very exciting! I don't know what I find more interesting - that the robot is able to catch it so seamlessly, or that, like humans, the robot must make constant minor adjustments to keep it balanced. I seem to recall hearing somewhere that constant minor adjustments are in fact much easier to program than a single perfect state.
posted by rebent at 4:17 PM on February 21, 2013


BlackLeotardFront: But the military would love one of these to go toss grenades into 4th-story windows, don't you think?

For grenade throwing, Cooperative Quadrocopter Ball Throwing and Catching (also from ETH Zurich, also using external cameras and processing) is also relevant.
posted by James Scott-Brown at 4:20 PM on February 21, 2013


I don't think I'd call that an "inverted pendulum." I would describe it as "a stick."
posted by charlie don't surf at 4:21 PM on February 21, 2013 [10 favorites]


I'd like to see Zoe catch and balance a pole on her nose.

Dude, she's the one programming the quadcopters to do it for her.
posted by yoink at 4:23 PM on February 21, 2013 [1 favorite]


You all realize this can only lead to military drones with weapons-grade devil sticks, right?
posted by Sys Rq at 4:26 PM on February 21, 2013 [5 favorites]


It sounds like quadrocopters have also learned to narrate YouTube videos.
posted by eugenen at 4:26 PM on February 21, 2013 [1 favorite]


I'm not sure they should be doing this in a room with windows.
posted by salishsea at 4:34 PM on February 21, 2013


I'm not sure they should be doing this in a room with windows.

Pretty sure they're working off a Mac.
posted by Joey Michaels at 4:36 PM on February 21, 2013 [4 favorites]


You all realize this can only lead to military drones with weapons-grade devil sticks, right?

Pointed sticks.
posted by charlie don't surf at 4:36 PM on February 21, 2013 [2 favorites]


I'd like to see Zoe catch and balance a pole on her nose.

I'd like to see Zooey do it.
posted by eugenen at 4:38 PM on February 21, 2013


The inverted pendulum probably means that ball on the stick is very heavy compared with the rest of the stick, and they're using that to simplify the calculations and to make the stick's motion more stable.
posted by hexatron at 4:40 PM on February 21, 2013 [2 favorites]


See, I watch this and all I see is "Cyberdyne Systems: The Early Years"
posted by percor at 4:46 PM on February 21, 2013 [4 favorites]


It seems to me that in this scenario, two weighted points on the stick are exactly equivalent to distributing the weight evenly throughout the length the stick.
posted by charlie don't surf at 4:47 PM on February 21, 2013


You all realize this can only lead to military drones with weapons-grade devil sticks, right?

Drones are taking jobs away from honest fire twirlers!
posted by Charlemagne In Sweatpants at 4:50 PM on February 21, 2013


I don't think I'd call that an "inverted pendulum." I would describe it as "a stick."

It's 3 sticks and two joints. I guarantee you can not balance one in the palm of you hand.
posted by GuyZero at 4:54 PM on February 21, 2013


The talk of inverted pendulums had me expecting some kind of robotic ball and cup game. Glad they can't do that yet!
posted by ceribus peribus at 4:55 PM on February 21, 2013


Oh geez, my bad, those aren't joints. I thought it was something like this. Still, not easy.
posted by GuyZero at 4:57 PM on February 21, 2013


Why do you say it has joints? I don't see anything bending. I suspect that the balls are movable weights, to allow them to (a) move the stick's centre of gravity and (b) give it a greater or lesser moment of inertia by moving them away from the centre or back towards it.
posted by Joe in Australia at 5:00 PM on February 21, 2013


Holy crap--balancing a triple pendulum with joints would actually be theoretically possible: have a look at this video. I doubt you could toss and catch it, though.
posted by yoink at 5:09 PM on February 21, 2013 [2 favorites]


The "inverted pendulum" is a description of the combination of the 'copter and the stick. An inverted pendulum is a physical structure where the center of mass is above the fulcrum.
posted by sonic meat machine at 5:19 PM on February 21, 2013 [3 favorites]


Why do you say it has joints?

yeah, they're just weights or tracking spheres for the visual processing system.
posted by GuyZero at 5:23 PM on February 21, 2013


these aren't the drones I'm looking for and not the jetpacks I was promised.
posted by Lipstick Thespian at 5:29 PM on February 21, 2013


So how soon can we call bookmakers to play the odds on droneball games?
posted by Burhanistan at 5:41 PM on February 21, 2013


If the robots take caber tossing from us, we will truly have nothing left.
posted by dephlogisticated at 5:41 PM on February 21, 2013 [4 favorites]


"we apply learning algorthms to the system in order to improve its performance over time"

OH COME ON WHY WOULD YOU EVEN DO THAT
posted by dephlogisticated at 5:50 PM on February 21, 2013 [7 favorites]


I want squads of drones with laser pointers on them to paint and triangulate targets of interest.
posted by fraxil at 5:55 PM on February 21, 2013


I can't wait for Cirque de Quadrocopter Overlords.
posted by vverse23 at 6:37 PM on February 21, 2013 [2 favorites]


charlie don't surf: "It seems to me that in this scenario, two weighted points on the stick are exactly equivalent to distributing the weight evenly throughout the length the stick."

Not really. The moment of inertia would be vastly different.
posted by notsnot at 7:53 PM on February 21, 2013 [1 favorite]


You all realize this can only lead to military drones with weapons-grade devil sticks, right?

No, what this leads to is "I, for one, welcome our new Quadrocopter overlords!"
posted by TDavis at 9:36 PM on February 21, 2013


Ya, this is pretty awesome!

yoink: Holy crap--balancing a triple pendulum with joints would actually be theoretically possible: have a look at this video. I doubt you could toss and catch it, though.

When I was studying this, I'm pretty sure my Linear Systems prof. talked about some lab where they had a septuple inverted pendulum (google pronounces septuple wrong!). Here's a paper about a quintuple. As long as the moment of inertia of each link is significantly different from all the others, the math/theory is quite simple. The problem is that significantly different is like half or double, and it is a challenge to have enough actuator authority and dynamic range to effect things of mass x and mass 64x at the same time.

And on that note:
hexatron: The inverted pendulum probably means that ball on the stick is very heavy compared with the rest of the stick, and they're using that to simplify the calculations and to make the stick's motion more stable.

Well, the math isn't changed by having a heavy ball at one end. It might make it easier to actually build a system--you might get away with less dynamic range and less actuator authority--well, on second thought, I don't really think having a heavy end would make much difference at all..

Anyway..
These guys do a double pendulum swing up in every permutation (link 1 and 2 together, link 1 then 2, link 2 then 1), which is kind of cool.
posted by Chuckles at 11:11 PM on February 21, 2013 [2 favorites]


Oh wow, when the maverick sentient machines assemble their configurable weapons platforms it'll be one helluva show!
posted by TwelveTwo at 11:55 PM on February 21, 2013


they only do it once, I was expecting some continuous juggling. How many out-takes were there?
posted by memebake at 12:45 AM on February 22, 2013


OK, now I'm curious: Remember the awful movie from back in the 1980s...Blue Thunder? Something like that. It featured a "silent helicopter" and Roy Scheider, if I'm remembering right.

Is such a thing even theoretically possible? Assuming you could completely silence the mechanical pieces, including the motor(s), is it possible to spin a copter's blade slowly enough to not make any, or minimal, noise and have it stay in the air?
posted by maxwelton at 2:11 AM on February 22, 2013


No, but at the altitudes that large surveillance drones typically fly at and their relatively low engine power, they are effectively silent to observers on the ground. Not quadrocopters or anything like that, but the MQ-9 and similar.
posted by thewalrus at 5:06 AM on February 22, 2013


I fosee a day when swarms pass through our residences uninvited, unannounced, and unstoppable. For our freedoms.

And corporations who send out those swarms will be eligible for the Medal of Honor.
posted by surplus at 7:58 AM on February 22, 2013


charlie don't surf: "It seems to me that in this scenario, two weighted points on the stick are exactly equivalent to distributing the weight evenly throughout the length the stick."

It's not. Imagine a stick that weighs 5 pounds, with practically all the weight in a lump in the middle. Now imagine another stick that weighs 5 pounds, with practically all the weight in two equal lumps at the ends. Imagine grasping them in the middle and rotating them back and forth. The first one will be easy. The other one will be hard because it develops a lot of momentum at the tips, so it's hard to reverse directions.

The uniformly-distributed weight stick is a third scenario that would be in between these two.
posted by adamrice at 9:12 AM on February 22, 2013 [1 favorite]


I'm not convinced, adamrice. Sure, I can see that when a stick is thrown, there would be a difference in rotational momentum in a center-weighted stick vs. a stick with weights at the end. But I checked the picture and the two spheres seem to be equidistant from the center and the ends, so they're at like .25 and .75 of the length. I figure that's roughly going to replicate the stick with evenly distributed weight.

But there is no real way to tell what the purpose of those spheres is, unless we ask the researchers. They could just as likely be position markers, to make it easier for computer vision systems to capture the position of the stick.
posted by charlie don't surf at 5:40 PM on February 22, 2013


The 'copter is not part of the inverted pendulum. And rotational inertia has nothing to do with the definition of an inverted pendulum. An inverted pendulum is basically anything that can be balanced on end that will fall over with the slightest touch. What makes the inverted pendulum interesting from a dynamical systems viewpoint is that its equlibrium position (balanced on its end) is unstable, such that any perturbation in the position will cause the pendulum to move away from equilibrium (i.e. fall over).

What makes the inverted pendulum interesting from an engineering standpoint is that the controllers need to turn an unstable equilibrium into a stable one by providing constant perturbations that counteract the pendulums motion away from its unstable equilibrium. This is a classic problem in control theory. What makes this reasearch unique is that the physical controllers providing the stability are quadcopters.
posted by grog at 8:27 AM on February 23, 2013


« Older The Australian Raid On SuperDaE And How A Prank Ov...  |  This might change things.... Newer »


This thread has been archived and is closed to new comments