Small but deadly?

February 1, 2012 5:52 AM

A swarm of nano quadrotors.
posted by secretdark (69 comments total)

Dog pod grid!
posted by speedo at 6:00 AM on February 1, 2012 [2 favorites]

Those are "nano" in the sense that this is "mini".
posted by DU at 6:01 AM on February 1, 2012 [4 favorites]

Seriously everything GRASP does just gives me a big old brain boner.
posted by elizardbits at 6:06 AM on February 1, 2012 [1 favorite]

Well that was both fascinating and terrifying. Sounds like a swarm of locusts.

And if you really must be pedantic, I would probably classify those as micro-UAVs. Nano-UAVs are more housefly-scale.
posted by backseatpilot at 6:08 AM on February 1, 2012 [1 favorite]

Well, I know what I want 16 of for my birthday!
posted by Mister_A at 6:15 AM on February 1, 2012 [1 favorite]

I expect one of these to be giving me parking citations in a few years.
posted by device55 at 6:16 AM on February 1, 2012 [10 favorites]

a REAL "nano" UAV can not be seen by the naked eye but can fly up your nose into your brain and whip it like a milkshake!

That said... That's pretty cool, but why do they have to throw them to get them airborn?
posted by HuronBob at 6:22 AM on February 1, 2012

They don't have to throw them, I don't think. That was just to demonstrate that they can recover from arbitrary configurations.
posted by DU at 6:25 AM on February 1, 2012

Needs more "Ride of the Valkyries".
posted by Curious Artificer at 6:27 AM on February 1, 2012 [2 favorites]

"So what's the public reaction like to our incredible, tiny, futuristic machines, the mere existence of which has implications that are vast, unseeable, and bordering on terrifying?"
"They say we're using the word 'nano' wrong."
posted by secretdark at 6:29 AM on February 1, 2012 [36 favorites]

Swarm? Please.
posted by The White Hat at 6:30 AM on February 1, 2012 [5 favorites]

Note to self put more stats into ranged weapons in the future.
posted by The Whelk at 6:30 AM on February 1, 2012 [7 favorites]

Soon buying birdnetting will get you put you on a list.
posted by 445supermag at 6:31 AM on February 1, 2012 [8 favorites]

They need to give them tiny trombones and flutes - they'd be great in a parade.
posted by skyscraper at 6:32 AM on February 1, 2012 [2 favorites]

Actually with some manipulation of rotor blade pitch and motor speed, they might be able to get them to harmonize their angry hornet noises while staying in formation.
posted by skyscraper at 6:37 AM on February 1, 2012 [1 favorite]

SEVENTY-SIX BUZZING ROBOTS IN THE BIG PARADE.
posted by The Whelk at 6:40 AM on February 1, 2012 [2 favorites]

I bet I could train them things to steal beer.
posted by Mister_A at 6:40 AM on February 1, 2012

I think they're smiling at me.
posted by SteelyDuran at 6:43 AM on February 1, 2012

New! Raid® Brand Insect Spray, Now Effective Against Nano-Copters*!


* - excludes Nano-Copters operated by US Federal, State and Local Government Agencies

posted by kcds at 6:44 AM on February 1, 2012 [1 favorite]

Am I the only one who has a strange desire to play Space Invaders?
posted by Fizz at 6:45 AM on February 1, 2012

mostly i kind of just want to cuddle them and whisper sweet nothings of totalitarian world domination into their little roboears
posted by elizardbits at 6:51 AM on February 1, 2012 [2 favorites]

Every time I see things like this or BigDog, I think "Cool!".

Then I think, "Someday these things will hunt humans."
posted by bitmage at 6:54 AM on February 1, 2012 [3 favorites]

Something something manhacks something.

Nothing a good crowbar won't fix.
posted by fight or flight at 6:55 AM on February 1, 2012

We are a long way from seeing this kind of behavior from these devices in less-controlled systems. You may recognize the cameras above the featureless white sheet as part of a motion capture system (think Gollum from the LOTR movies). While the "nanocopters" have stability control software onboard, I am almost certain that the actual motion planning is being done by an external system, using the output from the mocap system rather than the robots' own estimation of their location. Each robot has mocap markers (I couldn't tell if they were retroreflective dots, or if the two blue lights on each robot were IR LEDs for mocap purposes) -- either way, they'll have to solve a pretty intense SLAM (simultaneous localization and mapping) problem to do this in anything other than a white room ringed with mocap cameras. Not saying it's not an impressive piece of tech and control theory, just that it's not quite as scary-impressive as it initially appears.
posted by Alterscape at 6:55 AM on February 1, 2012 [9 favorites]

I wasn't scared until they went through the window-sized aperture...
posted by Ron Thanagar at 6:58 AM on February 1, 2012 [3 favorites]

Alterscape: That was my question - how much of the intelligence is onboard? Whenever I see things like BigDog (above) my first thought is "good Lord walking is incredibly hard - no wonder I suck at it".
posted by The Bellman at 6:59 AM on February 1, 2012

I have seen the future and it is small.

This creeps me the Hell out. The surveillance implications are terrifying.
posted by kinnakeet at 7:01 AM on February 1, 2012

The last figure eight formation reminded me of horse archers. Death by a thousand tiny little missiles, and if you try to approach the formation is always just out of reach.
posted by Meatbomb at 7:05 AM on February 1, 2012

reminded me of the school if fish sequence in Finding Nemo
posted by marvin at 7:06 AM on February 1, 2012

of*
posted by marvin at 7:07 AM on February 1, 2012

under my totalitarian dictatorship i will use these to deliver candy and puppies to my most loyal minions

and to slaughter my enemies obvsly
posted by elizardbits at 7:08 AM on February 1, 2012

Alterscape: I disagree that they're not as "scary-impressive" as they sound. Even within the realms of a heavily sensor-filled environment the implications are fascinating to consider. This team of quadcopters assembling a structure do so autonomously and within 'controlled environment' that was set up apparently relatively easily within an art institute. How hard would it be to throw together a 'controlled environment' somewhere else? And how accurate does this 'controlled environment' need to be, really? Instead of a vast sensor net as GRASP / UPenn have, would 5 or 6 Kinects provide enough information?
I understand 'impressiveness' is subjective but I think the implications of the scale, the autonomy, and the flight-capability of this are very, very impressive.

(FWIW: I've built a quadcopter before and know a little about onboard / offboard control.)
posted by secretdark at 7:09 AM on February 1, 2012 [1 favorite]

You stay down by day, but at night you can move around.  You still have to be careful because the HKs use infrared.  But they're not too bright.  John taught us ways to dust them. 
posted by nathancaswell at 7:16 AM on February 1, 2012 [6 favorites]

Friendly reminder: please spay or neuter your nano quadrotor!

NO, SERIOUSLY.
posted by argonauta at 7:18 AM on February 1, 2012 [4 favorites]

Actually with some manipulation of rotor blade pitch and motor speed, they might be able to get them to harmonize their angry hornet noises while staying in formation.

Now I will not be happy until I hear a chorus of quadrotors humming Flight of the Bumblebee.
posted by Edgewise at 7:28 AM on February 1, 2012 [3 favorites]

What is the flight time for one of these? Or even for the "macro" quadroters that are used for filming sports like BMX racing?
Is is possible to configure them to automatically return home and recharge, like a Roomba?
What is the maximum control range? How well do they perform outside? Could I get a pod of these and have them perform overwatch on my farm?
posted by Old'n'Busted at 7:52 AM on February 1, 2012

What's impressive about the GRASP lab's work is that they've got this incredibly stable control loop that lets them send the quadcopters through crazy maneuvers. Take a look at their earlier "aggressive maneuvers" video for some extreme examples, including quadcopters flying through tossed hula hoops*.

The missing piece here for me is to see how they do in an outdoor environment. One of the big problems with quadcopters outdoors is dealing with changing wind conditions and gusts. That'll be even worse for these tiny copters since they can't produce a lot of force to counteract a strong wind.

*Want to know what's super-impressive about that? Notice that they fly through the hoops as they're falling. They don't wait until the hoop is at near-rest at the top of its arc, oh, no. That's too simple for these robots. They want their job to be hard.
posted by sgranade at 7:54 AM on February 1, 2012 [3 favorites]

In the future we all carry around Gust Guns to destablize the quadcopters with wind.
posted by The Whelk at 8:03 AM on February 1, 2012

I thought Gust Guns were only for peaceful applications, like sailing catamarans!
posted by Mister_A at 8:13 AM on February 1, 2012

Well, it was a good run guys. But clearly humanity is over. But maybe Petman will save us
posted by delmoi at 8:21 AM on February 1, 2012

In the future we all carry around Gust Guns to destablize the quadcopters with wind.

If it makes you feel any better about us avoiding our dystopian UAV future, one of the autonomous helicopter projects I was working on was delayed for a while because of light rain.
posted by sgranade at 8:22 AM on February 1, 2012 [4 favorites]

HERF guns and emp-bombs, kids, start soldering. Until they manage to get hardened chips small enough to run these things, then we're proper fucked.
posted by Slap*Happy at 8:24 AM on February 1, 2012

This creeps me the Hell out. The surveillance implications are terrifying.

Don't be afraid, at least not for the time being. Quadcopters have flight times in the single digit minutes. Also the smaller they are, the shorter their range (where the battery makes up a larger share of the payload).

Is is possible to configure them to automatically return home and recharge, like a Roomba?

I don't see why an inductive charger couldn't be added. Get on it GRASP!

What is the maximum control range?

That depends on the radio. Typical R/C stuff is good for a few hundred yards, but you could technically outfit these with satellite or cell-phone radios that would make the control range nearly unlimited.

As someone mentioned upthread, the copters in GRASP's lab operate in a small controlled environment. You couldn't do this kind of formation flying without the external motion-capture cameras doing the tracking.

How well do they perform outside?

I'd like to know that too. Small, light craft usually can't handle much wind, but these things are so maneuverable that they could maybe get around that limit with some fancy control software.

Could I get a pod of these and have them perform overwatch on my farm?

You can buy large commercial quadcopters for exactly this kind of work.
posted by Popular Ethics at 8:25 AM on February 1, 2012

On reflection, I think both secretdark and sgranade have excellent points, and maybe I'm being too dismissive above. The control systems these guys have (I hadn't made the connection between the "aggressive maneuvers" video and this one) is indeed intensely stable. I am curious how much it depends on mocap systems as well, though.

Most quadcopters use IMUs with some mix of gyros (which report change in angular velocity), accelerometers (which report acceleration), magnetometers (compasses, more or less) and occasionally other things like pressure altimeters and IR horizon sensors (I know the Draganflyer uses them; not sure about other folks these days). All these things accumulate error over time, and you end up doing lots of kalman filter weirdness to compensate. Adding the mocap system gives them an absolute position reference that updates at 100+hz, which makes life significantly simpler, mathematically.

To address the "controlled environment" question -- I set up and operate motion capture systems for research (not with quadcopters). These systems use anywhere between 3 and 32 IR cameras to capture multiple views of the IR markers, and then use linear algebra to calculate the 3D position of each marker (and then more linear algebra to determine correlations between frames and which object is which). All the cameras need to be mounted rigidly and connected via cables. The portable systems I work with are typically mounted to an aluminum truss system (a la stage rigging). Once the cameras are mounted and connected, the system needs to be calibrated by waving a known set of markers around inside the capture volume, which helps the system work out the calibration. The system only "sees" the IR-reflective (or emissive) markers -- the cameras do a threshold calculation to black out everything dimmer than a specified brightness level and take the remaining points as the set of markers to track. The folks I work with can set up a 10' cubic capture volume in 3 or 4 hours, and we don't do it frequently (ie, if we did nothing but set up/break down the system, we could probably do it faster). It does require access and running around with ladders, etc. I'm fairly confident this kind of system couldn't just be set up in, say, a public square, without impractical prep time.

Systems like Kinect recover depth without markers by analyzing the distortion of a known pattern as it is projected on an unknown surface. This is far easier to set up (one camera) but much harder to recover position data from quickly, since the computer vision system needs to do all kinds of fun pattern recognition, rather than just a simple threshold, to extract objects. The implication is that multi-camera IR-based optical motion capture can always run at a higher refresh rate on equivalent hardware.

I have seen some neat Kinect+UAV demos, for instance this (youtube). I think we have a long way to go before this kind of approach is fast enough to keep up with the kind of systems shown in the aggressive maneuvering video. One limitation you'll notice from the video I linked is that the copter is operating in a static environment -- extracting moving objects like people or other quadcopters from the sensor data, so they can stabilize against just the fixed parts of the environment is a big challenge. These approaches can be brought together, with time and research, and that IS a little scary.
posted by Alterscape at 8:28 AM on February 1, 2012 [10 favorites]

Quadcopters have flight times in the single digit minutes. Also the smaller they are, the shorter their range (where the battery makes up a larger share of the payload).

This is where we wait for the other jackboot to drop...

Wireless power.

...there it is.
posted by Slap*Happy at 8:31 AM on February 1, 2012 [1 favorite]

Am I the only one who has a strange desire to play Space Invaders?

No! One quadcopter per pixel, please. I'd even be willing to be that there's a way to vary the RPMs of the motors to reproduce the 'marching' sound effect...
posted by the painkiller at 8:37 AM on February 1, 2012

I think their system, especially the "nano" (which is really more like 'mini' -- are they saying their full sized ones will be a billion times larger? Even volume wise that would make them a thousand times larger on a side).

But as computers get smaller and smaller, you'll probably eventually have kinect style laser sensors the size of a current cellphone camera. Which means you could put a bunch of them on something like this.

Power density is the big issue, though. Batteries on these things probably don't last very long. Maybe you'll need something like liquid hydrogen. For military applications you could even do something exotic like Americium or some other radioactive isotope powered sterling engines (I remember hearing that a few grams of that stuff in a sphere puts out like 150 watts of energy - you could potentially create micro nuclear reactors, with the downside of no real possibility of containment but definitely the possibility of meltdown.

Wireless power.

I remember hearing about that a couple years ago, but if you check out how large those transmitters were -- well they're pretty huge. Not really practical for --- well anything at the moment.
posted by delmoi at 8:40 AM on February 1, 2012

I expect one of these to be giving me parking citations hunting down political dissidents in a few years.

The next logical step from current drone aircraft campaigns, seems like.
posted by aught at 8:48 AM on February 1, 2012

Am I the only one who has a strange desire to play Space Invaders?

I was thinking that the upcoming IRL version of Total Annihilation will be AWESOME!
posted by Mister Fabulous at 8:48 AM on February 1, 2012

Given that flocking algorithms have been well-studied for some time, I wonder if anyone's tried to program them into 'copters.
posted by primer_dimer at 8:54 AM on February 1, 2012

Actually with some manipulation of rotor blade pitch and motor speed, they might be able to get them to harmonize their angry hornet noises talk while staying in formation.

FTFY.

also:
MetaFilter: I wasn't scared until they went through the window-sized aperture...
posted by sexyrobot at 9:00 AM on February 1, 2012

Adding variable pitch props, or swashplates, would remove part of what makes quadcopters so easy to build (in terms of hardware, anyway) -- the four brushless motors are the only component that actually move. Everything else is software! Probably easier to just hang a loudspeaker off of one, if talking is what you're after.
posted by Alterscape at 9:07 AM on February 1, 2012 [1 favorite]

A swarm of nano quadrotors.

I believe the proper term is a curse - a curse of nano quadrotors.
posted by schoolgirl report at 9:28 AM on February 1, 2012

It seems like much of the difficulty in operating in random outdoor spaces could be solved with a few carriers/AWACS. Especially if you had lead time you could send out a few shoe box to milk crate sized control units (desguised as anything common in the area). Those units could provide cameras and calibration. In urban areas a sophisticated unit could tap into mains power (available on many light posts; desguise your recharge units as cameras) to recharge the quads.
posted by Mitheral at 9:29 AM on February 1, 2012

...to do this in anything other than a white room ringed with mocap cameras. Not saying it's not an impressive piece of tech and control theory, just that it's not quite as scary-impressive as it initially appears.

You do recall that cities are often blanketed with CCTV already, right? Even if you had to do motion capture to control them, that would be easily achievable especially in defined areas *cough*zucotti park*cough*.
posted by odinsdream at 9:41 AM on February 1, 2012

When they get the saw blade attachment, we're all screwed.
posted by polywomp at 9:45 AM on February 1, 2012

Fizz: "Am I the only one who has a strange desire to play Space Invaders?"

Reminds me more of Homeworld.
posted by jiawen at 9:59 AM on February 1, 2012

Previously on MeFi.
posted by charlie don't surf at 10:09 AM on February 1, 2012

You can buy large commercial quadcopters for exactly this kind of work.

That's the one I was obliquely referring to regarding "BMX videos"; I was hoping that there was a more longer-running system available that can be automated to provide an "eye in the sky" like camera point. Right now the only viable solution I can come up with a tethered balloon, with would make people think I'm having a party or selling used cars.
posted by Old'n'Busted at 10:13 AM on February 1, 2012

Surveillance implications, surveillance schmiplications.

Your revolution will be distinctly harder when your government has swarms of floating unmanned guns that can strafe in formation and track you by heat profile.
posted by Slackermagee at 11:27 AM on February 1, 2012 [4 favorites]

I am interested in your ideas and would like to subscribe to your newsletter.
posted by mikelieman at 12:23 PM on February 1, 2012

Bah. I'm not worried.
posted by The otter lady at 1:14 PM on February 1, 2012

If six of them form an evil pair of eyes, ten are left to make a mouth. That's enough to to rise into the air and lip-sync "Greee... tiiings... ... ... Wweee... aare... Yyy....graa...muuul" when someone enters the room.
posted by springload at 5:15 PM on February 1, 2012

That was cool and scary. Yay/Boo.
posted by Divine_Wino at 7:48 PM on February 1, 2012 [1 favorite]

I, for one, welcome our new quad nanoverlords.
posted by skippyhacker at 8:52 PM on February 1, 2012

Amazing and altogether horrifying.
posted by IvoShandor at 4:48 AM on February 2, 2012

I recently wrote in some other FPP that UAVs were not the threat that people believed them to be yet. A surveillance state would still be difficult with the limits of current UAVs. But I described my nightmare, based largely on the previous UAV FPP from these guys. Imagine not a single UAV watching you, but a huge network of UAVs, perched throughout the city in every nook and cranny. Think about that UAV in the video, stuck to a piece of velcro, like a bat. Now imagine one of those networks on every city block, waiting to be called into action. Imagine the surveillance target is watched by a dozen drones simultaneously from multiple angles, then as he moves to another block, another set of drones moves into position to continue surveillance. That idea was based on the previous video.

Now imagine a fleet of networked UAVs in formation. Don't imagine the formation so small as this demo video. Imagine that formation covering a city block or two. Imagine several formations, one in flight, another perched in their charging stations. They rotate, one in flight while the other charges. The formations provide continuous surveillance. Multiple formations could wait in their perches to be summoned to different surveillance tasks. Think of them like dozens of flocks of birds, some perched, some flying.

These engineers are implementing the surveillance state. They have all the basic designs finished except one. This video shows how to maneuver the formation through an obstacle. Now they only have to implement "flocking," how to maneuver around an obstacle that moves through a formation. For example, how to keep in a formation when an obstacle like a radio tower or tall building is in their flight path. Flocking algorithms are already well known, the solution is trivial.

This is the problem with engineers. Do they not even consider the implications of what they are building? Or do they just build something because it would be cool to see a bunch of UAVs flying in formation? Someone ought to force them to watch the movie Real Genius and then smother them with Jiffy Pop.
posted by charlie don't surf at 9:06 AM on February 2, 2012 [1 favorite]

Drone Pilot Discovers River of Meat Blood
posted by homunculus at 10:55 AM on February 2, 2012

Also relevant: "Although the FAA allows hobbyists to fly model airplanes for recreational purposes, that authority does not extend to operators flying unmanned aircraft for business purposes," the Air Division of the Los Angeles police department reminded the California Association of Realtors.
That could soon change. The FAA plans to unveil a new set of rules this year that will cover the burgeoning interest in flying commercial and private UAVs."
posted by Popular Ethics at 10:38 AM on February 3, 2012