Watch a swarm of flying robotic drones construct a tiny building

January 16, 2011 7:26 AM

Watch a swarm of flying robotic drones construct a tiny building. In the future, construction workers will be a buzzing, mildly disturbing haze of mechanical diligence.
posted by Tom-B (54 comments total)

Impressive stuff. I followed the link back the guys who made them. This scares me more.
posted by dougrayrankin at 7:32 AM on January 16, 2011 [4 favorites]

"We tell the quadrotors what structure to build and they figure out the assembly plan and then build it."

This was the interesting part to me
posted by Redhush at 7:34 AM on January 16, 2011

I'll be more impressed when they can sexually harrass passing vehicles.
posted by Joe Beese at 7:34 AM on January 16, 2011 [10 favorites]

Do you suppose those robotic drones are really that, or is there someone off camera with a remote control?
posted by crunchland at 7:36 AM on January 16, 2011

no, they won't
posted by kitchenrat at 7:43 AM on January 16, 2011 [1 favorite]

Let me know when they can build sexbots.
posted by Brandon Blatcher at 7:45 AM on January 16, 2011

construction workers will be a buzzing, mildly disturbing haze of mechanical diligence.

I just had this weird image of robotic construction workers taking a break and yelling down at the women walking by. But with mechanical diligence.
posted by twoleftfeet at 7:47 AM on January 16, 2011 [2 favorites]

The Aggressive Maneuvers for Autonomous Quadrotor Flight video is probably a better indicator of what these things will be used for.

"We tell the quadrotors what structure to build and they figure out the assembly plan and then build it."

Redhush: This was the interesting part to me

That was the vague part to me. I think they're fudging the concept of "autonomous" a bit. The highest-rated comment on the above video puts it this way:

Ok so as far as I have understood so far. Theres a 20 cam vicon state estimation system providing location and positioning information for the UAV and the obstacles. The information is sent to the main processor outside of the UAV which estimates the trajectories and all the heavy math and sends flight data to the UAV which performs the maneuver.

This system has an amazing level of accuracy for positioning and UAV stability but strictly speaking it is not an autonomous drone.

posted by mediareport at 7:57 AM on January 16, 2011 [3 favorites]

Vuvuzelarific!
posted by fairmettle at 7:59 AM on January 16, 2011 [1 favorite]

Nevermind, found DIY instructions online.
posted by Brandon Blatcher at 8:03 AM on January 16, 2011 [1 favorite]

Make sure you're grounded first dude.
posted by elizardbits at 8:08 AM on January 16, 2011 [1 favorite]

Great for precision 3D placement, but not necessarily great for actual construction.
posted by DU at 8:12 AM on January 16, 2011

I doubt they're actually that hard to build. The secret sauce is the programming. But really what you're seeing is the effects of mores law. Cheaper and more importantly lighter computers.
posted by delmoi at 8:13 AM on January 16, 2011

no, they won't

I can see applications for drones in large, well-funded projects - most building will still be done by individuals and small firms w/modest budgets, though.

Since human labor will likely be ever cheaper and more disposable, I'm guessing the construction sites of the future will be "mildly disturbing" in a more traditional way.
posted by ryanshepard at 8:19 AM on January 16, 2011

Half the quadrotors are just hovering there watching other ones work. I guess there are still some constants to the universe.
posted by mendel at 8:23 AM on January 16, 2011 [18 favorites]

Very impressive positioning, regardless: but how do they achieve it, and can they do it in real world situations?
posted by Segundus at 8:23 AM on January 16, 2011

Most people don't realize how far we've already gone in this direction. The days when a home construction site would start with pallets of 2-by lumber and a bunch of guys with hammers and saws are pretty much gone. Instead, nowadays trucks arrive with entire wall and roof truss sections prefabricated in a factory, where yep lots of the labor is done by robots and is computer controlled. All the on-site guys do is stick the modules together to frame the building. Framing a large house can take less than a day, and I recently watched a fairly large apartment complex go up in just a couple of weeks.

The autonomous flier method promises an interesting change in direction; go back to the pallets of lumber but automate the assembly on-site instead of tooling a factory. This would save a lot of transportation and custom factory tooling costs and probably the rental of the crane that's necessary to handle prefab sections. If the bots are cheap enough, I could see that offsetting their cost.

And of course you get lots of advantages with the automation; the bots don't get tired, don't need to take breaks, don't get heat stroke, don't fall off the roof, don't make mistakes, and can probably even see in the dark. They can more easily and more safely get to awkward spaces than a human worker could. And they can document everything they do.
posted by localroger at 8:29 AM on January 16, 2011 [6 favorites]

can they do it in real world situations?

Only if you can carefully light your construction site with IR flood lights and carefully position 20 IR cameras around the scene at all angles to read the position reflectors.
posted by Rhomboid at 8:36 AM on January 16, 2011 [4 favorites]

I think the main goal of this research group is youtube hits and comments on gizmodo.
posted by tempythethird at 8:42 AM on January 16, 2011 [2 favorites]

I think the main goal of this research group is youtube hits and comments on gizmodo.

See also: adding fuel to the collective metafilterian nightmare of quadrocopter-powered velociraptor attacks.
posted by kurosawa's pal at 8:48 AM on January 16, 2011 [2 favorites]

It's impressive to see, and very worthwhile to research, but hard to tell if the particular application will ever be economical in ordinary environments.

The limiting factors would probably be how economical the flying can be made, and whether the AI elements can scale up.

AI is notoriously susceptible to "combinatorial explosion", where things that work great on small experimental efforts don't work at all when sized up to something practically useful.
posted by philipy at 9:00 AM on January 16, 2011

Great, robot drones building housing for people who no longer work because robot drones are building housing for people...
posted by tommasz at 9:01 AM on January 16, 2011 [4 favorites]

Gibson was right; pretty soon all our buildings are going to be built by nanobots.
posted by KingEdRa at 9:01 AM on January 16, 2011

The future has all the automation we expect, but the ownership of it is problematic.
posted by yesster at 9:02 AM on January 16, 2011 [1 favorite]

Todo: Buy shotgun to defend future self from semi-autonomous quad-rotor UAV's.
posted by humanfont at 9:12 AM on January 16, 2011 [1 favorite]

WHAT'S UP SKYNET, HOW YA DOIN
posted by orville sash at 9:20 AM on January 16, 2011 [1 favorite]

I think they're fudging the concept of "autonomous" a bit.

I'd read "autonomous" as "no human in the loop, except maybe in a purely "executive" capacity giving very general orders."

It seems to me a fully autonomous system. But the quadrotors are slaved to the system's brain, with little capacity for independent action.
posted by ROU_Xenophobe at 9:24 AM on January 16, 2011

Big deal. I could build that.
posted by blue_beetle at 9:24 AM on January 16, 2011 [5 favorites]

mediareport: "The Aggressive Maneuvers for Autonomous Quadrotor Flight video is probably a better indicator of what these things will be used for.

"We tell the quadrotors what structure to build and they figure out the assembly plan and then build it."

I'd like to see them communicate with the quadrotors in Aargh!Tect-ese
posted by symbioid at 9:27 AM on January 16, 2011

And humans get one step closer to making themselves obsolete.
posted by oddman at 9:28 AM on January 16, 2011

The first thing this made me think of was the early classic building program SHRDLU from 40 years ago.
posted by buzzv at 9:31 AM on January 16, 2011

heh, a Karo covered sheet should fix these frisky robotic fly-abouts. if not, a sexy way to enjoy a sunday.
posted by clavdivs at 9:34 AM on January 16, 2011 [1 favorite]

Well, on a real building it would probably use less energy to have robotic cranes do the lifting, but I could see the work of lifting, placing and welding steel frames and prefab concrete parts being done largely by robots.

There's reasons you don't see construction sites buzzing with helicopter lifts right now, and they're very good reasons.
posted by zoogleplex at 10:02 AM on January 16, 2011 [2 favorites]

The last shot made me spit coffee out of my nose. They know we're watching!
posted by Popular Ethics at 11:23 AM on January 16, 2011

I have a great idea for a docking station where these robots can automatically swap batteries and continue working indefinitely. Do you think I can get a job at UPenn? (No seriously. I'll even take grad student wages to play with these toys).
posted by Popular Ethics at 11:25 AM on January 16, 2011

We can't even fully automate auto assembly lines, and that's without flying gyrocopters. Having been in the building trades, from houses and condos to processing plants and high rise structures, I can't see anything more practical than what can be prefabbed and installed by humans. There are too many instances where human judgement is required, even with prefab. Installing girders and beams might be one thing, but concrete stressing, post tension cables, re-bar height, pan decking, temp support forms, there's just too much. And to me that's the simplest type of high rise structure, and it says nothing of other trades like plumbing, electrical, and HVAC, which often have to work in conjunction with ironworkers and carpenters because of the comparatively delicate nature of their installations.
posted by l2p at 11:36 AM on January 16, 2011

l2p: I don't think anyone is suggesting the building trades are in any immediate danger - except possibly those in the magnetic lego industry. In the future though, if we continue to improve battery power and artificial intelligence, you will probably see these quadrotors buzzing around everywhere.
posted by Popular Ethics at 11:56 AM on January 16, 2011

l2p, the thing that struck me watching these was fault tolerance. 2x4's warp and bend, they shrink in dry air and swell in humidity. Truckloads of the wrong material arrive on site. The surveyor screwed up. It rains. The wind blows something over. Thieves steal stuff.

There are sooooooooooo many things involved in construction that screw up every other thing in the pipeline. The actual assembly, the construction, is the easy part.

The quadrotors are awesome as hell, no doubt. And this is just the beginning. But it's going to be a long time before we see anything like this in the building trade.

Also, I was picturing the grad students putting together the magnetized building blocks. Every piece had to be cut very precisely, every piece had to fit. I wonder how many times they screwed up, just making the pieces, much less the quadrotors. I am curious about the black joint pieces on the ends of the grey tubes - I wonder if they fabricated those too?
posted by Xoebe at 12:06 PM on January 16, 2011

This project is actually a practical demonstration of something James P. Hogan wrote about more than 30 years ago in The Two Faces of Tomorrow. No, it's not ready to be used to build real buildings yet, but it's only a matter of time. The significant thing in this demo is that no human programmer had to tell the worker bees "now pick this up and put it here;" instead, the human told the brain "build me something that looks like this" and the robots did the rest. That's actually quite sophisticated and well on the road to doing lots of other things we assume humans are necessary for now.
posted by localroger at 12:08 PM on January 16, 2011

There's a proof of concept in there, no doubt, localroger, which accounts for some of io9's breathlessness, anyway. But some of us look at videos like this and can't help but see the obvious limitations (and fudging and hype) as well as the possibilities. It's worth keeping both the breathlessness and the skepticism in mind. "It's just a matter of time" kinda hand-waves away the things Xoebe mentions.
posted by mediareport at 12:20 PM on January 16, 2011 [1 favorite]

Can you really call three a swarm?

And those blocks looked suspicously magnetic to me.
posted by ComfySofa at 12:21 PM on January 16, 2011 [1 favorite]

I've seen lots of tech predicted that did not come to pass, and lots that was not predicted that did come to pass. I don't think anybody really had an idea even 20 years ago of how powerful and ubiquitous computers would be. I can clearly recall an era when, as a fairly well educated person, I'd have argued at length that the very idea of non-supercomputers running faster than 10 to 20 MHz was absurd. Yes, megahertz. Or the idea that anything short of a government or multinational corporation would own a gigabyte of memory of any type. If you had told me that I'd one day have eight gigabytes on my keychain I would have laughed in your face.

The point of doing proof-of-concept work like this is to separate the pie in the sky stuff that seems likely but won't happen from the mundane stuff that seems impossible but does. I work in industry and I have seen the problems that arise in real-world implementation. These robots are not ready to rise to those challenges, but the demo here is showing that at least some of those challenges can in fact be met. Considering the ubiquity of computing power nowadays I'd guess that we are only power sources and actuators away from having a full scale version of this demo working with real construction materials.

Now, power sources have been a big failure mode for real tech compared to SF, but full scale implementations could be expected to use things like gasoline engines. I'd guess the real problem is the fragility and expense of actuators, the state of the art for which are now hydraulic. The actuators on the demobots are obviously not capable of grabbing a 2x4, dealing with the fact that it isn't straight, and nailing it in place.

But that doesn't mean that someone isn't building a system that could do those things right now. I am suspecting that something more like electrical servomotors with a lot of reduction gearing would be more practical, especially for airborne workers. Or possibly some chemistry based solution that just hasn't been invented yet will make powerful moves directed by small voltages as practical for robots as they are for animals. Considering that such systems are obviously physically possible (I'm using one example to type this), and considering how far machines have come in my lifetime, I would consider it foolish to bet against the machines perfecting a skill that a demo like this has shown to be at least theoretically within their capability.
posted by localroger at 1:01 PM on January 16, 2011

And those blocks looked suspicously magnetic to me.

The voiceover explicitly said that they're magnetic.

So, it's not impressive in terms of the complexity of construction, but I could see an unmanned space mission using something like this, with pre-manufactured building components.
posted by XMLicious at 1:43 PM on January 16, 2011 [1 favorite]

I wonder if you could use these to pick fruit.

Also, building stuff in the future almost certainly won't be done by actual flying robots, but rather crane type things or even humanoid robots.
posted by delmoi at 1:45 PM on January 16, 2011

I wonder if you could use these to pick fruit.

Sure. The rotors will slice off the fruit-bearing branches, and then the wheeled models will seek out and detach fruit on the ground that match selected color and shape parameters.

Sure you'll have to plant new trees after every harvest, and the majority of fruit will be discarded, but the IMPORTANT thing is Human workers! Will be! Obsolete!
posted by happyroach at 2:50 PM on January 16, 2011

I wonder if you could use these to pick fruit. --- or, after we've killed all the bees with our wifi and cell phone signals, maybe we can use them to pollinate them.
posted by crunchland at 3:03 PM on January 16, 2011 [1 favorite]

Todo: Buy shotgun to defend future self from semi-autonomous quad-rotor UAV's.

What are you going to do when the UAV is smart and fast enough to detect aimed weapons or dodge bullets and is armed with a laser?

Me? I'm stockpiling microwave ovens. The bigger and higher the wattage the better. I figure the robots will try to ban 'em around the time 2020 rolls around and I'll make mint on the black market when people figure you can make a sort of slow EMP rifle out of them and fry robot brains. Plus they're just the ticket for warming up a tasty ratburger while on the run.
posted by loquacious at 6:28 PM on January 16, 2011

Using the warehouse sorting and supply matching/delivery of RMT and Kiva to build the crate, the traffic management and freighting possible with the robotic google car and this robotic drone building technology, we will have autonomous industrial development within a few decades.
posted by AndrewKemendo at 7:32 PM on January 16, 2011

I could see an unmanned space mission using something like this, with pre-manufactured building components.

So that would have to be the rocket version, right? Then we're all definitely dead.
posted by imperium at 8:22 PM on January 16, 2011

Maybe this tech will develope into something to take away the pain of never getting the jet packs we were promised...
posted by Redhush at 9:59 PM on January 16, 2011

Probably a phantom from the cold war, I saw these things building bridges over rivers and canyons. Possibly workable even from the exact kind of parts shown.
posted by Goofyy at 7:08 AM on January 17, 2011

Where did I put my old tennis racquet?

And why isn't "racquet" recognized by spell-check?
posted by Faint of Butt at 7:39 AM on January 17, 2011

Autofac, an early SF treatment of self-replicating machines.

You will be more likely to click if I also say: "by Philip K. Dick (1954)".
posted by Herodios at 11:38 AM on January 17, 2011

Metafilter: a buzzing, mildly disturbing haze

sorry
posted by hattifattener at 12:14 AM on January 18, 2011