MIT Drone Finds Its Way Using Kinect Vision 77
garymortimer writes "This MIT multicopter is able to fly in GPS denied environments by creating a 3D map of its surroundings on the fly (no pun intended) based on point clouds generated by a Kinect. Also pretty handy for avoiding trees and other obstacles outside at low level. This processing is onboard, unlike other systems that depend on motion capture rigs."
ROFL Copter (Score:4, Funny)
Remote Objective Forensic Lead Copter.
Re:ROFLMFAO Copter (Score:2)
Remote Objective Forensic Lead MIT Funded And Open-source Copter - Well, we can hope.
Re: (Score:2)
Humor takes a dive
No, but that will be the case with the Kinect submarine.
Thank you, thank you! I'll be here all week!
How much would this have cost... (Score:3, Insightful)
before the Kinect came out?
Frankly, I'm pretty sure most of this stuff *already* existed before the Kinect came out. I'm just wondering exactly how much Microsoft mass producing the hardware saved the researchers on having to make their own devices.
I imagine it's quite a bit since we went from "hardly any talk about robot vision except about how goddamned hard it is" to "A NEW ROBOT/CAMERA/WHAT HAVE YOU USING THE KINECT AS ITS EYEBALLS!!!!!" every week once the Kinect was out for a month.
I'm complaining, yeah, but I'm also trying to put things into perspective. These aren't leaps and bounds of tech, they're just leaps and bounds of people actually making use of the damned hardware since they don't have to sell their second child to pay for it.
Crowdsourcing I guess? Instead of 2-3 sterile environments trying to figure out robot vision algorithms, now a bajillion institutes and hobbyists are working on the problem. I guess one way or the other, this is a big step in the fields of autonomous robotics and so on. Now if only we could work on the goddamned speech recognition software... The hardware for that's been around for ages and yet people haven't been messing about with it as much as they do with the Kinect.
Where's the algorithm? (Score:3)
These aren't leaps and bounds of tech, they're just leaps and bounds of people actually making use of the damned hardware since they don't have to sell their second child to pay for it.
Crowdsourcing I guess? Instead of 2-3 sterile environments trying to figure out robot vision algorithms, now a bajillion institutes and hobbyists are working on the problem
Working on *which* problem exactly?
I don't see any of those people working on artificial vision algorithms. What they are doing is using a "good enough" vision algorithm which comes bundled with the hardware to develop applications for artificial vision.
Re: (Score:2)
Re: (Score:3)
How much would this have cost before the Kinect came out?
A lot. Please see an earlier comment I made [slashdot.org] on the exact same issue: prior to the launch of the Kinect, if you wanted an accurate depth map on a robot, you had to use LIDAR sensors. Not only are those expensive, but they're both heavy and somewhat unreliable (since they have moving parts in the form of rapidly spinning mirrors) which rules out mounting them on a copter. There are ways to estimate depth from 2D images alone, but then accuracy suffers immensely. With Kinect you get the best of both worlds.
Re: (Score:3)
Actually 2 years back a quadcopter were flying in the same hall [diydrones.com] using LIDAR. I think I found a price of ~$5000 for the units they used, but my memory might be an order of magnitude wrong.
You can get along way with 2D images, a lot further than you think, of course with this tech it's easier.
Re: (Score:3)
The best LIDAR you can buy for the size is probably the Hokuyo [hokuyo-aut.jp] UTM and URG. Definitely small enough and light enough to mount on a copter. They'll set you back $2000-$6000 though. To get a 3D image out of them you need to pivot them on the y-axis. They also generate a lot of heat, and suck up a lot of energy.
Crowdsourcing I guess? Instead of 2-3 sterile environments trying to figure out robot vision algorithms, now a bajillion institutes and hobbyists are working on the problem.
Not really. The people who have been doing the amazing research are still doing it; the really exciting stuff is still coming out of CMU, MIT, Stanford, etc. It's just that now the hobbyist can implemen
Re: (Score:2)
strangely enough, you can blame Microsoft for that!!! They got the technology from L&H in a technology swap, incorporated speech recognition into word etc. and then sat on it and never bothered to do anything with it... meanwhile L&H went belly up as there was no reason for those needing speech recognition software to buy it separa
Re: (Score:2)
They also bought up Entropic ..then sort of open sourced it, then abandoned it.
Nuance has bought up most of the speech recognition companies and they have a fairly strong product, but there just haven't been any big leaps in the field since HMM's were first used.
Re: (Score:1)
Yeah, they should be punished for it...
I don't buy it (Score:3)
This MIT multicopter is able to fly in GPS denied environments by creating a 3D map of its surroundings on the fly (no pun intended) based on point clouds generated by a Kinect.
No pun intended = "I came up with a weak pun, but I don't want people to blame me for it"
Re: (Score:2)
Re: (Score:3)
Re: (Score:1)
Nobody cares if your puns were intended.
I agree no one cares if puns are intended or not but the link means that poster is a far bigger douche and idiot than anyone who actually says their pun is intended or not. Seriously, if someone pointing out their puns is enough to get you this cranked, you're already a cherry on a loser sundae.
The phrase "no pun intended" makes me want to band saw my dick off just in case I'm the last man alive and I have to risk repopulating Earth with some moron's incapable vagina. There's no such thing as an unintentional pun; the act of typing the phrase "no pun intended" makes it intentional. If your pun truly wasn't intended, then why didn't you erase it and write something else, asshole?
Seriously, this guy is such a douche, he ignorantly and very successfully argues everyone should go out of their way to make puns, and point them out, just so we can see this douche mutilate himself.
Aside from that,
Re: (Score:1)
So you're saying he's above the bar to which he measures others. That alone invalidates your position. If he is above his own remarks, and lame bemusement, doesn't that mean he's just a blow hard who doesn't have anything to contribute in the first place? That's rhetorical...which brings us full circle and back to my original post.
Re: (Score:2)
Tip of the hat to Microsoft (Score:4, Insightful)
Re: (Score:1)
Lets not forget MS didn't invent it, they wrapped in a package and sold it. I don't think they could really have a hope stopping any API.
What corporation would go after a completely home brown API package for a physical device?
Re:Tip of the hat to Microsoft (Score:4, Insightful)
Well, they didn't actively encrypt or obfuscate it. So, at least they decided to play nicely and let people develop for this.
Well, I know you're not new here ... but, really? I think Sony would sue people into oblivion for something like this at the drop of a hat. It seems to be increasingly the norm for some corporations to more or less say that it is illegal to use hardware they sold you in a way they don't approve of. Seems to be Cuecat did it several years ago.
And, corporate hand-wringing aside, I think it's really cool that people are using the Kinect to make things like this. From the sounds of it, this is opening up lots of interesting avenues for researchers to be able to build cool things.
I for one, welcome our new fully autonomous 3D navigating multicopters. :-P
Re: (Score:2)
First off, I don't need to do research to give my opinion ... I really do think that they'd be quite likely to trot out the lawyers under most situations because, in my opinion, Sony is a bunch of litigious bastards.
Sony is giving you a program you put on your PS3 which acts as a server whi
Re: (Score:1)
Having your own opinion is fine and dandy.... You are not, however, entitled to your own set of facts... You can connect a PS3 controller (be it a DS3 or a Move) to your PC, and no one is going to particularly give a damn about it. Why? Because a) That's not going to affect the money they get from game licensing, which is where they get most of their revenue, and b) they've been fine with letting this happen for years.
The times that they are litigious bastards are when someone does something that's doing s
Re: (Score:2)
Move is nothing more than a wand and a webcam. There's nothing in it that's really too much different than a Wii (if anything, it's an evolution of the Wiimote's IR camera sensor).
People have been doing image recognition prior to this for ages now, and any depth mapping via cameras has involved dual cameras for stereoscopic vision.
Sony's Move really doesn't offer researchers too much to play with that they don't already have - recognizing a glowing colored ball in 3D space? We can already do it (and it's ev
Re: (Score:2)
If it could easily do VGA (without heavy interpolation) then that would suggest it can detect depth with a point resolution at or even near the infrared sensor resolution. This can't be the case since there can only depth information in each projected point. Since there *must* be black space between the projected points to be able to detect a point (zero black would be a white image!), this means you could never be more than some fraction of the sensor resolution. Even if you're doing sub-pixel detection, y
Re: (Score:1)
I for one, welcome our new fully autonomous 3D navigating multicopters. :-P
But how fast can it make the calculations for the jump to light speed?
Re: (Score:1)
Sony?
Re: (Score:2)
Lets not forget MS didn't invent it, they wrapped in a package and sold it. I don't think they could really have a hope stopping any API.
What corporation would go after a completely home brown API package for a physical device?
sony?
Re: (Score:2)
It seems we are not talking about the same Microsoft here. The company I know and don't trust will ony care about how much money it makes after they make sure nobody can use it to compete with them on any of their markets.
And yes, their API are only 'open' if you use Windows. I'm still surprised they didn't go into a court after the people that published the really open API. They are in fact doing something nice (in the 'non-bad' definition of 'nice') and unusual.
Re: (Score:1)
I can not think of any cases where they took someone into court for 'using their stuff in a strange way'. Try to steal from them and they drag you into court in a heartbeat. They usually are fairly up front about it 'use our documented API and we will support you, dont use it and we might change it in v2 so be ready for that'. They even usually bend over backwards to make things backwards compatible. Name one other company that does that? Not even many open source projects can claim that.
They just make
Re: (Score:2)
The 'Thou shall not benchmark our products!' Microsoft? Yes, they've sued companies for releasing benchmarks of their products, for releasing interoperating kits that let their products talk with concurrent ones, and are known to put small business out of the market by threatening to sue for the weardest reasons.
They are doing something unusual with the Kinetic.
Re: (Score:1)
I've always had a terrible distaste for Microsoft for too many reasons to list.
BUT, I have to say that I am impressed with Microsoft on their approach of this one. All around, I really look at this and think that this is the attitude they should have adopted a very, very long time ago, and they produced a very outstanding product.
I'm happy to give them a sincere 'pat on the back' on this and try to find ways to positively reinforce more of these wonderful devices.
No pun intended (Score:2)
Puns are droll. There's no need to apologise for them, unless they are truly poor. There's no need to point them out - it's like explaining a bad joke.
Re: (Score:2)
Puns are awesome. I hope I don't ever get to a stage of life where I am too pretentious to have a sense of humor.
You are correct in that you shouldn't point them out. Clever ones stay hidden so only people who are thinking about what they are reading can find them.
Re: (Score:2)
And dont call me Shirley.
Re: (Score:1)
The very best puns annoy the audience with how bad they are.
(Partially) misleading summary. (Score:4, Informative)
According to TFV the "multicopter" uses onboard processing to find reference points between successive video frames, which it then uses to determine how fast and how far the drone moves.
However, the actual map generation and navigation is handled by a separate computer.
Re: (Score:2)
So why didn't Kinect and Wii come from MIT? (Score:2)
Here's a question: All these university researchers are using Wii controllers and Kinect devices to do research. How come they didn't invent this stuff themselves? Is it because they couldn't think in mass-market terms so their solutions were overly complex and expensive?
Now here's another question: Why don't Nintendo and Microsoft make developers kits for their devices sans game console? Or even better, make the open source (I can dream)?
Re: (Score:3)
It's because the researchers couldn't build factories in China to pump out their equipment by the million. Mass production, economies of scale and all that.
If you hired a team of engineers to invent and build you a one-off car, how much do you think it would cost?
Re: (Score:3)
Depends on the car. Also it depends on what that car could do that other cars couldn't, say something like "fly", and how cheaply it could be made and how the usage could be expanded to markets not quite the intended one. Prototyping is expensive, and I'm sure that Microsoft's prototypes of the Kenect were very expensive, but they had a plan to sell millions of the thing, making it less expensive for
Re: (Score:2)
It is not that they couldn't think in mass market terms. It is that they couldn't reach a mass market. And that does make their solution overly expensive, but not more complex.
Now, about your second question, I have no idea. I hope somebody will comercialise stand-alone versions of the chips that go into those devices. For the Wii it already happened, you don't need to use a WiiMote anymore, you can simply buy the chips.
Re:So why didn't Kinect and Wii come from MIT? (Score:5, Insightful)
Here's a question: All these university researchers are using Wii controllers and Kinect devices to do research. How come they didn't invent this stuff themselves?
They did, long ago. This sort of thing is usually developed as a demonstrated concept in an academic lab. Proof of concept devices are ungainly, expensive, and incomplete, but they show that the academics' idea works and has potential for further development. Then companies take that public domain knowledge and make products from it. These are much cheaper and better packaged. Now the products are part of our technology culture, so they're natural tools for the next round of academics to use as tools for further innovation.
Is it because they couldn't think in mass-market terms so their solutions were overly complex and expensive?
Now here's another question: Why don't Nintendo and Microsoft make developers kits for their devices sans game console? Or even better, make the open source (I can dream)?
Microsoft is releasing Kinect for PC with an SDK supposedly intended to enable this sort of development. They chose to leave the output from the existing Kinect interface unencrypted, too, resulting in the already large homebrew scene demonstrated in storied like this.
Re: (Score:2)
Which makes sense when you make your money from selling the hardware. If I'm a hardware maker, I want everyone able to use my hardware so I sell more units. So I create the SDK and say "here ya go" and then provide support for it. I know that's why at my last company we ended up using certain hardware over others because the ones we chose had good SDK, good documentation, and finally good development support. For one label printer manufacture, that resulted in at least 13500 units of their label printer
Can't wait! (Score:2)
I imagine we'll be seeing this used in autonomous vehicles pretty soon. Having four kinects on board with some basic processing could definitely replace some or all of the bulky LIDAR systems.
Re: (Score:2)
I think a few kinetic driven vehicles can't work near each other. So, ok, they could be used in autonomous vehicles, but the applications are quite limited.
In a few years I expect vision systems based on natural light to be way cheaper than they are now, and by then somebody will probably make a low power chip.
No, range and lighting conditions too limited (Score:2)
I imagine we'll be seeing this used in autonomous vehicles pretty soon. Having four kinects on board with some basic processing could definitely replace some or all of the bulky LIDAR systems.
Won't work. The Kinect's structured light approach only works when you can overpower ambient light with the Kinect's little infrared source. That's indoor or night only, and short range. LIDAR units work in bright sunlight because, for a few nanoseconds per cycle, they're brighter than the sun in their portion of the spectrum.
Short range LIDAR units aren't bulky. They're smaller than the Kinect. [mesa-imaging.ch] For more range, you need bigger collecting optics.
Multiple-camera systems can now do stereo reasonably well
what will they come up with next! (Score:2)
What did you say, MIT made a roflcopter?
Genius!
Haha, lol Kinect (Score:2)
Just don't move too fast or it will never see you.
The Kinect's update speed is ludicrously slow for a flying machine's collision avoidance system. A bird can potentially fly past the device between frames, and at high speeds there's no way it could keep up object avoidance.
Good luck, but you'll probably have to build your own.
Better than the Google van! (Score:2)
Maybe Google can replace the Google van with Google drones that use this technology. Can you imagine seeing a dozen of these flying through your neighborhood?
BIM (Score:1)
Can I get this for my Roomba.... (Score:2)
Re: (Score:2)
Laziness application (Score:1)