DARPA Wants Help Building a Drone That Flies Like a Hawk 42
DillyTonto writes DARPA has put out a call for ideas on how to build a fast, autonomous, maneuverable UAV that can fly up to 45 mph, navigate without assistance from humans or GPS into and through buildings that are a labyrinth of stairwells, small rooms, narrow hallways and terrorists. DARPA wants this drone to fly like the bird in this awesome hawk POV video that shows it shooting through gaps narrow enough it has to tuck its wings to get through. If you can watch the video without thinking of the forest moon of Endor, there may be some movies you should watch over the holidays.
Yes. (Score:2)
Let us all help the Sardaukar build a better ornithopter.
For the Empire!
Re: (Score:2)
Really cool footage.
power (Score:1)
Re: (Score:3, Interesting)
Bird brains don't calculate the physics from first principles, and that's not really necessary even for a robot. The key is a fast feedback loop: am I falling? Do that thing that increases the lift from my wings a bit; am I going up too much? Decrease lift; am I going to break my wings off on that tree that's coming up in 0.6 seconds? Increase lift a bit for 0.4 seconds then tuck wings in; etc. etc. I think the key difficulty here will be realtime object recognition, made harder by rapidly changing lighting
Re: (Score:2)
Current miniature LIDAR is too heavy and uses too much power. I would guess IR (kinect) but even that is pretty damn hefty.
Re:power (Score:4, Interesting)
There are present day computers powerful enough to calculate the physics for that
That is only needed during the design and testing, when a lot of CFD simulations will be needed. But once the bird is deployed, it will just use a lookup table. When you are walking down the street, you don't use physics to calculate the exact length of stride to optimally place your foot. Instead, you just take a step more-or-less like the last one, and then compensate any over/under stride on the next step.
Re: (Score:2)
I'm sure that it will use a lookup table, but it's also going to have to build those tables dynamically because due to the nature of mechanical devices. 1) no two are identical and 2) they wear while in use, especially while running near the edge of materials technology, further exacerbating point #1. You really do need a learning control system if you're not in a perfect world, or doing something hilariously easy — which this isn't.
Re: power (Score:1)
Take the act of catching a pop-fly ball. One of the heuristics that is believed (through experiments) to be used by humans is to maintain the angle of the ball with a fixed point, such as the horizon. If the angle closes, run forward. If it grows, run backwards. If you continue doing this then you position yourself to where the ball will land without calculating parabolas, speed, etc.
Brains learn and acquire all kinds of simple heuristics like this. It's fascinating.
You can program many of these heuristics
In other words... (Score:2)
...it is going to require TWO Gumstix.
Oh, good (Score:3)
Just what we need. When they can stamp this out for pennies, they will truly be able to spread 'democracy' to every corner of the earth.
Re: (Score:2)
Alas, we get an abundance of cool tech from military exploits, and this may just be the day's Polyanna cheer, but I see us slowly evolving into a society of post-militaristic peoples. Discounting the Cold War and its few proxy wars in third world nations, no large war between superpowers has occurred in nearly 70 years.
Globally, poverty levels and deaths of children under five have steadily declined. HoHoHo... it's not
impressive hawk. I know just the guy (Score:2)
That hawk video is impressive. I know just the guy for the job, smart as heck and he works on machine vision and drones, but early proposals have to be done January 6th.
Re: (Score:3)
This is typical DARPA MO. I don't know when the RFP was issued, but they are known for ridiculously short lead times - because the reason the RFP exists at all is because someone has a working (or damn close to it) prototype.
Re: (Score:1)
I suspect that many of the military-funded RFPs are like that. I was part of two teams that submitted proposals for a CDMRP (Congressionally Directed Medical Research Program) operated by the DoD for the US Congress. Details of the winning proposals were made public. Our two submissions were superior in every way to the winning proposals, from meeting desired features of the proposals, to levels of expertise, to realistic milestones, to publication records, to institutional support and resources, to vari
Re: (Score:2)
Hawk in slo-mo. http://imgur.com/gallery/Q3njr... [imgur.com]
They're joking right? (Score:2)
We can barely build devices that can navigate a house at a crawl (stairs, furniture, etc) and they want to create something that can navigate thousands of random obstacles at high speed? Real world environments (changing light levels, leafs, webs, wind, etc) are going to play havoc with anything that they do build.
Re: (Score:2)
They're not joking about wanting it.
Building it might still be a ways away.
Re: (Score:1)
Only until the hawks start shooting back.
Hawks tend to have MUCH better aim than people.
Damnit! And my falcon is just done! (Score:1)
Why didn't I build a hawk... :-(
Why not a gull (Score:2)
Why not have a drone that flies like a seagull, after all they are more common, and would be less likely to be noticed.
(and still deadly if you have enough of them (The Birds)
Re: (Score:1)
Only if the enemy is hiding in a trawler.
...or McDonalds.
What do you get the agency that has everything? (Score:2)
Re: (Score:2)
Probably for the same reasons that NIH doesn't produce torpedo. Just a guess.
Prototype is already built (Score:4, Informative)
Having working in the incestuous realm of DARPA, what I can tell you is that somebody already has a proof of concept pretty much ready to go. Since I have to be billed as "fair" they put out the RFP (with a ridiculously short deadline - which is no problem for the folks with the almost working prototype). This company is in cahoots with the DARPA PM, who in all probability worked at that company in the not too distant past.
This is how it is at DARPA.