Tom's Looks at Two DARPA Grand Challengers

skeeball writes "As a follow-up to this article, Tom's Hardware has a behind the scenes article on two of the teams competing in the DARPA Grand Challenge 2005. "The Defense Advanced Research Projects Agency (DARPA) hosted the first Grand Challenge Project last year, offering a reward of $1 million. This year, the prize money has been doubled, making the competition all the more interesting.""

Comments

Article Link

[Kozz]

Too bad the submitter didn't Link the Article itself.

Article link

[rdwald]

Right here.

They also pictured the wrong vehicle

[mtrisk]

I know for a fact that the Hummer pictured is the newer H1ghlander, which CMU acquired after the first race. Their first vehicle, Sandstorm, is an older 1986 military hummer with the top shaved off. Both vehicles will be competing in the upcoming challenge.

The Line up is not complete!

[locokamil]

Just so all you geeks out there know, the final lineup for the DARPA GC has not been decided yet.

Several teams with extremely competent designs will be site tested by DARPA officials during the week of August 15th.

Keep your eyes on the Princeton University team (disclaimer: I'm heavily involved in developing software and lasers for them). We barely missed the cut in April, but we're gearing up for the second round of qualification tests in August. We've taken an approach very different from the other teams (we love to hate on CMU and Stanford for their bloated budgets and hardware), insofar that we've refused to let our budget rise over $40000. Furthermore, our work is done ENTIRELY by a team of six undergraduates, three of whom are freshmen (I'm the only senior on the team).

Is this a shameless plug for the Princeton team? Hell yeah. But I just felt that it should be known that there are people in this competition who are trying to THINK their way out of the maze instead of BUYING their way out of it.

That's so Tom's Hardware

[Animats]

That's so Tom's Hardware. "7 Pentium M CPUs!", and no word about the algorithms. They could have at least said more about the sensors. Actually, everybody's sensors suck. The radars can't profile terrain, the LIDAR units are only line scanners, the stereo vision systems have trouble locking up on dirt, and the vision systems are a long way from being intelligent. True 3D LIDAR is coming, but not this year. The Grand Challenge rules prohibit the use of the best available 3D LIDAR system, because it was developed with Government funding and wasn't available by August of last year.

So we have a line-scanning LIDAR on a tilt head, like CMU, which is an adequate but bulky solution..

We have two industrial Pentium 4 machines running QNX, on our Grand Challenge entry, along with five Galil programmable motor controllers. We have room for 3 CPUs, but the compute load fit on two of them, so we took the third one out.

Technically, QNX was an excellent choice, but because few people know it and many don't want to learn it, using it has made recruiting difficult.

Re:That's so Tom's Hardware

[Animats]

Why is the vision processing so poor?

Because, despite decades of work, vision processing of unstructured scenes still sucks.

There are things that work in computer vision. You can do stereo, if the image has strong edges in it. You can pick out big moving objects. You can find the horizon. You can work out your own positional movements from video. You can find faces, align, and recognize them, sort of. You can find known objects in any orientation (which is very useful in industrial systems.) You can follow roads.

Beyond that, not much works.

How did they manage to write

[btempleton]

A long article, full of photos and pointless details, and yet after a dozen pages tell me nothing?

Amazing.

Some Errors

[kf6auf]

A couple of corrections:
Two Teams Compete for Best Robot Car in DARPA Challenge
I would just like to point out that the headline is off by over an order of magnitude! We here at Caltech and many other people at many other schools are also competing to have the best robot car too!

The vehicles are given no more than 10 hours to complete the 176-mile route, which will be kept secret until the beginning of the race.
The map is given to each of the teams several (3?) hours prior to the start of the race. One result of this subtle difference is that teams can program a general path into the vehicle and have it deviate from it only as necessary instead of just popping the DVD into the computer and having the computer do everything.

Something that people should keep in mind is that many schools are using the program as a learning experience and solely out to win the competition, but provide their students with not only a limited budget, but make them do things themselves even when it might be cheaper and undoubtedly easier to simply buy premade parts elsewhere. The use to the military will not be the machines that are built with all sorts of fancy equipment and sensors that Tom's Hardware liked to talk about, but the algorithms and techniques that are used to guide the vehicles.

Scott

Re:The developments won't be used for "defence".

[antispam_ben]

From Google.com:
the best defense is a good offence: 600,000
the best offence is a good defense: 242,000

I'm surprised the second one give so many in proportion to the first even though it's over 2 to 1, but it's surely because all the words, regardless of order, appear on so many pages. Redoing with quotes:

"the best defense is a good offence" about 1,940
"the best offence is a good defense" "about 91"

Yes, "the best defense is a good offence" wins again, this time by over an order of magnitude.

And watch out, because We Arrogant Americans are more offensive (all puns intended) than ever. Someone knocked down Our Towers, and We're pissed.