MIT takes on DARPA's Urban Challenge

MIT -- long known for winning all sorts of competitions involving modern technology -- has entered into another heated contest which will test its mettle against a wide-variety of opponents... and that contest is the DARPA Urban Challenge. DARPA, who we know and love for its fantastic flights of scientific fancy (see the shoot-through shield and laser-guided bullets) has posed a challenge to contestants to create an autonomous auto (AKA a self-controlled vehicle) which can traverse an urban landscape (such as city streets) all by its lonesome. To create such a vehicle, a team at MIT has taken a typical Land Rover, outfitted it with 40 CPU "cores", high-end GPS receivers, inertial sensors, laser scanners called LIDAR (light detection and ranging), highly sensitive odometers, and a slew of video cameras. The team hopes to pool all of these disparate sensing technologies into a cohesive whole which will imbue their vehicle the preternatural ability to operate on its own in an urban setting. These are truly exciting times to be a car.

[jd @ Sep 4th 2007 1:30PM]

but will it be able to slam on the horn and flip off the driver when someone cuts it off

[Frank @ Sep 4th 2007 1:37PM]

WTF - This isn't even close to being news. Stanford already won the contest. Not to mention that it was the second year they held the competition. Why does MIT get special treatment and get paraded around just for entering the race?

[furby @ Sep 4th 2007 1:46PM]

Think you've become a little confused Frank. Yes, Stanford did win the DARPA unmaned vehicle contest but this is however the URBAN challenge which has yet to be held for the first time.

MITs looks to be the most technologically beefy. Uber geekness

[Frank @ Sep 4th 2007 1:58PM]

I understand what you mean, but the URBAN challenge is the exact same thing with different tasks. Instead of going around rocks and cliffs, not they must navigate STOP signs and merging traffic. Yes, I understand that the challenge has been upped, I'd be upset if they didn't make it harder. However, this team doesn't really seem to have employed anything new to merit being mentioned.

I don't mean this to bash MIT at all, I just wondered why it should even be an article.

[Jon @ Sep 5th 2007 7:21PM]

Have you ever tried programming a robot to do any of those things? I have. Completely different situations. In the first Grand Challenge, the vehicles could basically go anywhere they wanted to as long as they didn't get stuck. It didn't matter if to go around a rock you went an extra 50 feet in one direction. The Urban Challenge is different. In the Urban Challenge if you take an extra 50 feet to go around something you're driving through a Starbucks.

That is why this is a different contest than the Grand Challenge. As to why the article features MIT, I assume it is because MIT just announced that they are participating in the Urban Challenge, where they had not previously. You know, it was something different, new. Hence the name "news".

[Karim @ Sep 4th 2007 2:28PM]

Actually, this will be the *third* DARPA autonomous vehicle challenge. Nobody finished the first time (2004), and Stanford's "Stanley" won the second time (2005).

Despite their technology and robotics expertise, MIT didn't even enter the previous two challenges, which is why it's kinda newsworthy that they're in this one. According to the article, the only reason they're in it now is because a bunch of grad students banged on the table and said they wanted to play, too :-)

Seems like a lot of these vehicles are festooned with LIDAR from SICK AG....

[Frank @ Sep 4th 2007 2:49PM]

SICK is the biggest manufacturer of LIDAR units. 1 year before the second Grand Challenge, Carnegie Mellon put in a purchase to SICK for way more units than were necessary in order to put them on back-order. Other teams were seriously crippled by the lack of physical units for months. Gave them a nice head start over a majority of the teams.

[Karim @ Sep 4th 2007 2:56PM]

Thanks Frank, didn't know that.

Seems like Bad Form to trip up one's opponents like that! No wonder CMU lost -- putting the SICK LIDARs into backorder must have given them bad car-ma. *groan*

[MasterCKO @ Sep 4th 2007 4:55PM]

wow, that's some dirty pool from Carnegie Mellon. Too bad those suckahs lost. I hope that it taught them a lesson.

[Grindboy @ Sep 4th 2007 1:40PM]

I'm liking the "disable" button, seems were not going to be over thrown by this robot!

[JJ @ Sep 4th 2007 2:19PM]

That only works if you can get to the button before it gets to you!

[Shane @ Sep 4th 2007 3:30PM]

"In Soviet Russia, car drives you!"

[Adam @ Sep 5th 2007 12:23PM]

Yeah, I would have put the button towards the back of the car. I wouldn't want to have to get in front of it if it is running out of control.

[Dustin @ Sep 4th 2007 2:00PM]

Well, perhaps if they replaced all the electrical components, they can keep it running for more than a few miles.

[Andrew Koyfman @ Sep 4th 2007 2:03PM]

...or you can read the full story and see some cool pictures of the car on the xconomy site

http://www.xconomy.com/2007/09/04/mit-plans-to-win-darpa-robot-car-challenge/

[Richard @ Sep 4th 2007 2:08PM]

LIDAR is the child of Laser and Radar...

Bred for its skills in magic.

GOD!

[kiwhan @ Sep 4th 2007 2:10PM]

with all that cpu's I guess they will put ac to a good use.

[Raymond @ Sep 4th 2007 2:20PM]

They better make sure that they don't use any Sony car batteries in there! :p

[Admiral @ Sep 4th 2007 2:39PM]

wasn't LIDAR something from Space: Above and Beyond?!

[Zach @ Sep 4th 2007 3:55PM]

Actually, the article has it wrong. The car itself was built out by a team of undergraduates from Olin College (http://www.olin.edu). Olin students did the hardware design, build, purchase and integration. MIT was responsible for software including perception, controls and planning. Oh, and a handful of CANBus interface boards.

[Carol @ Sep 4th 2007 4:47PM]

Sure hope I win a prize today:)

[Carol @ Sep 4th 2007 4:48PM]

Thanks for sending and wish me well....

[Alex @ Sep 4th 2007 5:29PM]

NC State's entry (a Lotus coupe) looks cooler.

[macbrain @ Sep 4th 2007 6:42PM]

Georgia Tech's Entry (see http://www.sting-racing.org/ )
is a Porsche Cayenne, and it looks pretty awesome too. I've met some of the people working the GT team and have checked out the Cayenne first-hand.
I look forward to seeing who has the best firmware and software in the final reace, because thats who is going to win. Everyone is using similar systems for sensing and driving.
http://www.sting-racing.org/

[Thomas @ Sep 4th 2007 5:29PM]

NC State University's Lotus Elise DARPA urban challenge vehicle FTW!!!

[lordofthefobs+engadget @ Sep 4th 2007 7:57PM]

I was actually on the team for one of the competing team. (we actually made to the NQE)
I have to disagree with frank that this is same as the other two challenges since this is held in urban setting (thus URBAN Challenge). In previous challenges each vehicle ran through the course one at a time, but this one, they will be required to run along side with each others as well as vehicles driven by human to create variety of situations for them to react to. Also in first two challenges, there was single line course with dense GPS dots to guide the vehicles, but in URBAN Challenge, it's a map of a city and the vehicle must find it's way from one GPS position to the other which are sparsely placed. Quite different situation, especially from programmer's perspective.

Though i agree that this isn't news at all since I worked on the project more than 8 months ago and I'm pretty sure MIT didn't just decide to join the race.

[Zach @ Sep 5th 2007 12:09AM]

MIT and Olin started working on a vehicle in July of 2006.