Drive, Stanley, Drive

by Kieran Healy on October 8, 2005

The DARPA grand challenge is a 175-mile race for autonomous vehicles—cars or trucks that drive themselves. It’s currently underway out in the Mojave desert in Nevada. The teams in charge of the vehicles were told the route early this morning, and the vehicles set off a few hours later. The course is tough, with obstacles and sections (like tunnels) where it’s impossible to use GPS devices. Last year, the challenge was a bit of a disaster, with no team managing more than a few miles, and many vehicles failing completely. This year only a few have dropped out and three have already covered almost 100 miles of the course. The DARPA Challenge Homepage has live updates of all the vehicles. There are three main contenders: H1ghlander (a Hummer H1, you see) and Sandstorm are both run by Red Team Racing, based at Carnegie Mellon and sponsored in part by some big defense contractors like Boeing and Harris. The other challenger (currently running H1ghlander a close second) is Stanley, a modified Volkswagen Touareg run by a team from Stanford. Confirming an observation Dave Barry makes somewhere (about how men are able to sit down in front of a TV showing a tennis match between two anonymous Eastern Europeans from the 1980s and instantly begin supporting one of them), it took about 30 seconds for me to become a strong Stanleyite.

Of course, DARPA is kind of a hit-and-miss agency: sometimes it helps invent the Internet, sometimes suitcase nukes or microwave-based riot-control/torture devices. So the Grand Challenge can be seen either as the precursor of safe autopilot for cars or the embryo form of one of these. The technology behind the vehicles is pretty cool. It reminds me (as the film Apollo 13 reminded me) of an old Punch cartoon of two hairy, flea-bitten cavemen standing in front of the gorgeous cave paintings of Lascaux. “Art, art, art,” says one to the other. “When are we going to get some engineers?”

{ 9 comments }

1

Brackdurf 10.08.05 at 5:58 pm

Well then, you’re in luck, as by my guess, around 5:00 est, Stanley was the unofficial winner (it was my favorite too). I think Stanley actually passed at least one of the Red Team trucks along the way. Oddly, the three lead cars–the two Red Team trucks and Stanley–seemed to be all within about five minutes of each other (they start staggered, but those were the first three starters, I believe). The odd thing about following the race on the clever though slightly convoluted grandchallenge.org flash site is that I kept thinking, dang it, Red Teams 1 and 2, go a little faster! You’re only a couple minutes behind the lead, with millions of dollars on the line. Can’t you just speed it up a bit and take a few risks?! But of course the robots can’t see each other, and can’t (yet) do that extra little spurt that makes human (or horse) competition so fun to watch, and so much more likely to end up with a photo finish.

2

Darcy 10.08.05 at 6:03 pm

I’m *at* CMU, and I’m a Stanleyite, too. There’s something unsatisfying rooting for a team that has the largest financial backing and years of practice in the challenge.

3

Brackdurf 10.08.05 at 6:22 pm

Correction–there seems to be an issue with counting in time when the robots were “paused” to prevent passing each other… No official word on the winner, and the hard thing seems to be both subtracting the pause times, and taking into account the cost of time to accelerate back to speed (which depends on the number of pauses, not just to total paused time). Who would ever think a robot race would descend into technicalities…

4

Kieran Healy 10.08.05 at 9:37 pm

This better not be a sneaky way to make sure that the Red Team win, is all I’m saying. Stanley rules.

5

Ruchira Datta 10.08.05 at 10:13 pm

What’s up with this Gray Insurance Company team, which also seems to have been one of the few to finish? Anyone heard of them before?

6

Happy Expat 10.09.05 at 7:30 pm

Looks like Stanley won. Congratulations! Good thing UC Davis didn’t enter a vehicle …

7

Maynard Handley 10.10.05 at 2:49 am

DARPA provides both good and bad because DARPA is basically the MITI, the (high-tech) industrial planning department of the US govt. America being populated by the certifiably insane, the fiction has to be maintained that the US govt does not do such things, and that DARPA is purely in the business of killing people ever more efficiently.
I imagine one day, just as the Republicans have begun so much of the other nonsense they spout, they will look at DARPA, conclude that it really ought to be *only* in the killing people business, and the US will lose one more of the various props that kept it a going concern.

8

me2i81 10.10.05 at 11:32 am

Artifical Intelligence research was for years kept afloat by the military’s burning desire for killer robots. They eventually had to settle for shoot-by-wire. The Grand Challenge is all about moving technology along to the point where autonomous transport vehicles could be bidded out for contract. TerraMax finished in twice the time as the rest of the finishers, but did it with a vehicle that could actually transport something.

9

paul 10.10.05 at 3:01 pm

The federal government has tried on at least two occasions that I can think of to limit the DARPA (aka ARPA) role in not-explicitly-military research. In about 1970 this clobbered a bunch of university computer-science research, with the rather surprising result of fostering a bunch of long-term R&D at various high-tech companies that had effective monopolies and hence money to burn. In the 80s, the reining-in of DARPA made a significant contribution to AI Winter, wiped out a bunch of research into interesting parallel computer architectures, and led to the development of several (for a time) highly successful wall street quant firms.

Strange tricks history plays.

Comments on this entry are closed.