I didn't think I'd have to write this one for another 5 years. Last year, nobody won - or even finished. This year, we have a winner. Four winners and five finishers, actually, though only the 1st place team collected the reward:
The challenge is to build a robotic vehicle that can successfully navigate a (132 mile) course across the Mojave Desert in less than 10 hours without human intervention, with a $2,000,000 prize to the winner.... Challenge vehicles must autonomously traverse the challenge route without human interface or control of any kind. The vehicles must perform general route selection and navigation to follow the Challenge route. Vehicles must sense their environment to perceive terrain features, ground conditions, obstacles, and other Challenge vehicles. They must intelligently control their speed and direction so as to avoid or accommodate all of the above. And they must do these things quickly....
As it happens, Stanford/Volkswagen's Tuareg "Stanley" beat the Carnegie-Mellon University Red Team's two Headless Hummers and the Ford Escape Hybrid run by plucky New Orleans-based Team Gray. Oshkosh Truck's huge TerraMax, meanwhile, used a slow-but-steady strategy until its last-leg sprint and finished in a little under 13 hours.
When asked for a quote, the winner reportedly said:
"1001001001000000111000001110111011011100110010101 10010000100000011110010110111101110101001000000110 0001011011000110110000101110"
(Hat Tip: VeganBob - and see translation)
Lots of good coverage and insights on Slashdot, with a very good liveblogged coverage of the race entire at TG Daily and an article by Popular Science.
So, why is this significant, other than being kind of cool?
DARPA is a Pentagon agency, and they've become very interested in these capabilities because the US military is looking to field "Mulebots" that can carry cargo in the field around 2010-2015 as part of its ambitious Future Combat Systems program. Smaller ground robots are already playing a large role in dealing with IEDs, tests are underway re: having small flying UAVs carried, launched from, landed on, and refueled by land robots, and South Korea says it has developed a quasi-robotic vehicle that will be able to reload mobile artillery without the danger of being shredded by shrapnel.
We're a long, way from the Terminator here - the "easy" things are hard enough, and remote control rather than anything like autonomous action is the overwhelming rule. But it's a definite military trend to watch, and the DARPA Grand Challenge moved them much further along for a minimal expenditure of dollars.
Personally, I'm surpirsed a South Florida team didn't win. If you're down there, you can see all kinds of large vehicles cruising the roads without any visible driver.... More seriously, as our population ages and safe driving becomes a bigger issue, the idea of a vehicle with "emergency driver assist" (read: "take-over") features may become more and more attractive to more and more people. Stanley may have descendents that help us all.
Then again, he may be able to help right now - digitaldc:
"Stanley would be the perfect vehicle for agressive driver rehabilitation. Replace the agressive driver's vehicle with Stanley and they can do nothing but scream obscenities as it plots its course to their workplace at a blazing 19.1 miles per hour. Three weeks of this and the agressor becomes as docile as a lamb."
I bet it would work.
I'll submit that 19.1 miles per hour is pretty good for the Challenge, given how much of the course was offroad. And I'll close with one more insightful point, along the lines of an Armed Liberal's "PowerPoint Is A Distraction: The Shining Kids of Carl Hayden High" that was posted a while back. Humankind704050:
"When you look at the results, and you see two colleges with virtually unlimited resources and millions of dollars spent on their vehicles, huge corporate sponsors and engineers at their beck and call from Boeing to Catepillar, who finished, and then this dinky little Team Grey from a suburb of New Orleans, with a splintered development team as a result of the Hurricane Katrina disaster, and they FINISHED just behind the big guys, leaving other heavily-funded vehicles in the dust.
Relatively speaking, a small indy group, even if their time was a tad slower than CMU or Stanford, essentially put those three teams to shame when you compare the resources they had available to them."
Yes. Armed Liberal:
"If we're about anything at Winds of Change, I think we're about that capital, about the spirit that creates, that restlessly looks for new paths - whether through the historic hatreds and distrust that we are all subject to as humans, the gridlock of modern interest-group politics, the problems of energy, of the environment, of poverty and oppression."
And...
"This post is about the bottomless pool of human talent. And about the fact that it's everywhere - sprouting up even when it's not tended and nurtured as deeply as it should be in some places. People long to create, they long to make, they dream of improving the world. We just have to look, and be willing to see it. We need it. It's this capital - the capital of imagination and work - that will sustain us and that we need to grow."
As da man sez, "Indeed."
I've gotta admit I'm impressed. I would have bet against any of them finishing, given last year's results and (even more) DARPA's ability to be as evil as they wanted with the course. That much difference over 18(?) months is a real testimony to learning curves and lots of practice.
There is, however, one challenge they did not face: traffic. Have to crack that one if we want robotized logistics vehicles potentially sharing the 'roads' with everything from civilian pickup trucks to goat herds. Maybe Grand Challenge 2 should be from the Nevada high country to LAX. Unfortunately, the liability insurance premium would exceed the prize...
Awesome! This is great news. And it's really exciting that so many finished the course.
Onward to Mars ...
No drivers?! What will Pierre Pettigrew do?
Wow. I thought it would be 2 - 3 years before we had a finisher.
Well, as Clarke said...
I was surprised and impressed myself. I was even more surprised to learn, over the weekend, that there was a Congressional mandate (flowing down to the Pentagon, and thence to DARPA) to get a third of the military ground vehicles replaced by 2015.
That's... ambitious. But then, so was finishing the DARPA race. Apparently, public humiliation ("You guys didn't even make it twenty miles?!") is a powerful spur to academics. (Academics, don't take offense. If I weren't addicted to my standard of living, I'd be back in school myself.)
As Tim says, traffic is going to be a whole 'nother nut to crack, as well as the related challenge of operating near pedestrian traffic. All we'd need is a documented case of one of these things running over a hapless (or actively provocative!) civilian in South Korea to derail the whole effort. But these, too, are probably solveable problems.
For one thing, the early operative vehicles don't need to be deployed in Iraq, or even South Korea. They can be developed and deployed for use at domestic installations where higher pedestrian discipline can be enforced until the problems are worked out. Likewise, expanding the program will be easier if local governments can allow longer hauls only over specific roads at specific times, for instance, base to base transport only at night in clear weather conditions until safety is proved.
One obvious benefit to a mature effort would be a greatly increased logistics efficiency... and logistics is one of the unsung cornerstones of American military power. The fun part is to consider, ten or fifteen years down the road, the combination of automated vehicles with agent-based logistics systems such as Couggar, Ultra*Log, and others. In a nutshell, these projects are intended to relieve some of the intellectual burden associated with logistics planning and execution, the same way that a fleet of robotic vehicles relieves the physical burden. These projects fascinate me.
A non-obvious benefit is that, as these technologies have the potential to seriously enlarge the "tooth to tail" ratio, they not only increase efficiency, but may bring down recruitment targets over the long haul.
And the civilian applications to this are immense.
Fear of litigation and greedy lawyers will make sure that this never happens on civilan roads.
Starting with the 60s we could make automated hwys roads safer than humans could drive. A jury or judge would not recognize that. Only that an accident occured. Not that many did not.
So this is only a military thing until the legal system grows up and becomes adult.
Tie that with these two "advances":
Sniper Detector
Sentry Gun
Fear of litigation, greedy lawyers, and the insurance industry will require this to happen, once the safety records are consistently documented as being better than human performance.
That won't be in 2015, but it may be in 2025.
Imagine also what Wal-Mart can do with technology like this. In some ways, they are already aiming for it on the computational end. They are aiming for software that automatically stocks (or at leasts, suggests) more burgers and charcoal for football weekends, for instance.
www.cajunbot.com
GEAUX CAJUNS!!!
Run one on the Paris - Dakar raid, give it a real test.
Meanwhile, there are also efforts to create more aware flying UAVs that can take advantage of thermal lift... or coordinate complex missions like hunting in packs and killing enemy missiles.