jawfish

DARPA Grand Challenge 2005
john posted in technologies on May 28th, 2005

The DARPA Grand Challenge is back and this time things will be different. The equipment that’s driving around the desert is going to change your life, and soon.

DARPA, the mil-tech juggernaut, may be creepy, impenetrable, and a big waste of your money ( if you are off-shore, that would be a big threat to your country, ahem). However, DARPA research has benign outcomes too, like the Internet. Well maybe you have some mixed feelings there, as well. I didn’t say change for the better necessarily, just a big complex change.

The Grand Challenge is a test of robot unmanned vehicles over unknown, rough terrain. It is really very difficult, as a human driver would have to work pretty hard to complete the course. Last year none of the competitors had completed basic testing and as result most were incapable of even starting the course. This year Stanford, and best of 2004, Carnegie-Mellon, are claiming to have breezed through a preliminary obstacle course test. Both entries claim to have done extensive real-world testing before the official preliminary test in May.

The Challenge is a time trial over a surprise course. No radio communication is allowed with the vehicle, although there were elaborate safety measures at the first test. The course will be like the first one, with off-road, gravel and paved sections. The vehicles not only have to maneuver to stay on course, they have to handle the large rocks, steep inclines, and gullies typical of an off-road test.

The Mojave is the likely location, which makes it the center of two 21rst century breakthroughs: robotic vehicles and commercial spaceflight.

I have no doubt that the test will be passed eventually, and the military will get its robotic land vehicles to go with the airplanes.

My guess is that a successful traverse of the course, a proof-of-concept, is an event equivalent to the flight of the Wright Flyer or Edison’s carbon-filament light bulb. This time though its not the mechanical technology thats revolutionary, it’s the software.

The decision-making software, the driver’s brain, is the hard part. Insects can cross rough terrain with far more agility than our bots today, and it’s going to be a long time before an ant’s level of efficiency is achieved. It’s not clear to me in civilian or military applications that large ground vehicles can be left alone without supervision, anyway, so inefficient, and just plain dumb bots, may be combined with a radio link. That’s basically remote control. The test calls for full autonomy, though, and there won’t be any connection to the vehicles as they try to pick their way across miles of desert.

What’s the significance of autonomous vehicles?

First, an autonomous vehicle is a ground-based robot like R2D2 or C3PIO with a jeep-like body.

That means that the hard part, the software, is directly adaptable to lots of other shapes - a small biped like a chicken, a small or large tank with crawler treads, a wheelchair, a snake, a lawn mower, a snack-cart, or a dump truck.

All science fiction you say, just magazine drivel. Well take a look at two commercial prototypes: the Toyota and Honda robots. The Toyota robot

can play a trumpet, sort of, and the Honda robot

can walk quite well and shake hands. Ho hum you say, but you have to understand, thats Good Enough to be very useful. We don’t need robot musicians or baseball players, we need cleaners, haulers, and DARPA needs substitutes for human soldiers. We aren’t talking Star Wars, we are talking smart tools, and substitutes for humans doing dangerous work. Controlling the military ramifications of this is a subject for another Jawfish post.

Military: at the moment the Army is bogged down fighting an MP’s war in Iraq. Soldiers are forced to go house to house in very dangerous conditions, and to drive convoys over roads that contain deadly IED mines. I’d guess reducing or eliminating the drivers in convoys would be attractive. Solutions to the house-to-house situation are more technically interesting, and they, I think, have the greatest potential for destabilizing weapons development.

Without being any sort of expert, you’d think that a house-searching robot would need to be able to approximate a man’s agility, though a chimpanzee or hummingbird would be much better. But maybe a simpler vehicle could be used for reconnaissance, as a way of protecting the soldiers. In any case, its small and maneuverable that’s needed. There are already bots on the market that have the agility to climb stairs, and generally manuever inside buildings. But, they still lack brainpower. Look at the IRobot site for some examples of military bots, already in production. IRobot says the PackBot will be autonomous by the end of 2005.

Civilian: Its really hard to believe that large general-purpose, and therefore dangerous, autonomous vehicles will be let loose. One failure that sends a robot jeep headed across a freeway will produce public panic. So my guess is that small, semi-disposable, and fenced uses are more likely. Industries with high labor costs and low-skill requirements will be the first target. There’s already the Roomba vacuum

Lawn mowing also comes immediately to mind.

Assuming a golf course would benefit from cutting the grass at night, with one instead of four groundskeepers on duty, you can see a model. Robot lawn mowers could run a fixed map of the course, with predetermined work areas and no-go boundaries, and a simple electronic fence like those used for dogs to contain them. One worker could watch and maintain them. The mowers would have to be smart enough to avoid obstacles, like a dead mower, or tree limb. Here is a prototype mower:

How about agriculture? Is the labor cost so high that flocks of robot chickens could tend row crops, or autonomous combines could harvest wheat on the giant industrial farms of the West? Maybe so, especially if there were savings on pesticide use with the robot chickens, and the ability to work 24×7 on the big farms.

If I am right, and not just indulging in Popular Mechanix fantasy, then there will be many uses we can’t foresee. The Roomba, Aibo, and Robosapiens are worth a look for an idea of what inexpensive bots can do.