Slashdot Mirror
$2 Million on the Table for DARPA Urban Challenge
coondoggie writes "The contestants: Thirty-five driverless vehicles. The goal: to navigate an intricate faux-urban environment quickly. The prize: $2 million for the fastest qualifying vehicle. 'The National Qualification Event will take place this weekend ... DARPA says its third-annual Urban Challenge program has the lofty goal of developing technology that will keep soldiers off the battlefield and out of harm's way. The Urban Challenge features autonomous ground vehicles maneuvering in a mock city environment, executing simulated military supply missions while merging into moving traffic, navigating traffic circles, negotiating busy intersections, and avoiding obstacles.'" I'll be cheering, as long as the creepy robot bear isn't participating.
Caveat I haven't had a chance to read the entire challenge to see if it answers this question. I'm in Iraq I have better things to do, sort of...
Does the vehicle have to be one piece? Specifically can it launch a UAV to provide a top down view of the street? This could be then used to avoid crowds (or head towards them), get around dead ends, and generally navigate the cities. The imagery we have is often horribly out of date and roads have moved, stopped existing, or new ones have popped up.
I think having an eye in the sky dedicated to the vehicle could be a tremendous asset.
I'll meet you at the intersection of "Should be" and "Reality"
One of the major side effects of the DARPA Grand Challenge series is that the supporting hardware has become much better. You can now buy most of the major components off the shelf. GPS/INS/compass/odometer navigation units are a few thousand dollars, rugged, and work well. When the first Grand Challenge was announced, the off-the-shelf solution cost about $170K and required 4U of rackmount space, with air conditioning. CMU actually used that in the first round.
LIDAR units have improved enormously in the last two years. Last time around, everybody just had single-beam line scanner LIDARs, usually from SICK, except for Team DAD, who built a multibeam scanner that worked but wasn't rugged enough. This time, the major players have multibeam LIDAR units from Velodyne or Ibeo. Velodyne's unit has 64 lasers on a spinning drum. Now you can image your entire environment in 3D at 5Hz.
Controlling the vehicle is easier, too. There are now cars available with electrical power steering and brakes, and one can tap into those systems to drive. And there are at least three vendors selling gear for remote/autonomous driving of existing cars.
So now it's almost entirely a software problem. You don't burn so much time and effort building sensor and actuator systems.