October 6, 2005 The DARPA Grand Challenge National Qualification Event (NQE) Is finished and 23 robotic vehicles have been selected to compete in the Grand Challenge final event this coming Saturday, October 8, in the Mojave Desert near Primm, Nevada. The finalists will traverse a rugged desert course that features lakebeds, narrow desert roads, tight turns, tunnels, gateways and treacherous mountain passes. The actual course will not be revealed to teams until two hours before the event begins at approximately 6:30 a.m. (PDT). The team whose vehicle traverses the entire course the fastest in under ten hours will win US $2 million. Dan Christian attended the NQE and filed this report. Dan will also be reporting for Gizmag from what promises to be one of the most significant automotive races in history.
The chosen finalists, in alphabetical order, are:
The finalists were selected at the conclusion of an intense eight-day semifinal known as the National Qualification Event (NQE), held at the California Speedway at Fontana, Calif., from September 28 through today. Forty-three robots of many shapes and sizes took turns pursuing a series of 2.2-to-2.7-mile courses designed to resemble desert conditions. The semifinalists had been chosen from among 195 original applicants; over the past several months, DARPA narrowed the field to the 43 semifinalists through a series of qualifying activities, including in-person site visits by DARPA officials to assess the capabilities of the various robots.
The DARPA Grand Challenge is an off road race like no other. There is only one vehicle category and it includes SUVs, trucks, military vehicles, ATVs, custom built racers, an 8 ton truck, and a motorcycle. The first vehicle to drive up to 175 miles in under 10 hours gets two million US dollars. The race isn't really about size or power, but it's about intelligence. There are no drivers for these vehicles. Not on-board, not back in some control room, none. They have to autonomously handle paved roads, dirt roads, (open) gates, stream crossings, switch backs, and under passes. All they get is a list of GPS "way points", speed limits, and how far they can safely stray from the given positions.
Now before we start, you have to understand why DARPA wants vehicles driving themselves across the desert. DARPA has posted the US$2 million prize and run the competition twice now and the motivation is quite simple. DARPA wants to be able to ship supplies to the front lines without risking soldier's lives. If a robot truck hits a landmine, you just order another one. No one would be hurt, so it's not a big deal. Trucks can be repaired to as good as new – people can’t.
But in order to achieve DARPA’s objectives, this robot vehicle has to be smart about its job. There may be obstacles blocking the path, or the positions may not be in the middle of the road. The robot has to "see" the obstacles and road. Then it has to find a safe way to the next way point without hitting anything or going out of bounds. Fifteen robots attempted the race last year, but none of them got further than 7.5 miles and most of them didn’t get out of sight of the starting chute.
There are 43 robots competing in the National Qualification Event (NQE) for the chance to compete in the real race. Only the top 20 teams will start the Grant Challenge Event (GCE). Each robot will be followed by a DARPA-driven chase truck that can tell the robot to stop, go, or completely shut down. This is for safety purposes only. They will only stop the robot to avoid congestion, and only shut it down to keep it from hurting something off the race course. The robot has to avoid hurting itself if it wants to finish the race.
As of Saturday, 13 teams had finished the second round NQE course, four teams followed the course perfectly, and five teams had times of under 10 minutes.
Carnegie-Mellon University (CMU) has two robots in the competition: Sandstorm (#19) and H1lander (#25). Sandstorm is a highly modified military Hummer and went the farthest in the 2004 race. H1lander is a new commercial H1 Hummer (see the pictures, it isn't quite "new" anymore). Both vehicles have already done excellent runs with excellent times. CMU's robots are the ones to beat.
Stanford's "Roadrunner" (#3) is the strongest challenger so far. They are the only other team to have two perfect runs and excellent times. They didn't compete last year, but they seem to have made up for lost time.
Desert Tortoise (#15) from Intelligent Vehicle Safety Tech has shown some fast runs, but has been having trouble handling parts of the course and some obstacles.
Some robots are still being tuned up. DEXTER (#28) from Team ENSCO and Golem 2 (#18) from The Golem Group didn't finish the first qualification corse, but had a good runs on the second one.
There are another 37 robots that I don't have the time to elaborate about in this report. Each one has a different approach to solving the intelligence, speed, and durability demands of the race. DARPA set out to inspire the next generation of engineers and scientists. They succeeded in capturing the hearts, minds, time, and wallets of hundreds of people. Off road robot racing has indeed proven to be a "Grand Challenge".
Tune in next week for the race results. See the Grand Challenge web site for the details of the race, including links to all the teams.
Full disclosure: I also work for Google, which is one of the CMU team sponsors. I used to work for Red team leader "Red" Whittaker. I went to school at CMU and Stanford. I might be a tad biased, but that won't change who crosses the finish line first....
See the stories that matter in your inbox every morning