A group of Cornell students gave a new meaning to the term “backseat driver” as it spent the last year creating Titan, a car that can literally drive itself. The 32 undergraduate and graduate students engineered the autonomous vehicle to compete in last weekend’s DARPA Grand Challenge, a 131.6-mile race with a two million dollar prize, held in Primm, Nevada.
Cornell’s vehicle crashed into a bridge nine miles into the course on Saturday and was eliminated, but it performed well up to that point and in qualifying rounds. Team Cornell was among 23 finalists culled from nearly 200 entrants from universities, high schools and private engineering and aeronautical companies. The vehicle underwent eight days of qualifying events in desert conditions at a speedway in California, where it competed against 42 other robots, before it was admitted to the finals on Oct. 5.
DARPA, the Defense Advanced Research Projects Agency, is the research organization of the Department of Defense. It set up the Grand Challenge in response to a Congressional mandate that the Pentagon make one-third of all military land vehicles autonomous by 2015.
The intelligence gathered from the Grand Challenge race will help create vehicles that can transport supplies or conduct reconnaissance in dangerous places like bomb-lined highways in Iraq. Titan and the other finalist vehicles could traverse desert terrain, wade through water, cross ditches and barbed wire and navigate dangerous obstacles.
Isaac Miller grad, a member of the computer science team, said vehicles created for the Grand Challenge could be used for humanitarian needs as well. Instead of having Red Cross workers driving along dangerous roads, he said there could be “robotic hospital vehicles” providing medical aid in war zones.
“I would rather have a bunch of autonomous vehicles driving around on a battlefield than I would have myself driving around there,” he said.
Grand Challenge Program Manager Ron Kurjanowicz said the race combined many elements necessary to create successful military projects.
“The DARPA Grand Challenge is a truly powerful mix of American ingenuity, team spirit, competitiveness, entrepreneurship, engineering and computer science,” Kurjanowicz stated in a press release.
Team Cornell used a Spider Light Strike Vehicle, an all-terrain vehicle that has been tested in combat situations, as Titan’s foundation. Then the sensor, mechanics and computer science teams installed a Global Positioning System unit that can read terrain accurate to 10 centimeters, sensors, stereoscopic vision and an artificial intelligence system to create a map of the vehicle’s surrounding area, a “unified model of the world as we see it,” Miller explained. The AI also relates commands to the engine and brakes and can coordinate three-point turns.
Members of Team Cornell spent the summer doing practice runs in and near Mojave, California, working out kinks in the vehicle’s approach to cliffs and uneven terrain. In order to make sure the vehicle survived to race day, there was always a driver who could take control “as soon as the AI did something stupid,” Miller said.
Practice courses were usually run at less then 10 miles per hour so the sensor team could observe the vehicle’s behavior upon approaching small or large obstacles. The actual Grand Challenge was run at speeds averaging near 20 miles per hour and the team’s only control was the initial “start” command.
Titan had been “performing perfectly” for the first nine miles of Saturday’s race, according to the team’s blog, but it made a sharp turn upon approaching the bridge and hit a guardrail when a DARPA chase car overrode the system to stop the vehicle. A new GPS mode had gone into effect as soon as the vehicle approached the bridge and Titan veered to avoid what it thought was a wall after realizing it was six feet higher than it initially measured.
Aaron Nathan ’06, who worked on sensors, said DARPA’s chase vehicle was “trigger happy” and “very nervous about it going off the bridge.” He said his teammates were confident the vehicle would have been able to correct its position had it been given more time.
The winner, Stanford University’s vehicle Stanley, completed the course in six hours, 53 minutes. Four other teams also finished the course.
“The competition was very, very fierce,” Miller said. Before Saturday’s race, he said Carnegie Mellon and Stanford were “far and away the ones to beat” with tremendous financial resources available. Miller said Carnegie Mellon hired industry professionals to work on their vehicle. Titan, on the other hand, was “most definitely a student-run project.”
The Cornell team raised nearly $450,000 in donated equipment and cash from sponsors including ST Kinetics, General Motors, Northrop Grumman, the College of Engineering and Texas Instruments.
Titan is currently being shipped across the country and will return to Ithaca within the next few weeks. It is still in working condition and, Nathan said, has no structural damage. “There is a lot of research that can be done with the vehicle still,” he said.
DARPA has not yet announced another race for next year, but Nathan said it would be beneficial for the field to hold another competition.
“Although there was a winner, there’s a lot of progress that can be made by continuing the event,” Nathan said.
Archived article by Melissa Korn
Sun Senior Editor