Georgia Tech/SAIC Sting 1 vehicle reaches semifinals
Nov 6, 2007
The blue Porsche Cayenne pulls up to a four-way intersection and stops. After it continues through the junction, it approaches a vehicle stopped in its lane. The Cayenne checks to make sure there are no cars approaching in the opposing lane, passes the stopped car and returns to its original lane.
This scene may sound normal, but this is no ordinary Porsche Cayenne-it thinks for itself and requires no driver. This autonomous vehicle was designed by the Georgia Institute of Technology in collaboration with Science Applications International Corporation (SAIC) for the Defense Advanced Research Projects Agency's (DARPA) Urban Challenge.
Georgia Tech's vehicle, named Sting 1, did not qualify for the final challenge during the National Qualifying Event (NQE) held from October 26-31 at the urban military training facility located on the former George Air Force Base in Victorville, California. Sting 1 finished as one of 35 teams that made it to the NQE.
"As a first-time entrant, the team has done an outstanding job making it to the semifinal round of the world's most challenging robotics competition," said Tucker Balch, team lead and associate professor in Georgia Tech's School of Interactive Computing in the College of Computing.
With six cameras, eight computers, Doppler radar and infrared laser radar on board, Sting 1 was designed to operate without any human intervention and obey California traffic laws while performing maneuvers such as merging into moving traffic, navigating traffic circles and avoiding moving obstacles.
The road to California began in the summer of 2006, when Georgia Tech and 88 other teams signed up to participate in this year's Urban Challenge.
"Georgia Tech didn't compete in the two previous Grand Challenges, but SAIC did," added Balch. "Their experience helped us develop software that could have enabled a robot to place well in the previous challenges and then we took it further with additional capabilities necessary for the Urban Challenge."
The Georgia Tech team, consisting of researchers in Georgia Tech's College of Computing and College of Engineering and the Georgia Tech Research Institute (GTRI), chose the Porsche Cayenne as their vehicle and in August 2006 began to install computers that would drive the car automatically.
Eight computers networked together through two high speed networks were programmed to know the rules of the road. This included knowing how to stay in a lane, how to overtake another car, how to make turns in city traffic, how to maneuver the waiting patterns at an intersection, how to merge into traffic and how to behave in a parking lot.
According to the racing team, the car really had to think for itself.
"When moving forward, the car usually ignored obstacles that were in its planned path," said Tom Collins, electronics lead and GTRI principal research engineer. "But when obstacles were detected, the car would plan and execute a different route."
SAIC engineers developed methods for visual lane detection and tracking. On unpaved dirt roads, the colors of the road and non-road areas were modeled to identify a path, adapting over time as lighting or surface colors changed. On marked paved roads, a camera kept the car in its lane by detecting the typical white and yellow lines that mark a driving lane. If the vision system was unable to find a lane, the car used lasers to follow the curb. Ten laser range finders sent out infrared laser beams that constantly scanned to provide Sting 1 with an accurate measurement of the distance to any objects, such as curbs and other cars.
At intersections, the team used laser and radar sensors to see other waiting or approaching vehicles. Six off-the-shelf Doppler radar systems used to detect moving objects allowed the car to see as far as two football fields away in all directions. Cameras helped guide the car through the intersections and onto new roadways.
"We had to guarantee that there was at least a 10 second window that would allow us to pull out onto a road, accelerate and get up to a reasonable speed without cutting someone off," noted Henrik Christensen, principal investigator for the team and director of Georgia Tech's Robotics and Intelligent Machines Center.
The researchers tested their car for months in the parking lot behind the Centergy One building in Technology Square on the Georgia Tech campus. They also utilized the Georgia Public Safety Training Center in Forsyth, Ga. on weekends to test the ability of the car to maneuver in an urban environment.
The Urban Challenge is the third in a series of DARPA-sponsored competitions to foster the development of robotic ground vehicle technology without a human operator, designed for use on the battlefield. Safe operation in traffic is essential to U.S. military plans to use autonomous ground vehicles to conduct important missions and keep American personnel out of harm's way.
Georgia Tech researchers are already thinking about life after the Urban Challenge.
"We've already talked about expanding this work to other areas," said Vince Camp, hardware lead and GTRI senior research engineer. "We're looking forward to using the technologies in applications such as autonomous lane striping for the Department of Transportation."
Challenges like this also aim to improve safety in vehicles consumers purchase. Some high-end vehicles sold today have backup sensors that alert the driver to obstacles and can parallel park without driver assistance. There are also systems that will alert a driver that is approaching a car in the same lane too quickly or if a driver is leaving the appropriate lane.
"These types of systems will help us become better drivers, but it's probably going to be a decade or so before we see fully autonomous vehicles," said Christensen. "At some point, though, drivers will realize that their cars are probably much more aware of what's going on around the car and are better equipped to deal with a situation than human drivers."
DARPA awarded a first-place prize of $2 million to Carnegie Mellon's Tartan Racing Team. Second and third places went to teams from Stanford Univesity and Virginia Tech.
Research News & Publications Office
Georgia Institute of Technology
75 Fifth Street, N.W., Suite 100
Atlanta, Georgia 30308 USA
Media Relations Contacts: Stefany Wilson, College of Computing (404-894-7253); E-mail: (firstname.lastname@example.org) or Abby Vogel, Research News & Publications Office (404-385-3364); E-mail: (email@example.com) or Kirk Englehardt, Georgia Tech Research Institute (404-407-7280); E-mail: (firstname.lastname@example.org).
Writer: Abby Vogel