Virginia Tech's entry in DARPA's Urban Challenge is moving forward to the qualifying rounds, thanks in part to a custom-designed drive-by-wire control system and unique navigation software that makes the vehicle's driving decisions almost human.
"VictorTango," a team of Virginia Tech engineering and geography students, is among 36 semi-finalists selected by DARPA (Defense Advanced Research Projects Agency) to vie for the $2 million Urban Challenge prize. Qualifying rounds begin Oct. 26 at the former George Air Force Base in Victorville, Calif., and the final event will take place Nov. 3.
Urban Challenge teams are attempting to develop vehicles that can maneuver a 60-mile course of simulated military supply missions in less than six hours -- with no human intervention allowed past the starting line. The vehicles will have to obey California traffic laws, merge into moving traffic, navigate traffic circles, negotiate intersections, and avoid a variety of obstacles.
VictorTango has converted two Escape hybrids donated by Ford Motor Co. into autonomous vehicles by outfitting them with a "drive-by-wire" system, a powerful computer system, laser scanners, cameras, and a GPS (global positioning system), said Patrick Currier, a mechanical engineering (ME) graduate student. The students dubbed the vehicles "Odin" after the chief god in Norse mythology.
"The drive-by-wire system allows the computers to control the throttle, brake, steering, and shifting and to drive the vehicle," Currier said. "This system was custom developed by the team and is unique in that it is completely hidden from view, enabling Odin to retain full passenger capabilities."
TORC Technologies LLC, a company in Virginia Tech's Corporate Research Center founded by alumni of the university's robotics program, has worked with VictorTango to develop the software for the vehicle's computer system.
VictorTango and TORC developed Odin's sophisticated navigational software, which is modeled on human behavior. "To successfully navigate in an urban environment, Odin processes all of the sensor information, classifies the situation, and then chooses a behavior, such as passing another vehicle, staying in the lane, or parking," Currier explained.
This "human-like" system makes Odin capable of choosing the best course out of millions of possible courses, he said.
The team has outfitted Odin with four computers that perform specialized sensor processing and hardware interface tasks and two powerful servers that provide the primary computing power.
Three laser scanners mounted on the vehicle's bumpers can scan a combined 360-degree field of view 12.5 times per second to detect obstacles. "These scanners are capable of detecting and tracking cars at a distance of up to 100 meters and are Odin's primary method of detecting other vehicles," Currier said.
Four more laser scanners are mounted on Odin's roof rack -- two to detect small obstacles such as curbs and potholes and two to check the vehicle's blind spots when it changes lanes or merges into traffic.
Two cameras mounted on the roof rack serve two purposes -- they enable Odin to sense its location and identify its proper position in the traffic lane, and can also positively determine if an obstacle detected by the scanners is another vehicle.
Odin's GPS has been coupled with an inertial measurement unit and wheel speed sensors to measure movement in all directions. "This system provides Odin with accurate position, even if the GPS signal is temporarily lost," Currier said.
"Odin is now capable of driving on a marked road, following moving traffic, passing stopped vehicles, handling four-way intersections, and performing three-point turns," Currier said. The vehicle is being fine-tuned so that it can merge with moving traffic, pass moving vehicles, and park.
Currier is one of 10 graduate students on the Virginia Tech team, which also has included as many as 50 undergraduates.
The students are guided by four faculty advisers, three of them from Virginia Tech -- professor Alfred Wicks and assistant professor Dennis Hong of the College of Engineering's ME department, and geography department chair Bill Carstensen of the College of Natural Resources.
The team's founding adviser, Charles Reinholtz, a former Virginia Tech Alumni Distinguished Professor of ME and engineering education and now the chair of ME at Embry-Riddle Aeronautical University in Florida, continues to work with VictorTango.
In October 2006 VictorTango was one of only 11 "track A" teams chosen to receive $1 million contracts to develop autonomous vehicles capable of conducting simulated military supply missions in an urban setting. In all, 89 teams from universities and industry entered the competition in 2006.
In addition to the $1 million from DARPA and the two Escapes from Ford, the Virginia Tech team received a $100,000 grant from Caterpillar and additional sponsorship from National Instruments and several other corporations.
VictorTango qualified for the semi-finals during a site visit DARPA judges made to Virginia Tech earlier this year. The team's vehicle successfully demonstrated its fully autonomous capabilities, driving a road course and interacting with human-driven vehicles.
DARPA is sponsoring the Urban Challenge as a more sophisticated follow-up to the two Grand Challenge competitions, which were held in 2004 and 2005 in the Mojave Desert. Virginia Tech competed in both of those contests, and the university's two entries placed eighth and ninth in 2005.
"The Urban Challenge will be far more difficult to navigate than the open desert in the Grand Challenge," Currier said. "In the Grand Challenge, the vehicles followed a GPS 'bread crumb' trail and the obstacles they maneuvered around were static. In the Urban Challenge, vehicles must obey the rules of the road and avoid moving traffic."
Materials provided by Virginia Tech. Note: Content may be edited for style and length.
Cite This Page: