The DARPA Urban Challenge 2007 — It's so much fun!
June 19, 2007
images
“I love hard problems,” says Seth Teller, one of the Team MIT leaders for the urban (autonomous vehicle) DARPA Grand Challenge (DGC), having just faced the critical stages of preparations for a site visit by a team of DARPA observers.
The hope of Team MIT and the 52 other groups participating in the DGC is to be among the top 30 teams that pass the site visit to compete in a National Qualification Event (NQE) October 21-31 and survive to compete as a finalist for the Urban Challenge on November 3, 2007. DARPA will award $2 million, $1 million and $500,000 awards to the top three finishers that complete the 60 mile urban course within the six-hour time limit.
‘Solving hard problems’ is an understatement in taking on the latest DARPA Grand Challenge event.
Team MIT leaders and members see three types of uncertainties in meeting the urban autonomous test: uncertain input –the relationship of the provided environment with the actual driving environment; uncertain sensing—the relationship of the available sensor data to the actual static and dynamic surroundings of the vehicle; and uncertain actuation—the relationship between the commanded vehicle motions and the vehicle’s actual physical progress. In short, Team MIT has focused on three major areas for development and mastery: sensing, planning and control.
To top off the challenge, Team MIT’s entry into the DARPA (urban) Grand Challenge comes following the first two events held in 2004 and 2005, in which MIT did not compete. Both team lead John Leonard, professor of Mechanical Engineering at MIT and Seth Teller, perception team leader, agree that the MIT team was eager to face this third event because of the complexity inherent in urban environments (as opposed to a road race in the desert) and the potentials for growth and development far into the future.
Team MIT was well into the running for the DGC by Oct. 2, 2006 when DARPA selected it along with 10 other entering teams in a DARPA-funded accelerated track. See the DARPA Oct. 2, 2006 press release.
Teller looks beyond the DGC toward the impact of future autonomous vehicles citing the potential to significantly benefit public safety by saving thousands of lives each year in more efficient and controlled driving conditions. Lives will not be the only thing saved by autonomous driving. Autonomously controlled vehicles will reduce gas usage for example, by enabling coordinated highway convoys and urban traffic merging. This in turn will lead to greatly increased human productivity as human drivers are freed from the stress and cognitive burdens of commuting.
Just as planes now engage in autopiloted flight, vehicles on the ground will develop the capability not only to guide themselves, but (unlike current airplanes) to execute collision avoidance behaviors. Cars approaching each other at intersections will exchange trajectory information to allow for appropriate reactions completely free of the need for human engagement.
For Teller, the most rewarding aspect of this project is being able to pursue an ambitious, high-risk effort with a large multidisciplinary and talented team. In fact, the multidisciplinary nature of the MIT team is what makes it stand out in comparison with the 52 other teams now vying for the challenge. As John Leonard describes it, “Team MIT …draws on the expertise of various faculty for ‘distributed decision making’." Co-PIs on Team MIT include Professor Jonathan How (Aero/Astro) and Professor David Barrett (Olin College). Other MIT faculty members involved in the project include Professor Emilio Frazzoli(Aero/Astro), also member of Laboratory of Information and Decision Systems (LIDS), EECS Professor and member of CSAIL Daniela Rus, and other members of the MIT community.
Team MIT also benefits from technical support offered by the Charles Stark Draper Laboratory, the MIT Lincoln Laboratory and BAE Systems. Undergraduate and graduate students have been drawn in with related classes (6.142 and 6.897) in Fall, 2006.
As Team MIT nears its site visit, they continue to develop the capabilities of the team vehicle, Talos II, a Land Rover LR3 that is essentially a mobile machine room complete with supercomputer and drive-by-wiring enclosed in a nice red exterior. At the site visit, vehicles are evaluated on their ability to navigate a test course including a four-way intersection, obstacle avoidance, lane following, navigating u-turns, stopping at an intersection and traversal with other moving vehicles .
Team MIT acknowledges the multiple sources of support that have come from DARPA, the MIT School of Engineering, the Computer Science and Artificial Intelligence Laboratory (CSAIL), the MIT Department of Aeronautics and Astronauticsl, the MIT Department of Electrical Engineering and Computer Science and the MIT Department of Mechanical Engineering, the MIT Lincoln Laboratory, the MIT Information Services and Technology, the CSAIL TiG (The Infrastructure Group), the Charles Stark Draper Laboratory, the Ford-MIT Alliance, the Franklin W. Olin College, BAE Systems, Quanta Computer, the Nokia Research Center, Mobileye, Delphi, Applanix, Acumentrics, Drew Technologies, Advanced Circuits and the South Shore Tri-Town Development Corporation.
Additional links:
MIT News Office, July 5, 2007: Drivers Unwanted: MIT 'Robocar' takes a spin
DARPA News Release-Oct. 2, 2006
DARPA News Release-Dec. 8, 2006
EECS Home Page | Site Map | Search | About this page | Comments and inquiries welcome