The DARPA Urban Challenge 2007 was the third event in the DARPA Grand Challenge series, organized by the Defense Advanced Research Projects Agency (DARPA) to advance autonomous vehicle technology. Held on November 3, 2007, at the former George Air Force Base in Victorville, California, the competition required vehicles to complete a 96-kilometer urban course in under six hours while obeying traffic regulations and interacting with other traffic. The event marked a shift from the off-road desert courses of previous challenges to a mock urban environment, testing vehicles' ability to handle intersections, merging, and obstacles.
Tartan Racing, a collaboration between Carnegie Mellon University and General Motors, won the competition with a modified Chevrolet Tahoe named 'Boss', securing the $2 million first prize. Stanford University's team, partnered with Volkswagen, took second place and $1 million, while VictorTango from Virginia Tech, with TORC Robotics, finished third and received $500,000. The event demonstrated significant progress in autonomous driving, with six vehicles completing the course out of 11 finalists.
Background and DARPA's Role
DARPA, a research agency of the U.S. Department of Defense, initiated the Grand Challenge series in 2004 to spur development of fully autonomous ground vehicles. Congress authorized cash prizes to encourage innovation, with the goal of making one-third of military ground forces autonomous by 2015. The first challenge in 2004, held in the Mojave Desert, saw no vehicle finish; the best, Carnegie Mellon's Sandstorm, traveled 11.78 kilometers. The 2005 challenge saw five vehicles complete a 212-kilometer desert course, with Stanford's Stanley winning $2 million.
The Urban Challenge extended the series to urban environments, requiring vehicles to navigate roads with traffic, obey stop signs and traffic lights, and merge into lanes. The competition was open to teams worldwide, provided at least one U.S. citizen was on the roster. Over 100 teams registered initially, with 11 selected for Track A, which received $1 million in funding from DARPA, including major universities and corporations like CMU, Stanford, and Oshkosh Truck.
Competition Format and Rules
The Urban Challenge course at George Air Force Base included a 96-kilometer route with intersections, traffic circles, and obstacles. Vehicles had to complete the course in under six hours while following all traffic regulations. Teams were given sparse waypoint maps, which some, like Tartan Racing, enhanced with additional extrapolated waypoints for better navigation. The competition tested not only basic driving but also decision-making in dynamic traffic scenarios, such as merging and yielding.
Track A teams received substantial funding, while Track B teams, including many smaller groups and universities, competed without that support. The selection rationale for Track A was not publicly explained by DARPA. The event attracted diverse participants, from high schools to corporations, reflecting a broad interest in autonomous vehicle research.
Tartan Racing's Victory
Tartan Racing, led by Carnegie Mellon University professor William 'Red' Whittaker, developed 'Boss', a Chevrolet Tahoe equipped with a suite of sensors including LIDAR, radar, and cameras. The vehicle used sophisticated algorithms for perception, planning, and control, enabling it to navigate the urban course safely and efficiently. Boss completed the course in approximately 4 hours and 10 minutes, with an average speed of about 22 miles per hour, demonstrating robust performance in traffic.
The team's success was attributed to extensive testing and a modular software architecture that allowed for rapid iteration. Tartan Racing's approach to enhancing waypoint maps and its focus on handling edge cases in urban driving were key factors. The victory highlighted the potential of Artificial intelligence and Machine learning in autonomous systems, though the competition relied more on classical robotics and computer vision techniques than modern deep learning.
Impact and Legacy
The Urban Challenge accelerated research in autonomous vehicles, influencing subsequent developments in the field. It demonstrated that autonomous vehicles could operate in complex environments, paving the way for later initiatives like the DARPA Robotics Challenge and the Subterranean Challenge. The event also spurred commercial interest, with many participating teams and researchers later contributing to companies like Waymo and Tesla.
In the years following, autonomous vehicle technology advanced significantly, with Deep learning and Neural network approaches becoming central. The Urban Challenge's emphasis on real-world navigation and safety remains relevant to current efforts in Artificial intelligence and robotics. The competition's legacy is evident in the growing deployment of autonomous vehicles in controlled settings and the continued investment in Machine learning for perception and decision-making.
Conclusion
The DARPA Urban Challenge 2007 was a landmark event in autonomous vehicle history, showcasing the feasibility of urban driving without human intervention. Tartan Racing's victory, along with the achievements of other teams, demonstrated that with sufficient funding and technical innovation, autonomous vehicles could handle complex traffic scenarios. The competition's success led to increased funding and research in the field, contributing to the rapid progress seen in the following decade. As of 2025, autonomous vehicles are being tested in various cities, and the foundational work from the Urban Challenge continues to influence both academic research and industry development.