DARPA Grand Challenge 2005: Stanley crosses the desert (2005)

On 8 October 2005, the Stanford Racing Team won the second DARPA Grand Challenge with Stanley, a Volkswagen Touareg modified for autonomous driving. Led by Sebastian Thrun, the vehicle finished the 212 km (132 mile) desert course near the California-Nevada border in 6 hours, 53 minutes, 58 seconds, at an average speed of about 19.1 mph. Stanford received the $2 million prize on 9 October 2005. Four other teams also completed the route within the ten-hour limit.

After a race with no finisher

The 2004 Grand Challenge had offered a $1 million prize, but no vehicle finished the desert route; Carnegie Mellon University’s Sandstorm covered the greatest distance at roughly 11.8 km. The 2005 event attracted stronger sensor and planning stacks. Stanley used LIDAR, cameras, and onboard mapping to stay on course without human steering, a public demonstration that machine learning and robotics could handle harsh outdoor conditions.

A milestone for self-driving research

The challenge was run by the DARPA, the US agency that had also cut AI budgets during earlier winters (article). Winning teams combined computer vision, trial and error in the style of reinforcement learning during development, and robust engineering. Thrun and colleagues later helped launch Google’s self-driving car effort. The event is often cited as a turning point that convinced industry that autonomous driving was an engineering problem worth billions, years before deep learning dominated perception stacks.

Why it matters

Stanley’s run linked military funding, academic robotics, and commercial ambition in a single televised result. It sits between statistical machine learning advances such as random forests and the ImageNet era that made neural networks central to vision again. The precise course time and distance come from Stanford’s technical report on Stanley (Thrun et al., 2006).