The encyclopedia · Software & IT · Strategic decision · 2004–2007
DARPA paid $2m to whoever could drive a desert course — and seeded self-driving
A fixed $1–2m prize replaced cost-plus contracts; 195 teams entered, and in 2005 Stanford's Stanley finished the 132-mile course and won.
DARPA · Stanford Racing Team
The solution
In 2004 DARPA offered $1m to the first team that could drive an autonomous vehicle 142 miles across desert from Barstow, California, to Primm, Nevada, in under ten hours. Fifteen finalists tried; none finished — the best managed 7.5 miles.
The failure was the point: it created a community of engineers, students, off-road racers and hobbyists who had never worked for the Pentagon. A second Grand Challenge followed 18 months later with a $2m prize.
In October 2005, five vehicles completed the 132-mile course in southern Nevada. Stanford's Stanley won in 6 hours 53 minutes, and a third challenge (Urban Challenge, 2007) added traffic and rules.
Why it worked
- Fixed prize means the government pays nothing for failure.
- Open entry pulls in talent the vendor pool would never include.
- A hard, observable goal (finish the course) prevents specification gaming.
- Each round's public failure taught the whole field what to fix.
What can be applied
A prize turns procurement into an open tournament: pay only for a demonstrated outcome, and let outsiders you never knew existed self-select into the race.
Aftermath
The Grand Challenges are credited with turning autonomous vehicles from a lab curiosity into an industry; by 2014 defense and commercial programs were proliferating, and DARPA launched follow-on challenges in robotics and cybersecurity on the same model.
Sources
spotted an error? The archive wants to know.