Stanley wins the 2005 DARPA Grand Challenge

Background

DARPA, the U.S. Department of Defense agency that funds high-risk research, designed the Grand Challenge as a race for vehicles that drive themselves. Nobody sits behind the wheel and nobody steers by remote control: the car has to read the terrain with its own sensors, decide where to go, and turn the wheel itself. The format was deliberate — a public contest with a fixed date and a cash prize, meant to pull a slow laboratory problem into the open.

In 2004 the best vehicle completed just 11.78 km before it stopped ^[2]. That figure became the reference point for everyone who entered again: it said both how far a self-driving car could get in the desert and how much was still missing.

The race

The second DARPA Grand Challenge began at 6:40 am on 8 October 2005 ^[2]. Teams entered from 36 U.S. states and 4 foreign countries, 195 in all ^[2]. The course ran 212 km (132 mi) through southern Nevada ^[1][2], and the rules were unforgiving: a vehicle that stopped, hesitated too long, or left the route was out.

Only five vehicles finished ^[1][2]. DARPA's own account gives the same numbers — 195 teams entered, and five vehicles completed the 132-mile course ^[1]. Most of the field, with all its sensors and software, never reached the end.

Stanley

Stanford University's entry, Stanley, finished first. DARPA records its time as 6 hours and 53 minutes ^[1]; Wikipedia's results table lists 6:54 and first place for the Stanford Racing Team ^[2]. The two accounts differ by a minute, which changes nothing about the outcome.

The more striking number is how far the whole field moved. Of the 23 finalists, all but one went further than the 11.78 km managed by the best vehicle in 2004 ^[2]. The distance between a car that can drive a little and a car that can finish 132 miles was closed within a year.

The prize for this running had been raised to USD 2 million, and it was claimed on 9 October 2005 ^[2].

Why it mattered

The race gave the field a public, measurable target. A desert course is not full driving — there is no traffic, no pedestrians, no city streets — but it puts perception, planning, and control together outdoors, for hours, with nobody to take over when something goes wrong. In practice, many separate pieces of software and hardware have to agree with each other continuously, or the vehicle stops. Reaching the finish at all was the barrier; winning was the extra.

The people involved carried the work outward. Google's official blog later described Mike Montemerlo as the software lead for the Stanford team that won the 2005 Grand Challenge ^[3]. Events like this one mark the point where self-driving moved from laboratory demonstration to a problem that universities and companies worked on in public.