DARPA Grand Challenge ｜ reinvent the wheel #9

Twenty Years Later, How DARPA’s Wild Robot Race Rewired the Future of Driving

On a brisk March morning in 2004, under the wide-open skies of California’s Mojave Desert, a strange parade began to form. There were fifteen vehicles at the starting line—each one bristling with spinning sensors, welded brackets, and wires that seemed to dangle by the grace of God or duct tape. One looked like a Humvee that had swallowed a satellite dish. Another was a motorcycle that refused to fall over. Engineers and graduate students hovered like pit crew and parents, anxiously clutching laptops, antennae, and half-finished sandwiches.

This was not a joke, though it looked like one.

This was the inaugural DARPA Grand Challenge—a government-sponsored robot race, dreamed up by a Pentagon agency with a history of wild bets. The rules were stark: build a vehicle that can drive itself 142 miles through rugged desert terrain, without a human behind the wheel, without a remote control, without any guidance at all. Just software, sensors, and steel.

Oh, and the winner would get 1 million dollars.

In hindsight, that million-dollar prize seems quaint, almost ironic, given that the challenge would go on to seed a hundred-billion-dollar industry. But on that dusty day near Barstow, with the sun beating down and the caffeine supply running low, the future of autonomous driving looked more like a science fair than a revolution.

Then the race began—and everything fell apart.

Act I: The Desert is Unforgiving

None of the fifteen vehicles completed the course. In fact, none came close. The best performer, a heavily modified Humvee named “Sandstorm” from Carnegie Mellon University, made it just 7.4 miles before veering off course and getting stuck on a rocky incline. Smoke poured from its wheels as its navigation system tried to brute-force its way out of a ravine.

“Sandstorm”

Another vehicle made it less than a hundred yards before it flipped over.