DARPA Challenge brought lab tests out into the real world, and started a movement
December 19, 2016 @ 12:01 am
Unless you are living in an underground bunker and trying your best to ignore the auto industry, you are aware that the car business is now hurtling madly toward a future called "autonomous driving."
Automakers around the world, tech companies that want to be automakers, auto parts makers, municipal planners and even operators of delivery fleets and taxi companies all are marching toward a reality of cars and trucks that will steer themselves, brake when drivers are not paying attention, change lanes by themselves and, eventually, pick you up at your house and take you to work with nobody at the steering wheel.
You may have wondered: How in the world did all this start?
The answer surprises many.
There was a Big Bang on a Saturday in November 2007, and chances are you missed it.
"That was the moment," agrees Red Whittaker, a leading robotics professor at Carnegie Mellon University in Pittsburgh who has spent his career exploring and patenting ways to automate mining, farming and industrial vehicles.
"That day in 2007 was the moment when concepts that had been around for years suddenly came out of the laboratory and into the world. And unless you were aware of the decades of research that had been going on, the whole thing probably came as a complete surprise to you."
It was on a closed Air Force base in Southern California that day that the autonomous-driving industry was born. The U.S. Department of Defense -- as in the Pentagon -- had invited a couple hundred of the country's most advanced transportation science and computer problem-solving thinkers to conduct a 60-mile obstacle course race -- primarily to make a point. The Defense Department wanted to demonstrate that it was possible, practical, safe and maybe even financially attractive to make automobiles drive through a city with nobody behind the wheel.
As the country's "real" auto industry geared up for its annual holiday season blowout sales, the California gathering was largely overlooked. But in the months and years that followed, future-minded thinkers realized that the race had been an auto industry game changer.
Into the light
The Nov. 3, 2007, contest was not the first of its kind. The Defense Department for years had been trying to nurture research on self-guided machines at big universities, including the Massachusetts Institute of Technology, Stanford University and Carnegie Mellon. Robots were seen as the next great battlefield tool in America's arsenal. Robotically guided vehicles might be used to approach suspicious cars in war zones. Robot devices might creep through caves or over mountain hideouts or along the seafloor looking for enemy combatants.
An independent Defense Department organization launched by the Eisenhower administration, called the Defense Advanced Research Projects Agency, or DARPA, was, by the 2000s, bent on bringing advanced transportation thinkers out of their laboratories and university computer rooms to compete in the light of day. The government had even granted DARPA special power to give away prize money to sponsor public contests to do it.
In 2004, DARPA offered a $1 million prize to challenge participants to get a driverless vehicle to complete a 150-mile trek through the Mojave Desert. But like new Marine Corps recruits collapsing in basic training, not one of the participants finished the course, and no one claimed the prize.
A year later, DARPA tried again, this time offering $2 million to challenge teams to complete a 132-mile Mojave course with driverless vehicles. And this time, five teams completed the race. A Volkswagen Touareg converted by Stanford University into a self-driving vehicle called Stanley won.
Two years later, DARPA doubled down, moving the contest to a mock urbanscape where driverless vehicles would have to navigate lifelike city traffic and unpredictable pedestrians.
Stanley, a modified VW Touareg, was Stanford University's winning entry in the 2005 DARPA Grand Challenge. Stanford race participants went on to populate the industry's autonomous-drive push.
Less cheerful times
At that late 2007 moment, the auto industry was not a happy place. Auto sales were falling. Pickups and SUVs, which had made the industry rich for a decade, were taking it on the chin as gasoline prices rose and the housing market teetered into trouble. The U.S. economy was looking sketchy. Americans were losing their jobs, and General Motors was losing money.
Among the teams for that race, Stanford was back with its VW partners. An MIT team had partnered with Land Rover. Virginia Tech worked with Ford Motor Co. Carnegie Mellon worked with GM.
"There was a little more awareness this time around," recalls Whittaker, who directed the Carnegie-GM team. "You could tell that some people were starting to watch this activity a little closer. In the previous DARPA races, it had really been an off-road competition with a lot of rugged vehicles, like a dune buggy, an old Hummer and an autonomous motorcycle. But for the 2007 race, it was really about urban driving. It was really about ordinary passenger cars for the first time."
Strolling through the racing grounds that Saturday were individuals who illustrated the growing awareness. Steve Wozniak, co-founder of Apple Inc., visited with the teams and introduced himself. At that time, Apple had no car venture.
Also in the crowd, observing the teams, were Google founders Larry Page and Sergey Brin. Google also had not declared its interest in automobiles. Microsoft, a company that had been at least supplying software products for Ford and other automakers for nearly a decade by then, also had representatives strolling the field.
Carnegie Mellon's modified Chevrolet Tahoe, winner of the 2007 DARPA race that lit the fuse for today's autonomous driving frenzy
Carnegie Mellon's converted Chevy Tahoe won the race, and Whittaker's team accepted the $2 million. But Whittaker philosophically muses that "there was the prize money -- and then there was the larger prize. We proved the technology worked out on real streets and could comply with traffic laws and compete with real cars."
"We ignited the industry's interest," he says. "OEMs had to start asking, "How can we not participate in this technology?'"
Immediately, many companies saw it as a ground-floor opportunity. Race participants from that day became hot properties for computer tech companies and automakers that suddenly glimpsed a future of driverless cars. If a college professor knew how to make a sedan stop by itself to avoid an obstacle in the road, there were new fortunes to be made. If there were university computer science technicians who knew how to convert that behavior into computer code, they were valuable employees.
"A handful of companies immediately saw amazing potential in all this, and they began hiring everybody they could from academia," says one race participant who was recruited to help Internet giant Google get into the automated vehicle business and asks not to be identified. "But that's the great part of this story. An idea emerged from outside of the traditional auto industry, and companies like Google with their very fast-moving and creative culture got it. And as a result, the idea took flight far faster than it would have otherwise."
Google said as much itself three years later, in 2010, when the company revealed that it was planning to produce self-driving cars. A revealing blog post connected its work directly to DARPA.
"To develop this technology, we gathered some of the very best engineers from the DARPA Challenges, a series of autonomous vehicle races organized by the U.S. Government," the public statement said. "Chris Urmson was the technical team leader of the [Carnegie Mellon] team that won the 2007 Urban Challenge. Mike Montemerlo was the software lead for the Stanford team that won the 2005 Grand Challenge. Also on the team is Anthony Levandowski, who built the world's first autonomous motorcycle that participated in a DARPA Grand Challenge, and who also built a modified Prius that delivered pizza without a person inside."
The Google blog was attributed to a "software engineer" named Sebastian Thrun. But the academic world knew Thrun was far more than a software engineer. A renowned researcher in computer learning, the German-born Thrun was recruited to Google soon after running Stanford's racing team for DARPA in 2007. His team's converted Volkswagen Passat named Junior came in second behind Whittaker's Chevy. Thrun's Stanford team also had developed Stanley, the DARPA 2005 winner.
Thrun stepped down to an advisory role at Google in 2014.
But the degree to which academic talent was hired away after DARPA was a surprise, Whittaker says. He saw a number of colleagues and graduate students recruited to auto industry pursuits from Carnegie Mellon's Robotics Institute, where he has worked for 36 years. Urmson, for instance, an assistant research professor, had served as director of technology on Carnegie's winning DARPA 2007 car. He was recruited to direct Google's self-driving car project in early 2009. He left that Google post in August.
"I've spun off some 20 companies and organizations over the years from what we do here," says Whittaker, 68. "This is how it works in academic circles. But I really didn't see this one coming."
After the 2007 race, Whittaker, who drives home every day to a large farm where he raises some 200 head of cattle, turned his focus back to his earlier research work of automated industrial and farming vehicles. His career research has led to the development of massively large automated rock-hauling trucks, now being used in an Australian mine. His patents include the technology to let tractors plant and harvest rows of crops without a driver.
He believes that hugely important applications need to be developed outside passenger cars to automate dangerous vehicle jobs in construction, mining, farming and emergency fields. Those fields will be just as revolutionized as passenger cars by the advent of driverless technologies, he says.
Meanwhile, the autonomous Big Bang continues to pull talent out of universities such as Carnegie Mellon.
This year, the San Francisco-based ride-hailing company Uber opened an advanced technologies center in Pittsburgh, near Carnegie's campus. Uber hired away some 40 people from Carnegie's staff to bring driverless-car know-how to what is currently a taxi company.
John Bares crossed over from Carnegie Mellon to Uber. A research scientist who came to the university as an undergraduate, stayed to receive his Ph.D. and later became director of the National Robotics Engineering Center at Carnegie, Bares left last year to become director of Uber's new tech center.
"These people are pioneers, and this is inevitable now," Whittaker says. "What you start just keeps spreading out in new ways. You don't know where discoveries will take you. Do you think that when the Wright brothers first got their plane off the ground they had any idea that it would result in the technology to compare airline ticket prices online?"
There was a telling moment at the conclusion of the ceremonies of that November 2007 race, some participants recall.
The director of DARPA at that time, Tony Tether, was presenting the prizes and congratulating the teams. A team member asked him, "Tony, when is DARPA's next race, and what will it be next time?"
"There isn't another race," they recall Tether answering. "Mission accomplished. You've achieved what we hoped to achieved. That's it.
"Our job was to start the ball rolling," he said. "Now all of you have to go out and make something out of it."