Buoyed by their stunning win of a desert "Grand Challenge" last year, Stanford University robotics experts have begun work on a car that will drive itself from San Francisco City Hall to downtown Los Angeles in October 2007.
The original car, dubbed "Stanley," is facing retirement and "will likely wind up in a technology museum," one of its key architects, computer science Professor Sebastian Thrun, told the Weekly as he showed Stanley off in advance of a public presentation Thursday evening.
His talk on Stanley's "brains" and the historic desert race will be at 7:30 p.m. at the Hewlett Teaching Center on the Stanford campus -- it is free and open to the public. Information is on the robotics Web site: http://robots.Stanford.edu .
Thrun said he has long been interested in various aspects of robotics and artificial intelligence, but became interested in robotic cars when the first DARPA Grand Challenge was run in 2004 and none of the vehicles completed more than 5 percent of the course.
"To me it was a no-brainer that a car could do better," he said.
Thrun also shares a longer-term vision -- say five or 10 years -- of robotic cars driving commuters to work, allowing their non-driver occupants to reclaim some of their commute time for productive activities or just relaxation.
As for the SF-to-LA run next year, Thrun said one big difference is that a human being will be sitting inside the car, which has yet to be given a name, as a safety measure. Stanley was on its own when it completed the 132-mile Mojave Desert run last Oct. 8, netting the $2 million award for Stanford's Computer Science Department.
"From a futuristic perspective, it's infinitely safer to have someone in the car," Thrun said, adding that the California Highway Patrol would likely have something to say about an unmanned vehicle cruising around the state.
"We have no intent to kill anybody."
The non-driver human would keep hands off the wheel except in an emergency situation, Thrun said. Just as today's cruise-control systems disengage when someone taps the break pedal, the robotic systems would disengage if someone took hold of the steering wheel.
The new, traffic-ready robotic car, will also be a Volkswagen, but instead of the burly SUV, four-wheel-drive model it will be a lighter station-wagon, probably a Passat, Thrun said. VW is a partner in the robotic-car venture, along with several other firms. And it will have to travel at highway speeds rather than the 25-mph crawl Stanford achieved across the desert to ace out more than 100 other vehicles. Stanford's team focuses primarily on the software and integrating it into a functional system.
The long-term implications of robotic-assisted commuting are huge, Thrun believes.
"The average American worker spends an hour a day in the car, being non-productive. The amount of (productivity) money lost is estimated at $61 billion a year. This is a huge opportunity. It's going to be a multi-billion industry," he said. The average is far higher for Palo Alto area commuters and Bay Area in general, he acknowledged.
In addition, robotic-assisted commuting could actually make the roads safer, he said.
"Last year, 42,000 people died in accidents -- 15 times as many as died in 9/11. If you could make driving 10 percent safer, you could save more than 4,000 lives a year," plus eliminating the suffering, lost time and costs of damage in far more injury accidents.
Many components of a robotic vehicle are already making their way into vehicles, he noted, citing emergency braking warning systems.
Stanley cost about $500,000 to assemble, Thrun said, and much of the funding for his traffic-ready successor has already been lined up.
Much of the $2 million Desert Challenge Award will go to create a permanent endowment fund "so there will always be a 'Stanley Student' at Stanford," Thrun said.