If global wireless communications, fighter drones, and Martian space probes aren't proof enough that the kitschy 1950s version of the future has come to pass, some folks from UC Berkeley have yet another piece of evidence: cars that drive themselves.
The technology has advanced to the point that even a college kid can build it. To wit, in a campus courtyard one recent evening, a knot of Cal students was watching a race between teams from a "mechatronics" design class. Wearing T-shirts and shorts or sweats, the young onlookers stood on one side of a yellow caution-tape barrier. On the other, black tape marked a looping track on the courtyard tiles. A model car the size of a large shoebox and topped with a mass of wires and switches began driving around the track. There were oohs and cries of "Nice!" from the crowd as it scuttled around the course at a good speed, following the tape with impressive precision. Its builders watched nervously as they waited for their test vehicle to complete its run -- there was little else to do, since the humans were out of the loop.
While the race was merely an academic exercise -- the teams were competing for a grade -- it was a demonstration of know-how similar to that being used to build real self-driven cars, once solely the fantasy province of sci-fi novels. Replace the abovementioned tape course with special freeway lanes embedded with magnets -- and put buses, big rigs, or real cars in place of the model ones -- and you've got the idea.
These robot vehicles are ready to hit the highway, Cal researchers say, and a few already have. Caltrans and UC Berkeley demonstrated self-driving cars on a freeway near San Diego in 1997 as part of a national "automated highways" consortium, and tested self-driving buses in 2003. Putting such vehicles into regular use, they say, could drastically cut down on traffic congestion, and consequently air pollution, in the Bay Area and Los Angeles. "The technology is there now," says Cal professor Karl Hedrick, former head of California Partners for Advanced Transit and Highways, or PATH, a statewide program based at the university.
Here's how it would work: The driver takes the car as far as the freeway, where it enters a highly exclusive carpool-type lane. At that point, an onboard robotic pilot takes over, sensing tiny magnets embedded in the asphalt. The car joins a tight "platoon" of robot vehicles that follow one another closely and synchronize speeds via radar and radio communication. When a vehicle reaches its desired exit, the car exits the platoon and the special lane, and the driver takes over once again.
These tight convoys, researchers say, will lead to reduced congestion and fewer accidents, plus higher fuel efficiency, since the speed of the cars is constant and each is shielded from excessive wind drag by the car just in front. During the 1997 demo, Hedrick recalls, the drivers sat back in their seats as eight specially adapted cars maneuvered themselves into a platoon along modified highway lanes. "They never touched the wheel," he says. The automated buses tested a year and a half ago worked in essentially the same way and were demonstrated on the same stretch of highway, where the magnets remain.
Robot features already can be found in some regular cars. "Adaptive cruise control," for example, monitors surrounding traffic and adjusts the cruise speed accordingly. This system, optional in high-end Mercedes models, works just like the robot cars' platooning system, Hedrick notes. The only difference is that adaptive cruise control isn't intended, or equipped, to work entirely without driver cues. Another recent option on some cars is "lane departure warning systems," which sense when a car is veering from its course.
But perfecting -- or nearly perfecting -- the technology is a far cry from marshaling the political will to spend money on such "futuristic" technologies, or the social willingness to relinquish control of our cars to robots. Forgetting the substantial cost of dedicating special freeway lanes, the specter of computer-driven vehicles raises a host of questions, from who bears responsibility for system failures to how insurance companies would view the technology. "There's the technological aspect and then there's the political, legal, social aspect," notes Ron Fearing, professor of the mechatronics class. "The technological side is probably closer to being solved than the political, legal issues."
In the Berkeley class race, not long after the road-hugging speedster elicited oohs and aahs of appreciation, another model car hit the track. As its builders looked on with increasing dismay, it sped away from the start line only to waver wildly, before swerving off-course and knocking over a handful of the soda cans and water bottles used to flank the track. Extending the analogy between real and miniature robot cars illustrates some of the obstacles to adoption of the real-world models. Imagine the headlines when the first computer glitches lead to freeway mayhem, injuries, or fatalities.
Researchers working on real robovehicles maintain that automated buses, trucks, and passenger cars would make the roads safer overall because they would cut down on the driver screwups implicated in the majority of car accidents. "Ninety percent of accidents are due to human error," says Pete Hansra, a Caltrans engineer. "Computers don't take drugs, computers don't feel sleepy, so the computer can control the car and we can reduce a lot of accidents."
Caltrans funded research on so-called "automated highways" up until late 2003, but pulled the plug soon afterward due to budget constraints, says Hansra, who notes that the priority is the "day-to-day stuff" like potholes. Still, he says the agency plans to return to such projects as the economy improves and Caltrans' budget grows. That funding would go toward automated buses first, he suspects, though the agency also would be interested in backing work on automated cars.
Steven Shladover, a researcher at the UC Berkeley PATH program, also expects automated buses to be the first application of robot-car technology. One early use would be precision docking, where a robopilot would kick in as a bus pulls up to a stop. This would allow it to get close enough to a platform to allow passengers, including wheelchair users, to enter across a flat surface. Auto-driving buses also could drive along special busways in platoons, like those demonstrated in 2003. Besides reducing traffic, fuel use, and accidents, Shladover explains, the buses drive precisely enough that their dedicated lanes need be barely wider than the buses themselves, minimizing the space needed to accommodate them.
Self-driving buses are an easier sell to politicians, industry, and the general public than automated cars. After the buses, we can probably expect to see automated big rigs driving in efficient platoons, Shladover says. It may be quite some time, though, before the average driver gets the chance to set foot in a robocar.
But commuters who dream of leaning back and reading the paper while their cars navigate morning traffic could help speed things up a bit, according to Hedrick, the former PATH director. "It's going to take a city like Los Angeles that says, 'We can't survive like this anymore,' and they go to the federal government and say, 'We want automated lanes,'" he says. "I always thought the first application would be in areas where the congestion is so bad -- LA and San Francisco are likely candidates -- that the people start to demand something."
Perhaps the real question, then, is whether drivers are prepared to hand the wheel over to a computer. To hazard a guess: Maybe one day in the future.