My Chauffeur, the Robot
It has been going on for more than five years now in Mountain View, California, hometown of Google Inc.: cars with odd-looking devices like metal buckets on their roofs, painstakingly negotiating city streets. They don’t speed, they don’t tailgate—they just make their cautious, unhurried way through town.
These customized Priuses and Lexuses are Google prototypes for driverless cars. Technicians have been testing them since 2009, and, according to Google, they’ve traveled more than 700,000 accident-free miles on Nevada and California streets and highways, getting ready for their introduction to car buyers within the next five years, the company says.
The high-profile global Internet company has leaped into the burgeoning robotics field with an energy that’s drawing the rest of the auto industry with it, experts say. “Google, along with Tesla with its electric cars, has provoked much faster change than the traditional manufacturers would have brought,” says University of Texas engineering professor Kara Kockelman, a leading authority on driverless vehicles.
Safety experts and those stirred-up auto manufacturers are now all in. They’re betting that in the next decade, the trend toward robotics will revolutionize people’s traveling habits, save billions of dollars in costs related to congestion and damage from traffic accidents, and prevent thousands of deaths. The Center for Automotive Research, an independent nonprofit organization in Ann Arbor, Michigan, has proclaimed driverless technology as “the engine pulling the auto industry forward.”
As for the long-term effects on the auto industry, nobody’s sure. There are, as usual, assumptions that exciting new products will stimulate sales. But there have also been predictions that the new technology just might prompt a wholesale reduction in car sales, with motorists buying into dynamic robot-driven ride-sharing programs and fleets of driverless cars actually reducing the total number of cars on the road.
The trend toward robotics will revolutionize people’s traveling habits, save billions of dollars in costs related to congestion and damage from traffic accidents, and prevent thousands of deaths.
Not Quite Ready for Liftoff
In any case, Mountain View residents, including employees of the many high-tech companies nestled in Google’s neighborhood, say the cars with the buckets—holding the all-seeing LIDAR systems that use radar and laser-focused imaging to scan roadways ahead and behind, gathering 1.5 million bits of data per second—are virtually ubiquitous in town.
“They’re all over the place,” says Eugene Robinson, an editor at Mountain View-based Ozy.com, a news website. “But they’re not driverless. They seem to have drivers.”
That would be because California regulations now require a “driver” to baby-sit the driverless car when it’s sharing the road with other vehicles.
But it’s also because Google—and the dozen or so car manufacturers furiously trying to come up with their own versions of the driverless car—have yet to reach that crucial point of liftoff, when human drivers are able to relinquish all control to the machines. They’ll let the autonomous car handle a slow-moving situation, but with a human driver ready to take over when something unpredictable happens. Or they’ll send an unmanned vehicle down a course that has been mapped out with electronic guideposts—“like a rollercoaster on track,” says Mary Cummings, a former Navy fighter pilot who is now a Duke mechanical engineering professor.
But turn the car loose, unsupervised, in city traffic? “We’re a long way from that,” says Cummings, who is helping Google develop the artificial intelligence that driverless cars will require.
The AV Is Here
Even so, the autonomous vehicle, or AV, as researchers call it, is no longer a science fiction fantasy. The AV is here. At least, big pieces of it are.
Since 2011, Heathrow Airport in London has had a fleet of “pods,” unmanned vehicles carrying up to four people and their luggage, to transport passengers from parking lots to terminals at speeds up to 25 mph. Mining company Rio Tinto has been experimenting with robotic trucks since 2008. It now has more than 50 driverless trucks hauling ore at four Australian mines. At the same time, car manufacturers are bringing components from AV prototypes—like adaptive cruise control, where the driver sets the speed but the machine overrides him if he gets too close to the vehicle in front—into their new models.
Google recently unveiled a Volkswagen Bug-sized car some observers have dubbed the “koala car,” which is so automated it doesn’t contain a steering wheel. Google plans to build 100 of them by mid-2015, presumably for transporting blind or disabled people in low-traffic situations, though the company is not yet sharing details on the project.
To be sure, there are some daunting challenges that must be overcome before fully autonomous vehicles start hitting the roads. One is teaching the computers that run the cars to think more like humans. That’s what a lot of researchers like Cummings are focusing on now.
“When you’re driving in a fog and you see a bridge, you understand what it means, even though you just see the edge of the bridge,” Cummings says. “In your mind, you extrapolate what the entire bridge looks like. You have the ability to construct a speedy mental model based on sparse information. Computers are just not that good at it yet.”
Teaching It to be ‘Social’
Google officials acknowledge that the company’s car has a ways to go before it’s ready for real-world driving situations. They say that much of what they’re doing with those jaunts through Mountain View and elsewhere is trying to teach the car to recognize “social signals,” such as those involved in changing lanes, merging onto freeways or behaving correctly at four-way stops. There are also challenges to overcome in bad weather, especially snow or fog, which may retard the accuracy of the LIDAR system.
Also remaining to be worked out are liability issues. Who’s responsible if there’s damage from, say, an AV rear-ending another car? Is it the software’s failure or perhaps a mechanical breakdown? Or could the “operator” of the vehicle, the person who pushed the buttons to get the car moving, be somehow liable? In short, who’s to blame in an accident if the vehicle is the driver? Google is well aware that it would likely be at the top of the lawsuit list in such cases.
In fact, the most compelling argument for the widespread use of AVs is safety. About 33,000 people die every year in the United States in traffic accidents, and more than 2 million are injured. The death toll worldwide is about 1.2 million. Safety experts say that, in terms of driving safely, robots have it all over us humans.
“Under the best of circumstances, human reaction time [to a road hazard] is 0.2 to 0.5 seconds,” Cummings says. “The AV can actually detect a situation far earlier than humans.” Then there’s the matter of focus (“Unfortunately, humans are not really good at paying attention for long periods of time,” she says), to say nothing of bad human habits like drinking or texting while driving. According to the National Highway Traffic Safety Administration, more than 90% of all crashes are caused by human error, with 40% of fatal crashes attributable to alcohol, drugs, fatigue or other all-too-human distractions, to which driverless cars would presumably be immune.
Two researchers for the Eno Center for Transportation, Kockelman and Daniel Fagnant, analyzed the potential savings that autonomous vehicles would bring from crashes avoided, reduction in travel time, saved fuel and possible parking benefits. The estimated potential payoff: about $2,000 a year per AV and as much as $5,000 when “comprehensive costs” from accidents (costs from pain and suffering by crash victims and their families) are figured in.
Even more important is that with sufficient market penetration, AVs have the potential to eliminate most traffic deaths in the world, experts say. “That’s like curing one of the world’s most dangerous diseases,” says Brad Templeton, who has worked as a Google consultant on driverless cars and founder of one of the world’s first dot-com companies, ClariNet Communications Corp. All this, of course, assumes that these systems operate flawlessly and their internal computers do not crash. Understandably, those assumptions may make the market nervous. Can anyone recall a new computer or mobile device without problems?
The Race to Driverlessness
Once developers convince regulators and the public that they have bug-proof systems, it seems clear that significant benefits will accrue not only to the public, but also to the auto industry—and consequently to the metals industries.
Aside from Google, right now Audi, BMW, Cadillac, Ford, GM, Mercedes-Benz, Nissan, Toyota, Volkswagen and Volvo are known to be testing “driverless systems.” Some are close to being marketable.
Already the race to driverlessness has produced adaptive cruise control and electronic stability control (with electronic sensors feeding information to the braking system so that it selectively applies the brakes to avoid skids). Soon to come, according to industry experts, are automated lane-changing programs, parallel parking systems and automated valet parking. That last feature involves cars parking themselves at a stadium or a mall and then, after being given a signal, returning to pick up a driver.
And Audi expects to have its new traffic jam pilot ready for new models by 2018. This is a stress-avoidance system that allows the driver to turn over the controls to the car itself during slow-moving traffic. The program, which the company demonstrated in 2013 at the annual Consumer Electronics Show in Las Vegas, handles the bumper-to-bumper maneuvering while the driver relaxes, perhaps pushing back his seat for a few minutes of meditative quiet, while the other drivers around him grind their teeth with frustration.
The technology works in freeway traffic conditions up to 40 mph. At that point, the car alerts the driver that it’s time for a human to take over.
It’s the kind of technology that harried urban drivers dream about, says Audi of America Communications Manager Brad Stertz. “Imagine going around the Beltway [in Washington, D.C.] with the car doing the bulk of the really tedious driving, like moving 10 or 15 feet at a time,” he says.
Clearly, newfangled technology like this is likely to sell cars. That’s good news perhaps for the U.S. auto industry, which has not quite recovered from the 2008 recession, despite having sold about 16.5 million vehicles last year. In the pre-recession 2000s, levels reached 17 million six times. Canadian sales totaled about 1.8 million last year, up from about 1.7 million.
Industry experts say that most auto manufacturers have done market research studies, though they’re not making them public. Some U.S. academics have done studies of their own, notably the University of Michigan’s Transportation Research Institute, which last year studied consumer attitudes toward AVs in China, India, Japan, the United Kingdom, Australia and the United States.
They Want It, but Will They Pay?
The researchers found that in all of those countries, most respondents knew about the new technology and wanted it in their cars. But, except for respondents in China and India, a majority was unwilling to pay extra for it. In the United States, 44% said they were either very interested or moderately interested in owning a self-driving vehicle. But only 25% said they’d be willing to pay $2,000 extra for the new technology (way below what the market would require for the AV systems), while more than 50% said they were unwilling to pay anything extra.
But turn the car loose, unsupervised, in city traffic? “We’re a long way from that.”
Kockelman and her University of Texas students conducted their own survey in tech-savvy Austin. They found that, depending on what an AV could do, respondents there were willing to pay on average from $4,000 to $7,500 for the new technology (the latter for the full AV treatment, with drivers turning over the controls to robotic systems), though about 15% said they wouldn’t pay anything. The AV’s time has arrived, Kockelman suggests. “The car manufacturers know that if they don’t catch this wave, they’re out,” she says.
For manufacturers, though, there could be some troubling long-term implications for their bottom lines in all of this. By most estimates, if there were currently an AV in automobile showrooms, it would probably sell for about $80,000, which is way above what most car buyers are willing to spend. And consumers might quickly gravitate to car-sharing programs, which would operate like Car2Go and Zipcar, with the added benefit of a home pickup.
“Mostly, people don’t want to have to buy these assets or manage them,” Kockelman says. “So they’d give up their second car … and eventually their first.”
In a study of consumer habits, the Texas researchers found that “Americans have many more cars than they need to serve current trip patterns in most locations.” Fewer than 17% of newer cars owned by private households are in use at any given time on an average day, the researchers found, and fewer than 10% when older cars (more than 10 years old) are included.
“The car manufacturers know that if they don’t catch this wave, they’re out.”
Who Needs Their Own Car?
Hence, car sharing as an efficient alternative to owning a car. Kockelman says it’s “like buying by the glassful instead of owning the whole distillery.” But established car-sharing services like Zipcar wouldn’t provide the service that a self-driving car, activated by a mobile app, would. An AV could be prompted to pick up a rider at his front door, drop him at his destination and return to take him home, all without the hassle of finding a parking spot. With service like that, Kockelman and Fagnant suggest, who needs to own a car?
But should we expect plummeting sales of new cars with the arrival of AVs? Not necessarily, Kockelman says. The new paradigm may actually promote interest in driving (with driver-friendly road conditions and greater access to cars for the disabled and elderly) and a consequent boost in demand.
Cars of the AV future may be “smaller and nimbler,” Kockelman says, which could be less profitable for manufacturers but “a win for the environment.”
There are plenty of obstacles to overcome before the United States and other countries become fully robotized on the road, says Michael Toscano, former chief executive and president of the Association for Unmanned Vehicle Systems International, an advocacy group for AVs. There’s a longstanding car culture, for example, which glamorizes the freedom of driving wherever you want to go. But a closer look at American car driving might more clearly expose its flaws, Toscano says.
“I guarantee that people who have to sit in traffic an hour or two every day don’t enjoy driving,” he says.
“We’ll always have NASCAR, because people like watching human beings go around a track at 200 mph making left turns,” Toscano adds. “But as for my commute of 17.8 miles, where I go through 51 traffic lights, give me a driverless car.”
Edmund Newton is a Washington, D.C.-based writer, formerly of the L.A. Times, Newsday and the New York Post, as well as the former managing editor of New Times-Broward Palm Beach. He has written for, among others, The New York Times, Time, People, Daily News Sunday Magazine, Black Enterprise, Ladies’ Home Journal, Essence and Audubon.