Here Come Self-Driving Cars
The Technology Is Nearly Ready, but Are We?
The notion of self-driving cars is at least 70 years old, and Google showed it to be a real-world possibility earlier this year with well-publicized tests of its own autonomous vehicles. Indeed, in the past six months just about every major automaker has announced some type of autonomous or semi-autonomous vehicle project.
So the issue isn't whether technology can deliver self-driving cars. The questions instead are when they will be ready for market and, more importantly, when the market might be ready for them.
In addition to some complex questions about insurance and liability involved in cars that drive themselves, there's the issue of whether car buyers, who have been steeped in the belief that cars are vehicles of personal freedom, will now be willing to give up the wheel. What's the problem that self-driving cars would solve for consumers? Why would we want them?
Self-driving cars could lead to substantial decreases in automobile accident rates. They've been trending downward. The Census Bureau counted 5.5 million auto accidents in 2009, the last year for which national data is available, versus a peak of 6.7 million in 1995. But cars that take human driver error out of the equation could result in a much faster rate of decline, automakers and safety experts say.
Surrendering the Wheel
At some point "it will be silly to continue driving, because your reactions are not as fast and you are not as safe" as a computer-controlled car loaded with advanced safety systems, says Nady Boules, head of General Motors Corp.'s electric and control systems research lab.
That's not an argument that's likely to win over the driving enthusiast who would never think of ceding control to a computer and a bunch of electronic sensors. But if car owners think seriously about the possibilities, even the most die-hard "don't take my steering wheel" types might find occasions when letting the car drive itself makes sense.
Driving a twisty mountain road is fun, for instance, but it might be nice to read the paper or play a hand of cards with fellow passengers while the car's electronics take care of daily commuting chores or a long, mind-numbing highway trek.
In addition to greater safety and freedom from boring drives, automakers also cite improved fuel economy as a reason to bring on self-driving cars. And there's a market-driven reason they're interested: Automakers want to ensure that they don't lose two important groups of car buyers.
John Hanson, head of safety, environmental and quality communications for Toyota Motor Sales U.S.A., says that baby boomers, a significant part of the nation's car-buying market, "are getting older and their reactions are slowing, but they still want personal mobility." And younger people aren't as interested in driving, he says. "They are more interested in social interaction and connectivity." Autonomous driving, he says, can serve the needs of both groups.
Drivers Still Needed
Still, we're far away from the day when you can load the grandparents in the backseat and tell the car to take them to their doctor appointment. Autonomous cars that won't let the driver take control are possible, but no one is yet seriously considering that scenario, says Ford Motor Company technology spokesman Alan Hall.
That could change "far out in the future with a real intelligent transportation model," he says, but Ford's belief is that as things stand now, "the driver always needs to be engaged in the driving process.
"The driver is still the driver," Hall says.
That belief is shared by other automakers and, it seems, by regulators. It certainly is reflected in the widely publicized decision by Nevada lawmakers to "legalize" driverless cars. In fact, the state's law merely set up regulations for testing such vehicles on Nevada highways. It explicitly requires two licensed drivers in each vehicle, ready to take control if the autonomous systems fail.
Cars Could Be Ready by 2025
People who get paid big bucks to know about these things say that the technology behind autonomous driving has largely been developed now. The remaining technical hang-ups are miniaturization, cost containment and system integration.
GM's Boules says the self-driving car can be ready for the mass market by "sometime in the next decade."
Daimler research chief Thomas Weber is even more optimistic. The technical issues will be resolved soon and preconditions for autonomous driving will be in place by 2015, he told interviewers earlier this year.
Google has been testing its fleet of self-driving cars for several years. The vehicles have collectively run up more than 300,000 miles on real streets and highways without an accident, although there are reports that the human monitors who sit in the driver seats have had to intervene a few times.
Carmakers don't see huge technical barriers, either. "We could make one now," says GM's Boules.
Are People Ready?
But who wants them? As with so many things, a military application might lead the way.
"The military is interested in self-driven vehicles because they are a way of pulling personnel out of harm's way," says autonomous driving researcher J. Christian Gerdes, who heads the Center for Automotive Research at Stanford University (CARS). "But just three years ago, you wouldn't have heard these conversations among automakers."
A big unknown is whether consumers will want this technology once it is ready for the market. Like iPads and bacon-flavored ice cream, cars that can drive themselves could fulfill a need most people don't yet know they have. Ford, for example, "is still very early in the process of getting usable feedback from consumers," says Hall.
"There's still a lot for us to learn in the evolution and rollout of driving assistance technologies, including listening to customers about how they value these technologies and what they want to see happen," Hall says. Ford wants to know from consumers how autonomous driving technologies "improve and take away from the driving experience."
Safety agencies and automakers also will have to put on a concerted consumer education drive to get people comfortable with the idea that the car can drive itself, Boules says.
In the meantime, companies like GM "are spending a lot of effort and resources to try to understand driver behavior in the instances when the driver is relinquishing command, or taking it back," Boules says.
Self-Driving at Pikes Peak
To see what robotics are best for, and where humans have the edge, researchers at Stanford teamed with the Volkswagen Group and Oracle Corp. a few years ago to outfit an Audi TTS sport coupe with autonomous driving gear and send it up Colorado's Pikes Peak.
Cars reach the summit via a 12.42-mile stretch of road that has about 160 turns and rises 6,500 feet from just about halfway up the 14,100-foot mountain. No easy Sunday drive. Yet the test car made the climb sans driver and without incident in 2010, and did it at a real-world average speed of just under 30 mph, versus about 44 mph for a professionally driven stock TTS.
Stanford continues to build and test self-driven racecars. Gerdes says the idea is to use them to push technology to the edge, and to learn from how they fare without drivers in control.
Cost Is a Barrier
If you could order up an autonomous car today, what you'd get would be expensive and eye-catching — but not in a good way. Google's test cars have what looks like a metal pail mounted on the roof. Inside is an array of 64 laser sensors that create real-time, 360-degree maps as the car moves through traffic, helping its brain locate it precisely and identify surrounding objects.
The car has a few dozen additional sensors as well. They give it awareness of traffic and other conditions in its immediate area. That information enables the control software to make driving decisions. The price tag for that rooftop array alone cost $70,000 when it was initially deployed in 2007 at the automated vehicle trial sponsored by the Defense Advanced Research Projects Agency (DARPA), says GM's Boules.
It might be cheaper today — Google isn't talking. But the array still is expensive and looks goofy on top of a car. Google can get away with that, says Boules, but GM and other automakers can't.
"It's not well-integrated into the car design," Boules says. "To replace all that with other sensors integrated into the car that can achieve the same results and be cost-efficient is a tall order. We need coordination among suppliers, systems integrators, universities and the auto industry."
Still, he says, "DARPA showed it can be done" and ongoing tests and trials of autonomous driving systems by BMW, Daimler, GM, Toyota, Volkswagen Group,Volvo and others continue producing advances that are hastening the arrival of the driverless car.
Sorting the Legal Issues
Building competitively priced vehicles with autonomous driving capabilities is only part of the battle, though. Developers will have to persuade motorists — and regulators — that the systems work flawlessly. And myriad legal issues must be resolved.
Persuasion will be the job of safety advocates and automakers' marketing departments, using the results of years of testing to show consumers that self-driven cars will work. The legal issues, however, will have to be tackled by lawyers, insurers, lawmakers and the auto industry.
One place where the issues already are being contemplated is Stanford University, where Bryant Walker Smith is trying to come to grips with how society might best ready itself for the driverless vehicle.
Smith, a research fellow at both CARS and Stanford's Center for the Internet and Society, says that the key issues are insurance and legal liability, operating regulations, data security and privacy.
A New Insurance Approach
On the liability front, Smith says he sees autonomous vehicles bringing a shift in the way insurance is applied and paid for.
Instead of individual motorists purchasing separate policies for each vehicle, he sees manufacturers insuring the cars and consumers paying the premiums through higher car prices.
One thing Smith insists is not a problem is the question of whether such vehicles are legal or not. Technology progresses because new technologies are automatically legal in the U.S., he says. Lawmakers have to outlaw something to make it illegal.
Still, vehicles are regulated and driverless vehicles will be no different. Several states — including California, Florida and Nevada — are already preparing the ground. That leaves 47 states to go.
Faulty Humans, Forgotten Skills?
It's not a lack of technology that will keep self-driven cars off the roads for years to come, say Boules and other experts. Instead, their introduction will most likely be slowed by the legal complexities inherent in regulating them.
Stanford's Gerdes says there still is a technology issue to consider: Computers are only as good as the humans who program them. "If you say putting a computer in place of the human driver makes a car safer, you've got to remember that the computer is susceptible to human error."
Gerdes also wonders if people's driving skills will atrophy with increased reliance on automation. He suggests that ongoing driver training programs should accompany automated driving.
"Driving is a combination of reflexes, informed judgment and experience-informed decision- and prediction-making when we need to avoid or get out of problem situations," he says. Until the computers get all that down pat, "it would be a mistake to think that all this is really ready to go."