Headlines like to scream about self-driving cars coming to your garage because the media likes to hype things that are shiny and new. But if you actually sit down and talk to the experts about the car of the future—and not just the whiz-bang stuff but the whats and whens and hows—like we did in the run-up to the 2016 New York Auto Show, they’ll calmly explain that they’re not even remotely close to selling you a car that drives itself. Sure, GoogleX has put two million miles on self-driving cars. But what does that mean, exactly?
Mary Cummings, a professor of robotics at Duke University who discussed the topic in front of Congress in March, put it like this: “While I applaud this achievement, New York taxi cabs drive two million miles in a day and a half.”
In other words, driving cars under 25 mph around Mountainview, California, and select communities in Texas is a far cry from the real world, where cars would need to communicate with one another and with surrounding infrastructure—not to mention seamlessly and safely share the road with cars piloted by sentient meatbags. Lots of autonomous miles racked up by very few vehicles does not replicate this to scale.
In fact, as impressive as the GoogleX program and others like it are, they’re just the tip of the autonomous iceberg. Toyota’s John Hanson, who’s in charge of communicating the Japanese automaker’s R&D program, says a truly self-driving car is still a ways off because automakers have “already done the easy stuff—all that’s left is what’s really hard.”
The hard stuff that’s left is what automakers are really cranking on at the moment—where money and resources and brainpower are directed. It’s exciting stuff, and potentially problematic, because nothing ensures these separate entities are pushing in the same direction, to say nothing of actually working at cross-purposes. No one—no, not even Tesla—has answers, yet. But these are the questions that are furiously being addressed.
WHO SETS THE STANDARDS?
Audi’s working on deals to let your car park itself: In Somerville, Massachusetts, the company has a memorandum of understanding to help that tech-focused city build a garage for self-parking cars. Some cars can already parallel park themselves, or pull themselves into regular spaces. What’s coming are cars that, after you’ve dropped them at a garage entrance, could crawl through a multi-level structure, find an open spot, and occupy it. Electric or hybrid-electrics would park themselves in bays with ground-based inverters so no attendant would be needed to see to charging. Audi says that, in that kind of building, each space would be roughly two square meters smaller—and if you eliminate multiple elevators and stairwells, make narrower lanes, and excise pedestrian paths, the resulting garage can accommodate about 60 percent more cars within the same footprint. In an urban environment that amounts to less prime real estate used for parking. Automated parking is also more time-efficient for drivers, whether that’s based on a “smart” system that lets you reserve a spot on the street or in this type of smart garage.
But there are a few problems. First, there’s no agreed-upon protocol for a garage like this to use. How would other brands with their own, proprietary automated systems function in the Audi garage?
And who decides on such standards in the first place? In mid-March, 20 automakers agreed that automated emergency braking would be incorporated by 2022 on everything they make. Then Toyota went a step further, promising the tech across almost all of its brands by the end of 2017. That’s a good thing but it also points to the fact that, when it comes to many future technologies, carmakers aren’t made to adhere to any standards—likely because those standards don’t yet exist, whether you’re talking about how a car parks itself, the specific technology underpinning that capability, or what features an autonomous or semi-autonomous car should or must have.
What’s the standard for car-to-car communication? And who’s going to police all this?
Last month the U.S. Department of Transportation said its own safety standards are outmoded, concluding, for example, that under current law a self-driving car would be legally required to feature a steering wheel and brake and throttle pedals—even though the high-profile car Google wants to create would have none of those things. The DOT admits it will have to move faster while the auto industry, which often disagrees with Google on things autonomous, actually thinks they have a point here.
Ken Washington, Ford’s VP of Research and Advanced Engineering, says that deciding today what a future autonomous car should look like is premature, saying “it muddies the waters.” Washington adds: “One interesting business model is an autonomous car you don’t own. You wouldn’t want a customer who doesn’t own that car jumping into the front seat and start taking over control and doing things that aren’t designed into the service.” But he also notes that at this point, the car of the future might be both autonomous and drivable. Even that has yet to be decided.
In fact, deciding an overall “look” for the car of the future is a small part of a much larger and more complicated set of questions. How must autonomous cars drive? What’s the standard for car-to-car communication? What about car-to-infrastructure? And who’s going to police all this?
“Traditionally in the U.S., states have regulated driver behavior and the Feds have regulated the machine,” says Brad Stertz, who considers such red-tape matters in Washington, D.C. on behalf of Audi. Stertz says states like California have opened up a lead on the Feds by creating their own rules governing Silicon Valley’s autonomous-vehicle experiments, and that’s upended the traditional model.
That’s one reason why Google wants the Feds to step back in and make their own rules—so there’s some sort of road map to follow. Stertz notes that “nothing makes it more difficult for automakers than a patchwork of systems and regulations across multiple markets.”
Of course, that negates that possibility that the rules the Feds do come up with—like, say, autonomous cars must have a traditional steering wheel—prove unsuited or even antithetical to a self-driving vehicle.
HOW DOES THE INDUSTRY PROGRESS TO THE NEXT LEVEL?
You probably already own a car with stability control, and you probably don’t think about that system as a type of autonomy. But it is. According to the Feds, there are four gradations of driver assistance, all the way up to Level 4, where a car can drive itself entirely. Right now we’re at Level 2; here, your fancy S-Class or Q7 not only flashes warnings if you try to switch into an occupied lane, it can also stop, steer (as long as you leave a hand on the wheel), and decelerate or accelerate in traffic, without any input from the driver.
The next phase, Level 3, is a sticky, muddled area where few carmakers really want to go at the moment.
The trouble isn’t the machine—many car makers think they’re only a few years out, if not closer, in terms of the tech—it’s the person at the wheel. Toyota’s Hanson says the next level is “when the machine isn’t just processing but anticipating it’s situation.” (The Feds define Level 3 as a vehicle that will “…enable the driver to cede full control of all safety-critical functions under certain traffic or environmental conditions.”) At that level, he says, “we’ve got to trust it to make a life and death decision, that the machine’s ‘got it.’”
But if the machine determines it doesn’t “have it,” the Feds say it must cede control back to the driver—who may, in a moment of false trust, be busy sexting with his girlfriend.
Chuck Gulash at Toyota’s Collaborative Safety Research Center in Michigan is employed to attack just this kind of problem. “We’re looking at facial expressions of drivers and experimenting with warnings that signal if somebody is not looking forward,” Gulash says, but notes that such “state study,” which tells researchers if you’re cognitively able to drive, is still in its infancy.
“It’s still a very difficult topic,” he says. “We still don’t know exactly what does and doesn’t work.”
And that’s just Level 3, covering a car that can only drive itself in limited circumstances, like on a highway. Reaching Level 4 is in some ways almost easier, says Ford’s Washington.
“Right now we don’t want to put a vehicle on the road that gives the customer more confidence than it deserves. Level 4 is harder than Level 3, but right now we’re saying it’s still the responsibility of the driver to be in control. Don’t get used to parking assist or lane keeping. Pay attention.”
It’s still a very difficult topic. We still don’t know exactly what does and doesn’t work.
But it’s human nature that we do get used to our electronic nannies—just ask anyone who’s lost his sense of direction thanks to a built-in GPS. So what’s the tipping point, the moment where we lose even basic skills? Duke’s professor Cummings cites a litany of issues even with Level 2 assistance. There’s boredom, because the driver doesn’t need to be fully engaged, up to what she calls “mode confusion,” where the driver is unclear on whether he’s being assisted by the vehicle. In that way, the jump to all-autonomous, all the time, might actually be simpler—but you still have to get through Level 3 first.
CAN CAR COMPANIES MOVE AWAY FROM AN OWNERSHIP MODEL?
Ford has a pilot leasing program in Austin, Texas, that allows up to six customers to go in together on a two-year lease for a single vehicle. A related ownership app called FordPass will let co-owners know when their shared car is in use, or free, and where it’s parked. And in London, Ford is loaning electric Focus models with a Zipcar-like program called GoDrive. Meanwhile, Audi invested $28 million in a car rental “club” called Silvercar, a mixture of Zipcar and Car2Go that only rents Audi A4 models but extends the short-term borrow time up to days instead of just hours. The rental company currently operates in major cities like Dallas, New York, L.A. and San Francisco.
Counterintuitively, Washington says that car sharing, lease sharing, and “club” models all create an advantage for automakers because it puts newer technology in front of active drivers far more quickly than the traditional “buy and hold” scheme of car ownership. That’s crucial, says Stertz, because many customers don’t really understand new technologies.
“We’ve learned that most people have no idea what an electric car is about,” he says. “The biggest battle is to get them to experience an electric car like an A3 and see how it will work in their lives.”
This helps explain why the prospects are so enticing to automakers. If car companies could phase out the need to service ancient iron, eventually moving to a lease- and share-only model, their fleets would be far easier to maintain, to upgrade (even on a modular basis), and, in theory, everyone would experience newer, higher-quality cars without having to worry about maintenance.
It’s not about the rules of the road—you can program those.
A sharing model also lets carmakers onramp toward autonomy more quickly, because drivers become familiar with the latest technology as it evolves instead of being forced to make huge technological leaps every five or eight or ten years. Dealers might also dig this model because it gives them an in to the Uber/Lyft ride-sharing economy, which is now being left on the table. And of course, they would still perform profitable maintenance work on the cars customers are borrowing from the OEMs.
WILL THE CAR OF THE FUTURE BE ABLE TO TALK—AND LISTEN?
Gadi Amit heads up San Francisco design firm New Deal Design. He’s been involved in some big Silicon Valley projects, from helping usher in the first Fitbit to making a car charger that’s actually beautiful. Amit has also been hired by the car world to work on often secretive projects, such as a new vehicle he can’t even allude to but says will debut in about a year. According to Amit, the biggest blind spot amid the furor over autonomy is the role of a car’s, for lack of a better term, “body language.”
“You know the guy who’s left blinker is on, but he’s actually turning right, and somehow it doesn’t cause an accident?” he asks. Amit says that mishaps in those situations are avoided because the other people who share the road with that guy—pedestrians, cyclists, and other drivers—all subconsciously read the attitude of the vehicle, taking cues from the position of the wheels and even reading the driver’s face to see where he’s looking.
“It’s not about the rules of the road—you can program those,” Amit says. “It’s all of this negotiation besides [those rules]. It’s an omission from the discussion that’s hiding in plain sight.”
Amit says that even if we can create cars with sensors and programming to interpret this subtle and intuitive language—Toyota’s Gulash and Ford’s Washington confirm that this is in fact being studied—autonomous cars will need to signal their next moves overtly and unambiguously.
That’s easier said than done. Those signals, like brake lights and turn signals, will need to be universal, instantly understood, and adhere to a recognized hierarchy. (Think of the wave-through that tells another driver to take her turn at a four-way stop, even if the indicating vehicle arrived first.)
Amit warns that we shouldn’t swoon over technological showmanship. He works in the tech industry and says that it’s great at bells and whistles, but is also known for some spectacular crashes.
“I don’t see technology companies acting that responsibly,” Amit says. “Carmakers understand that for this to work it’s not just about systems, it’s about agreeing to a lot of standards and a lot of study.”
Amit thinks it’s all very enticing, but it’s just not coming today. Or tomorrow. Or even in 2025.
“It’s coming,” he says. “But it’s going to take decades.”