Self-driving cars are the future of transportation, maybe not tomorrow or even in 10 years, but at some point between now and when teleportation comes along to render all vehicles obsolete. Whether the autonomous future is something you’re looking forward to or not, it’s less a matter of opinion these days and more of an inescapable inevitability. Even so, industry consensus tends to end with the belief that there will indeed be self-driving cars.
Questions about deployment timelines, accident liability, legislation, infrastructure, and the overall effect on the automotive landscape continue to be hotly debated. One stubborn fact in particular gets glossed over during all these discussions of our automotive future: There are about 230 million light vehicles now on American roads with zero or very limited self-driving capability. If sales of new cars, crossovers, and trucks remain healthy through 2018, another 16 or so million will join them. Ditto 2019, and so on. It seems unlikely that any government fearful of revolution will force Americans to scrap all these cars suddenly en masse. So how these traditional (i.e., human-controlled) cars integrate and interact with semi-and fully autonomous ones is a pressing question for engineers developing the automobile, version 2.0.
“There’s clearly going to be a period of time where autonomous vehicles and standard vehicles will be on the road together,” says Richard Barone, vice president of transportation for the Regional Plan Association, an urban research and advocacy group. Barone recently helped pen a report on how cities might prepare for the continuing rollout of autonomous vehicles. Although he’s the first to admit there’s no way of knowing how brief or prolonged this period of coexistence will be, there’s one thing he’s sure of: “It’ll be up to AVs to adapt and respond to the randomness and rationales of human drivers.”
So far, as we enter what might be called the early Atari 400-era of autonomous cars, things are hit-or-miss in this department—with a lot more press coverage going to the “hits.” In March, the technology suffered a major public-confidence crisis when an autonomously driven Uber taxi in Tempe, Arizona, fatally plowed into a woman pushing her bicycle across the street. It didn’t help that an in-car video showed the human operator, ostensibly there to prevent such a tragedy, deeply engaged in texting right before the incident. Or that the car’s sensing systems were clearly unable to detect the pedestrian in the dark. Uber was forced to suspend its much publicized autonomous testing as the public fretted over a clearly unperfected technology.
Whether it’s an autopilot-engaged Tesla Model S plowing into a parked fire truck at 65 mph or a fully autonomous shuttle in Las Vegas freezing as a semi slowly backs into it, there’s clearly cause for concern when it comes to robo- and non-robo-car coexistence. Especially among the public. In January, even before the Tempe fatality, a Caravan public opinion poll showed that 64 percent of respondents were worried about sharing the roads with driverless cars. This level of unease was uniform across the entire U.S. and didn’t change based on residence or political party affiliation. It also echoed the results of earlier polls. A recent study by the Pew Research Center revealed deep public skepticism about driverless cars, with 56 percent of respondents saying they simply wouldn’t ride in a self-driving vehicle.
These fears might discount the hundreds of autonomous vehicles already on the roads that don’t routinely smash into other cars or people, but they do underscore the importance of perception. The fact is, a higher degree of trust will almost certainly be needed for the public to truly embrace any level of vehicle autonomy. And that’s where Anca Dragan comes in.
At her laboratory at the University of California, Berkeley, the roboticist and artificial-intelligence researcher is devising new ways for self-driving cars to better react to (and coordinate with) pedestrians and human drivers on the road. As you might imagine, this is not an easy task. As the 17,249 police-reported automobile crashes that happen every day hint at, human drivers are susceptible to everything from distraction to exhaustion. We also tend to alter our driving habits and patterns based on something autonomous vehicles aren’t great at recognizing: emotional states.
“One idea we’ve been using to recognize and predict aggressive human driving has been that even though the resulting behaviors may seem irrational, you can actually explain them simply as a shift in driving objectives,” Dragan explains. For instance, aggressive drivers often stop prioritizing safety and start caring a lot more about efficiency. They might start purposefully being uncooperative as part of their objectives. These are all things that, with the help of algorithms, cameras, and sensors, Dragan can capture, detect, and use to anticipate future behaviors. It’s not always perfect, she says, but it’s good enough for a car to respond appropriately.
One of the trickiest problems Dragan is working on has less to do with autonomous cars adjusting their driving behavior to humans and more to do with how humans might adjust their behavior to autonomous cars. For some drivers and pedestrians, simply recognizing an autonomous vehicle on the road can produce some undesirable behaviors, whether it’s more aggressive or more timid driving. Both can be dangerous. And to lessen these potentially hazardous reactions, Dragan has come up with an unconventional solution: Autonomous vehicles need to be more aggressive. “Right now, many of them operate like an overly cautious 15- or 16-year-old who’s just learning to drive,” she says. “They seem timid and scared, which does not engender trust.” Program robo cars to be a little more assertive—a little more like how an experienced human might drive—and trust will follow, she says.
But building public confidence in autonomous cars and ensuring smoother, safer interactions on the road are only part of the solution. There’s also the matter of determining where and under what circumstances those interactions will take place. And that brings us to the often contentious issue of regulations.
Some interested parties, like Bob Lutz, the outspoken former vice chairman and head of product development at General Motors, predict that once the standard-to-autonomous-vehicle transition gets going, everyone will have five years to get their manually driven cars off the road or sell them for scrap. Others aren’t so convinced. Steven Sherman, a senior engineer at automotive advisory company Novation Analytics, notes that the National Highway Traffic Safety Administration has already determined it won’t impose blanket federal laws regarding AVs, instead leaving regulations largely up to the states, which currently decide what gets a license plate. “The use cases and general requirements people have of their cars in the U.S. are extremely diverse,” Sherman says. “To assume we’re going to have a one-size-fits-all rule for autonomous vehicles in New York City and rural North Dakota? I don’t think that’s realistic at all.”
Barone, the urban researcher, agrees. “People will still be able to own and operate their manually driven cars. They’ll still be able to hold on to their old Mustangs and Corvettes and their vintage Mercedes,” he says. “They’re not going to just outlaw cars.” What he does see happening, at least over the next 10 to 20 years, is the imposition of luxury taxes for fossil-fuel-burning vehicles and increasing restrictions about where and when you can drive a human-piloted car.
The owner of a Shelby 427 Cobra might be restricted to certain lanes on the highway or have to equip the car with sensors for autonomous vehicles to see and interact with. Or possibly cars will follow the path of general aviation, in which older light aircraft are currently being made to upgrade their avionics to integrate with next-generation GPS-based air-traffic-control technology. The Cobra buyer, if he or she wishes to drive into a city with an autonomous-vehicle traffic network, might need to carry a universal GPS transponder that reports to the network where the Cobra is, in what direction it’s headed, and at what (possibly shocking) speed it’s moving, so that autonomous cars tuned into the network can act accordingly.
No doubt there are privacy issues, but as time and technology march on, regulators might make allowances for vintage-vehicle owners only if they’re willing to give an inch here and there. Either way, it seems likely that society will find a way to integrate yesterday’s cars into the flow of tomorrow’s traffic, and our Cobra pilot will not be relegated to driving solely on a track or in the countryside.
Whatever form regulations do take in the future, it’s clear we’re approaching an inflection point within the auto industry and—if you believe current generational trends—with basic attitudes toward cars and car ownership. The future is never easy to predict, but there’s one fact that seems undeniable at this point: Cars with steering wheels aren’t suddenly going to vanish from our roads in a poof of regulatory smoke. Indeed, they might never fully disappear.
“I can’t imagine a time when there won’t be any manually operated cars on the road,” says Barone, who points out that there are still people who rely on horses to get from point A to point B. Other insiders, like Sherman, go even a step further. “In the end,” he says, “it’s possible that autonomous vehicles will open up the roads more for enthusiast drivers. You’re taking the rush-hour-traffic people—the basic-transportation drivers who really don’t care what they’re driving—and moving that entire group out of the left lane and into autonomous shuttles and ride-sharing fleets.”
If Sherman is right, our driverless future could paradoxically end up being a boon for those who truly love to drive.