DETROIT — The heavily hyped, highly anticipated era of the self-driving car is closer than you think — and as far off as ever.
Within a month or two, showrooms will welcome the first of a parade of new mainstream models that can be equipped with advanced safety technologies that take control of the vehicle for the driver in certain instances. They will bring to average consumers the kind of autonomous driving capability that has been mainly confined to high-end, optioned-up premium and luxury vehicles.
The 2015 Toyota Camry, for example, will be available with adaptive cruise control and collision avoidance systems that can brake to stop or slow the car if a crash is imminent, technologies now offered primarily on Lexus models. By 2017, Toyota plans to make those and other near-autonomous technologies available across its namesake brand.
Other automakers are moving in the same direction. Ford’s 2015 Edge crossover will have a technology package that includes more than a dozen sensors around the car: cameras, radar, ultrasonic.
Then, in the following two to three years, those kinds of technologies will become more capable and likely fall in price. In that time frame, General Motors plans to launch a Cadillac with its Super Cruise system, which can take control of steering, acceleration and braking in stop-and-go traffic or at higher speeds.
Not far behind: automated emergency braking, which will stop a car anytime it’s headed for a low- or moderate-speed frontal crash, whether or not the cruise control is engaged.
The emerging consensus among industry experts and executives gathered here for last week’s Intelligent Transport Systems World Congress was that by the time today’s adolescents start buying their first new cars, the term “fender bender” is going to be on its way to obsolescence.
“A collision-free society is within reach,” Frank Paluch, president of Honda’s U.S. r&d arm, said in a presentation at the ITS conference.
For all the optimism, though, the ultimate goal — the Google-eyed vision of a car that makes its own way to a destination while the “driver” sits back without hands on the wheel or eyes on the road — remains distant and uncertain.
True, autonomous-driving technology is near a tipping point, as long as it is designed to assist a still-engaged driver. The promised safety benefits are commanding premium prices on option packages. That’s why such rapid progress is expected in the new cars that will hit the streets in the next few years.
But once the discussion turns to cars that drive themselves in place of the driver, many roadblocks suddenly come into play: legal issues, technological challenges, cost considerations and infrastructure requirements.
“It’s time to get real about the challenges,” said Delphi CEO Rodney O’Neal, who is retiring next year. “Do we really want to remove the driver from the vehicle?”
Indeed, on the sidelines of last week’s conference, Honda demonstrated a technology-packed Acura RLX sedan that navigated its way through a short highway loop around Detroit, merging into traffic, changing lanes, braking and accelerating on its own — mostly. The car did much of the driving, but a Honda engineer at the wheel had to intervene on a few occasions when traffic slowed suddenly or other cars cut in front of the RLX.
Among the most obvious questions are the legal issues. Current laws require drivers to be in control of their vehicles at all times. A handful of states, including Michigan, California and Nevada, now allow the testing of autonomous vehicles on their roads, but new laws would have to be passed, likely in all 50 states, to allow consumers to own and operate fully autonomous cars — a process that would surely take years.
The issue isn’t dependent only on U.S. legislation. International standards and accords also would have to be struck.
While it is now possible to produce and market cars that can control themselves in stop-and-go traffic or while cruising steadily at highway speeds, much of the technology for fully autonomous vehicles isn’t ready to make it out of the lab.
Part of the reason is that truly self-driving vehicles require far more technology: many more sensors, radar and cameras; more advanced software and algorithms to process and compare their signals; faster processors to handle the number-crunching; and highly accurate, dynamic maps to put the accumulated data in real-world context.
Honda’s autonomous RLX, for instance, had an array of sensors built in but also needed a 2-D scanner, a rotating rooftop gyroscope and a robust GPS to track its location and the location of other vehicles in real time.
French supplier Valeo demonstrated the self-parking system that BMW now uses on its X3 crossover. Its four cameras can see 360 degrees around the vehicle and are augmented by ultrasonic sensors. While it is powerful enough to guide the vehicle into a parking space, it falls well short of the requirements for controlling it at traveling speeds.
“If you’re talking about the highway, then you’ve got to process 10 times the data and do it in a tenth of the time,” said Bruce Beyer, a Valeo executive at the show.
Some of the technological burden could be eased by vehicle-to-vehicle communication, or V2V, small wireless transceivers that enable vehicles to talk to one another and relay information about their position, direction and speed to help avert crashes.
GM plans to make V2V technology standard in its 2017 Cadillac CTS, and federal auto safety regulators last month proposed mandating it. But it will be years before large numbers of cars have V2V and perhaps a decade or more before the large majority of cars on American roads can talk to one another.
Until that happens, the safety benefits of V2V will be limited. “If nobody else by two years from now puts V2V on a vehicle,” quipped John Capp, director of GM’s global safety strategy, “the first car off the line will be waiting for another CTS to talk to.”
No room for error
Another issue is redundancy. If a car is going to take complete control from the driver for long periods of time, any autonomous system has to have backup systems that can instantly resolve any conflicting data from different sensors or jump in if any part of the technology fails, said Steven Dellenback, director of r&d at the Southwest Research Institute in San Antonio.
Moreover, for automobiles moving at the speed limit or more, the millions of lines of computer code can’t have the kind of glitches that consumers will tolerate in smartphones or laptops. “You have to prove there’s no bugs,” he said.
In his presentation at the ITS conference, Dellenback showed a camera’s view of a distant deer on a roadway. Standing sideways, the animal was easily recognizable. But when shown facing the camera, its outline looked much like that of a pedestrian. If such an image is hard for humans to distinguish, “that’s a tall order” for software, sensors and computer chips, he said.
Automakers and their technology partners still have a great deal of work to do to get all the sensors, chips and algorithms required for automated driving to work in concert.
Eric Blumbergs, a senior engineer at Honda R&D Americas, pointed out that autonomous cars will need a much more complex and expensive electrical system than what is used in cars today. Under each hood would be a high-speed network of supercomputers. “We need a whole new level of processing and networking within the car,” he said.
Another big hurdle is, of course, cost. Not only is more advanced technology needed, but it must become significantly less expensive than what is available today. Jon Lauckner, GM’s chief technology officer, said the cost of radar and lidar, a similar technology that uses laser beams instead of radio waves, has to fall “to about $40 [each] from hundreds” of dollars.
Still more work is needed outside the car, on the transportation infrastructure. Many sessions at the ITS World Congress looked at the potential impact of having vehicles communicate with infrastructure such as traffic lights or systems that monitor traffic jams and accidents. Such technologies, known as V2I, could help even out the flow of traffic and reroute drivers around congestion.
But that likely would require action and spending by federal and state governments, a dim prospect given the tightness of budgets and the gridlock in Congress.
And finally, before cars can take over full control from drivers, automakers and technology companies have to know more about driver behavior.
“How will people behave it they take their hands off the wheel?” asked GM’s Capp. “It doesn’t take too long for drivers to be bored.”