Inside Volvo's self-driving car: Improving driver safety without the driver

Watch this: Volvo's self-driving car test ride through the streets in Sweden

05:04

It's a rainy day in Volvo's hometown of Gothenburg. Lingering showers churned from the deep forests of Sweden pass overhead, drenching the road. And the S60 sedan we're in. The driver pays little attention to the rain, continuing to talk in an animated fashion with his hands. Both hands. He could be talking with his feet too, if he'd like, as right now they're equally unnecessary.

The black sedan with green graphics in which we cruise is not a typical Volvo, though other than the teal vinyl there's little on the outside to distinguish it from a car you could buy today. A pair of white, hockey puck-size discs on the roof the only bits of flare that differentiate this from any of the company's other cars, and indeed that's part of the point. This is Volvo's take on a self-driving car, a version of which the company pledges to have on sale by 2017.

You don't need to check the calendar to know that's not far away. In an industry once typified by five-year refresh cycles and a slow, begrudging adoption of technology, three years seems like an impossibly short runway from which to launch a car that can drive itself. Or it would be, had Volvo didn't have something of a running start.

Volvo self-driving car in Sweden (pictures)

+5 More

See all photos

The path to automation

Look at the self-driving test vehicles from Ford or Google and you'll see machines festooned with additional sensors, eyes, and ears that make up for the somewhat blind nature of their base autos. Yes, Volvo's car also has some extras on the roof, but those extra GPS modules are simply an extra safety blanket during the testing of this car. They're not really needed.

Volvo already manufactures cars that have all the laser, radar, sonar, and visual sensing equipment needed for autonomous driving. It makes up the company's City Safety program, currently available in the US as part of a $2,100 technology package. A forward looking camera and laser scanner are built into a pod on the windshield, tucked behind the rear-view mirror, while a radar system lives in the nose, hidden beside the company's unapologetically masculine logo.

"The car assists the driver, warns the driver, and then automatically brakes and steers away," said Anders Eugensson, Volvo Cars' director of government affairs. Eugensson is a slim man with cropped, steel gray hair, ice blue eyes, and the friendly, soft-spoken manner that seems common among Swedes. The functionality he describes largely stays out of the way, only making itself known immediately before an impending accident -- a distracted pedestrian or oncoming car wandering into your lane.

The sensor package that enables City Safety is just the latest of a long list of safety innovations that reach back to the beginning of the company. Laminated glass, three-point seatbelts, side-impact airbags, whiplash-preventing headrests... all things that Volvo invented or adopted as standard equipment well before the rest of the industry.

Eugensson calls this the "step-by-step development of driver safety." Autonomous driving is the next step. "Now we say we want the car to do everything. That's quite a bit more complex." There's still a long way to go. A lot of training for the car, a lot of learning for the engineers, and still more technology needed. Many of Volvo's cars already make use of wireless connectivity, but direct vehicle-to-vehicle and vehicle-to-infrastructure communications are the final missing technological pieces. These will enable the cars to know of poor road conditions, disabled cars, and road maintenance without having to see for themselves. (These communications, using the 802.11p standard, will eventually be mandatory in US cars.)

Even ignoring the technical details, there are still many missing pieces. The first is the answer to the big liability question: Who's at fault when a self-driving car crashes? While all the other manufacturers are busy shrugging their shoulders, Volvo has made its position on this quite clear: when the car is being manually driven, the driver is at fault in an accident. But, if the car is in autonomous mode and causes a crash, Eugensson said Volvo will take responsibility. "It will be difficult to sell if the driver is still liable. It gives a false promise." One needn't be a talking lizard to know this should result in cheaper insurance premiums.

Building consumer trust is the next piece of the puzzle, something the Swedes are still pondering. "When someone sees a driver reading, is that going to be accepted? This is one of those problems we don't have an answer for yet." Part of that is thanks to the lack of any sort of certification or testing for autonomous driver competence. Every driver on the road needs to pass a test to get a license. Will an autonomous car have to do the same? That's another question where Eugensson is lacking an answer. "Crash testing, active safety testing, you can come up with a generic test procedure. How do you do this here?" They're working on it, though, Eugensson indicating he was recently discussing this very problem with officials in the United States.

The reasons

If there are still questions about insurance and liability, and if nobody's sure today's drivers will even trust the tech, why bother developing self-driving cars at all? It'd be mighty easy to sit back and let the competition do the hard work, then quickly spin up a similar system relying on the knowledge gained by everybody else.

For Volvo, it's part of a larger effort, something the company calls Vision 2020. Within six years, Volvo wants its cars to be so safe that not a single owner is killed or seriously injured. Not one. Worldwide. Nobody's mandating this policy, no government would pass legislation requiring completely crash-proof cars. This is all on Volvo's shoulders, and that means fixing the most dangerous part of any car on the road: the driver.

"Human error is behind almost all crashes," Eugensson said. It's at least partly responsible 95 percent of the time, either thanks to negligence (drunk driving, distracted driving, falling asleep, etc.) or simply because a driver failed to avoid a preventable accident. If you can eliminate driver error you can eliminate nearly all accidents.

But safety isn't the only reason. There's convenience, too. "If we can offer time we think that is very important to our consumers," Eugensson said. More time for getting caught up on email, interacting with the family, or reading a book -- but not napping. For now, at least, Volvo will require that drivers stay attentive and ready to take over driving duties at all times. The company is testing an infrared camera system plus other sensors that will detect a driver's current state. Should the driver fall asleep or suffer a medical emergency the car will be able to come to a safe stop.

And then there are the infrastructure benefits. Autonomous cars can drive far closer together, meaning more cars on existing roads and narrower lanes in the future -- just two meters in width, down from today's 3.6. Parking garages, too, could be improved. If the car parks itself then no concessions will be needed for humans. Ceilings can be lower and parking spaces smaller. Instead of driving to the garage, you'd just get out on the street and let the car park itself. Then, when it's time to leave, hit a button on your smartphone and wait for the car to come to you, KITT-style.

The drive

I'm stuck in the passenger seat, sadly. Despite repeated inquiries Volvo's representatives would not allow me to drive -- or sit in the driver's seat, as it were. From the shotgun position I have a fine view of the bevy of diagnostic sources that festoon the center console. Dozens of serial and Ethernet ports to the left of my knee, capable of streaming enough telemetry to sate the appetite of the most desperate of data addicts.

In the center console, where you might normally rest a drink, lies nine toggle switches and obtuse labels like "DSPACE" and "FLIGHTR," all flanking the big red "EMERGENCY STOP" button that's become ubiquitous in these sentient machines. Press that and, well, you can guess what happens.

Only about 35 miles of road around Gothenburg are approved for the system to work, so the driver manually navigates us through the heart of the city before merging onto the highway. Once at speed the system is enabled with a quick press of a button on the steering wheel, just like the cruise control. But with this you can not only take your feet off the pedals, you can take your hands off the wheel.

And that Volvo's engineer does, gesturing towards the sensors on the roof, the laser scanner behind the mirror, the traffic outside, the broody clouds above, and in general talking with his hands so effluently that I can't help wondering whether he's been instructed to ensure the camera-toting chase car can tell for sure that there are, indeed, no hands on the wheel.

Jazz hands aside, it's a pretty laid-back voyage. When possible, the car tracks at the speed limit exactly (signs identified by a forward-facing camera), though the driver can manually force the car to go faster. Of course, it also backs off from slower-moving traffic. It can't, however, change lanes as of yet, a bit of programmatic know-how that's yet to be fully implemented. This means the prototype must be manually maneuvered around a lane closure.

By 2017, though, Gothenberg will have a functioning vehicle-to-infrastructure system, meaning such impediments can be broadcasted out to the cars on the road so that they can safely anticipate and avoid them, even before they come in to view.

So it's a bit of a boring, underwhelming loop around the city for the most part. But then, driving on the highway should be boring. At least in this case the driver was free to do whatever he liked -- reading a book, checking in on Facebook, or even fielding a constant stream of inane questions coming from a journalist seated to his right.

Coming soon

Volvo's system is what's considered Level 3 autonomous driving. That is, able to navigate along a given road safely all by itself. Level 4 is the sort of system Google is working on, where you can punch in an address and then curl up in the back seat for a bit of a nap if you like. Volvo pledges to lease 100 of these Level 3 cars to the citizens of Gothenberg in 2017. (Other cities should come soon after, with LA and Shanghai mentioned as high-value targets.) Google, too, hopes to have its system on the roads by the end of the same year, but with no automotive partners announced it's not difficult to be skeptical.

Indeed, Volvo's Anders Eugensson is. That jump from Level 3 to 4 is "going to take a long time." How long? "Two decades" he says at first, then, after a few moments pondering, says it could perhaps come a bit sooner -- but certainly not in the next three years.

This is something of a race, with technological domination becoming an increasingly important part of the marketing message of most automotive brands. Having your self-driving car cross the finish line first will certainly be worth a healthy and boisterous advertising campaign. However, this is not a development process that can be rushed.

If these cars make a mistake, if their developers miss a corner case or if a sensor fails and the system doesn't notice, the results could be disastrous. Self-driving cars could save the lives of thousands, but should they cause even a single serious injury it would not only be a great tragedy, it would knock consumer adoption of these systems back by years. The engineers have to be totally sure, and that's going to take time.

And so, should you visit Gothenburg, keep an eye out for a black and green S60 sedan lapping the city. No longer boxy, but safer than ever, and getting smarter by the mile.