A Lexus drives down the eight-lane highway outside Palo Alto, California, in heavy rush-hour traffic. Except for the rotating cylinder perched on its roof like an oversized tin can and the word "Google" on its doors, it looks like any other car. In reality, though, it's a search engine on wheels.
The Lexus steers itself down the highway all by itself. The man in the driver's seat -- Dmitri Dolgov, a software engineer for Google -- never actually touches the wheel.
Dolgov explains what the car can do, which turns out to be quite a lot. It can steer, accelerate and brake automatically; it surveys its surroundings with cameras and uses radar to measure the distance to the car in front of it; and its laser scanner -- the cylinder affixed to the roof -- monitors objects in all directions.
"See?" Dolgov asks, pointing as a car swerves in front of the Google vehicle from the right. There's no need for Dolgov to intervene. The robotic car has identified what is happening and gently brakes until there is once again a proper distance between the cars.
With its 12 vehicles, Google has the largest known test fleet of self-driving cars. All together, the Internet giant has covered over half a million kilometers (300,000 miles) in these robotic vehicles, most of it on California's public roads and highways. The cars have driven through Los Angeles, around Lake Tahoe and down the famous hairpin turns of San Francisco's Lombard Street. They have become so reliable, in fact, that Google is now taking SPIEGEL out for a demonstration.
Self-driving cars, long dismissed as a utopian pipe dream, are rapidly reaching the stage where they will be ready for the market. "We're not talking about 20 years here, but more like five," says Sebastian Thrun, initiator and director of Google's project.
Five years until the first driverless cars hit the streets? It sounds like just any of the other science-fiction ideas that seem to percolate out of the manically creative world that is Google headquarters. But could it be that the company is about to show the automobile industry what the future of mobility looks like?
In truth, however, the real surprise here is something else entirely: Everything Google can do, carmakers already do as well -- they just don't talk about it as openly. In one European Union-funded research project, Volvo successfully drove a convoy of five vehicles that only had a human driver in the lead car. BMW recently sent a robotic car on a two-hour drive from Munich to Nuremberg. And Volkswagen and a research team from Stanford University have caused a stir with their driverless Audi sports car, which that has been zipping around US racetracks.
Gradually Automating Cars
Although Google doesn't enjoy a monopoly on the field, its prominent position allows it to exert pressure on others and demonstrate the feasibility of the idea. The auto industry isn't missing the technology needed for the next revolution in mobility. It lacks the guts to put that technology on the market.
"The necessary technology for autonomous cars is already in place," confirms Lothar Groesch, an expert on safety technology. Groesch, 66, who holds a Ph.D. in physics, has spent most of his career working in development for Daimler and Bosch, the automotive parts giant. But now his job as a freelance industrial consultant allows him to speak his mind freely, rather than being limited to what his bosses want him to say.
Groesch recently helped Bosch with its development of driver assistance systems. He quickly recognized that, when taken together, all of the instruments designed to assist drivers added up to a technology suite that will ultimately make it possible to liberate cars from their drivers.
The question is whether or not people will embrace it. Carmakers' greatest fear is that this development will rob the automobile of its magic, reducing the once all-powerful driver to a passive passenger.
But the fact is that this process has already been underway for a long time. It began 20 years ago with the introduction of Electronic Stability Control (ESC) systems that apply targeted braking to individual wheels so as to prevent skidding and make sure that overzealous drivers don't lose control while accelerating around curves. Of course, most cars offer a way for people looking to drive the old way to switch the ESC system off. But, says Groesch: "That's foolishness in terms of road safety." And engineers believe that there is less and less justification for having drivers the better automated vehicle-control technology gets.
Today's cars come with radar sensors and cameras that can recognize, for example, situations where a collision may occur if the driver doesn't react in time. These cars first sound a warning, then brake fully, although often not until it's too late -- the vehicle still crashes into the obstacle but at a lower speed. "The car could do better, but it's not allowed to," Groesch explains. "It would make a lot more sense to intervene earlier, without first giving a warning. The warning only wastes time."
Freedom vs. Safety
This is where car developers -- and lawmakers -- are forced to ask themselves some weighty questions. Is the experience of driving a car something worthy of preserving? Does it lose its allure when drivers are stripped of the freedom to drive their cars themselves -- and also of the freedom to cause accidents with those cars? How much blood is society willing to spill for the sake of our freedom to drive cars by ourselves?
No other invention in the history of civilian technology has caused as much harm as the automobile has -- not airplanes or electricity or even nuclear power. A person dies in a traffic accident somewhere in the world every 30 seconds, adding up to well over 1 million deaths each year. And the World Health Organization estimates that his figure will only continue to rise as more and more people in developing countries acquire cars. What's more, human error is the cause of almost every automobile accident.
As measured by capacity, commercial airplanes and trains are up to 1,000 times safer than automobiles. And the reasons are clear: Airplanes and trains are not steered by hundreds of millions of people who have received driver's licenses without any further verification of their character or intelligence. Instead, they are controlled by a much smaller group of experts trained for precisely this task.
Additionally, the controls for vehicles traveling by rail and air are largely automated. At this point, the primary function of the driver of a high-speed train is to regularly press a so-called "dead man's switch," which informs the automatically driven train that the driver is still awake and alert. The captain of a commercial airplane, meanwhile, turns on the autopilot shortly after takeoff, and only takes over the controls again shortly before landing.
What, then, qualifies an overtired traveling salesperson to manually drive his or her car 100 kilometers or more to get home through monotonous, steady traffic on a Friday evening? And is he or she really having any fun in doing so?
From Deserts to Drawing Boards
These are the kinds of questions Sebastian Thrun was asking himself even in his days as a computer science student in the German cities of Hildesheim and Bonn. Thrun, a 36-year-old genius in his field, took up a position as a professor at Stanford University, one of the most prestigious universities in the US, and is now a developer at Google. Thrun explains that he lost a friend in a car accident as a teenager and sees this as his motivation for turning to his field, computer science, as the key to accident-free automobiles.
Thrun first attracted attention in 2005, when he participated in a challenge run by the Defense Advanced Research Projects Agency (DARPA), the research wing of the US military. The DARPA Grand Challenge involved 23 driverless, automatically controlled cars racing through the Mojave Desert, in the American Southwest. The Stanford Racing Team, headed by Thrun, entered the competition with a converted Volkswagen Touareg and steered its way to victory on rough terrain. The first autonomous cars were off-road vehicles by necessity, because they still weren't able to orient themselves well enough to keep them safely on the road.
One young computer entrepreneur was a spectator at the DARPA Grand Challenge that day and took an interest in the new technology. Larry Page, a Google co-founder and something of a rock star in the Internet world, recognized the alluring possibilities of driverless driving. So he immediately hired Thrun and other key members of his team.
Google's fleet of self-driving cars has become an integral part of the mystique surrounding America's most successful Internet firm, which started out as something quirky and has since become rather uncomfortably omnipresent. So far, though, the company has only used its autonomous vehicles as a form of advertisement. It has no official plans to commercialize vehicles, nor does it give the impression that such plans are the works. "We have no need to open up a car company here," Thrun says.
What's more likely is that Google would seek out a partner, thereby giving the ailing American automobile industry with a badly needed technological boost. General Motors, Ford or Chrysler could eventually become a hardware supplier for Google's designs, something of a Foxconn for the automobile world.
Challenges and Progress
Dressed in a green fleece jacket, Thrun warmly receives visitors at Google's headquarters in Mountain View, California, not far from Stanford. Just outside the entrance, Google employees playing beach volleyball are living representations of the company's aura of eternal corporate coolness. The formalities required for entry into the building, on the other hand, are roughly as strict as those at the Pentagon, and the company spokesman urges visitors to hurry. Google doesn't have all day.
Thrun, the computer science professor from Germany, has become an important figure at Google. He helped create Street View, the company's controversial compendium of photographic images of front yards and houses from around the world. Street View, Thrun explains, was a useful exercise in preparation for the autonomous vehicle project, since the streets that self-driving cars travel will also need to be thoroughly photographed first, although with a focus on a different type of data. Google Maps' collection of images, in other words, cannot be directly used for driverless vehicles.
Google's self-driving cars draw on a detailed directory of every street, building and bridge, all of which is stored on computer servers. Cameras and laser scanners mounted on vehicles check the images they receive from their surroundings against what is in this directory. In other words, this system's precision and reliability rely entirely on computing power -- something that is increasing at a furious pace.
Moore's Law, a standard principle of computer science, effectively posits that processing power doubles every two years. The first Intel processor, built in 1971, had 2,300 transistors. Today's standard microchip holds over 2.5 billion. And as computers' processing speed increases, so does the reliability of robotic cars. Just a year ago, Thrun says, the test operators of these cars had to intervene an average of once every 8,000 kilometers (5,000 miles) to correct a mistake on the part of the automatic driving system. "Now we can drive 80,000 kilometers without having to intervene," he says.
That's impressive, but not yet a breakthrough. A human driver who made a serious mistake once every 80,000 kilometers wouldn't exactly be held up as a model driver -- and a computer that does so certainly won't be.
But Thrun tries to put things in perspective. His self-driving cars, he says, don't make careless mistakes. The cameras never ignore a red light, and the radar reliably prevents rear-end collisions. "In those areas," he says, "robotic vehicles are already better than humans."
The driverless vehicles are worse, though, at reliably identifying objects. "That's something we humans are incredibly good at," Thrun explains. He picks up objects from the conference table in front of him to illustrate his point: "Here, a telephone. A roll of tape. It's not something we have to think hard about."
The cars' cameras see these things just as clearly as the human eye does, but the computers take longer to assess whether or not it would be dangerous to drive over them. What this means is that a robotic car will slam on the brakes even when the object in question turns out to be just a cardboard box blowing down the street, because it can't immediately assess whether the box isn't actually a baby carriage. And if suddenly braking for a cardboard box in the road causes a collision with the car behind, who is liable?
"Robotic cars are still too polite, which means they can sometimes be a nuisance," admits Doglov, the Google programmer. The vehicles always err on the side of caution, braking for cardboard boxes and never cutting into a stream of traffic where each car is traveling closely behind the one in front of it -- something drivers sometimes need to do, because they can otherwise end up sitting for hours on a highway on-ramp, waiting for an opening.
But humans won't always trump computers even when it comes to distinguishing between boxes and baby carriages, one of humanity's last bastions in its competition against artificial intelligence. When Thrun says this technology will be ready to go into mass production in five years, he's assuming that computing power will continue to multiply -- and that's a realistic assumption. Though high, the standards that need to met are attainable.
Pioneering Steps at Daimler
However, this isn't the first time that computer technology and automotive engineering have clashed. These are two industries with two very different approaches to the issue of susceptibility to failure. If a home computer crashes, the user simply boots it up again. But if the same thing occurs to the system that controls a car's safety functions, it could be life or death.
"Before we certify that a new driver assistance system is ready for mass production, it needs to have completed millions of kilometers of test driving without any errors," says Jochen Hermann, head of development for this product sector at Mercedes-Benz. The Stuttgart-based Daimler subsidiary is viewed as the founder of the automotive industry, and still seen as the keeper of the Grail when it comes to safety technology. Many pioneering driver assistance systems, from anti-locking brakes to ESC to brake assist technology, all had their breakthroughs here in Mercedes-Benz's research and development division.
Thrun, the Google developer, also has the highest respect for the company. "Mercedes does beautiful work, absolutely," he says. Such comments would seem to imply that Daimler might well number among Google's preferred potential partners for implementing its technological ideas. But Daimler might not even need Google at all.
The company has also developed its own cars with self-driving technology, the first of which will hit the market this summer, when Mercedes-Benz launches the next generation of its S-Class luxury sedan.
The S-Class is the company's flagship model, and the one Daimler often uses to introduce its latest developments. Mercedes took a decisive step toward autonomous driving, for example, with the 1998 S-Class, which introduced Distronic, a cruise control system with sensors for measuring and maintaining the vehicle's distance from the car in front of it. The 2013 model will go a step further by being capable of steering itself, making it the first to fulfill all the criteria of fully automated driving.
It will do so, though, only under one specific set of circumstances: in congested traffic. When the vehicle is traveling at walking speed, the driver can choose to switch on cruise control and take his or her feet off the pedals and hands off the steering wheel. The car then does everything itself, automatically accelerating, braking and steering. Simple sensors and cameras to monitor lane markings, along with radar equipment to measures the distance from the car in front of it, are enough to allow the vehicle to perform these functions.
"Traveling at a crawl is a very easy thing to master," Thrun agrees. But the new Mercedes S-Class can do more than that. It corrects the driver's steering at higher speeds as well, although in this case the driver must remain at least minimally involved, by keeping his or her hands on the steering wheel. Pressure sensors check that the driver is doing so and after a few seconds of handless driving, an alarm sounds and the autopilot switches off.
Legal Road Bumps
Here, Mercedes is following much the same principle as the dead man's switch in a high-speed train -- and for legal rather than technological reasons. Indeed, Mercedes could allow its S-Class to drive on highways completely autonomously. But it doesn't believe it's allowed to do so.
The reason for this lies with the Vienna Convention on Road Traffic, a UN treaty signed on November 8, 1968. Article 8 of this document clearly states, "Every moving vehicle or combination of vehicles shall have a driver."
This same principle, carried over more or less verbatim, has found its way into the laws of the treaty's signatory nations. And, there, has remained, largely ignored for decades because the prospect of driverless cars was simply never an issue.
Added wording in these state-level laws further specifies that that "driver" is understood to mean a living being and not a collection of semiconductors. The German law on the subject, for example, reads, "Every driver must be physically and mentally capable of driving," before going on to talk about regulations concerning driving lessons. In other words, there can be no doubt here that "driver" means a person, not a computer.
Still, Ralf Herrtwich, who oversees the development of driver assistance and chassis systems for Daimler, believes there is "a lack of clear regulations." And, indeed, there is no doubt that technological developments in this area far have far outpaced amendments in the applicable laws. Even the new S-Class' autonomous driving capabilities in low-speed traffic exist somewhere on the edge of what the law allows, putting Mercedes' new luxury sedan in a legal gray area.
Thrun, the Google researcher, takes a less delicate view when it comes to the situation in the US. "No state in the US expressly forbids autonomous driving," he says. Nevada, in fact, expressly allows it and is currently working to establish more precise regulations. California and other states plan to follow Nevada's example.
Utopian? Or Feasible?
This open-mindedness on the part of the authorities ties in with certain economic policy interests. The computer industry is the last true bastion of the American Dream, the last economic sector in the US that is entirely intact and has the potential for boundless growth. Apple and Google are what General Motors once was, and Google's self-driving car is tantamount to an embodiment of the belief in digital progress.
Thrun, almost boyish-looking despite his 45 years, is perfect for his role as a major player in this pursuit. He took time off from his job at Stanford to found a free online higher-education portal called Udacity as a way of democratizing education, and he likewise sees his driverless car as serving a redemptive societal function. "Think of all the people who are blind or suffer from Parkinson's or Alzheimer's," he says. "Millions of Americans are denied the privilege of driving on health grounds."
In one advertising video, Google technicians have a man who is almost completely blind take the driver's seat of an autonomous Toyota Prius, which drives the man to do his shopping. To film this video, the car was surrounded by a police escort -- in the real world, it's still too soon for blind reliance on digital robotic driving.
What, then, is actually possible here and what is not? Designers in the automotive industry find Google's showy celebrations of its autonomous vehicles unsettling. Daimler developer Herrtwich, for example, finds it inappropriate to act as if computer-steered vehicles will soon be able to navigate through the fray of urban traffic. "City traffic is an utterly chaotic situation, and designing autonomous cars that can drive in it is not even one of our goals at this point," he says. "Autonomous driving in monotonous, steady highway traffic is a far more reasonable and feasible goal."
Still, even some conservative German designers take things considerably further. "The pace of development in electronics has often been underestimated," says Groesch, the industrial consultant. In another project with Daimler and Bosch, Groesch worked on developing airbag controls. In the beginning, it was widely believed to be impossible to develop sensors that would react quickly enough to deploy side airbags in time. Today, such airbags are an industry standard.
Reducing Mistakes to an Acceptable Level
That Groesch no longer considers autonomous driving a purely utopian vision is owed to one key piece of technology that has advanced by leaps and bounds: "Laser scanning," he explains, "is unbeatable at identifying traffic entering from the side."
Capable of up to 10 revolutions per second, these scanners fire 60 laser beams or more in a 360-degree arc around the vehicle. The beams are invisible to the eye and pose no danger to humans, but they strike objects and bounce back as pulses of light. From the time lapse between pulses, the computer can measure the car's distance from objects.
This allows the vehicle to establish a three-dimensional image of its surroundings to a distance of up to 100 meters (330 feet) -- a more comprehensive view than is possible with the human senses. Interpreting the information from the laser beams faster and better than the brain can interpret feedback from the eyes is only a matter of increasing computing power.
It's more than likely that, within the foreseeable future, autonomous cars will no longer make a mistake every 80,000 kilometers, but perhaps only every couple million kilometers. And someday they will outperform humans in every situation, even in chaotic city traffic.
The benchmark for developers in this field is ASIL D, under ISO 26262, an international standard applied to the safety of electronic and electric systems in automobiles. The standard stipulates that failures should be nearly eliminated -- a tall order when it comes to this sort of mobile technology. What happens when condensation or snow blurs a laser's lens, even if just for a few seconds? "Driving autonomously through Nevada on a sunny day is fairly easy," says Daimler developer Herrtwich. But in a flurry of snow in the mountains, the world looks very different for cameras and lasers.
Fading Interest in Driving?
How many mistakes can a robotic car make and still be unquestionably better than a human driver? There's no question that the first autonomous car to run over a person will receive more attention by far than the thousand human-driven cars that will undoubtedly do the same on the same day.
Manufacturers, meanwhile, expect to meet with fewer reservations toward self-driving cars from younger people, many of whom have little interest in cars anyway and would rather check their email during a drive than put in the effort necessary to turn the steering wheel. This is a generation that might even be grateful to let a computer take over as chauffeur.
Market researchers have observed a "trend toward de-emotionalizing automobility," in the words of Stefan Bratzel of the FHDW University of Applied Sciences, in Bergisch Gladbach, Germany. In one study, Bratzel found that fewer and fewer young people living in cities own their own car, and that many no longer even have a driver's license. How else can car manufacturers expect to reach these potential customers, if not with a car that takes care of the driving itself?
At the same time, developers in the field are coming to a bitter realization. "Driving a car will increasingly be seen as a waste of time," says Groesch.
At Mercedes' development headquarters in Sindelfingen, Germany, designer Hermann ponders for a moment, then says something that no longer sounds blasphemous even here in this temple to the cult of the automobile: "There are plenty of situations where I don't want to have to drive because doing so isn't any fun."