Automotive > Infotainment & Telematics Blog



Behavior the Biggest Barrier to Self-Driving Cars

by Roger Lanctot | Sep 16, 2015

I am just back from Beijing, China, where the people behind the steering wheels make Boston drivers look positively passive. In fact, the driving behavior in Beijing is bizarre enough to send even the most determined and talented Google self-driving car engineer back to the drawing board to draw up a new resume and find a new career.

The press has been full of stories lately regarding self-driving cars being stymied and stumped at intersections. But the combination of wrong way driving and illegal U-turns in Beijing raises the bar for meeting autonomous driving objectives on a global basis.

The Beijing experience highlights the fact that driving behavior is regional and personal in nature – especially considering that Shanghai drivers, while presenting their own challenges, are not nearly as creative as Beijing drivers.

China’s unique driving environment is a poser, but the Google car is getting a reputation for automated indecision at intersections right here in the U.S. where human drivers use tactics such as rolling stops and feints all in a driving context that includes pedestrians and bicyclists. In China, you have to factor in an abundance of three-wheeled vehicles (some of which look handmade) that behave like and compete with pedestrians and other motor vehicles on the road, on the side of the road and on the sidewalk.

But physical objects have predictable trajectories while human behavior quite often does not. This, then, is the fly in the autonomous driving ointment. Aside from the limits of the electronic horizon relative to the vehicle’s speed, human behavior is the element posing the greatest challenge to autonomy.

Drivers are simply unpredictable. In fact, driving behavior is as unique as a fingerprint, which means the advocates of self-driving cars are trying to read these driving fingerprints in real time even as they interpret in real time the changing physical world in which they operate.

In fact, robotic driving will ultimately erase driver individuality. But until that time arrives, the only predictable driver on the road will be the autonomous car, and other drivers can be expected to take advantage of those cars with near-miss passes and intersection fake-outs.

This is why experts, including myself and my colleagues at Strategy Analytics, foresee a solid decade of further development activity before full autonomy can be achieved. Drivers practically pride themselves on unpredictability – a useful attribute in a competitive driving milieu where at least some drivers are looking to gain an advantage in their attempts to arrive sooner at their destination.

We are pursuing autonomous driving, some believe, to remove the driver from the driving equation, believing, as some at the US Department of Transportation, the Insurance Institute for Highway Safety, the Alliance of Automobile Manufacturers and Google would have you believe: that 90%-95% of car crashes are attributable to human failure.

But the scientists working on vehicle autonomy generally do not take the blame-the-driver view of car crash causality. They see autonomy as a natural evolution of the driving experience toward something more automated – recognizing that some people don’t want to or don’t enjoy driving and others are or will be incapable of driving in the future.

The irony of the blame-the-driver-for-crashes-and-fatalities position is that the creators of autonomous driving systems are working to replicate the observational and decision-making capacity and functions of the human brain in their efforts to enable self-driving cars. Flawed though the human brain may be, its sensors and processing capacity are a difficult portfolio of functionality to reproduce… or surpass. (Just ask nVidia.)

In this context, Toyota’s announced approach makes the most sense. Keep the human in the loop.

Toyota has been going its own way for the past five years or more, eschewing the flavor of the week in advanced automotive technology. Toyota adopted telematics, but only half-heartedly and it has avoided Apple CarPlay and Google’s Android Auto completely, for now. (Don’t get me started on Toyota’s quixotic pursuit of hydrogen fuel cell technology.)

In the area of vehicle autonomy, Toyota has hired Gill Platt from the Defense Advanced Research Projects Agency, to direct a self-driving car development effort intended to focus on advances in artificial intelligence technology to make humans better drivers.

In press reports, Dr. Pratt described the two approaches favored by Toyota as “parallel” and “serial” autonomy - where parallel means the machine watches what the driver does, while serial means it replaces the driver.

Toyota, until recently the world’s largest carmaker, envisions cars of the future that will act as “guardian angels,” watching the driving behavior of humans and intervening to correct mistakes or avoid collisions when needed.

Dr. Pratt told the New York Times that Toyota’s goal was to keep the “human in the loop” in the car of the future and to ensure that driving remained fun. “A worry we have is that the autonomy not take away the fun in driving,” he said. “If the autonomy can avoid a wreck, it can also make it more fun to drive.”

He told the New York Times that Toyota intends to use artificial intelligence to actively correct driving errors or to compensate for the diminished eyesight or reaction times of older drivers.

Toyota isn’t looking to replace or displace drivers. Toyota is seeking to assist drivers in precisely the ways drivers expect to be helped.

If we have the technology to avoid collisions that might result from moments of inattention or distraction or the slip of a foot on an accelerator or brake pedal then why wouldn’t we implement those systems? In fact, we’d not only expect our cars to work that way, we’d expect our insurance premiums to decline appropriately.

Unlike Google, Toyota is not interested in removing the steering wheel and brake and accelerator pedals. It will be a decade or more before that makes sense. And until that time, until we can replicate the observational and processing capacity and capability of the human mind, humans will remain behind the wheel. From Toyota’s perspective driving is too much fun to be left to robots – and maybe too difficult.

Previous Post: Is WeChat Ready for Its Automotive Closeup? | Next Post: 'I Can't Believe It's Free'
Leave a comment