Driving to be obsolete?

kw:

In a story I read about 40 years ago, a suburban family has recently bought a new car. One weekend they decide to "take a drive", and the in-car navigation system directs them onto an ultrahighway. Once they enter the on-ramp, a system called Replectric takes over, an extra set of restraints grabs everyone, and the car smoothly accelerates to 300 mph. Panicked—this their first time on an ultrahighway—the father tries one thing after another to disengage the system and regain control, but can only do so once the navigator ramps them off the ultrahighway. They pull into the first car lot they see, sell the car for next to nothing, and pay a premium price for an older car. As they drive off, to take "surface streets" home, one salesman remarks to a newer colleague, "We see one every few days. A lotta folks have no idea what the newer cars can do."

The in-car navigation system became a reality just under 20 years ago. In another 10 years, or even less, will the car be able to drive itself? Nissan CEO Carlos Ghosn thinks so, as reported in this article in Forbes by Dan Bigman. I wonder, though, will we want the car to drive itself?

We spent Thanksgiving weekend in Oklahoma City with friends. On the way to a downtown appointment, we turned onto Highway 40, which runs left-to-right in the middle of this image. After we crossed May Avenue, seen left of center, the GPS navigator got lost. Starting in 2008, the highway was moved. At the upper right, you can see a graded area where the old road was.

The GPS database doesn't have the new highway in it yet, so it showed us driving across empty ground, and kept imploring us to "take the next left". There are no left turns off of 40, not for miles. The image above is from August 2012. The image below from May 2008 shows the old highway and the excavations just beginning for the new one. And I had downloaded the latest map just before we left (we took our own GPS navigator with us)!

Fortunately, the Google Maps database is more up-to-date. Being a confirmed paranoid, and also a map lover, I had made a set of paper maps before we went. We reverted to the paper to get to our destination.

By 2020 or so, will the navigator companies (TomTom, Garmin, Magellan et al) have solved their data lag? About 14,000 new miles of road were built last year. It sounds like a lot, but compared to the 4 million existing miles it is tiny, a third of a percent. From a data standpoint, though, it is a lot. Some new roads are many miles long, but others may be only a block or two. Thousands of database updates are required. That's why your GPS company wants you to pay $50-$100 yearly for map updates. They are paying a lot of people to keep up.

For a nation of driverless cars to work, we'll need nationwide cooperation between all the highway authorities and the navigation venders. You don't want hundreds of cars getting lost in the middle of OKC at 70 mph!

A second, more crucial element will be roadway recognition. The new Highway 40 in OKC is easy. The lane markings are very distinct, and the roadway has concrete sidewalls. A car with radar can at least orient itself by the sidewalls, and a vision system could follow the lane markings. But we've all been on roads that are hard for us to follow. It will be a long, long time before computer vision systems with automatic recognition (robot vision) get better than human abilities.

I foresee a long transition. Probably by 2020 a few pricey cars will have partial autopilot features, but in deference to aircraft, there will be a different name for it (probably not replectric!). A car might be able to do all the driving, but at first the "take me there" system will be engaged mainly on clearly-marked highways, until people learn to trust them. Such changes take a generation to work out. Children born in the next few years will grow up with cars that can take over much of the driving, and by the time they are drivers, they will think nothing of entering their destination into the navigation system, pushing the Go button, and falling asleep or doing FaceBook updates. I suspect a larger proportion of my generation will never get used to that, and will strive to maintain control. But by the time I am 80, in 2027, perhaps I'll have to use a self-driving car just to be allowed to have my own transportation. Some European countries already do not allow anyone over 70 to drive. My father gave up driving at the age of 89, and he is in better shape than most 80-year-olds.

Over a decade or two, highway fatalities will probably decrease. If they decrease a lot, that may motivate legislators to favor driverless cars, or even require their use after some changeover date. Some folks won't like that, but if the stats back it up, it will be inevitable. Enjoy driving while you still can!