Monday, December 02, 2013

Self-Driving Cars: More Bumps in the Road

In what is probably the most detailed reporting on Google's self-driving cars to appear so far, New Yorker staff writer Burkhard Bilger shows just how far the technology has advanced since the Defense Advanced Research Projects Agency (DARPA) held its first Grand Challenge race of autonomous vehicles in the Mojave Desert in 2004.  Nobody came even close to finishing that first race, but only a year later the lessons learned from the inaugural debacles paid off when five vehicles completed the 132-mile course.  Today, Google's fleet of self-driving cars regularly plies roads in California, where the legislature recently passed new licensing laws making it legal to ride in such a vehicle without actually driving it.  But as Bilger briefly points out, a lot remains to be done before you can reasonably expect to own (or at least ride in) a self-driving car yourself.  And in my opinion, technology is not the main stumbling block.

A couple of years ago, I wrote in this blog that I perceived at least two problems which stood in the way of self-driving cars:  unexplored technical problems that might arise if lots of them were on the road all at once, and the reluctance of drivers to hand over the wheel to a robot.  I now think that the first issue has probably been overcome (or easily can be if it arises), and the second issue will take care of itself as the technology becomes more available and peer pressure or necessity (would you rather be told you're too old to drive, or buy a car that can drive itself?) convinces reluctant drivers to hand over the keys to Cyborg. 

But Bilger touches on what I now believe is the single most important obstacle that might slow the spread of autonomous vehicles, at least in the U. S.:  the conservatism of U. S. automotive engineers. 

Bilger spoke with representatives of several car companies:  GM, Ford, Nissan, Toyota, Mercedes, and Volvo, among others.  Ford and GM continue to make incremental "driver-assist" options available, but don't seem enthusiastic about self-driving cars at all. Nissan is the only firm that has made a definite commitment to market a self-driving car, with a target date of 2020.  Mercedes is worried about what the currently-required laser dome on the roof will do to styling, and Volvo is concentrating on safety more than autonomy:  their goal is to make fatal crashes in a Volvo essentially impossible.  But whether a robot or a human drives the car is not their primary concern.  Toyota is still recovering from the controversial accusations that their cars were prone to sudden acceleration, and has paid out millions in legal costs as a result.  That firm is probably not eager to market a product that a few accidents could transform into another huge legal liability.

Here is what I think will happen.  In highly congested non-U. S. cities—Tokyo, Amsterdam, Berlin—auto makers will first market self-driving cars to people for whom car ownership is very expensive in terms of parking and driving aggravation.  Bilger makes the somewhat curious claim that once cars can drive themselves, most people will not feel the need to own one.  I for one fail to see the connection, except in circumstances where it is a positive pain to own a car, such as living in Manhattan. 

Google admits it's not planning to go into the car business.  But if it thinks Ford or GM is going to buy turnkey controls sold by Google and install them in their own products, they have not given sufficient consideration to the power of N. I. H.:  Not Invented Here.  Not only will the U. S. auto engineers be reluctant to hand over critical responsibilities for their products to a bunch of California geeks; the Detroit crowd recognizes that the whole idea of car ownership is tied intimately to the fact that you drive the thing, you don't just ride in it.

Most U. S. automakers sell cars by playing on the emotions of potential car owners.  The idea is "you are what you drive."  Drive a Dodge Ram?  You're a rough, tough guy who can climb mountains while carrying a ton of rocks—in your pickup.  And so on.  The psychological distance between the driver's seat and the passenger seat (even if you're still sitting behind the wheel) is vast.  A car that drives itself isn't a car anymore, it's a one-person bus.  And public transportation in this country is about as sexy as a roomful of old men playing dominoes. 

To sell self-driving cars, the U. S. auto companies would have to retool their whole way of thinking about how cars are sold.  Of course, if buying a car becomes a thing that only really rich people can afford to do (like keeping a chauffeur), and most cars become part of some public transportation network, the marketing job for the auto industry becomes much easier.  They will have to sell only to a few large municipal purchasing agents rather than to millions of individual car owners.  But except in a few quasi-European cities on the U. S. coasts, I simply can't picture this happening to any large extent.  People love their cars too much to let go of them, even if they no longer drive them. 

Perhaps we will go through another U. S. automaker shakeout, like the one that happened in the early 1980s as foreign automotive producers out-manufactured U. S. firms and took over huge tracts of market share.  If lots of people like the idea of not having to drive, but still want to own a car, Nissan will find out when they offer a truly self-driving vehicle.  Legislatures in states where the demand is high will take care of the licensing problem, and if U. S. carmakers ignore or downplay the self-driving car trend after foreign makes start selling, it's their funeral, along with the funerals of those people who die as a result of human-driver error—deaths that Google engineers claim can be reduced drastically once we switch to self-driving vehicles.  And that's another factor that may push U. S. auto manufacturers unwillingly into the self-driving-car business:  insurance companies.  If a large enough database of statistics shows that self-driving cars are, say, four times as safe on average as human-driven ones, insurance rates on the self-driving models will plummet, and people will have to pay more for the privilege of driving rather than letting the computer steer. 

Sooner or later, the sight of driverless cars will no longer attract the attention it does today.  But a lot of things will have to change first, and among the most important are attitudes of engineers, legislatures, and drivers themselves.

Sources:  Burkhard Bilger's article "Auto Correct" appeared on pp. 96-109 of the Nov. 25, 2013 issue of The New Yorker.  I addressed the issue of autonomous vehicles in my blog in this space on August 11, 2011.

1 comment:

  1. People in major cities are getting used to fleets of publicly owned bicycles -- pick one up at a communal lot close to where you live, pay a small fee, ride it to a communal lot or bike rack near your job, park it and forget about it. I can imagine something similar for tiny little cars being set up -- something just big enough for one or two people, or one person taking a cat to a vet, or one person bringing home groceries. This isn't a big reach beyond running special buses for the elderly and shut-ins, which many cities already do. Setting such things up with computer drivers and maybe designated traffic lanes in metropolitan areas would probably be straightforward.

    Of course, it might just about kill the regular auto industry. Yeah, it's fun running your 300-hp Ramcharger over dirt roads on the weekend, but what's your real incentive to buy such a beast when you're getting to work and shopping for five bucks a day in something like an automated golf cart?