RSS
 

How Google Can Bring The Driverless Car To The Masses

27 Apr

I first heard about Google’s self-driving car from a TED talk by Sebastian Thrun. In that talk, Sebastian talks about a friend who was killed in a car accident, and follows by explaining his vision to make cars safer by making them steer themselves.

Sebastian Thrun’s vision will be fulfilled when driverless cars are obviously safer than human controlled cars, and so become the norm. It appears the biggest obstacle is no longer technical, but regulatory. It’s not a matter of making a self-steering car that works, it’s a matter of persuading governments that automated cars do work, that laws should be changed to allow them, and having those laws worded in such a way that they don’t allow half-bit programmers kill people by putting buggy software into semitrailers.

I imagine there would have to be strong regulations governing self-driving car safety, something like the regulations covering air safety today. Providers of car automation software would need a licence. These licences could be revoked at the will of a body like the Civil Aviation Authority, which would exercise their authority in response to questions about a system’s safety. Crashes would be investigated as thoroughly as air crashes are today, by independent panels. Cars would have the equivalent of a black box, so that these panels could review the seconds leading up to a crash and the decisions made by the software. Perhaps the day will come when a single road fatality garners the same media attention as a plane crash does today.

So how do we get from here to there?

Laws are swayed by public opinion, and public opinion is swayed by dramatic events. The worst possible dramatic event would be a fatal accident caused (or simply not prevented) by an automated vehicle. How can Sebastian Thrun and his team generate the positive kinds of dramatic events that will make the public want these cars on the road? Here’s a few ideas.

Tell Everyone

Sebastian Thrun has already thought of one way to get these cars on the road – teach anyone who wants to learn how to make their own. You can register for an online course at udacity.com which will teach the basics of making a driverless car. If teams at Ford, Honda and BMW start to build systems like these into their cars and perfect them, they will certainly bring political pressure to bear in order to make them legal.

Taxis in India

Imagine if Google made a thousand driverless Tatas, sent them to Calcutta or Dehli or Mumbai, and put a thousand taxi drivers behind the wheel. Imagine if these cars started clocking up millions of miles a year with no accidents on those cities’ roads – that would make a powerful case that driverless cars are safe.

The Right Jurisdiction

Here’s a wacky idea : perhaps somewhere there is a state, province or country whose licensing laws are so badly worded that they allow software to get a driver’s license. I know it’s a bit far-fetched, but if Google’s car automation software passed a driver’s test and got a license, it’s on the road!

Car Safety Systems

This is the idea I think is the best.

Currently, Google’s car drives itself, but the autopilot cuts out as soon as the human takes the controls. Imagine if it worked the other way around – if the car needed a human driver, but the automation system cut in whenever it detected an upcoming accident. This could be marketed as a Car Safety System, and from what I’ve seen on the videos, it could be marketed now. It’s ready now.

The reasons why it’s ready :

  • While a single accident by a self-driving car would set the idea back years, nobody expects Car Safety Systems to work perfectly. Think of seatbelts, air bags and ABS – people get killed in spite of them, yet these technologies still become standard or required features on automobiles. If a self-driving car killed someone it would make headlines. If a car safety system failed, nobody would bat an eyelid.
  • For this reason, it will be years before Sebastian Thrun and his team can afford to put driverless cars on the market. It may be a decade before they save a single life. However, the same system, activating only when an emergencies arise, could begin saving lives in only a year or two.
  • What’s more, a Car Safety System in widespread use will provide far more data to Sebastian Thrun’s team than they can possibly get in a lifetime from their limited testing on California’s roads. Their car automation system will get tested in a huge variety of driving conditions and emergency situations. They’ll see how it acts in circumstances they cannot hope to simulate. Best of all, they’ll discover exactly when their system fails, but in a way that doesn’t destroy the technology’s reputation.
  • Eventually, there will emerge the kind of public-opinion-swaying amazing story that they need. A thwarted attempt at road rage. A drunk safely steered through a busy pedestrian crossing. A truck driver, asleep at the wheel, waking to find himself safely and legally parked. A Heart attack victim brought straight to hospital by his Car Safety System. Then, Sebastian’s team can tell the world “Yeah, well, our Car Safety System is really the same software as our Driverless Car, we’re just waiting for the government to make automated cars legal. By the way, here’s our suggestion for the way the Civil Automated Automotive Authority should run.”

So, Google should keep the driverless car as a high-profile research project, and in the background, market it – starting as soon as possible – as a car safety system. A Car Safety System would be easy to market:

  • Make it affordable. License it to car manufacturers – with no exclusive licenses – for $1000 per install in developed countries, or down to $50 per install or less in developing countries.
  • Imagine the fun publicity stunts they could do – “The world’s most boring demolition derby”, or “Top Gear stuntman fails to destroy Google Safety Car”

I’m sure there’s a large base of people who’d love to have an uncrashable car. I know I would.

When the video below was taken, Google’s cars had done 200,000 miles of automated driving. That sounds like a lot, but they actually need 1,000,000,000 miles to know if their automated cars are safer than humans. That’s 5000 times what they’ve managed to clock up so far, and then they’ll need another billion or two to convince regulators. This can only be accomplished by going to market, but one mistake by a driverless car would do far more harm than the extra miles of testing would do. It certainly seems to be that the quickest way forward is to market they technology as a Car Safety System. They’ll clock up far more miles, and start saving lives straight away.

Do it, Sebastian!

 
 

Tags: , , , ,

Leave a Reply