Update on Driverless Cars

A few months ago, I wrote about self-driving cars. At the time, the last information publicly available came from 2007, in which the DARPA Urban Challenge demonstrated the possibility of cars that could safely navigate through a city without human interference. At the time, only 6 of the 11 autonomous cars that competed in the challenge were able to complete a course in a makeshift "city" on an unused military base. The vehicles traveled extremely slowly (about 13 miles per hour) and the course was only 50 miles long. There were no tricks, surprises, or unusual circumstances...the vehicles just had to drive themselves and react to normal traffic.

I was excited at even this rudimentary amount of progress in 2007, so I was even more delighted when the New York Times provided an update on self-driving cars yesterday. The technology has progressed immensely in the last three years, much more quickly than I would have guessed. Google has secretly been testing autonomous vehicles, working with none other than Sebastian Thrun, the lead engineer of the Stanford Racing Team, which took second place in the DARPA Urban Challenge, and first place in the 2005 DARPA Grand Challenge. The cars have been driving on actual highways, city streets, and rural roads in California, navigating their way through real traffic. There is always a human sitting behind the wheel who has the power to override the self-driving computer, just in case something goes wrong. In the last year, the Google Cars have driven over 1,000 miles on the streets of California without any assistance, and 140,000 miles with only minimal human assistance. They no longer travel at crawling speeds; Google has programmed the speed limits of all the relevant streets into the system and the vehicles are capable of traveling at the speed limit. In the Google Car fleet, there has only been one minor mishap in the last year: when another driver rear-ended a Google Car at a stoplight.

The New York Times article implies that it will be about eight years before self-driving cars are publicly available. Given the unpredictable nature of technological roadblocks and legislative paralysis, I'm opting to be a little more conservative, standing by my previous estimate: they'll be on the roads no later than 2020.

The NYT is quite bullish on their prospects, implying (via technologists and futurists) that "they can transform society as profoundly as the Internet has." That may be a bit of an exaggeration, but only because the Internet has transformed so much of our society. I do not think it is an exaggeration to say that self-driving cars will fundamentally alter the way in which we design cities, will reduce the annual automobile fatalities nationwide from approximately 40,000 to approximately zero, will greatly reduce traffic and pollution, will help alleviate poverty by eliminating the need for most people to own a personal car (instead you could summon one to pick you up like a taxi, but available in non-urban areas and without the high fees), and will allow us to enjoy our commutes more by freeing up our time to do things other than watch the road.

PREDICTIONS:
By 2020 - Driverless cars are commercially-available and street-legal somewhere in the United States.
By 2027 - New driverless cars outnumber new cars requiring at least some human control, in the US market.
By 2035 - Driverless cars are widely perceived as safer than human drivers. Somewhere in the United States, it is illegal for humans to drive.