23 October 2017

The Atlantic: “Inside Waymo’s Secret World for Training Self-Driving Cars”

Originally developed as a way to “play back” scenes that the cars experienced while driving on public roads, Carcraft, and simulation generally, have taken on an ever-larger role within the self-driving program.

At any time, there are now 25,000 virtual self-driving cars making their way through fully modeled versions of Austin, Mountain View, and Phoenix, as well as test-track scenarios. Waymo might simulate driving down a particularly tricky road hundreds of thousands of times in a single day. Collectively, they now drive 8 million miles per day in the virtual world. In 2016, they logged 2.5 billion virtual miles versus a little over 3 million miles by Google’s IRL self-driving cars that run on public roads. And crucially, the virtual miles focus on what Waymo people invariably call “interesting” miles in which they might learn something new. These are not boring highway commuter miles.

The simulations are part of an intricate process that Waymo has developed. They’ve tightly interwoven the millions of miles their cars have traveled on public roads with a “structured testing” program they conduct at a secret base in the Central Valley they call Castle.

Alexis C. Madrigal

Fascinating inside look at the Waymo operation tasked with making autonomous cars a reality. The comparison with competitors in terms of real and simulated miles makes it evident how far ahead Google is on the software side. In my opinion, employing simulation is definitely the better strategy here, as physical cars would never be able to match the amount of simulated hours on the road; this also allows engineers to replay the same situation again and again and measure if and how the driving algorithm improves. Tesla is mentioned in the article as a strong competitor, because they are actively collecting data from their existing vehicles, but the company is limited by the number of cars sold (where they continue to have manufacturing issues) and the situations Tesla drivers are encountering.

Listen to the audio version of the above article

Overreliance on simulation can have its downsides though: it the model is wrong or not fully compatible with physical reality, results will be unreliable. Specifically, the article mentions three US cities that have been modeled for these tests; what about other countries, with different driving styles and social conventions? Will a car trained on these three cities be able to successfully navigate the crowded streets of India, or even Britain, with their left-hand traffic?

The directions in my phone are not pointed to an address, but a set of GPS coordinates. We proceed along a tall opaque green fence until Google Maps tells us to stop. There’s nothing to indicate that there’s even a gate. It just looks like another section of fence, but my Waymo host is confident. And sure enough: A security guard appears and slips out a widening crack in the fence to check our credentials.

This begs an almost-existential question: in today’s digital world, if a place is not found on Google Maps, does in really exist?

Post a Comment