Back to all posts

February 13, 2019

An update on Waymo disengagements in California

  • company news

Thumbnail
Each year California requires companies with AV testing permits in the state to report on their disengagements. These reports assist the state in its understanding of how well a self-driving car is doing its job.

Today the California Department of Motor Vehicles (DMV) published Waymo’s fourth annual disengagements report for the miles we’ve driven on California roads. Across the millions of urban miles we’ve driven on California roads, our disengagement rate dropped to 0.09 per 1,000 self-driven miles in 2018 (or 1 disengage per 11,017 miles self-driven). It’s a 50% reduction in the rate and a 96% increase in the average miles traveled between disengagements from the previous year (which had a rate of 0.18 disengages per 1,000 miles, equating to 1 disengage per 5,595 miles self-driven). That drop happened while our vehicles were driving a lot more California miles: 1.2 million miles in the state in 2018, more than tripling our 352,000 miles in 2017.

We believe the key to self-driving technology safely improving and scaling is through a robust breadth of experience and scenario testing, represented by a wider array of data points beyond disengagements alone. That’s why we have a multi-pronged testing program detailed in our Safety Report, and we follow that program in every state we test in. We are building the world’s most experienced driver by focusing on more difficult driving situations. At the start of 2018, Waymo self-driving cars had driven four million cumulative road miles around the country. By the end of 2018 we’d driven another six million miles, which means our self-driving cars have now covered 10+ million real-world road miles. The connection between our real-world miles coupled with our 7+ billion miles in our simulation is key to our improvement rate.

A lower rate of disengagements shows that our cars are getting better at recognizing and handling a wide variety of driving situations, including “edge cases” across the cities we’ve been testing in: those unusual situations that a human driver might see only once (or never) in a lifetime of driving. As we continue to expand new territory or introduce new skills, disengagements will occur. Disengagements in these cases are actually a good thing because they are the equivalent to discovering and solving an issue with our car’s capability. As we continue to expand our test fleet, we actively seek places that present challenges to give our fleet the opportunity to learn.

The way to keep learning and improving so this technology can scale is to give our self-driving vehicles as many robust and varied types of experiences and uncover as many of those edge cases as we possibly can. We’ll continue to refine our driver to make public roads safer and more accessible for everyone.