Advertisement

Self-driving car prototypes need less human help, data show

A row of Google self-driving Lexus cars line up at an event outside the Computer History Museum in Mountain View, Calif., in 2014.
(Eric Risberg / Associated Press)
Share

Self-driving car prototypes appear to be getting better at negotiating California streets and highways without a human backup driver intervening, according to data made public Wednesday by California transportation regulators.

The data reflect safety-related incidents reported by 11 companies that have been testing more than 100 vehicles on public roads, primarily in the Silicon Valley neighborhoods where the technology has grown up. The reports were made to California’s Department of Motor Vehicles, which posted them online.

The documents catalog the number of times from December 2015 through the end of November that humans took over control from a car’s software for safety reasons.

Advertisement

Waymo, as Google’s self-driving car project was recently rebranded, did far more testing than the other 10 companies combined.

Waymo reported that its fleet drove itself more than 635,000 miles with 124 safety-related “disengagements,” which must be reported when the technology fails or the backup driver takes control out of concern the car is malfunctioning.

The Google project’s disengagement rate was the equivalent of two incidents every 10,000 miles, a notable decrease over the prior year, when there were eight disengagements per 10,000 miles.

“This four-fold improvement reflects the significant work we’ve been doing to make our software and hardware more capable and mature,” Dmitri Dolgov, Waymo’s head of self-driving technology,Dmitri Dolgov, Waymo’s head of self-driving technology wrote in a blog post.

Waymo’s chief critic acknowledged the improvement, but John Simpson of the nonprofit Consumer Watchdog said the number of disengagements shows the cars still “simply aren’t ready to be released to roam our roads” without human backup drivers.

Advertisement