In a speech at an auto show in Detroit, the new CEO of Google’s driverless car division John Krafcik signalled a strong focus on “nothing short of full autonomy”.
But a recent report shows that this still might be a while off, with the autonomous mode on Google’s driverless cars switched off 272 times over a 14-month period due to “software failures” and 13 times to prevent an imminent accident.
The report was released by Google this week and reveals its autonomous vehicles have traversed more than 680,000km over this time, and incidents of “disengagement” are becoming less regular.
The number of interventions to prevent a “simulated contact” will be most troubling for the sceptics, with drivers having to take control of the wheel to prevent a collision.
“They are directed to take control of the vehicle as often as they feel is necessary and for a variety of reasons relating to the comfort of the ride, the safety of the vehicle or the erratic or unpredictable behaviour of other road users,” the report says.
“These events are rare and our engineers carefully study these simulated contacts and refine the software to ensure the self-driving car performs safely.”
Of these 13 incidents, two would’ve resulted in the autonomous vehicle hitting a traffic cone, but the others would’ve been more serious.
Google points out that the driverless cars are still very much in a “development phase” and its drivers always “err on the side of caution”, but the tech giant hopes to have the vehicles on sale to the general public by 2020.
In the speech at the Detroit car show, Krafcik emphasised the safety advantages of making the cars fully autonomous.
“Aiming for full autonomy not only reaches the most people, our team believes it’s also the safest approach,” Krafcik says.
“Having this audacious goal was what drew me to the Google self-driving car project.
“The industry has been making continuous incremental gains, but for self-driving cars to reach their full potential we need to focus on nothing short of full autonomy.”