Google has revealed that its autonomous cars have required human intervention to stop them crashing on 13 occasions during testing in California between September 2014 and November 2015. This and other data has been published by the technology company following a local regulator’s demand for the information.
Google wants to sell driverless cars that do not feature manual controls. A fleet of test models, which have manual controls and a driver prepared to intervene if necessary, have been covering California’s roads to collect data for Google.
Google car testing data
As ordered by the California’s Department of Motor Vehicles, Google has now revealed that its driverless cars covered 424,331 test miles between 24 September 2014 and 30 November 2015.
In that same time period, there were 272 cases when the software on a Google driverless car detected a “failure” that caused it to alert the driver and hand over control.
Furthermore, there were 69 occasions when the driver took manual control of a Google car without being prompted because they perceived there was a safety threat. It was worked out via computer simulations that in 13 of the driver-initiated interventions, a crash would have followed had the driver not taken control.
While two of these near misses would have involved hitting only a traffic cone, Google admitted the other 11 would have been more serious.
Google also said: “These events are rare and our engineers carefully study these simulated contacts and refine the software to ensure the self-driving car performs safely.”
The report indicates that a true self-driving car that can cope with all driving scenarios is still some way off. Google points out though that incidents where the driver takes manual control like the ones mentioned in its report are necessary to ensure all possible scenarios are encountered for in their research.