The safety of self-driving cars has just taken a positive leap forward. Google submitted a glowing Vehicle Disengagement Report to the California DMV this January. The safety of their Waymo self-driving cars leapt by 75%, from their results in 2015.
The report reveals that in 2016, human disengagement of autonomous cars was dramatically reduced. Of 635,868 miles travelled on public roads, a human driver needed to take over the wheel on only 124 occasions. In 2015 driverless cars traveled 424,331 miles on public roads. During that time drivers had to interrupt the automation 341 times.
Dimitri Dolgov, head of Waymo’s self-driving technology, claims this as a gigantic move forward. Technological advances in their cars aim to avoid the 94% human error that is the reason behind all car crashes. Of the 124 human disengagements, 10 resulted from the recklessness of other drivers.
Others are merely part of the testing process. Testers are perfecting the software deviations. They are monitoring and correcting unnecessary maneuvers that the vehicle makes. The purpose is to tweak any AI perception discrepancies.
Skeptics showing concern about self-driving cars bring up serious concerns. However, developers like Waymo incorporate these concerns into improving safety. In the past, testing never took place in extreme weather.
The AI had problems with recognizing things like temporary traffic lights, pot holes and harmless debris. The cars were failing to stop for police officers signaling for them to stop. Many times, autonomous vehicles would unnecessarily default into “extra caution” mode.
Keep reading …