Waymo uses machine learning to help self-driving cars see through snow
Wintry weather can wreak havoc on all manner of sensors and cameras.
Five minutes into a harrowing cross-state winter drive, I received a warning on the dashboard. All my front-facing sensors and cameras were obscured with ice, rendering the vehicle's numerous active safety systems disabled. Winter can make it extremely hard for sensors to do their jobs, but Waymo has a trick that my car didn't -- machine learning!
At this week's Google I/O conference, Waymo discussed how it uses sister company Google's developments in machine learning to help its self-driving vehicles navigate snowy climates. Snow shows up on Waymo's sensors as noise -- a giant purple cloud of object-obscuring noise. If a vehicle actually thought solid, unmoving objects surrounded it, the vehicle would stop and resume travel in the spring.
That's where machine learning comes in. It's able to filter out all that noise, and the result is an image that appears to pick up not only the vehicles parked on the street, but the curbs, as well. With the noise out of the way, Waymo's vehicle is free to continue its route. Let's just hope it's wearing winter tires. Safety first.
See also
Being able to handle inclement weather is one of the biggest hurdles in bringing autonomous vehicles to the public -- in addition to, you know, developing a safe, reliable AV in the first place and then convincing the public that it's safe and reliable.
As for your own car and its systems, there's not much you can do there. In the winter, ensure your various parking sensors and cameras are free of snow and ice, especially before setting out in the morning or beginning a long drive. You may also need to pull over midtrip to clear buildup, but it's worth the peace of mind if that's what you're after.
