Waymo shows how its cars deal with snow [Video]by Justin King
Machine-learning algorithms allow the car to filter out sensor noise caused by lidar reflections on snow flakes.
Waymo has shed a bit of light on how its cars deal with snow and other weather conditions with poor visibility.
Speaking at Google's annual I/O conference, Waymo CTO Dmitri Dolgov showed images of a snow-covered road and a cloud of sensor noise surrounding the Waymo car.
Engineers used deep learning to train the car's computer to ignore the sensor noise, caused by reflections off snowflakes and snow on the ground. The before-and-after photos show a much cleaner image of the surroundings after filtering has been applied.
Google's experience in AI has helped accelerate Waymo to become the apparent leader in bringing self-driving technology to market. In the early days, before spinning off from Google, the group brought in help from the Google Brain team to apply deep neural nets to the pedestrian detection system.
"Even in those early days, the results were remarkable?—?within a matter of months, we were able to reduce the error rate for pedestrian detection by 100x, making our system safer and more capable on the road," the company says.
The snow test was presumably conducted in Michigan, one of the company's satellite locations that are helping to prepare for a wider rollout beyond the sunny desert streets of Phoenix, Arizona.