FutureStructure

Waymo Self-Driving Minivan Hits Mountains to Train Software, Collect Snow Data

To create a fully driverless vehicle, capable of handling all driving situations, Waymo is training its autonomous minivans to operate in the snow.

by / March 28, 2017

Autonomous vehicles have overcome many challenges, including other drivers' erratic behavior, pedestrians crossing the middle of the street, and recognizing the difference between red and green lights. But one of the most challenging obstacles to overcome is the common snowflake.

Bad weather and snowfall cause several problems for self-driving vehicles: it can cover up road striping; lower visibility for streetlights and road signs; and stick to the surface of light detection and ranging (lidar) sensors — essentially blinding the "eyes" of the vehicle.

Waymo, the final form of Google’s self-driving vehicle team, announced on March 27 that its Chrysler Pacifica Hybrid minivans were taking this challenge head on. Waymo tweeted a picture:

The testing was done in South Lake Tahoe, Calif., about four hours from the company's Silicon Valley headquarters. Most of the 2 million miles the fleet has driven has been on city streets in Phoenix, Ariz.; Austin, Texas; Kirkland, Wash.; and Mountain View, Calif.

But driving in the snow is not just a problem for Waymo. EasyMile, which builds autonomous shuttles, also cites poor weather as a major challenge.

"Snow is an issue for us, because if it is heavily snowing, then the snowflakes will stick to the lasers on the vehicle. This would cause the vehicles to think that there is an obstacle,” said spokesperson Marion Lheritier. “Sometimes we have an emergency stop for snowflakes.”

Some companies have opened up more regular testing tracks and agreements with cities that regularly experience a snowy winter. Ford, through its partnership with the University of Michigan, does a large portion of its autonomous vehicle testing at Mcity. And Uber struck a deal with Pittsburgh last May to navigate its downtown.