While others flounder, Waymo has developed a metric for detecting and predicting the weather in the cities where it operates, for example, in foggy San Francisco.
It is a known fact that autonomous vehicles have difficulties navigating in adverse weather conditions. Since AV’s perception systems rely heavily on various sensors such as cameras, radar, and lidar to see the world around it, heavy rain, snow, and fog can make it more complicated. For example, wet roads can create reflections that confuse cameras, or ice can form on the lidar, preventing the sensor’s ability to send out laser points for measuring the distance to other objects on the road.
Alphabet owned Waymo, which aims to deploy fleets of robotaxis all over the USA, has been investing in weather research since its early days. Even though AVs still struggle to navigate in adverse weather conditions, Waymo has gotten a lot better at detecting and predicting the weather in the cities in which it operates. The company has even managed to create the first fog map for San Francisco. The map consists of millions of data points collected by Waymo’s fleet of AVs as they traverse San Francisco’s streets during a foggy weather. Combined with a special visibility sensor, Waymo creates a new meteorological “metric” which is then sent to “Waymo Driver”, the self-driving AI, to aid its decision-making.
Daniel Rothenberg, a trained meteorologist, and a member of the company’s weather team, said in an interview with The Verge:
“We’re describing our vehicles as mobile weather stations. And that’s exactly how they’re functioning.”
Waymo’s early testing was carried out in Arizona, with its ideal weather, mostly sunny and dry, and relatively flat geography. However, in the past few years, the company has expanded its testing to include more extreme conditions, including snowy Novi, Michigan, rainy Kirkland, Washington, and foggy San Francisco.
Traditionally, weather stations are considered the best source for real-time weather information. But since more of these stations are located at airports and can be imprecise in measuring local conditions, Waymo sees value in gathering more granular weather data to help inform its autonomous vehicles. Waymo’s fleet can track the progression of coastal fogs flowing from the Pacific Ocean. It can detect drizzle and light rains that lead to wet roads in situations that are invisible to the National Weather Service’s local Doppler weather radar.
“We’re understanding in real time the actual weather conditions that are impacting our vehicles in a really hyperlocal context,” Rothenberg said. “And that’s just something that hasn’t been done before.”
Ultimately, the final goal is to build vehicles that can safely drive through all types of weather conditions in a bid to outperform human drivers.
How, and if, AVs navigate in bad weather conditions is an open question and in terms of development of AI systems it is challenging. However, the approach of Waymo may not be as innovative as it seems. Essentially the system detects and predicts different weather conditions so that their robotaxi can avoid those areas, rather than navigate in the adverse weather. For example, a company called Sensible4 has already found a solution. According to them, vehicles that are powered with their technology can precisely locate themselves in all weather conditions and on the roads without any surface markings or other dedicated structures. I think that sounds like a cutting-edge technology, doesn’t it?
However, on first blush it might seem like Sensible4 technology would be hard to implement in real settings. Apparently, this is not the case. The technology from Sensible4 has been proven in real Scandinavian weather conditions with lots of snow, rain, and low temperatures. Ruter, the public transport authority for Oslo in partnership with mobility companies Holo, Toyota Motor, and Sensoble4 conducted a pilot study in Norway, where two autonomous vehicles with the Sensible4 technology drove for a period of one year. The result was that the solution from Sensible4 showed the viability of automated vehicles in inclement weather. It may be that Waymo is working on solutions like Sensible4, or maybe they see other problems with this solution. Regardless, if AVs are to be of any expansive social use they must be capable of navigating through weather, not just around it.
Written by Kateryna Melnyk,
RISE Mobility & Systems