Weather Still a Challenge to AVs
In the drive towards full autonomy, vehicle makers and solutions providers must tackle a nearly impossible variety of factors.
Some of these, often necessarily, are more attention-grabbing than others but there are numerous ones just as critical that don’t typically appear in auto technology headlines. One is the weather, which is somewhat odd because information about weather and its effect on road conditions has been crucial ever since humankind has climbed into a car. In our ‘assisted driving’ age, that need remains acute and experts say it will be more so as we move up the levels of autonomy. Fortunately, with increasingly robust vehicle-to-vehicle (V2V) and vehicle-to-everything (V2X) technology, weather data and forecasts can be delivered ever more effectively.
To explore this subject, TU-Automotive spoke to Bill Gail, CEO and co-founder of Global Weather Corporation, a company that concentrates on the delivery of weather data services.
Why is weather data analysis and dissemination critical for upper-level assisted driving and, ultimately, autonomous operation?
“Automated vehicles (lower or upper level) don’t operate well in poor weather conditions, such as fog or snowy roads. Many people believe it will be a long time before they can do so. Weather information can be a key element of safety-critical AV decisions, such as disengagement and operational domain designation – even simpler functions such as lane management. Weather information can be the difference between an autonomous vehicle (or vehicle function) that continues to operate properly and one that does not.”
Why is it important for V2V technology to include weather reporting?
“V2V can play an important role in weather information, particularly reporting of just-ahead road conditions. Vehicles that reliably report friction or slippery roads can report those conditions to vehicles behind them. Doing so requires in-vehicle processing to generate the relevant weather information but it is possible. One challenge is that vehicle sensors, other than temperature, don’t directly measure weather variables. For example, road friction sensors tell you that a road is slippery but can’t say the slipperiness is because the road is snowy or wet.”
By what means can a properly equipped car extract weather information – does it have its own sensors, or does it rely on outside data?
“Temperature sensors provide reliable (though not always highly accurate) information about air temperature. Other vehicle sensors tend to make ‘proxy’ measurements for weather variables. For example (as noted above), traction sensors make measurements that are related to road conditions but don’t tell you whether a road is wet or snowy. Similarly, wipers tell you something about rain but drivers use them differently so you can’t conclude that it is raining just because someone’s wipers are on. In many cases, data sent to the vehicle is more reliable and accurate than data obtained with vehicle sensors.”
How can weather data from external sources be delivered to the car?
“It is delivered by web services to the car, typically using cell service. V2V can be part of this network.”
Weather can change quickly, so how can we ensure that cars receive near real-time data on weather conditions?
“As cell coverage gets better and better, this problem gets smaller. For the great majority of people, it is not a problem today if their car is equipped to receive the information.”
Do car manufacturers realize the importance of robust weather data delivery to cars with ADAS features?
“They are starting to but they are doing so slowly. In part, many thought initially that all weather information could come from vehicle sensors. [There are] many companies that have learned this is not sufficient.”
Could weather technology be used for situations and applications unrelated to vehicles? What might be some examples?
“[Such] data is used in many situations by our existing customers. It is used in consumer weather apps, for analyzing retail demand, for managing wind and solar farms, for financial trading, for supply chain management, and much more.”
Where are we currently in terms of weather data-delivering technology, are we at perfect real-time delivery of such data? If not, what needs to occur before we get to that stage?
“A current-conditions estimate (what’s happening now, at all locations) or a forecast (same but for the future) is computed from a large set of input data sources. The data sources include weather model data from the US National Weather Service and international weather agencies, observations from weather stations, precipitation estimates from weather radars and, eventually, sensor data from vehicles. These are all generated at different times and with different update rates – some hourly, others more often.”
You point out that vehicle sensors don’t directly measure weather variables. Is it possible to develop sensors that can do this and what, if anything, is holding back this development at present?
“There are sensors that do this, sometimes referred to as Mobile Road Weather Information Systems (MRWIS). They are expensive, dedicated systems that are mounted on a vehicle. I was referring to the built-in vehicle sensors. It is certainly possible to improve existing sensors or add new sensors but you don’t see vehicle manufacturers doing this because of the cost.”
