Waymo is coming to Toronto.
asserted
Waymo → come → Toronto
Will its self-driving cars be winter-ready?
asserted
cars → drive → ?
Experts say seasonal challenges can be overcome, but require time and training
Waymo is coming to Toronto.
asserted
Waymo → say → Toronto
But the road to get there will be an often snow-covered one — and experts say that will take time and training data to overcome.
asserted
that → get → time
The U.S. company said Friday it plans to start testing the streets of Canada's biggest city, with drivers behind the wheel, in the coming days, as part of its "mapping and safety validation work.
asserted
it → say → work
"
The goal is for riders to be able to hail one of their driverless cabs by 2028 in the city, Waymo said.
asserted
Waymo → hail → city
The tests will take place at all hours of the day and night and across all days of the week, in an effort to make sure the technology can adapt to Toronto's "distinct driving conditions."
asserted
technology → take → conditions
The company didn't expand on what those conditions were.
asserted
conditions → expand → ?
But Steven Waslander, professor and director of the University of Toronto's Robotics Institute, says that likely means snow.
asserted
that → say → snow
"The big differentiator is really our winter," Waslander said of Toronto compared to other North American cities where Waymo is already up and running.
asserted
Waymo → say → cities
"The winter challenge is one that [Waymo] haven't demonstrated commercially yet.
asserted
Waymo → demonstrate → that
Waymo first hit the road in warmer cities with little precipitation like Los Angeles and Phoenix, Ariz., and most of the cities it offers ride-hailing services in today are much warmer and dryer than Toronto.
asserted
it → hit → Toronto
That's for good reason, according to Waslander, who says running autonomous cars in the snow comes with all kinds of extra challenges.
uncertain
running → accord → challenges
Autonomous cars rely on cameras, radar and lidar systems — which map the 3D environment using light — to situate themselves.
asserted
which → rely → themselves
All of that information is fed into a computer, which makes decisions about where to turn or when to hit the brakes, augmented by lots and lots of training data.
asserted
which → feed → data
Adverse weather conditions can pose a challenge for all of those systems, says Krzysztof Czarnecki, an engineering professor at the University of Waterloo.
asserted
Czarnecki → pose → Waterloo
Cameras and other sensors can get covered by snow and ice, making them less effective.
asserted
them → cover → snow
Heavy falling rain and snow can also reduce how well the driving system detects objects — as can snow covering cars or street signs.
asserted
snow → fall → cars
Cameras, radar panels and lidar sits atop a Waymo all-electric Jaguar I-Pace SUV in San Francisco, on Aug. 20, 2021.
asserted
Cameras → sit → Aug.
The systems help the autonomous vehicles navigate their environment.
asserted
vehicles → help → environment
Czarnecki says as autonomous vehicles are trained on more data, they get better at detecting obscured objects like snow-covered signs.
asserted
they → say → signs
But he says autonomous systems still struggle to make sense of them more than a human would.
asserted
human → say → them
Other drivers also change their behaviour when snow flies, Czarnecki says, making autonomous vehicles more likely to get "confused" by their actions or those of pedestrians.
asserted
vehicles → change → pedestrians
Adjusting to those kinds of elements takes time, as Czarnecki says the vehicles will need to be tested and fine-tuned in those environments.
asserted
vehicles → adjust → environments
That's why some autonomous cars already on the road in Canada don't run in poor weather.
asserted
cars → run → weather
U.S. startup Gatik, for example, which runs 50 self-driving trucks for Loblaws in the Toronto area, can drive day or night and handle complex driving situations.
asserted
which → run → situations
But it draws the line at heavy rain or snow.
asserted
it → draw → rain
Steve Waslander, a researcher at the University of Toronto who's been studying how autonomous vehicles function in wintery conditions, says those hurdles are big, but not insurmountable.
asserted
hurdles → study → conditions
What's needed are engineering solutions, including more data gathered by driving cars in poor weather which the Waymo system can be trained on.
asserted
system → need → which
"All of the existing techniques for detection, tracking, prediction, for planning, for control — they can all be applied," Waslander said.
asserted
Waslander → exist → control
"You just have to know that you're in these conditions and include that training information in the way you build the system."
Waymo has been testing its tech in snow-filled regions, like upstate New York and Michigan's Upper Peninsula.
asserted
Waymo → have → York
The company is also fully operating in Denver, Colo., and is conducting tests in colder, more northern U.S. cities like Minneapolis and Detroit.
asserted
company → operate → Minneapolis
According to the Waymo website, the company has logged tens of thousands of hours driving in snowy weather.
uncertain
company → accord → weather
The data from those blustery trips is used to train its autonomous driver system — the latest iteration of which has more than 100 million miles of autonomous driving experience — to adapt to all kinds of conditions.
asserted
iteration → use → conditions
"Today, we operate at below-freezing temperatures in the U.S., including with frost and hail, and we’re validating our system to navigate harsher weather conditions," Waymo said in an emailed statement to CBC News.
asserted
Waymo → operate → News
A resident attempts to unclog a few storm drains on a flooded street as a driverless Waymo taxi passes in the background, on Nov. 22, 2024, in San Francisco.
asserted
taxi → attempt → Francisco
Still, more winter driving data is needed.
asserted
data → need → ?
Waymo's product lead for weather told The Verge in 2025 that snowy conditions often made up less than five per cent of total driving data, though the company was using AI methods to augment and make progress on winter driving capabilities despite that.
asserted
company → tell → that
Waslander says with the expansions into so many other snowy markets will help collect that data.
asserted
Waslander → say → data
When Waymo arrives in a new city, Waslander says it usually deploys a fleet of 50 to 100 vehicles to map a city in all kinds of conditions.
asserted
it → arrive → conditions
…and 8 more, not listed.
The Aporia
