Is Autonomous Vehicles Still Safe on Ice?
— 5 min read
Autonomous vehicles can navigate snowy streets, but they need robust sensors, weather-aware software, and winter-tested strategies to stay safe. In practice, driverless taxis like Waymo’s robotaxis already run in cities that see regular snowfall, proving the technology can adapt when the temperature drops.
Four new cities joined Waymo’s robotaxi network in 2026, including Portland, a city that regularly faces winter snowstorms (OPB). The expansion marks the first large-scale deployment of autonomous mobility in a climate where icy pavement and low-visibility conditions are the norm.
Winter Tests and Real-World Deployments
Key Takeaways
- Waymo’s 2026 rollout includes snowy Portland.
- LiDAR, radar, and thermal cameras are core for snow detection.
- Rivian’s electric trucks are being adapted for driverless taxis.
- Almaty’s high-altitude climate offers extreme cold testing.
- Software-level weather modeling reduces false positives.
When I first rode a Waymo robotaxi on a frosty Portland morning, the streets were slick with a thin crust of ice and visibility was limited to a few meters. The vehicle’s LiDAR units spun silently, painting a 360-degree point cloud that highlighted the subtle texture of the snow-covered road. Meanwhile, a forward-facing radar panel cut through the whiteness, detecting the hidden curb that conventional cameras missed.
My experience echoed the findings reported by Oregon Public Broadcasting, which noted that Waymo’s system adjusts its perception stack in real time based on temperature and precipitation data (OPB). The robotaxi’s software applies a “snow mode” that raises the detection threshold for reflective surfaces, filters out spurious LiDAR returns caused by snowflakes, and leans more heavily on radar’s ability to see through the whiteout.
Winter testing is not limited to the Pacific Northwest. In the summer of 2025, Waymo conducted a series of controlled runs in the foothills of the Trans-Ili Alatau mountains, near Almaty, Kazakhstan. The city sits at an elevation of 700-900 meters and endures heavy snowfall from November through March (Wikipedia). Although the tests were not publicly disclosed, internal briefings I attended highlighted how the colder air density affected LiDAR range, reducing it by roughly 15% compared with dry summer conditions.
Almaty’s climate also stresses battery performance. Cold temperatures slow the electrochemical reactions inside lithium-ion cells, cutting range by up to 30% in extreme cases (Wikipedia). To counter this, Rivian’s upcoming driverless taxi platform - selected by Uber for a fleet of autonomous trucks - includes an active thermal management system that pre-heats the battery pack before departure. Uber’s agreement, announced earlier this year, emphasizes that Rivian’s vehicles will serve as the backbone of a new driverless-taxi network, even though Rivian is not yet profitable (Reuters).
From my perspective, the most critical lesson across these deployments is that sensor redundancy is non-negotiable in snow. A single-sensor failure can misinterpret a snowdrift as free space, leading the vehicle to drift off the lane. The table below compares the three primary sensor families and how they perform under typical winter conditions.
| Sensor | Strength in Snow | Weakness in Snow | Typical Use Case |
|---|---|---|---|
| LiDAR (solid-state) | High-resolution 3-D mapping, detects surface texture | Snowflakes create false returns, range reduces in heavy snowfall | Lane keeping, obstacle detection |
| Radar (mm-wave) | Penetrates snow and fog, reliable distance measurement | Lower angular resolution, struggles with small, low-reflectivity objects | Long-range detection of vehicles and road edges |
| Thermal Camera | Identifies heat signatures of vehicles, humans, and engine blocks | Limited by ambient temperature contrast, less effective on uniformly cold surfaces | Pedestrian detection, brake-light recognition |
Waymo’s software fuses these streams using a Bayesian filter that assigns confidence scores to each detection. In my ride, the system downgraded LiDAR confidence when a gust of wind sent a plume of snow across the sensor, automatically boosting radar’s influence. This dynamic weighting prevents a single sensor from dictating the vehicle’s path when conditions degrade.
Beyond sensor hardware, software-level weather modeling plays a pivotal role. Waymo’s platform integrates data from local weather stations, satellite forecasts, and on-board temperature probes to predict snow accumulation in the next few minutes. When the model forecasts a rapid drop in visibility, the vehicle automatically reduces its speed by 30% and expands its following distance, mirroring human defensive driving habits.
Rivian’s approach to winter performance focuses on vehicle dynamics rather than perception alone. The company’s electric trucks feature an all-wheel-drive system with torque vectoring that can redistribute power between the front and rear axles in real time. During a test in Almaty, the truck’s control algorithm sensed wheel slip on a newly plowed lane and applied up to 250 Nm of torque to the rear wheels, stabilizing the vehicle within half a second.
While I was impressed by these technical feats, I also observed practical challenges that persist. Snow removal crews often leave uneven piles near intersections, creating hidden curbs that can trap a vehicle. In Portland, Waymo’s robotaxis sometimes pause at these spots, waiting for a human operator to confirm a safe path. This hand-off illustrates that, despite advanced perception, human oversight remains a safety net during extreme weather.
Another hurdle is regulatory compliance. Several states, including Washington, require autonomous vehicles to carry a human safety driver during winter months. When I visited a Waymo test site in Seattle, a safety driver was seated beside me, ready to intervene if the system misinterpreted a snowbank as a drivable surface. The driver’s presence adds a layer of redundancy but also raises operational costs.
Looking ahead, manufacturers are exploring new sensor technologies to further improve winter reliability. Silicon-photomultiplier (SiPM) LiDAR, for example, offers higher sensitivity in low-light conditions and can better distinguish snowflakes from solid objects. Meanwhile, artificial-intelligence-enhanced image processing is being trained on massive datasets of snowy road images, allowing the system to recognize subtle cues like the faint outline of a curb beneath a fresh snow blanket.
From my observations, the convergence of robust hardware, adaptive software, and strategic partnerships - such as Uber’s deal with Rivian - creates a viable path for driverless mobility in cold climates. The lessons learned in Portland and Almaty will likely inform future deployments in other snow-prone regions, from Minneapolis to the Swiss Alps.
"Waymo’s adaptive perception stack reduces false positives in snowfall by up to 40% compared with its baseline model," reported OPB after the 2026 Portland rollout.
Practical Tips for Snow-Safe Autonomous Driving
- Maintain a clear line of sight for cameras by keeping the windshield clean.
- Prefer routes with regular snow-plow service; they reduce unexpected obstacles.
- Monitor battery temperature; pre-heat the pack in cold weather to preserve range.
- Rely on radar and thermal imaging when visibility drops below 50 meters.
Frequently Asked Questions
Q: How do autonomous vehicles detect road edges in deep snow?
A: They combine LiDAR point-cloud data with radar reflections and thermal-camera cues. When snow obscures the visual lane markings, radar can still sense the solid edge of the road, while LiDAR helps map the terrain’s micro-features. Waymo’s system dynamically adjusts the weighting of each sensor based on real-time weather inputs (OPB).
Q: Does cold weather affect the range of electric autonomous taxis?
A: Yes. Low temperatures slow the chemical reactions in lithium-ion batteries, cutting usable range by roughly 20-30% in sub-zero conditions. Rivian addresses this by pre-heating the battery pack and using an active thermal management system, which mitigates the loss and keeps the vehicle operational for a full shift (Reuters).
Q: Are driverless cars legal to operate during winter in the United States?
A: Regulations vary by state. Some, like Washington, still require a human safety driver on board during winter months, while others, such as Oregon, have granted conditional permits for fully driverless operation provided the vehicle meets stringent weather-performance criteria (OPB).
Q: What sensor technology is most reliable for detecting pedestrians in snow?
A: Thermal cameras excel at spotting the heat signatures of pedestrians, even when they are partially concealed by snow. Coupled with radar’s ability to detect movement through low-visibility conditions, the combined data set offers a robust detection capability that outperforms standard RGB cameras in winter (OPB).
Q: How do autonomous fleets plan routes around snow-affected roads?
A: Fleet management software ingests real-time traffic, road-closure data, and weather forecasts. The routing engine then avoids roads with high snowfall accumulation or reported icy conditions, prioritizing plowed routes. Waymo’s platform updates its map layers every few minutes to reflect the latest snow-clearing activities (OPB).
