2026-06-18 — views

Physical AI Weather ODD 2026 — Waymo LIDAR-Radar Fusion vs Tesla Camera-Only Rain and Fog: The All-Weather AV Benchmark

Waymo LIDAR plus radar compensates when cameras degrade in fog and rain. Tesla FSD is camera-only. Snow-belt cities are off-limits for all commercial AV today.

Article 187 in the Physical AI Benchmark Series — Physical AI Weather and Environmental ODD 2026: Waymo Multi-Sensor Adverse Weather Resilience vs Tesla Camera-Only Rain and Fog Sensitivity

Weather is not a secondary consideration in autonomous vehicle deployment — it is a primary gating variable for the Operational Design Domain (ODD). The ODD defines the specific conditions under which an AV system is certified and designed to operate safely. Rain degrades cameras, fog reduces visibility, snow covers lane markings and accumulates on sensors, extreme heat stresses battery and sensor hardware. The physics of how each sensor type interacts with weather conditions directly determines which geographic markets and seasonal windows any given AV architecture can serve commercially. This article is Article 187 in the Physical AI Benchmark Series. It benchmarks Waymo’s multi-sensor fusion (LIDAR plus radar plus camera) against Tesla’s camera-only approach across five weather ODD dimensions: sensor performance by condition, Tesla FSD camera limitations, geographic ODD implications, and a structured all-weather scorecard.

All figures labeled “(est.)” are derived from public disclosures, industry research, and analyst estimates rather than independently verified primary data.

Section 1 — Weather as an AV ODD Gating Variable

The Operational Design Domain defines the specific conditions under which an AV system is certified and designed to operate. Weather is one of the primary ODD dimensions, alongside geography, road type, speed range, and time of day. For sensors, weather is a physics problem — not a software problem — and the physics create structural advantages and disadvantages that software cannot fully overcome.

How weather degrades cameras. Cameras operate on visible light reflected from the environment. Rain introduces water droplets on the lens housing and in the air column between camera and scene. Water droplets scatter light, reducing image resolution, contrast, and effective range. Heavy rain generates spray from other vehicles that further degrades the image. Fog introduces a dense suspended water-droplet layer across the entire air column; effective camera range in dense fog can drop below 50 meters — less than one second of stopping distance at highway speeds. Snow accumulates on camera housings and lens covers; heating elements must be active to prevent ice formation. Direct sun glare, when the sun is at a low angle in front of the vehicle, can overexpose camera sensors and temporarily render lane markings and other vehicles invisible. These are not software bugs — they are physics constraints of visible-light imaging.

How weather affects LIDAR. LIDAR emits laser pulses and measures return time to build a 3D point cloud. Fog scatters LIDAR pulses — the laser energy is absorbed and reflected back by suspended water droplets before reaching distant objects, reducing effective range significantly in dense fog. Heavy rain produces false point-cloud returns as raindrops appear as obstacles. Snow accumulates on the LIDAR housing and requires active clearing or heating; falling snowflakes return strong laser pulses that mimic solid obstacles. LIDAR is, however, an active illumination sensor: unlike cameras, it generates its own light source and is completely unaffected by visible-light conditions including sun glare and total darkness.

How weather affects radar. Radar (millimeter-wave) is the most weather-resilient sensor of the three. Radar pulses penetrate fog with minimal attenuation, penetrate rain with minimal degradation, and penetrate light snow effectively. Radar provides velocity and range data across all precipitation conditions. Its primary limitation is angular resolution — radar cannot generate high-detail 3D maps of the environment the way LIDAR can. But for detecting the presence and velocity of objects in poor visibility, radar is the all-weather workhorse.

The case for multi-sensor fusion. No single sensor is ideal in all weather conditions. Camera excels in clear daylight; LIDAR excels in clear darkness; radar excels in all precipitation. Fusion — combining LIDAR plus radar plus camera — provides redundancy: when one sensor type is degraded, the others compensate. Camera-only systems have no such redundancy. When cameras are impaired by fog or heavy rain, a camera-only system has no fallback.

Current commercial AV geography reflects weather ODD. Every commercial driverless AV service today operates in mild or dry climates: Phoenix (desert, hot, dry, minimal rain), San Francisco (mild coastal fog but not heavy rain), Los Angeles (sunny, minimal rain), Austin (warm, some rain but no severe winter). No current commercial AV service operates in snow-belt cities — Chicago, Minneapolis, Boston, New York, Buffalo — precisely because weather ODD constraints make heavy-snow commercial driverless operation technically beyond the current state of the art for all companies.

Section 2 — Waymo Weather Performance by Sensor Type

Waymo’s sensor architecture (camera plus LIDAR plus radar) provides redundancy across weather conditions. The table below benchmarks Waymo’s fusion performance by weather type.

Weather condition	Camera performance	LIDAR performance	Radar performance	Waymo overall (fusion)
Clear and sunny	Excellent; best visibility	Excellent; full range	Good	Excellent; all sensors optimal
Light rain	Slightly degraded; water on lens reduces clarity	Slightly degraded; rain returns add noise	Excellent; unaffected	Good; LIDAR and camera degrade slightly, radar compensates
Heavy rain	Significantly degraded; spray and water on lens	Moderately degraded; rain returns increase noise	Excellent	Fair; camera and LIDAR impaired; radar compensates for velocity and range; Waymo may slow or limit ODD in heavy rain
Light fog	Degraded; effective range reduced significantly	Moderately degraded; fog scatters LIDAR pulses	Good	Fair; LIDAR and camera range-limited; radar maintains range
Dense fog (under 50m visibility)	Severely degraded; near-useless at short ranges	Severely degraded; fog scattering overwhelms signal	Good at range; short-range detail poor	Poor; Waymo likely pulls service in dense fog conditions
Light snow	Moderately degraded; snow on lens; reduced contrast	Moderately degraded; snow returns and accumulation	Excellent	Fair; camera impaired; LIDAR partial; radar good
Heavy snow	Severely degraded; whiteout conditions; lane markings buried	Severely degraded; accumulation on housing; white-out scattering	Good	Poor; Waymo has not attempted heavy-snow commercial operations; LIDAR and camera both severely impaired — this is why Waymo has not expanded to snow-belt cities
Extreme heat (Phoenix, 115F+ est.)	Performance maintained; cameras are electronic, heat within spec	Performance maintained with adequate cooling; LIDAR requires thermal management	Performance maintained	Good; Phoenix operations confirm Waymo can handle extreme heat; thermal management required
Night and low light	Degraded; cameras require illumination; headlights help	Excellent; LIDAR is active illumination, works in complete darkness	Good	Good; LIDAR enables night operation that camera-only systems struggle with

The heavy snow constraint. The table reveals Waymo’s structural limitation: heavy snow impairs both camera and LIDAR simultaneously, leaving only radar as a functional sensor. Radar provides velocity and range but lacks the spatial resolution required to navigate complex urban environments safely. This dual-impairment in heavy snow is the primary reason Waymo has not expanded to the snow-belt cities that represent the upper Midwest, Northeast, and Mountain West US population centers.

Section 3 — Tesla FSD Weather Performance: Camera-Only Limitations

Tesla’s Full Self-Driving (FSD) system uses cameras as the sole primary sensor for environmental perception (supported by radar on HW4 hardware for object detection). Unlike Waymo’s triple-fusion stack, Tesla’s camera-only architecture means there is no LIDAR fallback when cameras are degraded. The table below benchmarks Tesla FSD performance by weather type.

Weather condition	Camera performance	Tesla overall (camera-only)	FSD recommended action	Notes
Clear and sunny	Excellent	Excellent	Normal operation	Tesla cameras optimized for this condition
Direct sun glare	Severely impaired; bright light overexposes sensor; lane markings and vehicles can be invisible	Severely impaired	Driver must take over	Known FSD limitation; glare on sunrise and sunset roads has been associated with incidents in FSD research; NHTSA investigated sun glare issues
Light rain	Moderately degraded; water on windshield (wipers help); spray from other vehicles	Moderately degraded	FSD reduces speed; driver may need to assist	Tesla wiper activation plus FSD deceleration is current mitigation
Heavy rain	Significantly degraded; heavy spray, water on cameras, reduced visibility	Significantly degraded	FSD may disengage; driver must be attentive	Heavy rain is a known FSD limitation; camera-only has no radar fallback for spatial detail
Fog	Moderately to severely degraded depending on density; no active illumination to penetrate fog	Moderately to severely degraded	FSD reduces speed; dense fog may trigger disengagement	Without LIDAR or radar backup for spatial mapping, fog is a fundamental camera limitation
Light snow	Severely degraded; snow on cameras; lane markings buried; white environment reduces contrast	Severely degraded	FSD likely to disengage; Tesla recommends driver supervision in snow	Snow is a hard camera limitation; no LIDAR to detect road structure under snow
Heavy snow and blizzard	Unusable; complete camera impairment	Unusable	Driver must take full control	Tesla FSD does not operate in blizzard conditions; this is an acknowledged ODD boundary
Night and darkness	Degraded without ambient lighting; headlights help but limited range	Degraded vs daytime	FSD operates with reduced performance; driver advised to monitor	No active illumination means Tesla cameras depend on ambient light and headlights; LIDAR-equipped operators have active illumination advantage
Extreme heat	Performance maintained	Maintained	Normal operation	Cameras are electronic; heat within spec; battery range affected separately

The camera-only weather physics constraint. The fundamental limitation of Tesla’s camera-only approach in adverse weather is not a software problem — it is a physics problem. Cameras cannot penetrate fog the way radar can. Cameras cannot maintain effective range in heavy rain the way radar can. Cameras depend on reflected ambient or artificial light the way LIDAR does not. Software improvements can optimize how camera images are processed under degraded conditions, but they cannot overcome the physics of light scattering, absorption, and occlusion. The camera-only ODD is therefore bounded by the physics of visible-light imaging in adverse conditions.

Section 4 — Geographic ODD Implications: Where Can Each Company Operate?

The sensor-level weather performance differences translate directly into geographic ODD. The table below maps US geographies against the commercial viability of each architecture.

US geography	Waymo commercial viability	Tesla Cybercab commercial viability	Climate challenge
Phoenix, AZ (desert, hot, dry)	Operational since 2017; Waymo’s most mature market; dry climate is ideal for both camera and LIDAR	Good; camera-only handles dry desert well; extreme heat within camera spec	Extreme heat; minimal rain; ideal for both
San Francisco, CA (coastal fog, mild)	Operational; SF fog is light-to-moderate typically; LIDAR plus radar handle SF fog adequately	Moderately challenging; dense SF fog on certain mornings can impair cameras; bay-area microclimates vary	Coastal fog is Waymo’s real-world validation; Tesla would struggle in dense fog episodes
Los Angeles, CA (sunny, minimal rain)	Operational; LA’s dry climate is AV-friendly	Good; LA sunny climate ideal for camera-only	Minimal weather challenge; glare is primary concern
Austin, TX (warm, some severe weather)	Operational; Austin has occasional heavy rain and hail; Waymo multi-sensor provides resilience	Moderate challenge; Austin heavy rain events would impair Tesla cameras; hail can damage sensors	Heavy rain events and occasional severe weather are real ODD challenges for camera-only
Atlanta, GA (humid, rain, occasional ice)	Expected expansion (est.); Atlanta has occasional ice and snow events; Waymo LIDAR plus radar handle rain; ice events may trigger service suspension	Challenging; Atlanta ice storms (2 to 3 per year est.) would require FSD disengagement; heavy humidity plus rain is regular	Ice and heavy rain are challenges for both; Waymo more resilient in rain
Chicago, IL (heavy snow, ice, extreme cold)	Not yet attempted commercially; heavy snow plus ice is Waymo’s hardest challenge; LIDAR severely impaired by heavy snow	Very challenging; camera-only in Chicago winter is effectively non-operational for significant portions of December through February	Heavy snow is the hard limit for all current commercial AV; no company operates here commercially
Miami, FL (heavy summer rain, humidity)	Possible (est.); Miami rain events are intense but short; LIDAR plus radar plus camera fusion handles intermittent heavy rain better	Challenging; Miami’s intense summer afternoon thunderstorms could impair cameras significantly; frequent heavy rain is a structural challenge for camera-only	Miami’s rain frequency is a real camera-only challenge
New York City (rain, snow, complex urban)	Theoretically viable for rain and moderate weather; heavy snow is the challenge; dense urban environment adds sensor occlusion	Very challenging; NYC rain plus snow plus complex urban is camera-only’s hardest combination	NYC is the hardest AV environment in the US; neither company has announced commercial plans

The Sun Belt constraint as a strategic ceiling. The vast majority of US AV commercial revenue potential today is concentrated in Sun Belt cities with mild or dry climates — where Tesla’s camera-only approach works adequately. However, the snow-belt cities of the upper Midwest and Northeast represent approximately 40 percent (est.) of the US population. For national-scale commercial AV deployment, weather ODD is not an edge case — it is the primary geographic expansion constraint for every company in the industry.

Section 5 — Weather ODD Benchmark Scorecard

Weather dimension	Waymo	Tesla FSD	Cybercab (est.)	Edge
Light rain	Good (camera slight degradation, LIDAR slight, radar compensates)	Moderate; wiper plus deceleration mitigates	Similar to FSD	Waymo
Heavy rain	Fair (LIDAR and camera impaired, radar compensates)	Significant impairment; likely disengagement	Similar to FSD	Waymo
Fog	Fair (LIDAR and radar extend range beyond camera)	Moderate to severe; fog is a fundamental limitation	Similar	Waymo
Snow	Poor (severely impaired; not commercially deployed)	Severe; unusable in heavy snow	Severe	Roughly equal (neither operating commercially in snow)
Night operation	Good (LIDAR is active illumination, works in darkness)	Moderate; dependent on headlights plus ambient	Similar	Waymo
Direct sun glare	Good (LIDAR unaffected by visible light)	Severe; known limitation with incident history	Similar	Waymo
Extreme heat	Good (Phoenix proven)	Good (cameras handle heat)	Good	Roughly equal
Geographic addressable market	Temperate and desert and coastal fog validated; rain-resilient; snow cities are the expansion challenge	Dry and mild climates optimal; rain and fog are structural limitations; snow is commercial non-starter	Similar to FSD	Waymo (broader ODD across precipitation types)
Overall verdict	Waymo’s LIDAR plus radar plus camera fusion gives it a genuine multi-weather operational advantage that camera-only systems cannot match. The key dimension is redundancy: when one sensor type is impaired, others compensate. Tesla’s camera-only approach — which works excellently in clear conditions — has fundamental limitations in fog, heavy rain, and snow that cannot be fully mitigated by software alone. The physics of sensor-environment interaction are the constraint. However, the vast majority of US AV commercial revenue is concentrated in cities with mild or dry climates (the Sun Belt), where Tesla’s camera-only approach works adequately. Snow-belt cities are the true weather ODD frontier — and neither company has solved heavy-snow commercial AV operations.

The strategic implication. Weather ODD is not primarily a technology problem today — it is a market-definition problem. The companies that solve heavy-snow commercial AV operations will expand the addressable market from roughly 60 percent (est.) of the US population to 100 percent. Until that threshold is crossed, every commercial AV company is implicitly a Sun Belt business. Waymo’s sensor architecture gives it a broader multi-weather ODD within the addressable market; Tesla’s simpler and cheaper camera architecture means that solving snow commercially will require either adding non-camera sensors or accepting a fundamentally narrower geographic ODD for Cybercab robotaxi operations.

Note: All figures labeled “(est.)” are derived from public disclosures, industry research, analyst estimates, and reported data as of mid-2026. This article does not constitute safety certification or regulatory assessment.