Every robotaxi on the road today relies on a layered array of autonomous driving sensors to perceive its environment, predict hazards, and navigate without a human behind the wheel. But not all sensor suites are created equal. Waymo's vehicles bristle with lidar, radar, and cameras; Tesla's approach (where permitted) leans entirely on vision; and newer entrants like Zoox and Cruise (now restructured under GM) have charted their own paths. For anyone evaluating whether to step into a driverless vehicle — or simply wondering how these cars actually see the world — understanding the sensor hardware is the fastest route to an informed decision.
This guide breaks down the autonomous driving sensors powering commercial robotaxi fleets in 2026, maps where you can actually ride in one today, and examines the safety data behind the technology.
SAE Automation Levels: What "Self-Driving" Actually Means
Before dissecting sensors, it's critical to establish what level of autonomy we're discussing. The SAE International J3016 standard defines six levels of driving automation, and conflating them is the single most common source of public confusion.
| Level | Name | Who Drives | Examples (2026) |
|---|---|---|---|
| 0 | No Automation | Human | Legacy vehicles |
| 1 | Driver Assistance | Human (system assists steering OR speed) | Adaptive cruise control, lane-centering |
| 2 | Partial Automation | Human (must supervise at all times) | Tesla Autopilot/FSD Supervised, GM Super Cruise, Ford BlueCruise |
| 3 | Conditional Automation | System (in limited ODD; human takes over on request) | Mercedes DRIVE PILOT (limited markets, ≤40 mph traffic jams) |
| 4 | High Automation | System (no human needed within geofenced ODD) | Waymo One, Zoox (commercial service areas) |
| 5 | Full Automation | System (everywhere, all conditions) | Does not exist commercially |
Key distinction: Consumer systems like Tesla FSD Supervised, GM Super Cruise, and Ford BlueCruise are Level 2. The driver must remain engaged and is legally responsible. Waymo One and Zoox operate at Level 4 — no driver present, no steering wheel, no human fallback — but only within tightly geofenced operational design domains (ODDs). Level 5 autonomy does not exist in any commercial product as of 2026.
The Sensor Stack: How Robotaxis Perceive the World
Autonomous driving sensors fall into three primary categories, each with distinct physics, strengths, and blind spots. Commercial robotaxi operators combine them differently depending on their engineering philosophy.
Lidar (Light Detection and Ranging)
Lidar sensors emit millions of laser pulses per second and measure the time it takes for reflections to return, building a precise 3D point cloud of the surrounding environment. Modern automotive lidar units (from suppliers like Hesai, Innovusion, and Waymo's in-house Gen 5 system) achieve ranges of 200–300 meters with angular resolution fine enough to distinguish a pedestrian from a mailbox at 100+ meters.
Strengths: Centimeter-level depth accuracy; works in total darkness; unaffected by sun glare or shadow contrast.
Weaknesses: Performance degrades in heavy rain, fog, and snow (particulate scatter); high unit cost (though dropping — sub-$500 per unit for some solid-state designs in 2026 vs. $75,000+ for early spinning-top units); requires cleaning systems for road grime.
Cameras (Vision)
High-resolution cameras (typically 5–12 megapixels, running at 30–60 fps) provide color, texture, and semantic information that lidar cannot: reading traffic-light states, parsing road signs, detecting brake-light illumination on a car ahead. Tesla's vision-only architecture relies on 8 surround cameras; Waymo uses 29+ cameras as part of its multi-sensor fusion stack.
Strengths: Rich semantic data (text, color, signage); low cost per unit; compact form factor.
Weaknesses: Depth estimation is inferred, not measured (computationally intensive); performance drops in low light, direct sun glare, heavy rain, and fog; requires massive labeled training datasets.
Radar (Radio Detection and Ranging)
Millimeter-wave radar (77 GHz band) directly measures the distance and relative velocity of objects via the Doppler effect. It is the most weather-resilient sensor, functioning reliably through rain, fog, and snow that would blind lidar and cameras.
Strengths: Direct velocity measurement; excellent all-weather performance; low cost.
Weaknesses: Low spatial resolution (struggles to classify small or stationary objects); multipath reflections can produce ghost targets; limited vertical resolution.
| Operator | Lidar | Cameras | Radar | Other | Philosophy |
|---|---|---|---|---|---|
| Waymo (6th-gen) | 5+ units (custom) | 29+ | 6 | External audio receivers (sirens) | Redundant multi-sensor fusion |
| Zoox | 4 corner-mounted | 12+ | 4+ | Microphones, IMU | Purpose-built bidirectional vehicle |
| Cruise (GM, restructured) | 5+ (Hesai/Pandar) | 16+ | 6+ | Ultrasonic | Multi-sensor; paused ops since late 2023, limited 2025–26 restart |
| Tesla (FSD Supervised) | None (removed in 2022) | 8 | None (removed in 2022) | Ultrasonic (removed) | Vision-only neural network |
Note: Tesla's FSD Supervised is a Level 2 driver-assist system, not a robotaxi. It is included here for comparison because Tesla's camera-only sensor philosophy is the most prominent alternative to the lidar-inclusive approach used by every commercial Level 4 operator.
Where Can You Actually Ride in a Robotaxi Right Now?
As of early 2026, fully driverless (Level 4) commercial robotaxi services operate in a limited number of U.S. cities. Expansion is incremental and heavily regulated — announced plans do not equal live service.
| Operator | City | Service Type | Hours | Notes |
|---|---|---|---|---|
| Waymo One | Phoenix, AZ (metro) | Fully driverless | 24/7 | Largest continuous service area; ~250+ sq mi |
| Waymo One | San Francisco, CA | Fully driverless | 24/7 | Citywide access since mid-2024 |
| Waymo One | Los Angeles, CA | Fully driverless | Limited hours, expanding | Initial rollout 2024; coverage expanding |
| Waymo One | Austin, TX | Fully driverless | Limited | Launched 2025; geofenced zones |
| Zoox | Las Vegas, NV (Strip corridor) | Fully driverless | Limited hours | Purpose-built vehicle; short-route service |
| Zoox | Foster City, CA | Testing/limited | Invite-only | Not open commercial service |
Important context: Cruise, once the second-largest U.S. robotaxi operator (San Francisco), suspended all driverless operations in October 2023 after a serious incident in which a pedestrian was struck by a human-driven vehicle and then run over by a Cruise robotaxi that failed to properly respond. GM restructured Cruise through 2024–2025, and limited supervised testing resumed in select cities, but no open commercial driverless service has been restored as of early 2026.
How to Use a Robotaxi and What It Costs
Using Waymo One — the most widely available service — follows a familiar rideshare pattern:
- Download the Waymo One app (iOS/Android) and create an account.
- Enter your pickup and destination within the service area. The app shows whether your trip falls within the current geofence.
- Confirm and wait. Average wait times in Phoenix and San Francisco run 3–7 minutes during peak hours.
- Verify the vehicle. The app displays the car's make, model (Jaguar I-PACE or Zeekr for Waymo's 6th-gen), license plate, and a PIN you enter on the vehicle's exterior screen to unlock.
- Ride. No driver is present. An in-app "Help" button connects you to a remote support agent. The vehicle will not move until all passengers are buckled.
- Exit at your destination. Payment is automatic via the payment method on file.
| Trip Type | Waymo One | UberX | Lyft Standard |
|---|---|---|---|
| Short (~3 mi, city) | $8–$13 | $9–$15 | $8–$14 |
| Medium (~8 mi) | $16–$25 | $17–$28 | $16–$26 |
| Long (~15 mi, airport) | $30–$48 | $32–$55 | $30–$50 |
| Surge/Peak pricing | Moderate (1.2–1.5x) | High (1.5–3x+) | High (1.5–3x+) |
Pricing is broadly competitive with UberX and Lyft Standard for comparable trips. Waymo applies surge pricing during peak demand but typically at lower multipliers than human-driven rideshare. Tipping is not expected (no driver). Waymo periodically offers promotional credits for new users.
How Safe Are Robotaxis? The Real Data
Safety is the most scrutinized — and most contested — dimension of robotaxi operations. Here is what the available data shows.
Waymo Safety Data (through 2025):
- Total driverless miles: Over 30 million fully autonomous (no driver) miles as of late 2025, per Waymo's published safety reports.
- Injury-causing crash rate: Waymo's driverless vehicles showed an 85% reduction in injury-causing crash rates compared to human benchmarks in the same operating areas, based on a study published in partnership with the National Highway Traffic Safety Administration (NHTSA) data and independent researchers.
- Contact rates: Police-contactable incidents per million miles are significantly lower than the human baseline, though the comparison methodology is debated by some researchers.
Cruise incident (October 2023): The most significant robotaxi safety incident to date occurred in San Francisco when a pedestrian, already struck by a human-driven vehicle, was thrown into the path of a Cruise robotaxi. The vehicle braked but could not avoid running over the pedestrian, then dragged her approximately 20 feet while attempting to pull over. The California DMV suspended Cruise's permit, citing the company's failure to fully disclose details of the incident to regulators. This event led to Cruise's full operational shutdown and eventual restructuring under GM.
Context and caveats:
- Robotaxi miles are a tiny fraction of total U.S. vehicle miles traveled (~3.2 trillion annually). Statistical significance is still limited.
- Robotaxis operate in geofenced, mapped areas — often avoiding the most complex road scenarios (unpaved roads, construction zones, extreme weather).
- Minor contact incidents (e.g., low-speed bumping of stationary objects, cautious stops disrupting traffic flow) are more common than injury events and are sometimes underreported in aggregate safety metrics.
- For real-time safety data, the NHTSA Standing General Order requires all ADS-equipped vehicles to report crashes within 24 hours.
Current Limitations: Where Robotaxis Cannot Go
Level 4 autonomy means the system handles all driving tasks — but only within its Operational Design Domain (ODD). Step outside the ODD, and the vehicle cannot operate. Here are the practical limitations riders encounter in 2026:
- Geofencing: Service is restricted to pre-mapped zones. In Phoenix, Waymo covers ~250 square miles of the metro area — not the entire city. In San Francisco, coverage is near-citywide but still bounded. Request a trip outside the geofence and the app will refuse.
- Weather: Heavy rain, dense fog, and snow significantly degrade lidar and camera performance. Waymo will limit or suspend service during severe weather events. Phoenix's dry climate is one reason it was the first deployment city.
- Construction zones: Temporary lane shifts, missing lane markings, and flaggers directing traffic represent edge cases that may cause the vehicle to stop and request remote assistance, adding delay.
- Unprotected left turns: Early robotaxi generations struggled with unprotected left turns across oncoming traffic. While vastly improved in 6th-generation systems, these remain high-complexity scenarios that occasionally trigger conservative stops.
- Emergency vehicles: Robotaxis are programmed to pull over for approaching emergency vehicles, but the detection and response (especially for sirens at oblique angles or in dense urban noise) is not always instantaneous. Waymo added external audio receivers on its 6th-gen vehicles specifically to address this.
- Pickup/dropoff precision: The vehicle will not stop in travel lanes or unsafe locations. In dense urban areas, the nearest safe pull-over point may be a block or more from your exact requested address.
- No highway service (Zoox Las Vegas): Zoox's current Las Vegas service operates on a short, fixed corridor — not freeway routes.
The Regulatory Landscape: Who Approves Robotaxi Operations?
Robotaxi deployment in the United States is governed by a patchwork of state and federal authority:
- Federal (NHTSA): Sets vehicle safety standards (FMVSS) and investigates safety defects. NHTSA does not "approve" ADS software but can issue recalls and investigate incidents. The Standing General Order mandates crash reporting for all ADS-equipped vehicles.
- State DMVs/Regulators: Most states that permit driverless testing and commercial operations require permits issued by the state DMV or equivalent. California's CPUC (Public Utilities Commission) and DMV both play roles — the DMV issues testing/deployment permits, while the CPUC regulates commercial passenger service pricing and operations.
- Local governments: Cities can impose additional requirements or restrictions. San Francisco's Board of Supervisors has held multiple hearings pushing back on robotaxi expansion citing traffic disruption concerns.
- Key states with active frameworks: Arizona (permissive, no specific ADS legislation — Waymo's home base), California (structured permit system), Texas (statewide ADS law permitting driverless operation), Nevada (Zoox's Las Vegas operations).
There is no single federal "robotaxi license." An operator must navigate state permits, local agreements, and ongoing regulatory scrutiny in each deployment city.
Frequently Asked Questions
Is a robotaxi truly self-driving, or is there a remote human controlling it?
Level 4 robotaxis like Waymo and Zoox are genuinely self-driving within their geofenced service areas — no human driver is in the vehicle, and no remote operator is "driving" the car via joystick. Remote fleet support agents can provide routing guidance or authorize the vehicle to proceed past an obstacle (a process called "remote assistance"), but they do not steer, brake, or accelerate the vehicle in real time. The ADS handles all dynamic driving tasks.
How much does a robotaxi cost compared to Uber or Lyft?
Pricing is broadly comparable to UberX and Lyft Standard — typically $8–$13 for a short city trip, $16–$25 for medium distances. Waymo's surge multipliers tend to be lower than human rideshare during peak demand. There is no tipping. Promotional credits are frequently available for new users.
What happens if the robotaxi encounters a situation it can't handle?
The vehicle will come to a safe stop (typically pulling to the curb or stopping in-lane with hazards on) and request remote assistance from a fleet operations center. A remote agent will assess the situation and either provide routing guidance to the ADS or dispatch roadside support. Passengers can contact support at any time via the in-app help button or the in-vehicle screen.
Can I use a robotaxi in any city?
No. As of 2026, fully driverless commercial service is available only in Phoenix, San Francisco, Los Angeles (Waymo), Austin (Waymo, limited), and Las Vegas (Zoox, limited corridor). Expansion is gradual and requires regulatory approval in each new market. Announced plans for additional cities should not be treated as live service.
Are autonomous driving sensors affected by bad weather?
Yes. Heavy rain, fog, and snow scatter lidar laser pulses and obscure camera lenses, degrading perception. Radar is the most weather-resilient sensor, which is why multi-sensor stacks (lidar + camera + radar) provide redundancy that camera-only systems lack. Robotaxi operators will restrict or suspend service during severe weather to stay within their validated ODD.
Why doesn't Tesla use lidar for FSD?
Tesla's engineering philosophy holds that a vision-only system (cameras + neural networks) can achieve sufficient perception, arguing that since humans drive with vision alone, cameras should suffice. This approach eliminates lidar cost and complexity but sacrifices the direct, high-precision depth measurement that lidar provides. Tesla's FSD remains a Level 2 system requiring driver supervision — it is not a driverless robotaxi.
Understanding autonomous driving sensors is essential context for anyone evaluating robotaxi services, investing in autonomous vehicle technology, or simply deciding whether to hail a driverless ride. The hardware — lidar, cameras, radar, and the software that fuses their data — defines what these vehicles can and cannot do. In 2026, the technology is real, the safety data is encouraging but still limited in scale, and the service footprint is growing city by city under close regulatory oversight.
