How Driverless Car Sensors Work: Lidar, Cameras, and Radar Explained

What Makes a Robotaxi Actually Drive Itself?

Strip away the marketing gloss and a robotaxi is only as capable as the driverless car sensors feeding its onboard computer. Every autonomous vehicle operating without a human driver today relies on a combination of lidar, cameras, radar, and ultrasonic sensors to perceive the world — but not every company uses the same recipe. Understanding these sensor stacks is the key to understanding why a Waymo vehicle handles downtown Phoenix differently than a Tesla operating on FSD Supervised, and why some services expand faster than others.

This guide breaks down the sensor technology powering today's commercial robotaxi services, where you can actually ride in one, what it costs, the safety record so far, and the hard limitations that still constrain the industry.

SAE Automation Levels: Where Robotaxis Actually Sit

Before comparing sensor hardware, the SAE International automation framework is essential context. Consumer vehicles like Teslas on Autopilot or Fords on BlueCruise occupy a fundamentally different category than a Waymo Jaguar I-PACE picking up passengers in San Francisco.

The Sensor Stack: Lidar, Cameras, Radar, and Ultrasonics

Level 4 robotaxis need to detect, classify, and track every object around the vehicle in real time — pedestrians, cyclists, construction cones, emergency vehicles, debris — across 360 degrees, day and night. No single sensor type does this alone. Here is what each technology contributes:

Lidar (Light Detection and Ranging)

Lidar emits millions of laser pulses per second and measures the time they take to bounce back, generating a precise 3D point cloud of the environment. Modern automotive lidar units — such as those from Waymo's in-house "Laser Bear Honeycomb" or external suppliers like Hesai and Ouster — can resolve objects at 200+ meters with centimeter-level accuracy. Lidar works identically in darkness and daylight because it provides its own illumination.

Strength: Unmatched depth perception and spatial resolution. Lidar sees the exact shape and distance of objects regardless of lighting.

Weakness: Performance degrades in heavy rain, fog, or snow as laser pulses scatter off water droplets. Units remain expensive — typically $5,000–$15,000 per vehicle for automotive-grade hardware, though costs have dropped roughly 80% since 2019.

Cameras

High-resolution cameras (typically 8–12 per vehicle in a Level 4 stack) provide color, texture, and semantic information that lidar cannot: reading traffic-light states, parsing road signs, detecting brake-light illumination on the car ahead. Neural networks process camera feeds to classify objects and predict behavior.

Strength: Rich semantic data; low hardware cost; essential for sign/signal reading.

Weakness: Passive sensors — performance drops in low light, direct sun glare, heavy rain, or when lenses are obscured. Depth estimation from cameras alone (stereo vision or monocular depth) is less precise than lidar.

Radar

Millimeter-wave radar (typically 77 GHz) excels at measuring the velocity and distance of moving objects through Doppler shift, and it penetrates rain, fog, and dust far better than lidar or cameras. Most Level 4 stacks include 4–6 radar units for long-range forward detection and corner coverage.

Strength: Robust in adverse weather; excellent velocity measurement; relatively low cost.

Weakness: Low spatial resolution — radar struggles to distinguish closely spaced stationary objects and can generate false positives from roadside clutter (guardrails, overpasses).

Ultrasonic Sensors

Short-range ultrasonic transducers (12–16 per vehicle) handle close-proximity detection below ~5 meters — critical for low-speed maneuvering, parking, and detecting curbs or bollards near the vehicle body.

Where Can You Actually Ride in a Robotaxi Right Now?

Level 4 robotaxi service is real but geographically constrained. Each operator is limited to specific cities and, within those cities, specific geofenced zones approved by local and state regulators. Here is the current landscape as of early 2026:

Where can I use this robotaxi service right now? If you are in the Phoenix metro area, San Francisco, Los Angeles, or Austin, you can download the Waymo One app, enter your pickup and drop-off points, and a driverless Jaguar I-PACE will arrive — no human behind the wheel. Service areas are geofenced, so not every address within a city is covered. In China, Baidu Apollo Go operates at far larger scale, particularly in Wuhan.

Robotaxi Pricing: How Does It Compare to Uber and Lyft?

Robotaxi pricing varies by operator, time of day, and market, but the general structure mirrors rideshare — base fare, per-mile rate, and per-minute rate, with surge pricing during high demand.

How much does it cost vs. Uber/Lyft? On a per-trip basis, Waymo One pricing is broadly comparable to UberX or Lyft — sometimes slightly higher on the base rate, but the absence of tipping often makes the total cost similar or even lower. Waymo has also run promotional pricing and loyalty programs in select markets. The long-term thesis is that removing the driver (who accounts for roughly 60–70% of rideshare cost) will eventually make robotaxis significantly cheaper, but that inflection point has not fully arrived at current fleet scales.

Safety Record: What the Data Actually Shows

How safe are robotaxis? This is the question regulators, researchers, and potential riders ask first. The data is still accumulating, but several concrete data points exist:

The data suggests that Level 4 robotaxis, in their current geofenced operating domains, are involved in fewer at-fault crashes per mile than human drivers. However, critical caveats remain: robotaxi miles are driven overwhelmingly in mapped, geofenced urban areas at moderate speeds, not on rural highways or in severe weather. Direct comparison to the full spectrum of human driving is imperfect.

Every robotaxi operator is required to report crashes to NHTSA under Standing General Order 2021-01, which mandates reporting of any crash involving an ADS-equipped vehicle within one day if it involves injury, fatality, or significant vehicle damage. This provides a more complete dataset than exists for human drivers.

Current Limitations: What Robotaxis Still Cannot Do

What are the current limitations? Despite impressive progress, Level 4 robotaxis face hard constraints that riders should understand before booking:

• Geofencing: Vehicles operate only within pre-mapped, regulator-approved zones. Request a ride outside the boundary and the app will not dispatch. Expansion requires new mapping, testing, and regulatory approval — a process measured in months, not weeks.
• Weather: Heavy rain, snow, fog, and ice degrade lidar and camera performance. Most operators suspend or limit service during severe weather. Waymo operates in light-to-moderate rain in Phoenix and San Francisco but will pull vehicles during heavy downpours. Snow-belt cities remain largely off-limits for current Level 4 services.
• Construction zones: Temporary lane shifts, missing lane markings, and flaggers with hand signals remain challenging edge cases. Vehicles may reroute or request remote human assistance (teleoperations) when encountering unmapped construction.
• Unprotected left turns: Crossing multiple lanes of oncoming traffic without a dedicated signal remains one of the most complex maneuvers. Some operators restrict routes to avoid unprotected lefts entirely.
• Emergency vehicles: While robotaxis are programmed to detect sirens and pull over, real-world incidents (including the Cruise suspension event) have shown that complex multi-vehicle emergency scenes can overwhelm perception systems.
• Pickup/drop-off precision: Vehicles pull to the nearest safe curbside stop, which may not be the exact building entrance. Riders often need to walk a short distance.
• No highway operations (most markets): Current Waymo service is predominantly surface-street. Limited highway segments are included in some markets, but long intercity highway trips are not available.
• Passenger capacity: Waymo's Jaguar I-PACE seats up to 3 passengers (rear only). No wheelchair-accessible vehicles are widely available yet, though Zoox's purpose-built vehicle includes level-floor entry designed for accessibility.

Regulatory Landscape: Who Approves Robotaxi Operations?

Robotaxi deployment in the United States is governed by a patchwork of state and local regulations — there is no single federal license to operate a driverless service nationwide.

• Federal (NHTSA): The National Highway Traffic Safety Administration sets vehicle safety standards (FMVSS) and investigates crashes. NHTSA does not currently "approve" robotaxi operations but can issue recalls, demand data, and investigate safety defects. Standing General Order requires crash reporting.
• State DMVs and PUCs: California's DMV issues testing permits and its Public Utilities Commission (CPUC) issues commercial deployment permits — a two-step process that has become the template for other states. Arizona takes a lighter-touch approach, allowing testing and commercial operation with minimal permitting, which is why Waymo launched first in Phoenix. Texas, Nevada, and other states have their own frameworks.
• Local governments: Cities can influence operations through permitting, curb-access rules, speed restrictions, and service-area negotiations. San Francisco's Board of Supervisors has at times pushed back against robotaxi expansion over concerns about traffic obstruction and emergency-response interference.
• China: Baidu Apollo Go operates under municipal permits issued by city-level authorities (e.g., Wuhan, Beijing). The regulatory environment has been more permissive for scaling, allowing Baidu to deploy over 1,000 vehicles in Wuhan alone.

Is This Truly Self-Driving or Driver-Assist?

This is the question that causes the most public confusion, and the answer depends entirely on which vehicle you are discussing.

Level 4 robotaxis (Waymo, Zoox, Baidu Apollo Go): Yes, these are truly self-driving within their operational design domain. There is no human driver in the vehicle. The system handles all driving tasks — steering, acceleration, braking, navigation, and response to traffic conditions — within its geofenced service area. If the system encounters a situation it cannot handle, it will execute a minimal-risk maneuver (pull over safely) or request remote human guidance via teleoperations. The passenger is never expected to take control.

Consumer vehicles (Tesla FSD Supervised, GM Super Cruise, Ford BlueCruise): No. These are SAE Level 2 driver-assist systems. The human driver must remain attentive, hands on or near the wheel (or eyes on the road for camera-monitored systems), and ready to intervene at any moment. The legal liability for any crash rests with the human driver, not the manufacturer. Tesla's "Full Self-Driving" name has drawn scrutiny from NHTSA and consumer advocates precisely because it suggests capabilities the system does not possess.

Frequently Asked Questions

Can a robotaxi drive me anywhere in a city?

No. Robotaxis operate within geofenced zones that cover specific neighborhoods and corridors. In Phoenix, Waymo's coverage area spans much of the metro region (over 225 square miles), but in San Francisco and Los Angeles, the service area is more limited. Check the operator's app for real-time coverage maps.

What happens if a robotaxi gets confused or encounters something unexpected?

The vehicle will typically slow down and, if necessary, pull over to a safe stop. Many operators use remote assistance (teleoperations), where a human operator reviews the vehicle's sensor data and can approve a path or provide high-level guidance. The remote operator does not "drive" the car in real time — they confirm decisions the vehicle proposes.

Are robotaxis wheelchair accessible?

Current Waymo vehicles (Jaguar I-PACE) are not wheelchair accessible. Zoox's purpose-built vehicle was designed with level-floor entry and a ramp system, but the fleet is still small and not widely deployed. Accessibility remains a significant gap in the industry and a focus of advocacy groups and regulatory attention.

How do driverless car sensors handle bad weather?

Radar penetrates rain and fog reasonably well, but lidar and cameras degrade in heavy precipitation. Most operators will suspend service during severe weather rather than risk degraded perception. This is a key reason Level 4 robotaxis have not launched in cities with harsh winters — the sensor stack's weather limitations, combined with unmapped snow-covered roads, make reliable operation impractical with current technology.

Why doesn't Tesla operate a robotaxi service?

Tesla's FSD system relies on a camera-only sensor stack (no lidar, no radar in current vehicles) and is classified as SAE Level 2 — requiring driver supervision. As of early 2026, Tesla has not received regulatory approval to operate a driverless Level 4 service in any U.S. market. Tesla has announced plans for a "Robotaxi" platform (unveiled in late 2024), but no commercial driverless service is live. The camera-only approach has not yet demonstrated the redundancy and reliability that regulators require for driverless operation.

Who is liable if a robotaxi crashes?

In a Level 4 robotaxi, the operator (e.g., Waymo) assumes liability for crashes caused by the autonomous system — a fundamental difference from Level 2 driver-assist, where the human driver bears responsibility. Riders are not expected to carry special insurance for robotaxi trips. Operators carry commercial liability insurance and self-insure against system-caused incidents.