The Short Answer: 1925, but Not How You Think
If you are searching for when the first driverless car was made, the answer depends on what you mean by "driverless." The earliest publicly demonstrated remote-controlled automobile was the Linrrican Wonder (also called the "American Wonder"), a modified Chandler sedan guided by radio signals from a following vehicle. Engineer Francis Houdina demonstrated it through the streets of New York City on August 21, 1925, with no one behind the wheel. It was driverless in the literal sense — no human operated the controls — but it required a human operator in a trailing car sending steering and throttle commands by radio.
That distinction matters in 2026. Today's robotaxis from Waymo, Zoox, and Baidu's Apollo Go are not remote-controlled. They are SAE Level 4 autonomous vehicles: the onboard computer handles all driving tasks within a defined operational domain (geofenced city areas, specific weather conditions) with zero human intervention and zero remote operator steering the car in real time. The journey from radio-controlled novelty to AI-driven robotaxi took a century, and the technology stacks are unrecognizably different.
| Level | Name | Who Drives | Example |
|---|---|---|---|
| Level 0 | No Automation | Human does everything | Any conventional car |
| Level 1 | Driver Assistance | Human + one system (steering OR speed) | Adaptive cruise control |
| Level 2 | Partial Automation | Human supervises; system handles steering AND speed | Tesla Autopilot, GM Super Cruise, Ford BlueCruise |
| Level 3 | Conditional Automation | System drives in narrow conditions; human must take over on request | Mercedes Drive Pilot (limited markets, ≤40 mph) |
| Level 4 | High Automation | System drives fully within geofenced area; no human needed | Waymo One, Zoox, Baidu Apollo Go |
| Level 5 | Full Automation | System drives everywhere, all conditions | Does not exist in production (2026) |
Consumer vehicles with Autopilot, FSD (Supervised), or BlueCruise are Level 2 — the driver must remain engaged at all times. Robotaxis are Level 4.
A Century of Driverless Milestones
The road from 1925 to 2026 is marked by a handful of pivotal experiments, each pushing the boundary of what "driverless" meant:
- 1925 — Linrrican Wonder: Radio-controlled Chandler sedan driven through NYC traffic by Francis Houdina from a trailing car. First public demonstration of a car moving without a driver at the wheel.
- 1939 — Norman Bel Geddes / GM Futurama: At the New York World's Fair, the Futurama ride envisioned automated cars guided by embedded roadway electromagnetic fields — a concept of infrastructure-assisted autonomy, not onboard intelligence.
- 1958 — RCA / Princeton: Engineers demonstrated a car following buried roadway cables using sensors, achieving ~60 mph on a test track with no driver input. Still infrastructure-dependent.
- 1980s — Ernst Dickmanns (Bundeswehr University Munich): The VaMoRs (Versuchsfahrzeug für autonome Mobilität und Rechnersehen) project used computer vision and real-time image processing to guide a Mercedes van autonomously on closed roads at up to 60 mph. This was the first serious onboard-computer autonomy.
- 1995 — CMU Navlab "No Hands Across America": A semi-autonomous Pontiac Trans Sport drove 2,800+ miles from Pittsburgh to San Diego with the computer steering for ~98% of the trip (a human handled throttle and brakes). Used neural-network-based vision.
- 2004–2007 — DARPA Grand Challenges & Urban Challenge: The 2005 Grand Challenge was won by Stanford's "Stanley" (a modified VW Touareg). The 2007 Urban Challenge required vehicles to navigate city traffic, obey rules, and interact with other cars. Carnegie Mellon's "Boss" (Chevy Tahoe) won. These competitions directly spawned Waymo (from Google's self-driving car project, launched 2009 by Sebastian Thrun, who led the Stanford team).
- 2018 — Waymo One launches in Phoenix, AZ: The first commercial robotaxi service open to paying passengers in the United States. Level 4, operating within a geofenced portion of the Phoenix metro area.
- 2022–2024 — Cruise and Waymo expand: Cruise (GM-backed) operated in San Francisco, Phoenix, and Austin before a serious pedestrian incident in October 2023 led to its California permit suspension and eventual restructuring. Waymo continued expanding in Phoenix, San Francisco, and Los Angeles.
- 2025–2026 — Waymo, Zoox, and international operators scale: Waymo operates commercially in Phoenix, San Francisco, Los Angeles, and Austin. Amazon's Zoox operates its purpose-built bidirectional vehicle in Las Vegas and Foster City, CA. Baidu's Apollo Go completes millions of paid rides across multiple Chinese cities including Wuhan and Beijing.
Where Can I Use a Robotaxi Service Right Now?
As of early 2026, fully driverless (Level 4) robotaxi services available to the public in the United States are limited to specific geofenced areas. You cannot hail one everywhere in a city — they operate within defined boundaries, and the vehicle will refuse a ride that starts or ends outside the zone.
| Operator | City / Metro Area | Status | Notes |
|---|---|---|---|
| Waymo One | Phoenix, AZ | Fully operational, 24/7 | Largest U.S. service area; covers much of the East Valley |
| Waymo One | San Francisco, CA | Fully operational, 24/7 | Covers most of the city proper |
| Waymo One | Los Angeles, CA | Operational, expanding zones | Santa Monica, Hollywood, downtown core |
| Waymo One | Austin, TX | Operational | Central Austin geofence |
| Zoox | Las Vegas, NV | Operational (limited invite/public) | Purpose-built vehicle, no steering wheel |
| Zoox | Foster City, CA | Testing / limited rides | Pre-commercial service |
| Baidu Apollo Go | Wuhan, Beijing, Shenzhen, Chongqing (China) | Fully operational, millions of rides | Largest global robotaxi fleet; not available in U.S. |
How to ride: For Waymo One, download the Waymo app (iOS/Android), enter your pickup and drop-off within the service area, and a vehicle is dispatched — typically within 3–8 minutes. You unlock the car by tapping the app or entering a PIN on the rear-door screen. There is no driver, no steering wheel accessible to passengers, and no remote operator controlling the vehicle. Zoox rides are requested through a separate app and are currently available to invited users and select public riders.
How Much Does a Robotaxi Cost vs. Uber or Lyft?
Robotaxi pricing in 2026 is designed to be competitive with traditional rideshare — not dramatically cheaper, but comparable. Waymo One uses a pricing model similar to Uber/Lyft: base fare + per-mile + per-minute, with surge pricing during high demand. There is no tip expected (there is no driver to tip).
| Service | Estimated Fare | Tip Expected | Wait Time | Availability |
|---|---|---|---|---|
| Waymo One | $10–$18 | No | 3–8 min | Geofenced zones only |
| UberX | $9–$16 | Yes (15–20%) | 2–10 min | Widespread |
| Lyft | $9–$15 | Yes (15–20%) | 2–10 min | Widespread |
| Baidu Apollo Go (China) | $3–$7 (subsidized) | No | 2–5 min | Major Chinese cities |
When you factor in the absent tip, a Waymo ride often costs comparable to or slightly less than an UberX ride of the same distance once gratuity is added. Baidu's Apollo Go in China is heavily subsidized to build market share and is significantly cheaper, but that pricing is not sustainable long-term and does not reflect U.S. market economics.
How Safe Are Robotaxis? The Real Data
- Waymo: Over 40 million fully driverless miles completed by mid-2025 across Phoenix, San Francisco, and Los Angeles. Waymo has published peer-reviewed analysis showing its vehicles are involved in ~85% fewer injury-causing crashes per million miles compared to human benchmarks in the same cities, and ~57% fewer police-reported crashes. (Waymo Safety Report)
- Cruise (pre-restructuring): Completed ~4.5 million driverless miles before the October 2023 San Francisco incident in which a Cruise vehicle struck and then dragged a pedestrian who had been thrown into its path by another vehicle. California DMV suspended Cruise's permit; the company halted nationwide operations and later restructured under GM.
- Baidu Apollo Go: Reported over 100 million cumulative autonomous miles (including safety-driver and fully driverless) by 2025 across Chinese cities, with no reported passenger fatalities.
- NHTSA Standing General Order: The National Highway Traffic Safety Administration requires all Level 2+ and Level 4 operators to report crashes involving their automated systems. Data is published in periodic reports and shows that the vast majority of reported incidents involving Level 2 systems (Tesla Autopilot, GM Super Cruise) occur in highway scenarios, while Level 4 robotaxi incidents are predominantly low-speed urban contacts.
The honest assessment: Level 4 robotaxis have demonstrated a strong safety record in their geofenced urban environments, particularly at the speeds and road types they are approved for (typically under 45 mph, mapped streets). However, the total miles driven are still a fraction of the billions of miles human drivers log annually, and edge cases — construction zones, unusual road debris, erratic pedestrian behavior, severe weather — remain the frontier of autonomous driving challenges.
The Cruise incident of 2023 was a critical inflection point. It demonstrated that a robotaxi could encounter a complex multi-actor scenario (a pedestrian thrown into the vehicle's path by a separate human-driven car) and fail to respond appropriately — specifically, the vehicle stopped but then attempted to pull over, dragging the pedestrian. This led to heightened regulatory scrutiny and underscored that Level 4 autonomy is robust but not infallible.
Technology Stack: How Robotaxis Actually See the Road
Level 4 robotaxis rely on a layered sensor and compute architecture that differs fundamentally from consumer Level 2 driver-assist systems. Here is the typical stack for a current-generation Waymo vehicle (based on the Jaguar I-PACE platform with the 6th-generation Waymo Driver hardware):
- LiDAR: Multiple spinning and solid-state LiDAR units providing a 360° 3D point cloud of the environment, effective to ~500 meters. LiDAR works in darkness and is not dependent on visible light.
- Cameras: High-resolution cameras (visible spectrum) for lane markings, traffic signal state, signage text, and object classification. Waymo uses a multi-camera array covering 360°.
- Radar: Millimeter-wave radar for velocity detection of moving objects (other vehicles, cyclists) and performance in rain/fog where LiDAR and cameras degrade.
- Audio sensors: External microphones to detect emergency vehicle sirens and localize their direction.
- HD Maps: Centimeter-precision pre-mapped routes within the geofence. The vehicle localizes itself within the map in real time. This is a key limitation — the vehicle cannot operate in unmapped areas.
- Compute: Custom-designed onboard computer (Waymo's own silicon) processing sensor fusion, prediction models (what other road users will do next), and planning (trajectory generation) in real time.
Camera-only vs. LiDAR: Tesla's FSD (Supervised) — a Level 2 system — uses a camera-only approach (Tesla Vision), arguing that neural networks can extract sufficient 3D understanding from cameras alone. Every major Level 4 robotaxi operator (Waymo, Zoox, Baidu, Cruise) uses LiDAR as a core sensor. The industry consensus at Level 4 is that sensor redundancy — LiDAR + cameras + radar — is necessary for the safety case required to remove the human driver entirely.
Geofencing: Level 4 robotaxis operate within a pre-mapped, pre-approved operational design domain (ODD). The geofence is not just a GPS boundary — it includes mapped lane geometry, traffic signal locations, speed limits, and known construction zones. If a ride request would take the vehicle outside the ODD, the system declines the trip.
Regulatory Landscape: Who Approves Robotaxi Operations?
Robotaxi deployment in the United States is governed by a patchwork of state and local regulations, not a single federal framework:
- State DMVs and Public Utilities Commissions: In California, the California DMV issues autonomous vehicle testing and deployment permits, while the California Public Utilities Commission (CPUC) regulates commercial passenger service. Other states have their own agencies — Arizona has taken a permissive approach (Waymo's Phoenix operations began with minimal state friction), while Texas requires local coordination.
- NHTSA (Federal): The National Highway Traffic Safety Administration sets Federal Motor Vehicle Safety Standards (FMVSS). Current FMVSS assumes a human driver (steering wheel, pedals, mirrors). Vehicles like the Zoox purpose-built robotaxi (no steering wheel, no traditional controls) required NHTSA exemptions or a petition process to be road-legal. NHTSA also operates the Standing General Order requiring crash reporting from ADS-equipped vehicles.
- Local governments: Cities control road access, curb management, and can impose restrictions. San Francisco saw significant political tension between robotaxi operators and city officials over traffic obstruction incidents (vehicles stopping in travel lanes) before the CPUC voted to allow full 24/7 operations in 2023.
- International: China's Ministry of Transport and local municipal governments have issued permits for fully driverless robotaxi operations in cities including Wuhan, Beijing, Shenzhen, and Chongqing. Baidu's Apollo Go operates under these frameworks with significantly larger fleet sizes than any U.S. operator.
Current Limitations: What Robotaxis Cannot Do
- Weather: Heavy rain, snow, fog, and ice degrade LiDAR and camera performance. Waymo vehicles in Phoenix and LA benefit from predominantly clear weather. Operations in cities with frequent snow (Chicago, Boston, Minneapolis) remain limited or unavailable.
- Geofence boundaries: The vehicle cannot drive outside its mapped ODD. If your destination is 200 feet outside the zone, the ride is declined.
- Construction zones: While Waymo has improved handling of temporary lane shifts and construction barrels, complex or rapidly changing construction zones can cause the vehicle to stop and wait (sometimes blocking traffic) until a remote support team provides guidance — not remote driving, but high-level route advice.
- Unprotected left turns: Across multiple lanes of oncoming traffic at uncontrolled intersections remain a challenging scenario, and robotaxis tend to be conservative — sometimes excessively so, causing delays.
- Emergency vehicles: While audio sensors detect sirens, complex multi-emergency-vehicle scenarios (e.g., two fire trucks approaching from different directions) can cause the vehicle to stop in place.
- Highway driving: Most current U.S. robotaxi operations are limited to surface streets under ~45–55 mph. Highway-capable robotaxi service is limited, though Waymo has tested highway segments in Phoenix.
- Passenger behavior: No driver means no one to enforce seatbelt use, prevent vandalism, or handle medical emergencies in the vehicle. Interior cameras monitor the cabin, but intervention options are limited to stopping the vehicle and contacting emergency services.
Frequently Asked Questions
Is Tesla's FSD a robotaxi?
No. Tesla Full Self-Driving (Supervised) is an SAE Level 2 driver-assist system. The driver must remain attentive, keep hands available, and is legally responsible for the vehicle at all times. Tesla has discussed a future robotaxi network but has not launched a Level 4 driverless service as of early 2026. Conflating FSD with Waymo or Zoox robotaxis is one of the most common misunderstandings in autonomous driving.
When was the first driverless car made that could drive itself without remote control?
The first vehicle to drive itself using onboard computer intelligence (not remote control or embedded roadway infrastructure) is generally attributed to Ernst Dickmanns' VaMoRs project in the mid-1980s at Bundeswehr University Munich, which used computer vision and real-time processing to autonomously steer a Mercedes van on closed roads. The 1925 Linrrican Wonder was driverless but remotely controlled. The 1958 RCA experiments required embedded cables in the road. True onboard autonomy — the kind that underpins today's robotaxis — began in the 1980s.
Can a robotaxi take me to the airport?
It depends on whether the airport is within the geofence. Waymo One in Phoenix serves Phoenix Sky Harbor International Airport (PHX). In San Francisco, Waymo serves SFO. In Los Angeles, LAX service availability depends on the current geofence boundary. Check the app — if your destination is outside the ODD, the ride will be declined.
What happens if the robotaxi gets confused or stuck?
The vehicle is designed to stop safely (pull over or stop in lane with hazard lights) and contact a remote support team. The remote team does not drive the vehicle remotely — they provide high-level guidance such as confirming it is safe to proceed around an obstacle or authorizing a route adjustment. If the issue cannot be resolved, the ride is ended and the passenger may be offered a credit or alternative transport.
How many robotaxis are on the road in 2026?
Exact fleet sizes are not always publicly disclosed. Waymo is estimated to operate a fleet of 1,000+ vehicles across its U.S. cities. Baidu's Apollo Go fleet in China is estimated at several thousand vehicles across multiple cities. Zoox's fleet is smaller, numbering in the hundreds. These numbers are growing but remain tiny compared to the millions of rideshare drivers on U.S. roads.
Will robotaxis replace Uber and Lyft drivers?
Not in the near term. Robotaxis are geographically limited, weather-constrained, and capital-intensive (each Waymo vehicle is estimated to cost $150,000–$200,000+ in sensor and compute hardware). Uber and Lyft have partnered with autonomous operators (Uber has a partnership with Waymo in Phoenix and Austin, where Waymo vehicles appear in the Uber app) rather than being replaced by them. The likely medium-term model is a hybrid: robotaxis in dense, mapped, fair-weather urban cores; human drivers everywhere else.
