Robotaxi
Also known as: robo-taxi, driverless taxi, autonomous taxi, self-driving taxi
A robotaxi is a driverless vehicle that carries paying passengers on demand, usually booked through an app like a ride-hailing service.
They operate at SAE Level 4 within mapped service areas, with remote support staff for unusual situations. While Waymo now gives hundreds of thousands of paid rides a week in several US cities, robotaxis still struggle with expansion, public trust, and the cost of sensors and mapping.
Robotaxi, explained at three levels
At home
Think of an elevator: you press a button, it arrives with no operator, and it takes you where you choose within the building. A robotaxi works the same way across part of a city. You book it in an app, a driverless car arrives, and its computers steer, brake, and navigate using sensors and detailed maps, but only inside the area its company has mapped and approved.
At work
A robotaxi is an SAE Level 4 vehicle offering paid rides with no one in the driver's seat, inside a defined service area. The business case is removing the driver, the largest cost of ride-hailing, and running vehicles around the clock. Companies such as Waymo, Baidu Apollo Go, and Zoox run fleets supported by remote assistance staff, depots, and city permits.
At home
A robotaxi stacks several systems. Lidar, radar, and cameras build a 3D view of everything nearby. Software compares that view with a high-definition map to know exactly where the car is, down to centimetres. Prediction models guess what each pedestrian and car will do next, and a planner picks a safe path. When the car is unsure, such as at a blocked road, it asks a remote human for guidance.
At work
Running a robotaxi service is as much operations as software. Fleets need high-definition maps kept current, depots for charging, cleaning, and sensor calibration, and remote assistance teams who advise stuck vehicles rather than drive them. Each city requires permits and often data reporting. Economics depend on vehicle and sensor cost, rides per vehicle per day, and how many support staff each car needs. Waymo runs about half a million paid rides a week.
At home
Waymo publishes data showing far fewer injury crashes than human drivers on the same roads, over many millions of driverless kilometres. The catch is that robotaxis drive only in chosen areas and conditions, so comparisons need care, and rare failures, such as blocking emergency vehicles, still happen. In my opinion, the safety data is encouraging, but it deserves independent auditing before anyone assumes it holds in a Canadian winter.
At work
Most robotaxis use modular software stacks for perception, prediction, and planning, though Tesla bets on end-to-end neural networks using cameras only. Limits include the operational design domain, heavy snow that hides lane markings and coats sensors, and edge cases like police hand signals. Cruise's 2023 incident showed how one failure, plus poor disclosure, can halt a program. In Canada, provinces and cities would set permits. I'd plan curb space and data-sharing rules early.
Real-world examples
At home
Aisha visited Phoenix and took a robotaxi from her hotel to a restaurant. She found it strange at first to see an empty driver's seat, but the ride was smooth and cautious.
At work
Marcus is a transportation planner for an Ontario city. He studies robotaxi deployments in US cities to plan curb space, permits, and data-sharing rules before any service arrives locally.
How it compares
| feature | Robotaxi | Ride-hailing car |
|---|---|---|
| Driver | None, with remote support | Human driver |
| Service area | Limited mapped zones | Almost anywhere |
| Availability | Few cities | Widespread |
| Main cost | Vehicles, sensors, mapping | Driver pay |
Robotaxis remove the driver but trade it for a narrow service area.
For analysts
For cities and businesses, robotaxis are worth tracking but not counting on yet. I'd plan around pilots and permits rather than assuming quick, broad availability in Canada.
History & origin
The DARPA Grand Challenges of 2004 to 2007 kick-started modern self-driving research. Google began its self-driving project in 2009, which became Waymo in 2016. Waymo launched a public driverless service in Phoenix in 2020. Cruise, backed by GM, began driverless rides in San Francisco in 2022 but had its California permit suspended in 2023 after a pedestrian was dragged, and GM ended Cruise's robotaxi effort in December 2024. Baidu's Apollo Go expanded quickly in Chinese cities. By 2026, Waymo was reporting around half a million paid rides per week across more than 10 US cities, while Tesla ran a smaller service starting in Austin in 2025. In my view, robotaxis have finally proven they can work, but proving they can work cheaply everywhere is a different and longer race.