Waymo

Waymo is an American company that builds cars that drive themselves. Nobody sits at the wheel, and there’s no supervisor in the passenger seat either. The company grew out of a research project at Google and now belongs to the same parent group as Google, which is called Alphabet. Waymo does not sell these cars to private individuals. Instead, it uses them to run a taxi service: you order a car via a phone app, get in, and the vehicle takes you to your destination. In cities like San Francisco, Phoenix, and Los Angeles, this has been possible in regular traffic for several years now.

Why Waymo is considered the industry benchmark

Self-driving cars have been talked about for over a decade. Very many companies have tried and then given up. Waymo is among the few that actually carry paying customers without a human being able to intervene. That’s why the company often serves as a reference point in the news: it is used as a measure of how far the technology has really come.

Economically, a lot is at stake. In a classic taxi, the driver is the biggest cost factor. If that falls away, a ride could become significantly cheaper. At the same time, the cars themselves are expensive because they carry a lot of extra technology on board. Whether the business will ultimately pay off is an open question that investors debate a great deal.

On top of that comes the safety debate. Waymo regularly publishes accident figures and compares them with human drivers. Critics counter that such comparisons are difficult because the cars only operate in selected areas and in manageable weather conditions. Individual incidents, such as blocked intersections or collisions with cyclists, receive a lot of attention.

What the car sees and how it decides

A Waymo vehicle carries conspicuous structures on its roof. Inside are three types of sensors. Cameras deliver ordinary images, similar to a phone. Radar uses radio waves to measure how far away an object is and how fast it is moving. Lidar scans the surroundings with laser beams and creates a three-dimensional point cloud from this, i.e. a spatial image of the road.

These data streams come together and are processed by software based on machine learning. This means the programs were not written rule by rule by hand, but learned from huge amounts of recorded drives. They recognize cars, pedestrians, traffic lights, and lane markings. They then predict what these objects will probably do in the next few seconds. From this prediction, the system calculates a route and controls steering, acceleration, and braking.

One important point distinguishes Waymo from some competitors: the cars only drive on highly precisely surveyed maps. Every street in the operating area was previously mapped down to the centimeter. This makes the system more reliable, but also inflexible. A new city area cannot simply be unlocked via a software update. It first has to be mapped and tested, which takes months.

Waymo in the news and on the road

Waymo does not yet operate in Germany. Anyone who reads the name usually finds it in business news about Alphabet or in reports about the race for autonomous driving. Competitors include Tesla, which relies on cameras without lidar, as well as Chinese providers like Baidu. The shutdown of General Motors' Cruise robotaxi project was also big news in this context.

A common misconception: Waymo has nothing to do with the assistance systems found in ordinary new cars. A lane-keeping assistant or a highway autopilot requires that a human can take over at any time. With Waymo, there is no such human. Experts call this level 4 of automation. The difference is enormous, both legally and technically, even though both are often colloquially referred to as self-driving.