Skip to Content

Why Google’s Self-Driving Bubble Cars Might Catch On

The compact, speed-limited vehicles being tested by Google might have a better chance at success than automated versions of conventional cars.

Google’s announcement Friday that it will test small, pod-style autonomous cars on public roads might seem surprising to anyone enthusiastic about—or just familiar with–conventional cars. The vehicles look cute but hardly impact-resistant, and they have a top speed of only 25 miles per hour.

Cars like this one will start driving themselves around Mountain View, California, this summer.

But some experts suspect that the unconventional two-seater vehicle, known within Google as “Prototype,” represents a practical strategy to get fully autonomous cars into everyday use. Google still has significant work to do before its software can handle all the situations a human driver can. But it will be easier to build, test, and market small vehicles for limited environments than to craft autonomous cars that can handle everything from high-speed freeway driving to city streets, they say.

“There’s going to be an enormous market for small autonomous vehicles,” says Gary Silberg, an auto industry analyst at the consulting firm KPMG. He cites city centers, airports, campuses, and amusement parks as places where vehicles much like those Google is just starting to test could fit in. “From a market perspective, it’s a huge opportunity,” he says.

Google first unveiled its compact car design last year, in what seemed like a change in strategy from its effort to make conventional cars that were capable of driving themselves (see “Lazy Humans Shaped Google’s New Autonomous Car”). On Friday Google said that the new design will be unleashed on the roads of the company’s hometown of Mountain View, California, this summer. Eventually, up to 100 vehicles will roam the town’s suburban streets.

Prototype’s low top speed qualifies it for the less stringent vehicle safety standards the National Highway Traffic Safety Administration (NHTSA) applies to electric golf carts. The car must have lights, mirrors, and seat belts but is exempt from many of the crashworthiness standards and air-bag requirements of normal gas and electric vehicles. In Mountain View, it will be restricted to roads with a speed limit of 35 miles per hour or less.

That still gives the design a lot of scope as an urban taxi, says Bryant Walker Smith, an expert on autonomous vehicles at the University of South Carolina. “Almost the entire island of Manhattan between the expressways could be accommodated by a vehicle operating at 25 miles per hour,” he says. “Right there, you have several million people who could be serviced by a car like this.”

A slow, light car such as Prototype is also less apt to be involved in a catastrophic accident. Impacts are more likely to be gentle fender-benders rather than pile-ups, and there’s less potential to injure or kill pedestrians or cyclists. “A limited-environment low-speed vehicle will be technologically and socially viable sooner than a vehicle capable of operating anywhere,” says Smith.

But there are also drawbacks to focusing on such a limited vehicle. Tests of Prototype won’t give Google the experience with safety systems and crash tests it will need to design a more conventional autonomous car that can travel at higher speeds.

The Boston Consulting Group estimates that bringing full autonomy to market will cost car makers upward of $1 billion each over the next decade. That money will go to design prototypes, develop sensors and processing technologies, write integration software, and perform testing and validation. Apart from the navigation and decision-making algorithms, such systems are likely to be very different in a full-speed four-seater and a low-speed taxi.

Google also still has significant work to do on making its software capable of handling all road situations. The cars have recently gained some ability to cope with roadside joggers, police vehicles, and cyclists’ hand signals. But they still can’t reliably handle very rainy conditions or operate in areas that have not been mapped to centimeter-level accuracy.

The distinctive Prototype might help Google learn how other road users interact with fully autonomous vehicles. It is a Level 4 vehicle, an NHTSA classification applying to cars that require passengers to do nothing more than provide their destination.

At the moment, no one knows how people will react to a car with an empty driver’s seat. If people change their behavior in unpredictable ways, then Google’s software may have extra challenges. It can’t use social cues like eye contact or waving to other drivers and pedestrians as a human driver might to clear up a misunderstanding. However, California regulations currently require that all driverless test vehicles have a backup steering wheel, a brake pedal, and a safety driver ready to take over at all times.

Whatever Google’s ultimate aim, the company will continue to operate at least some of the 23 modified self-driving Lexus SUVs it already drives on public roads. They are able to drive anywhere Google has mapped in detail. In total, Google’s cars have covered 1.7 million miles around the Bay Area, including freeways and city and suburban streets.

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.