Skip to main content

Amazon expands scope of 'robotaxi' pilot

Zoox robotaxi
Zoox robotaxis are taking test drives in Las Vegas.

Amazon is continuing development of an autonomous ride-hailing platform.

The e-tail giant’s autonomous vehicle subsidiary Zoox, which it purchased for a reported $1.2 billion-plus in June 2020, is building on its successful 2023 pilot of a purpose-built self-driving robotaxi on open public roads with no manual controls or human safety driver.

Since that initial pilot in Foster City, Calif., Zoox has continued testing and developing its robotaxi technology in Foster City as well as in Las Vegas. As it prepares to offer autonomous robotaxi rides to consumers later in 2024, Zoox has been adding the following new features:

Geofence growth

Zoox has expanded its Las Vegas geofence, which is the virtual boundary the company operates its AVs in. This geofence spans approximately five miles from Zoox’s Las Vegas headquarters to the south end of the Las Vegas Strip along multiple routes. 

According to Zoox, the Las Vegas geofence is now larger and more complicated, with three-lane roads, required lane changes, unprotected right turns onto high-speed roadways, and double-right and left-hand turn lanes. 

Driving in these larger areas exposes the robotaxis to the busiest conditions they have ever encountered, according to Zoox, and provides data and learnings for the company as it continues to scale the technology.

Faster speeds

Zoox has also expanded the autonomous driving capabilities of its robotaxi in Las Vegas and Foster City to include driving at speeds of up to 45 mph, in light rain, and at night. While the robotaxi can reach 75 mph, Zoox started out its pilot driving routes with limits of 35 mph in 2023, and is now driving routes with a higher limit of 45 mph. 

The company says it has set internal safety targets and did not begin driving these routes until it could quantify those targets were met and consistently exceeded. Moment-by-moment, Zoox says the vehicle can adjust its behavior accordingly because its AI-based prediction and planner systems are working together. 

Zoox also applies AI-based learnings from AV testing in rain conditions in Seattle, cross-correlating it with our rain data from Las Vegas and Foster City. 

Night operation

For nighttime driving, Zoox robotaxi cameras need to be able to detect unclear shapes in darkness. This requires training and refining its machine-learning models, based on data from both its test fleet of AVs and its robotaxis. 

“There’s a ‘flywheel effect’ of collecting data, feeding it back into your models, and getting a better model with better precision and better recall,” said Marc Wimmershoff, VP, autonomy software, Zoox, in a corporate blog post.

This ad will auto-close in 10 seconds