After four years, Amazon subsidiary Zoox is taking its autonomous L3 vehicles from the automotive kiddy tracks of Las Vegas and the San Francisco Bay Area to the off-road rally course that is the streets of Seattle.
Self-driving cars and other autonomous vehicles have a great potential for making transportation faster, safer, and more convenient, but if the technology is to have any hope of becoming commonplace, it has to be able to operate outside of a handful of rigidly controlled roadways.
To achieve this, autonomous vehicles have to be developed and tested in the semi-chaotic environment of real-world cities. Zoox has already tested its cars in Las Vegas and San Francisco, and now the company is taking on the Mount Everest of urban driving: the city of Seattle.
Because Seattle is built around seven ice-age hills set on a pair of peninsulas on the shores of Puget Sound, it has long presented a challenge in terms of street planning, but over a century of growth and endless waves of “improvements” has turned the city’s streets into an even bigger challenge for motorists.
Not only are many of the roads, tunnels, and bridges too narrow to handle modern vehicles, they are also on steep grades in a strange mixture of straight paths, mazes, and odd turns punctuated by mini-roundabouts, and even ramps that go nowhere. Many of the roads are very badly maintained, the intersections are often confusing, and recent years have seen the addition of bike lanes and cafe seating to streets that aren’t wide enough to accommodate them, resulting in some scary lane sharing between motorists and cyclists.
On top of this, Seattlites have a reputation as not being the best drivers in the world and when rain or snow is added to the mix, YouTube-ready accidents like SUVs sliding down icy hills happen with depressing frequency. Not to mention that local cyclists have a reputation for treating stop signs as suggestions.
This may make taking a cab more attractive when visiting the Emerald City, but for Zoox it’s a natural laboratory for testing their fleet of Toyota Highlanders equipped with the company’s L3 hardware and artificial intelligence software. After opening its fourth office in Seattle in 2022, Zoox will begin a program of mapping out and geofencing the city and learning to safely and legally navigate the streets in different weather conditions.
For the upcoming test, the cars will use their latest software and sensors to map locations and drive around, but they’ll have a safety driver at the wheel to switch to manual if required. The data gathered will then be used for Zoox’s L3 and L5 fleets, and by the company’s Calibration, Localization, and Mapping, Simultaneously (CLAMS) and Zoox Road Network (ZRN) teams to produce 3D maps of Seattle that include bike lanes, speed limits, traffic lights, and other features.
“A smooth sea never made a skilled sailor, and it’s the exact same principle for our technology,” says Kai Wang, director of prediction at Zoox. “The challenges of the Seattle area will let us hone our software stack and ultimately improve the behavior of our vehicles.”