Start with the sensors, then design the rest: How Zoox built its robotaxi

These days, the hype is all about AI and robots, but almost a decade ago, the tech du jour was self-driving. You couldn’t swing a lanyard at CES for the latter half of the last decade without hitting a robotaxi; post-COVID, the number of startups has shrunk, but the technology has definitely matured. Go to the right cities—San Francisco and Austin, Texas, spring to mind—and you might see dozens of sensor-festooned vehicles among the downtown traffic. The pod-like robotaxis belonging to Zoox stand out. Other robotaxi developers are retrofitting existing vehicles like Hyundai Ioniq 5s with sensors and the computing power necessary for self-driving. Zoox, which was bought by Amazon in 2020, did that with its test fleet, but as it starts to offer ride-hailing services—currently in Las Vegas and San Francisco—it’s doing so with a purpose-built design that looks like it just drove off the set of a big-budget sci-fi production. “A robotaxi is not a car; it’s not a human-driven vehicle, and the requirements are wildly different, although it has to live in that world,” explained Chris Stoffel, director of robot industrial design and studio engineering at Zoox. It all starts with the sensors, each perched on a little ledge projecting from the top four corners of the robotaxi’s body. From up there, each has an unobstructed, high-level view, giving the Zoox robotaxi good situational awareness, particularly straight ahead. “Because we don’t have a traditional hood, we’ve optimized our frontal coverage in a way that would be nearly impossible on a retrofitted vehicle,” said Zoox director of sensor engineering Ryan McMichaels. “Not only do we do that for the maneuverability, but also the redundancy of the vehicle,” said Stoffel. “The hardware inside of the vehicle, it’s the same rack, it’s the same EDU on both ends, same battery pack—kind of split in both ends—two HVAC units. There’s a lot of redundancy built in there. It kind of got the kitchen sink of redundancy as we wanted to make sure this first product really could complete the mission,” Stoffel told me. “We’re picking people up and we’re dropping them off. How do we do that better than anyone else? The idea is to get into a spot that no one can or down a street or maneuverability that no one can, because we are really focused primarily on dense urban areas at the moment,” Stoffel said.