So Uber Wants Self-Driving Bikes and Scooters. Why? And How?

It’s not clear how to make a bike propel itself in the first place, since the e-bikes Jump and its competitors use are pedal assist—they augment foot power with motors, which aren’t that powerful on their own. BRENDAN SMIALOWSKI/Getty Images

Really, it was only a matter of time before somebody thought to combine today’s hottest transportation trends: shared electric scooters and autonomous driving. Over the weekend, Uber reportedly unveiled a micromobility robotics division at a robotics meetup in the Bay Area. Though the company declined to confirm or comment on the new addition, the division will reportedly explore how to make the scooters and shared bikes it’s now deploying alongside its cars capable of riding themselves.

The key questions, then, are the classic ones—why and how.

The “why” makes, well, a touch of sense. Self-driving e-scooters and bikes would be able to mosey themselves over to battery-charging stations, or wheel themselves into maintenance depots, or redistribute themselves to exactly where users need them to be. That would be helpful, because operations and logistics can be an pricey pain for tech-enabled scooter- and bike-share companies. Figuring out where and when scooters need to be deployed, and when and how they need to be charged and repaired, is hard enough. But getting human beings out into the field, and doing the actual work of collecting, charging, and maintaining the vehicles has proven costly for the companies. According to investor documents provided to The Information showing internal company numbers from last spring, the e-scoot company Bird spent almost half its gross revenue per ride paying individual contract workers to charge their scooters, and another 14 percent of gross revenue per ride to repair them.

LEARN MORE

The WIRED Guide to Self-Driving Cars

So yes, it would save companies a tidy sum if they could cut human labor out of the picture. “Operating expenses are the biggest thorn in the side of the companies,” says Richard Branning, the CEO and founder of the startup Sweep, which provides operations and logistical services to scooter firms. (Sweep runs operations for Spin, the scooter startup acquired by Ford last year.) But he argues that even the most advanced robots might not be able to extricate themselves from the pinches that shared bikes and scooters can get themselves into. “Let me know when a scooter is behind a fence and there’s a dog biting at it, when it can hop that 10-foot fence and get back on the road and zoom itself back to a charging pod,” he adds.

Indeed, making anything drive itself in the real world is really hard. Beyond a few demo rides, no company has been willing to take the human safety operator out of its cars, which is another way of saying that nobody has cracked the self-driving problem yet. And while scooters and bikes don’t match the speed or weight—and thus the potential danger—of cars, they bring their own challenges.

First, these are two-wheeled vehicles, and they tend to fall over when they don’t have a human using their weight and movement to keep them upright. Beyond adding extra wheels or heavy, complex tech like a gyroscope, there’s no obvious way to ensure they don’t tip, or to get them back up off the ground when they do. It’s not clear how to make a bike propel itself at all, since the e-bikes that Uber's mobility subsidiary Jump and its peers use are pedal-assist: They augment foot power with motors, but those alone aren’t powerful enough to move the whole bike.

Second, scooters will have to operate on the sidewalk or on the street. That means they’ll have to deal with vehicles, intersections, and walking humans. Maybe even the occasional rogue squirrel.

That stuff is hard, but achievable, according to Kevin Peterson, the CTO of sidewalk delivery-robot company Marble. The bigger hurdle he sees is cost. Any self-driving vehicle needs sensors to perceive the world and computers to turn that data into decisions about how to move. Even if you manage to do this without lidar—which is the most expensive sort of sensor and costs in the thousands-of-dollars range—and use a combo of sensors and computer vision techniques instead, you’re making the vehicle more expensive.“I don't know of a safe autonomy system that would be cheap enough to put on a scooter,” Peterson says, suggesting an ambitious $50 as a ceiling.

Trouble is, a key advantage of bikes and scooters is that they are cheap, around $500 per scooter. And these shared vehicles live outside and take a beating—even when people aren’t hurling them into local bodies of water or stuffing them up into trees. On top of the cost of the system, Uber would have to put more money into hardening an expensive sensor and computing suite, or replace the scooters’ and bikes’ cameras and other bits more often than they’d like.

OK, so coaxing scooters and bikes to drive themselves would be very difficult and very expensive. The undertaking is still somewhat on brand for Uber. CEO Dara Khosrowshahi has positioned the company as the “Amazon of transportation.” He wants Uber to be able to get you anywhere, and bring you any meal, no matter your travel mode. He also has decided against nixing the company’s expensive self-driving research and development efforts—Uber wants autonomy to be part of its very long term future.

As the company gears up for its initial public offering this year, of course self-driving scooters and bikes will be part of its narrative.

  • The prime challenges for Amazon's new delivery robot
  • The VA wants to use AI to prevent kidney disease
  • As tech invades cycling, are bike activists selling out?
  • Trump, a Russian agent? The alternative is too awful
  • Nike's new self-lacing basketball shoe is actually smart
  • 👀 Looking for the latest gadgets? Check out our picks, gift guides, and best deals all year round
  • 📩 Hungry for even more deep dives on your next favorite topic? Sign up for the Backchannel newsletter
Read also:   One Year In, Uber's CEO Has Bigger Problems Than Travis