The most impressive thing about the Uber trip from the Midwest to Southern California wasn’t that the truck drove itself the 344 miles across Arizona. It was what happened when two men named Larry and Mark met at the western edge of the Copper State. Larry, the trained safety driver, had spent the autonomous voyage watching over his robot. Mark was freshly arrived from Los Angeles in a conventional truck.
Each unhitched their trailer and hooked up to the other’s. Mark drove his new pile of cargo to its final destination. Larry headed back east, flipping his semi into autonomous mode.
This recent meet and greet, which Uber described today, marks the ride-hailing giant’s latest step into the world of long-haul trucking: a step that depended not just on self-driving tech, but on the logistical firepower to make it work for real.
Uber has been talking about trucks since it acquired Anthony Levandowski’s self-driving big-rig startup, Otto, in 2016. (Between Levandowksi’s role at the center of Uber’s bruising legal fight with Waymo and a trademark lawsuit, Uber dropped the Otto name last year.) And it’s far from the only company trying to teach robots to drive semis: Volvo, Daimler, Waymo, Tesla, and several startups have declared similar intentions.
That’s because, trucking is, as they say in the Valley, ripe for disruption: Americans are shipping more and more stuff. There aren’t nearly enough drivers to move it all, and the shortage will only get worse over time as Amazon continues to disrupt (there’s that word again) traditional retail stores. And while it’s hard to get vehicles to drive themselves in complex cities, the trucker’s domain—the freeway—is easy to master. Just stay in your lane and brake before you hit the guy in front of you.
The question is not, then, can the technology work, nor is it whether people will pay for it. It’s what happens when the truck leaves the highway and must negotiate intersections, avoid squashing pedestrians, and navigate through industrial yards. The consensus among the companies trying to take trucking robo is that all that stuff is simply too hard. So they have cooked up a few workarounds.
One startup, Starsky, wants to separate vehicle and human: People in call centers would remotely operate the trucks on surface streets. Uber is going with what we call the bar pilot model. A human in a regular truck picks up the shipment, takes it to a transfer hub next to the highway, and passes it over to the autonomous truck. The robot (with a licensed driver behind the wheel, for now) does the simple highway cruising. When the truck eventually exits, the handoff runs in reverse, and another human takes control for last few miles—like the bar pilots who lead ginormous container ships into and out of port. Someday, Uber’s transfer hubs could be dedicated facilities with their own on- and off-ramps. For now, its trucking team is using weigh stations off the I-40 in Topock and Sanders, Arizona, at either end of the state.
That’s the essence of it, but Uber imagines a much more intricate operation. This is where the meeting of Larry and Mark comes in: Each had a load of cargo to haul on each leg of their trip, so neither wasted time or money moving an empty box.
The point of Uber’s robo-trucking scheme isn’t just to safely get piles of sneakers and diapers and dog food from warehouse to freeway to warehouse. It’s to do it with the logistical efficiency. When that human driver brings a shipment to the robotruck, Uber wants him to have a newly arrived shipment to bring back with him. The goal is to keep every vehicle, no matter who or what is working the gas and brakes, to be moving cargo—and making money—at all times.
“I think the theory works,” says Jason Parker, who works on trucking at Flexport, which helps businesses organize their shipping efforts. But the realities of the trucking market could impede the flow. There are times and places where cargo tends to head in one direction: A lot more stuff leaves California on trucks than enters it (because so much comes in by ship), for example. Some companies and independent drivers own their trailers, and won’t much like seeing them toted away by a robot. Questions of liability get knotty when one trip involves as many as three companies, and non-human drivers.
To make it work at all, Uber needs exquisitely detailed knowledge of cargo movement patterns. This is where Uber Freight comes in. The service, which launched in late 2016, connects truckers with things that need trucking, the way UberX connects Prius drivers with San Franciscans fed up with BART. Like most Uber products, it’s a bid to supplant the traditional middleman, here the brokerage firms that in some cases still rely on an ancient practice known as calling someone on the telephone.
Torching an aging, inefficient industry to gin up more revenue is always a solid move. But now, Uber Freight can play a much more important role—in all likelihood, the role it was born to play. After more than a year of gobbling data on where cargo goes and when, Uber has reached a point of logistical savvy where it can send shipments in two directions. It may not sound like much, but it’s a necessary first step toward the whirling, perfectly choreographed, nationwide ballet of efficiency Uber has in mind.
The team has a few technological roadblocks left to mow down. For example, Woodrow says his team is figuring out how to ensure the trucks can leave the right lane when passing a pulled over emergency vehicle, as required by law in some states. And like human truck drivers, the semis need a good sense of how humans in passenger cars are likely to act around them. The answer, often, is stupidly—following too closely, or cutting the big boys off.
That’s the easy stuff, relatively speaking. Making the robot is one thing. Keeping it on the job—24/7—is another.
Robot Overlord Red Carpet