Oct 2, 2018 9:00 AM

A Self-Driving Truck Startup Keeps It Simple, Stupid

Ike, just out of stealth, is licensing its software stack from the automated delivery robot company Nuro.

Image may contain Vehicle Transportation Truck Train Fire Truck and Tire

If anything has changed between today and the halcyon days of 2016, it’s that those building and marketing self-driving tech are now less ... promise-y. The robots are still coming, the software developers and hardware mavens and balance-sheet-wielding CEOs insist. But more and more, they emphasize that this work is hard, the problems varied, the risks manifold, the regulations slow in coming. Even Waymo, the putative leader in the industry, flush with Alphabet funding—which plans to launch a commercial service this quarter—is having trouble teaching computers to be competent drivers.

Which is why the team at Ike would like to make the thing as easy as possible. “Our core features are descoping and focus,” says Nancy Sun, a cofounder and the startup’s chief engineer. Also, robot trucks.

The cofounders of Ike, which emerges from stealth today, have their share of experience trying to crack self-driving. CEO Alden Woodrow, CTO Jur van den Berg, and Sun have worked at most of the major AV players: Google, Apple, Otto, and Uber. But this time—after all fled Uber’s self-driving truck project, which shuttered this summer—they'd like to keep their project elementary. As elementary as a 15-ton computer that drives itself can be.

The team picked trucks over cars so they can focus on relatively simple highway driving. (Ike's namesake is Dwight Eisenhower, who signed the interstate system-spawning Federal Aid Highway Act into being.) No pedestrians, no cyclists, (hopefully) clear lane lines: Everything’s easier on the highway. And they're serious about staying there.

“We do not want to do a single right turn off the highway,” says van den Berg—even right turns are technological complications. Instead, he envisions Ike's trucks pulling into roadside transfer hubs, where humans drivers will climb in and pilot the rigs to their final destinations. It expects to start testing these trucks on California highways in a matter of months. (As of today, two red big rigs sit in the garage of its San Francisco offices.)

Ike inputs its own sensor data—in this case, lidar data collected on the I-101-380 interchange south of San Francisco—into Nuro's software.

Ike

To get itself on the road, Ike is taking a shortcut—a big one. Instead of building its own autonomous vehicle software stack, it's licensing one from Nuro. The two-year-old self-driving startup focuses on delivery robots; the company launched its first pilot project with the grocer Kroger in June.

The nascent self-driving space is littered with “partnerships”—Waymo with Avis and Fiat Chrysler, Lyft with Aptiv, Toyota and Uber. But Ike and Nuro have got something different going on. “This type of licensing arrangement seems rather new in the AV space," says Raj Rajkumar, an electrical and computer engineer who studies autonomous technology at Carnegie Mellon University. It might be a sign that once-unlimited autonomous vehicle startup funding is now limited, and that companies are getting creative with money.

Or that they're getting creative, period. The automated vehicle sector is still pretty new, especially to answering important questions like How hard is it to build a self-driving truck, really? and How do we make money off this, anyway?

“We’re still pretty early on in the industry, and this agreement shows the technology is not so easy to develop,” says Erik Gordon, a professor at the University of Michigan's Ross School of Business who studies entrepreneurship and tech commercialization. “It’s still valuable and it’s not something you can just go down to the Micro Center and buy.”

In exchange for its software, Nuro has taken a minority stake in Ike. The companies say there is now a “complete fork of the code base” between the two, meaning Ike won’t rely on Nuro for software updates or improvements. (A software demo in Ike’s Dogpatch offices found a company’s engineers tooling around in a program called Nuro Vision. They’ll change that name soon, van den Berg says.) And the companies' aims are dissimilar enough, both teams stress, that they won’t compete, at least not for a long while.

Nuro will give Ike access to its localization, perception, prediction, and planning software that helps AVs “see” and “understand” what’s going on around them, which Ike will have to adapt to trucks. For example, it will have to create and then implement a sensor setup that will help heavier trucks moving at highway speeds to see and then anticipate the actions of the vehicles around it at much farther distances than its littler Nuro counterparts ever will. But the team is most excited about Nuro’s software architecture, which it says gives its engineers an easy-to-use foundation for its testing. Sun estimates the license will save Ike about two years (and 50 to 60 employees’ worth) of work.

That saved time and talent is important, because Silicon Valley is flush with automated trucking startups. Startups Peloton Technology, Embark, and Starsky Robotics have been testing (and raising money) for years. Waymo got into trucks in 2017. Two other groups of Uber alums have launched their own trucking ventures in just the past few months: Former driverless car program head Anthony Levandowski is back with Kache.ai, and former software lead Don Burnette has Kodiak Robotics.¹ For all the questions about technology, these companies share another common problem: an uncertain regulatory picture. Federal rules for self-driving trucks got blocked up in Congress last year, and their future is uncertain. Someone will have to figure that out before these robo-truckers can become businesses.

At least one group is confident that even a Johnny-drive-lately like Ike can get succeed: Nuro. Licensing was never part of its roadmap, says cofounder Dave Ferguson (himself a Google alum). He knew and trusted the Ike team, and particularly van den Berg, who has been a friend and collaborator for more than a decade. “Trucking was an area we were really interested in,” Ferguson says. “But every time we considered doing it, we decided it was going to be too much of a distraction for Nuro.”

So here’s another path for a small robotics company to win at self-driving: Help launch another, similar-but-different company out into the world. The automated vehicle crowd is competitive; might as well have another friendly robot in the race.

¹ Correction appended, 10/2/18, 11:20 AM EDT: A previous version of this story misidentified Don Burnette's automated trucking company. It is Kodiak Robotics, not Kodiak Ventures.