The truck of the future may soon be silently gliding up behind you on the interstate. And don’t expect to see a driver in the cab.
IF YOU WANT to see the current state of trucking, there’s no better place than Interstate 80 in Pennsylvania–a black ribbon of commerce that extends from the Delaware Water Gap across the forested Pocono Mountains through the formidable Allegheny Basin. The highway is generally thick with trucks, often riding two wide, to the annoyance of the passenger cars trying to pass.
MAKE YOUR WAY southwest from I-80 onto state Route 28 and you’ll roughly follow the Allegheny River, one of the nation’s earliest freight routes, until road and river take you to the Lawrenceville section of Pittsburgh. A former industrial neighborhood, Lawrenceville is now part of the city’s ever-expanding robotics district. There, in a new office development called Tech Hub, you’ll find an autonomous trucking (AT) startup called Locomation.
In Pittsburgh, Locomation’s Çetin Meriçli (left) and Tekin Meriçli are planning two-truck half-robot convoys.Photography by Amber Ford
The five co-founders are all former or current faculty members at Carnegie Mellon University’s renowned National Robotics Engineering Center–all world-class engineers, and immigrants, too. The CEO, Çetin Meriçli, is from Turkey, as is his brother Tekin, the CTO. Alonzo Kelly, chief scientist, is a Canadian; Venkataramanan Rajagopalan, head of autonomy, an Indian; Michael George, VP of engineering, an Aussie. Each has jumped off the academic train to ride an only-in-America opportunity in trucking, because they think it’s enormous–up to $250 billion for their slice of the AT market. Enormous but narrow, because they know that AT has a massive technical barrier to overcome. Locomation’s roboticists have concluded that Level 4 autonomy–the point at which the driver disappears from the cab–isn’t possible at scale, no matter what Elon Musk says.
At least not yet. “It is tempting to believe that it’s a unidimensional problem–just a tech problem,” says Çetin Meriçli, 45, a professorial type who seems almost surprised at having taken on the day job of CEO. In classical robotics, engineers are taught to program everything as an if/else statement to address all possible scenarios a device might encounter. That works in a limited setting, like a freight yard. But you can’t program for the infinite scenarios on our roads.
This is, he notes, an old problem. In 1998, two researchers from CMU drove cross-country while operating in autonomous mode 98.2 percent of the time. But that remaining fraction proved vexing–edge cases that were unforecastable. And though by 2017 the industry had dropped programming muscle in favor of machine learning, even the machines can’t inhale and analyze enough data to account for the real world. Which is why you don’t see robotaxis in 100 cities, or 50, or even five. The end of the decade is now the best guess for figuring out self-driving cars at scale.
So the focus today is on the doable–and freight fits that narrow frame, moving, as it does, along predictable paths like interstates for some part of the journey, yielding fewer potential variables. “It was apparent to us from the get-go that freight automation is a larger, more near-term application for autonomy,” says Meriçli. “In other words, it is a problem worth solving.”
The business case for Locomation and its ilk is thus to forget end-to-end automation and focus on practical solutions. Even if they’re transitional, they wager, the payoff will still be substantial. Some are focused on long-haul routes on major highways, others on the so-called middle mile, up to 500 miles or so. The last mile–which ends at your door–is more feasible for robots and drones than for trucks.
For Locomation, the strategy is to put the autonomy in the back seat–or rather in the trailing truck of a paired convoy. To them, only one entity has a CPU powerful enough to be the leader. “Humans are super good at identifying weird situations and reacting to them,” says CTO Tekin Meriçli, 40. The company’s Human-Guided AutonomySM seeks the best of both worlds: a two-truck convoy, a human driver leading, the trailing truck connected to the leader by some awesome tech.
In Locomation’s garage, which is more like a lab or a Formula 1 shop, new Peterbilt tractors are kitted out with light direction and ranging lasers, cameras, sensors, steering and brake controls, transceivers, and a proprietary software stack to run it all. Initially, the equipment will be retrofitted onto existing trucks; eventually, it will be configured in by truck makers. The two vehicles do everything in tandem, or at least try to. Should they become separated by traffic, the driverless truck’s autonomous system will, if all goes well, lead it safely back to the leader.
That may sound less sexy than driverless modes, but the dual format brings real efficiencies. There’s a 13 percent savings on fuel because of drafting–the lead truck shielding the follower from air resistance. More important, the paired trucks can operate 22 hours a day, as opposed to 11. How? There’s also a person in the second truck, but they’re not driving. They’re resting in a sleeping compartment, ready to make the return trip as lead driver.
Running two trucks allows the operator to haul twice the freight–profitable for the drivers, too, since they get paid by the mile. Locomation says it can boost clients’ bottom lines by 15 to 20 percent. In turn, Locomation gets an upfront payment and a monthly subscription fee for running the convoys. Beyond that, one of its newest customers, Stevens Trucking, predicts a 22 percent drop in greenhouse gases and 50 percent fewer “empty” miles–when a truck hauls air on its return. Stevens plans to set hubs 500 miles apart–say, from St. Louis to Detroit–with trucks running between them like a metronome: back and forth, back and forth. “My goal,” says Kenny Stevens, company founder and CEO, “is that Stevens will be the first to market in our lanes.”