Mention self-driving vehicles and people probably envision Waymo’s minivans or test models from General Motors’ Cruise or Uber, which have drawn the lion’s share of funding and media attention. But the impact of autonomous technology may be felt earlier in the vast and lucrative commercial trucking industry where competition is ramping up fast.
Waymo plans to enter the logistics business and has adapted the same technology that drives its robo-taxis to operate Class-8 semi-trucks. But a host of ambitious startups believe they’ve got a shot at taking on the industry leader in commercial trucking. In fact, TuSimple, which has raised $83 million from investors including Nvidia, thinks it may have an edge on Waymo when it comes to driving in heavy rain.
The San Diego-based company tests its robotic Peterbilt trucks in Tucson, equipped with laser lidar sensors, cameras, radar, computers and software that it’s created in-house, where it’s focused on perfecting long-range perception and the ability to drive in heavy rain. Intense downpours that occur in southern Arizona’s “monsoon” season create slick surfaces that make it particularly dangerous to operate and stop vehicles weighing up to 80,000 pounds with a fully loaded trailer.
A video just released by the company shows one of its trucks operating smoothly on public roads outside Tucson during an apparent heavy downpour. Shot from inside the cab of the truck, the TuSimple technicians on board don’t have to intervene as the truck contends with wet roads and other drivers pulling in front of it, a traffic light and having to stop at a rail crossing to wait for a freight train to pass.
Waymo also operates its robotic vehicles in light rain, but specifies in the recently published “Emergency Response Guide and Law Enforcement Interaction Protocol” guide for emergency crews that bad weather “including heavy rain and snowy/icy conditions” limit its functionality. A Waymo spokesman confirmed that that guidance remains unchanged.
Along with Waymo and TuSimple, companies including Daimler, Embark, Starsky Robotics, Kodiak Robotics and Kache.ai, a stealthy startup that’s apparently connected to former Uber and Google engineer Anthony Levandowski, all see enormous opportunity in trucking. A booming economy and our insatiable desire to have goods delivered rapidly have pushed annual U.S. trucking revenue to an estimated $700 billion, according to the American Trucking Association. What’s more, the industry is grappling with a shortage of human drivers, particularly for long-haul highway routes, that the ATA estimated was 50,000 people in 2017.
And unlike crowded city streets or suburban roads, operating autonomous trucks on highways is a bit less complex, since the odds of encountering tricky left turns, double-parked vehicles, bicycles, pedestrians, dogs and other obstacles are much lower.
Heavy rain capability may be a small edge right now, but TuSimple considers it an important one that results from a different approach to its software development and a greater reliance on high-definition cameras rather than lidar, which sees well in the dark but can be affected by rain.
“There are two related developments that allow us to operate safely under rainy conditions,” Chuck Price, vice president of product for TuSimple, told Forbes. “Our cameras are much less susceptible to rain effects than Lidar, and our perception system is trained to identify objects properly even when those objects are distorted or partially obscured by rain.”
Additionally, TuSimple’s “control logic accounts for strong crosswinds, which exert tremendous and sudden side loads on a tractor trailer. Thus we can remain safely in-lane, even in very strong wind conditions,” he said.
Three-year-old TuSimple was co-founded by Caltech-trained cognitive scientist Xiaodi Hou, who’s also chief technology officer and U.S. unit president. The company has said it’s generating a small amount of revenue hauling commercial loads through the Arizona desert on U.S. Highway 10 for customers it’s not naming. (It has a sister company in China that’s focused on creating robotic trucks for drayage service — hauling cargo at ports.) Operations in the U.S. scale up substantially from next year.
“By the end of 2019 I expect to have 100 to 200 vehicles,” Hou said in a recent interview. The precise number is hard to know for certain owing to big demand for semi-trucks from across the industry, meaning “the lead time for getting a new truck is 50 weeks right now,” he said.
Hou doesn’t have the typical computer science background that’s common at Waymo, Uber, Cruise and other companies developing artificial intelligence to drive vehicles. Instead, the approach he’s taken is heavily influenced by a background in neuroscience and cognition to create something approaching consciousness.
“Consciousness means a lot of connections to things that cross-validate, in which you can tell the difference between a dream and reality. That’s part of the very intriguing characteristics of consciousness. You can relate different things together, you can compare things,” he said. “What we are building is a consciousness within a very limited set of elements of perception in a truck. This will be my core design for building a safe autonomous driving system.”