Autonomous trucking startup TuSimple, which has the largest fleet of autonomous trucks testing on U.S. and Chinese highways, has reached a $1 billion valuation after just over three years in business.
The San Diego and Beijing-based company said Wednesday it closed a $95 million financing round in December 2018, bringing total cash raised to $178 million.
“We are focused on finding the global leaders in artificial intelligence, and TuSimple is ahead of the pack,” said Colin Xie, vice general manager of investment at Chinese social media company Sina Corp., which participated in the latest funding round with Hong Kong-based investment firm Composite Capital Management.
“The combination of technical excellence and an impressive leadership team has propelled the company to unicorn status,” Xie said.
TuSimple is pursing Level 4 high automation, in which a vehicle can drive itself with almost no human intervention, according to National Highway Traffic Safety Administration guidelines.
TuSimple has contracts with a dozen shippers and fleets, including Fortune 100 and large international companies.
“It’s one of the advantages trucks have over passenger-car automation. You can’t really charge people for doing validation,” Chuck Price, TuSimple vice president of product, told Trucks.com. “With trucks, we can charge competitive rates to haul.”
TuSimple runs three depot-to-dock routes within Arizona daily, including autonomous operation on surface streets. It plans to start hauling freight through New Mexico and to the Texas cities of El Paso, Dallas, Houston and Austin this year. It has amassed about 300,000 miles of autonomous driving, Price said.
Tu Simple expects that validation of its cameras and computer vision-based system will continue until late 2020 or early 2021. It is sharing what it learns with customers to gain their confidence in eventual driverless operation.
TuSimple does not plan to be a trucking company over the long term.
“Our business model is to operate the autonomy when it’s on the road,” Price said. “Once it’s dispatched, we take responsibility for assuring that truck gets to its destination.”
That includes remote monitoring for breakdowns, taking any disabled trucks off the road safely and sending a technician or a tow truck.
“We’ll be able to plug into the cameras to see what’s going on,” Price said. “Our research is done. We know how to do this.”
TuSimple is building and testing its own fleet because it does not want to have a third party, such as a supplier, responsible for validation. Price said.
Today, each TuSimple truck has a safety driver with a commercial driver’s license who can take over operation if needed. An engineer occupies the passenger seat.
TuSimple’s goal is for both to be out of the cab by 2020, reducing costs and helping address the shortage of 51,000 long-haul truck drivers estimated by the American Trucking Associations. But such a plan would require states to adopt regulations to allow for autonomous trucking. Insurance and liability issues also remain hurdles.
Driver wages and benefits accounted for 43 percent of the total operating cost of trucking in 2017, according to the American Transportation Research Institute. Autonomous systems increase the cost of truck equipment. But removing the driver would more than make up for the additional technology costs.
With full autonomy, the cost savings could reach 45 percent, according to a McKinsey & Co. report issued in December 2018. It would take “many years” for an autonomous fleet to replace the hundreds of thousands of conventionally driven trucks on the road, the consulting firm said.
The technology challenge of autonomous trucking is best addressed by cameras rather than light-emitting radar called lidar, TuSimple says.
Founder Xiaodi Hou, a Chinese engineer, invented a camera-centric perception system able to identify objects 1,000 meters, or 3,281 feet, ahead.
TuSimple uses lidar as a backup to its cameras for close encounters, like lane changes.
A TuSimple truck on display at the Consumer Electronics Show in Las Vegas in January had an array of nine cameras around the truck and trailer and two lidars in the side mirrors.
“Anyone using lidar as their primary sensor is not going to see the distances you need to see with a large truck going 65 miles an hour,” said Price.
Not everyone buys that.
“I don’t need a thousand meters,” said Roger Nielsen, chief executive of Daimler Trucks North America. Daimler announced in January plans to test its own Level 4 autonomous truck in the U.S. this year.
Professional drivers look 300 to 500 meters or 15 seconds ahead as a safe operating distance.
“You don’t need a thousand meters, but you want a thousand meters,” Price said. “We can see far enough ahead to make strategic decisions (like) should I be in this lane or the other lane?”