TuSimple's self-driving 'middle mile' truck solution
New trucking technology was brought into the limelight back in November when Tesla announced the electrification of the transportation service that hauls more freight than any other in the U.S. Of course, just about anything Elon Musk announces receives the highest amount of scrutiny. However, just because how goods are shipped across the country isn't regularly at the top of most people's minds doesn't mean there aren't companies out there who give trucking the fanfare it deserves. One such company, TuSimple, demonstrated its vision for the trucking industry at CES in Las Vegas last month.
Founded in 2015, TuSimple wants to create the first commercially deployable autonomous trucking platform. To that end its display marked the AI company's U.S. debut for its Level 4 autonomous truck. TuSimple says its vehicles are on track for commercial development in the next few years, with testing set to begin early this year in Arizona. Its goal is safer roads, cleaner air, and improved trucking efficiency to meet growing consumer demand.
The upcoming round of tests follows on the heels of the successful completion of a 170-mi (273-km) public road test of the Level 4 system that was on display at CES. That journey in July spanned San Diego to Yuma, AZ, while public road testing in China began in September.
TuSimple says it currently operates on four test routes and has driven more than 15,000 autonomous test miles. The testing slated to begin in Arizona this year will include a fleet that will run a 120-mi (193-km) route from Tucson to Phoenix. A second fleet will operate a route in Shanghai. During these tests, TuSimple will transport customer cargo between logistics centers to test the real-world implementation of autonomous freight technology and to gain practical insights. By the end of 2018, the company’s goal is to accumulate millions of test miles to solidify its arsenal of data and experience, which it will ultimately use to deliver a safe and reliable autonomous driving solution to the market.
TuSimple’s platform focuses on line-haul trucking, or the transportation of cargo between ports, plants, warehouses, and distribution centers. By integrating technology into this "middle mile" part of the supply chain, the company aims to address current industry challenges such as road safety and driver shortages, while helping to reduce carbon emissions with optimized driving.
The company says that its full-stack solution brings together technology for perception, localization, mapping, motion planning, control, and actuation. Its technology is able to detect and track objects at distances of more than 200 m (650 ft) through advanced sensor fusion that combines data from multiple cameras.
Interestingly, the localization technology does not use one of the emerging sensing technologies for autonomous vehicles: LiDAR. Instead, the system uses TuSimple's own high-precision visual positioning and multi-sensor fusion technology to assure the truck and trailer stay in-lane by achieving consistent, decimeter-level precision—down to about 4 in (102 mm)—even in challenging environments such as driving through mountainous areas and tunnels. The company's computer-vision technology has set records in the KITTI Vision Benchmark Suite and City Scape: three in object detection, two in object tracking, and four in road segmentation. Furthermore, the truck’s decision-making system dynamically adapts to road conditions, changing lanes, and adjusting driving speeds to maximize safety and efficiency.
HD maps used by TuSimple's systems are designed by the company itself to create what it calls a "highly customized, high-precision map." The map consists of lane-level navigation information as well as 3D geometry of the terrain.
Motion planning and control algorithms are designed to adapt to real conditions to reliably conduct maneuvers, such as adaptive cruise control, lane changing, merging in and out, avoid slow or stopped vehicles, and respond to emergency situations. To perfect its motion planning platform, TuSimple collected a massive amount of human driver behavior data to develop an in-depth understanding of the behavior of other drivers and vehicles on the road.
TuSimple uses Nvidia GPUs, Nvidia DRIVE PX 2, Jetson TX2, CUDA, TensorRT, and cuDNN to develop its autonomous driving solution. Nvidia GPU Ventures joined a group of investors led by Chinese social media company Sina to invest more than $20 million in the startup in August of 2017 as part of a Series B funding round. The company raised an additional $55 million of Series C funding in November.
The influx of cash was necessary to aid the growing company. With centers already located in San Diego and Beijing, TuSimple says it established a new testing facility in Tucson in July 2017. The company's plans are to expand to hire a production team of mechanical engineers, electrical engineers, and test drivers. This team will focus on the testing and production of a commercially ready system for autonomous trucks.
Collaborations are currently with Peterbilt in the U.S. and Shaanxi Auto Group in China. The company plans to scale its U.S. efforts to commercialize self-driving trucks by deploying additional Peterbilts. In the future, TuSimple says its trucks will be available through the company’s OEM partners with TuSimple providing autonomous fleet management services to support integration and ongoing maintenance. TuSimple plans to introduce commercial autonomous trucking services to the market in 2019.