Today TuSimple is operating three to five autonomous trips per day for customers across different routes in the U.S. state of Arizona. While the journeys are completely autonomous, for each journey there are two safety engineers on-board. One engineer is positioned behind the wheel and the other focuses on data analysis.
Investment in TuSimple has come from companies like Nvidia and Sina. The trucks use LiDAR (‘light detection and ranging’) technology for the navigation, a surveying method that measures distance to a target by illuminating the target with pulsed laser light and measuring the reflected pulses with a sensor.
Despite the importance of LiDAR, TuSimple relies mostly on cameras alone. Radar and LiDAR act as “secondary sensors”; the camera array allows the trucks to visualize the area 1,000 meters (3,280 feet) ahead of a vehicle. This is key to the core technology of the vehicle, given that most driverless systems can only see to an optimum level of quality up to 150 meters ahead of the vehicle.
TuSimple deploys Nvidia graphic processing units as part of its autonomous driving platform. The technology includes the NVIDIA DRIVE PX 2 AI car computing platform, Jetson TX2 AI supercomputer on a module, and CUDA parallel computing platform. Also featured as part of the system are the TensorRT deep learning inference optimizer and runtime engine, and the cuDNN CUDA deep neural network library.
TuSimple currently has twelve customers and the company says it’s seeing a growing acceptance of autonomous vehicle technology. The company has plans to expand its autonomous vehicle fleet from eleven to forty by the mid-way mark in 2019.
The company’s growth has been rapid. As recently as June 2017, TuSimple successfully completed the maiden 200-mile Level 4 autonomous vehicle test drive, which was from San Diego to Yuma, Arizona. By the end of 2019, the company is looking to run several autonomous trucks daily.
Speaking with TechCrunch, TuSimple founder, president and CTO Xiaodi Hou reiterated just how fast the company is growing and where it is heading next: “Exactly one year after debuting our prototype system at CES 2018, we’re now running up to five commercial trips a day in Arizona, expanding our fleet and moving quickly toward our goal of creating the first commercial self-driving truck.”
In terms of future plans, The Drive notes that TuSimple is aiming to market its suite of autonomous-driving technology to other companies. In terms of its own services, TuSimple is aiming to have 40 trucks in fully-autonomous operation by June 2019 (building up from its current base of eleven vehicles).
