Autonomous trucking company TuSimple recently held its inaugural AI Day event in Shanghai, delving into the complexities of Level 3 and Level 4 functionalities for self-driving trucks, and the critical need for enhanced reliability in such systems.
Wang Naiyan, CTO of TuSimple, said that these functionalities might not be as complex as assisted driving, but highlighted the need to improve the reliability of such systems.
One of the ways TuSimple has explored to achieve this is in its development of redundant architecture. The company has developed a full and comprehensive redundant architecture, entailing the duplication of all critical components and systems within the autonomous vehicle. If one component, such as a sensor or processing unit, malfunctions or fails, the backup component can take over to maintain the system’s utility.
Another challenge TuSimple addressed was the need to improve the radius of detection of its autonomous vehicles. While standard passenger vehicles take three to four seconds to change lanes, heavy-duty trucks demand seven to eight seconds for the same maneuver due to their sheer size and weight, with loads nearing 50 tons and lengths exceeding 20 meters. A perception distance of 400–500 meters is therefore needed for these vehicles to safely navigate on highways.
TuSimple has come up with a smart way to help their self-driving trucks “see” and understand their surroundings better. Special cameras are installed to capture views from different angles around the truck, enabling the detection and tracking of objects in 2D and cross-referencing them with other forms of spatial data to create a detailed 3D visualization of their environment. This technique allows the trucks to detect and recognize nearby objects more accurately, improving the safety and reliability of their operations.
RELATED ARTICLE
The company has also developed an algorithm called Bird’s Eye View which can reduce the computational overhead of its trucks’ self-driving systems. The algorithm is suited for long-distance detection, providing added efficiency in scenarios involving objects with overlapping heights, ensuring accurate detection without compromising performance.
TuSimple’s efforts have yielded positive test results. The autonomous driving trucks were tested to be capable of detecting white plastic bottles from distances exceeding 120 meters, enabling the vehicles to initiate safe avoidance and lane change maneuvers proactively.
In 2023, TuSimple started testing its autonomous trucks on the Tomei Expressway, one of Japan’s key freight corridors that connect the major cities of Tokyo, Nagoya, and Osaka.