TuSimple’s self-driving semi-truck is set to revolutionise the freight transport industry in the US.
Story by Team CV

Supported by Navistar, which announced recently that it is investing an undisclosed amount in the US-based autonomous vehicle start-up, TuSimple has developed a semi-truck that has the potential to revolutionise the freight transport industry in the US. The irony of the situation is that it is not the only one. There are others in the fray too, like Tesla, Nikola and many others. Working closely with Navistar, the company has developed an SAE Level 4 self-driving semi-truck that is scheduled to go into production by 2024. Poised to add a new dimension to the trucking scene in the US by combining the legacy of truck manufactured by Navistar with the domain knowledge and expertise in IoT, AI and areas of automation of TuSimple, the semi-truck employs Autonomous Freight Network (AFN). About self-driving trucks, AFN is an ecosystem which consists of autonomous trucks, digitally mapped routes, terminals and autonomous operations monitoring system called TuSimple Connect.
With emphasis on higher safety, reduction in transportation costs and carbon emissions, TuSimple is said to be currently testing the semi-truck at the prototype level. It has until now run the truck for about 15,000 miles in China and the US combined. With a foundation in computer vision, algorithms, mapping, and artificial intelligence (AI), TuSimple, which was founded in 2015, is aiming to bring SAE level 4 autonomous truck driving solution in the transport industry. It is looking at empowering the transport industry with an ability to allow freight to be moved with safety, improved cost-efficiency, and fewer carbon emissions. Expressed Xiaodi Hu, Founder President and CTO, TuSimple, that autonomous driving systems based on AI could be termed as the most complex human creation yet. “It is through such a creation that we are focusing on the commercial solution,” he mentioned. Arriving at the current stage after three years of intense research and achieving self-set goals, the company is now confident of employing its autonomous solution on a commercial vehicle.
Claiming to currently operate a fleet of 40 self-driving trucks in the United States, TuSimple is planning to demonstrate completely driverless operations in 2021. The Shipping freight autonomously for companies like UPS and McLane Company between Arizona and Texas, the current fleet, as per the regulation, has a driver onboard even though he may not intervene until the need to do so is called for. Said to be 10 per cent more fuel efficient than the manually driven trucks, the current fleet, with autonomous technology onboards, is involving a number of suppliers like ZF, which is the co-developer of cameras, LiDAR, radar, steering and automotive-grade central computer (ZF ProAI).

LiDar
The autonomous truck provides light detection and ranging sensors to give a detailed view of the surrounding area within the range of 200 m. The company’s system uses an array of perception and localisation sensors and data along with their proprietary deep-learning detection algorithms to detect and track objects in real-time and make pixel-level interpretations within the field of vision. According to TuSimple, a truck can achieve a decimeter-level of positioning accuracy even in a tunnel or under a bridge. This technology guides the vehicle along a safe and fuel-efficient route based on terrain and real-time road conditions. The sensor array is optimised for use on Class 8 tractors and provides redundant secondary and tertiary perception, detection, and tracking data. Light Detection and Ranging (LiDAR) sensors provide secondary perception and detection at medium- and short-range.

Radar
Sophisticated radar technology is used by TuSimple to detect and classify objects within a distance of 300 m. This radar tech provides enhanced visibility in adverse weather conditions such as dense fog, sandstorms, and heavy rain. Cameras equipped on the trucks offer perception and identification up to 1,000 meters away. TuSimple uses AI to improve safety by allowing the truck to see up to 1,000 meters away. It locates objects at night and brings massive amount of data to simulation to help the vehicle in handling any problem.

Prediction
The company claims that with 360-degree awareness, a virtual driver could see, recognise, and plan manoeuvre or action in a span of 30 seconds. Such a short action time helps to avoid problems, maintain efficiency and keep the vehicle out of danger. TuSimple has tested its autonomous system across its entire fleet day and night on the road carrying freight. Enabling a self-driving truck to improve shipping times for goods and materials and to make highways safer and less congested. The company uses internationally recognised standards recommended practises and guidance from the US Department of Transportation, Aerospace and Military to inform the design, sourcing, verification, and validation of every element of the self-driving system.

Operational Design Domain
The SAE Level 4 (L4) self-driving vehicle operates autonomously under the operational design domain (ODD), according to Milton. The ODD for an L4 self-driving system dictates trucks operating limitations and excludes scenarios and situations for which an effective technical solution has not yet been validated. ODDs can be extremely broad (for example, interstate highways in clear conditions) or extremely narrow. The ODD must be specific and use clear and unambiguous statements to define the limit. The ODD must also take into account road types and conditions, weather, topographic features, speed limits and traffic laws, as well as other jurisdictional regulations. TuSimple’s ODD includes highways and surface streets from depot to depot during night and day and during harsh weather. It also includes parameters for road types, geographic and topographic features, speed limits, and laws and regulations.

Cyber Security
To protect the automated driving system from malicious attacks, TuSimple has designed protection that will help prevent attacks and implement mitigation strategies to minimise the potential impact of any cyber-intrusion. The protection suite has been engineered with cybersecurity specialists to create security protocols that protect vulnerabilities in the self-driving system and all features, components, or tools that interact with the self-driving electronics. Company’s cybersecurity protocols are designed to isolate and protect all onboard systems that communicate with the outside world through a variety of strategies, including physical isolation, firewalls, intrusion detection and authentication.

Future plan
Working towards getting the AFN to start operating in three phases, TuSimple will start with Arizona and Texas routes later this year and the next year. Between the year 2022 and 2023, it will expand with Los Angeles to Jacksonville routes. In the third and final phase, the company will grow its network all over the nation (read US) between 2023-24. The success that TuSimple will gather, will be replicated in Europe and Asia (the Asian pilot will be in Shanghai). Wanting to provide a nationwide transportation network that consists of mapped routes connecting hundreds of terminals to enable efficient, low-cost long-haul autonomous freight operations, TuSimple will open a new shipping terminal in Dallas, according to Xiaodi Hu. With the AFN, it would launch TuSimple Connect, an autonomous operations monitoring system to ensure safe autonomous operations and allow customers to track their freight in real-time. With a post-money valuation of USD 1.095 billion from various rounds, TuSimple has raised around USD 178 million till date. Its list of investors includes Nvidia, ZP Capital, Sina, Composite Capital.