Developing a Production-Ready Autonomous Domain Controller with NVIDIA

By TuSimple's editorial team
Mar 04

Share

The autonomous domain controller (ADC) is a powerful computing platform essential for autonomous driving. It is where all the critical processing and decision-making of an autonomous driving system is performed. However, despite its importance to autonomous driving, the ADC is not yet commercially available in the market today.

By collaborating with our long-time partner and industry-leader, NVIDIA, TuSimple plans to change this. Together we intend to bring one of the first purpose-built, high-performance, production-ready ADCs for autonomous commercial operations to market. We plan to design the world’s most advanced ADC which will be engineered specifically for TuSimple’s Level 4 (L4) autonomous trucks. This ADC will be designed to incorporate the NVIDIA DRIVE Orin system-on-a-chip (SoC) and run TuSimple autonomous driving software.

What is an ADC?

An autonomous domain controller is a specialized computer that controls specific groups of functions, or “domains” of activity or features, enabling the entire autonomous driving software stack. An ADC streamlines the electronic and software architecture by reducing the complexity of computations.

The ADC receives information from sensors such as cameras, LiDARs and radars as well as other sources. It processes the data and constructs a digital representation of the surrounding environment. The ADC software then makes policy and planning decisions surrounding the actions the vehicle should take, and eventually controls the critical functions based on the planning decisions.

An ADC will be incorporated broadly within the TuSimple autonomous driving system (ADS) as follows:

  • Sensor Suite: The vehicle sensors, including cameras, LiDARs and radars, collect data and input from the environment and send that to the ADC.
  • ADC: Powering hundreds of trillions of operations per second (TOPS), the ADC enables rapid interaction between TuSimple’s perception system, planning systems and vehicle actuation controls.
    • Perception: The ADC receives input from the vehicle sensors which is then combined with data from our proprietary high-definition maps. The system then constructs a digital representation of the truck’s surroundings.
    • Planning: TuSimple’s autonomous driving software, located on the ADC, makes policy and planning decisions to determine appropriate driving decisions.
    • Control: Based on these policy and planning decisions, the ADC sends control signals to the truck’s engine, steering, braking, and other critical functions.
  • Actuation: Engine, steering, braking, and other functions are dynamically applied based on policy and planning decisions sent by the ADC.

A Customized Approach

We are working with NVIDIA on a highly customized hardware and software solution. After successfully conducting the world’s first driver-out semi-truck runs (without a driver or other human intervention) on open public roads, we understand what it takes to conduct driver-out commercial operations.

Given our track record and learnings, we plan to work directly with NVIDIA to ensure that our system requirements are directly incorporated into the ADC specifications, including compute power, operating systems, sensors, and other on-board software and hardware requirements. This will enable us to build a better product together.

This vertical integration of the ADC into our autonomous driving system will help us mitigate the risk of relying solely on a nascent supply chain. It will also enable fast iterations and development of different concepts and generations. The ability to iterate faster, because of our direct relationship with NVIDIA, relative to standard industry practice, should allow us to reach an optimal design earlier and, ultimately, deploy the solution within our Autonomous Freight Network more quickly.

Pioneering the Future

As an early investor in TuSimple, NVIDIA has been a longstanding partner and supporter of our technology and mission to bring the safest, most fuel-efficient, and low-cost freight capacity to the market. We are extremely excited about the potential this collaboration has in accelerating the development and commercialization of autonomous trucking and, ultimately, reshaping the $4 trillion global truck freight market.

Posted by TuSimple

Recent Blog Posts