Blacksburg, VA-based Torc Robotics, the independent commercial vehicle autonomy subsidiary of Daimler Truck AG, announced at Nvidia GTC conference that it is collaborating with Flex and Nvidia to develop its scalable compute system for autonomous trucks that it says is the first deployment of physical AI (artificial intelligence) for autonomous long-haul trucking at production scale. Along with Flex, Torc showcased the advanced capabilities of the joint solution on its demo truck at the GTC event last week in San Jose, CA.

“With Daimler Truck’s autonomous-ready Freightliner Cascadias with in-built redundancy, our work with Nvidia and Flex is already providing a stable and proven foundation for Torc’s autonomous vehicle technology,” said Peter Vaughan Schmidt, Torc’s CEO. “By leveraging Nvidia’s Drive AGX in-vehicle compute and DriveOS, along with Flex’s Jupiter compute platform, we are able to ensure a low-risk, high-confidence path to production that is able to seamlessly transition as Nvidia’s and Flex’s solutions continue to evolve.”

Physical AI is the core of its software stack that enables self-driving trucks in the real (physical) world to navigate their surroundings end-to-end and in real-time using radar, camera, and lidar sensors—a key enabler being the Aeva-supplied Atlas automotive grade 4D lidar. This allows trucks to make informed decisions about lane changing, braking, and obstacle avoiding to ensure safe and efficient autonomous driving operations.

“Showcasing Torc’s production-intent hardware, software, and AD kit on Daimler’s Freightliner Cascadia at this level created a significant wave of attention,” said Andrew Culhane, Torc’s CCO. “We’re thankful for the attention we got from the thousands of GTC attendees, reporters, and technology partners. It’s another proof point that we’re right where we need to be right now, leading the pack in autonomy.”

Torc collaborated with Nvidia on a multi-chip adaptable architecture that leverages Drive AGX, using the Drive Orin SoC (system-on-a-chip) and DriveOS operating system, and Flex for its Jupiter compute design platform and advanced manufacturing capabilities. The company says this provides a scalable high-performance production hardware and software platform based on the autonomous-ready Cascadia equipped with advanced technologies and redundant systems to support the future deployment of autonomous driving capabilities.

“Nvidia Drive AGX has been industry-proven in full production for automotive real-time applications at the edge,” said Rishi Dhall, Vice President of Automotive at Nvidia. “It delivers the high compute performance, low latency, and multi-sensor connectivity needed for Torc’s sophisticated autonomous trucking software, delivering robust perception, prediction, and planning for safe and reliable operation. Torc is on a clear path to scalable production for its commercial launch in 2027 and working toward a seamless upgrade to Nvidia Drive AGX with Drive Thor.”

Torc says that the Nvidia and Flex solutions adhere to its stringent size, performance, cost, and reliability requirements while meeting the total cost of ownership targets of its fleet customers pursuing non-stop, long-haul driverless trucking integration. The collaboration is said to provide a true SDV (software-defined vehicle) functionality because it is adaptable to ever-changing ODDs (operational design domains), including new lanes, routes, hubs, hardware and sensor configurations, operational rules, and road conditions.

“Our collaboration with Torc, Daimler Truck, and Nvidia illustrates how Flex partners across the full ecosystem to enable mobility companies to launch next-generation technology with greater resilience and speed,” said Mike Thoeny, President, Automotive, Flex. “We appreciate the trust Torc and Daimler Truck have placed in Flex through leveraging our Jupiter compute platform and advanced manufacturing capabilities to deliver autonomous long-haul trucking at scale.”

 

Autonomous product validation

In October 2024, Torc announced that it had begun advanced validation of its autonomous trucks without a driver in a multi-lane closed-course environment earlier in the year. Conducted at an operating speed of up to 65 mph to optimize fuel efficiency, the driverless product acceptance test underscored the company’s evolution to productization, positioning the company to scale and commercialize safe, robust autonomous trucking solutions by 2027.

Unlike a demo, this milestone highlights Torc’s entry into scalable product release, with the company’s applied AI technology, system architecture, production-intent embedded hardware, and safety engineering converging to shape a product that prioritizes true software best practices and safer roadways for all.

“This is a key moment in our mission to build a profitable, scalable business as the world’s leading autonomous solution,” commented Schmidt. “We observed impressive reliability in our repeated driverless runs, which leveraged Torc’s unparalleled embedded and integrated platform on Daimler Truck’s Freightliner Cascadia.”

Torc said that the product validation milestone shows its commitment to rigorous safety and maturity standards, marking a critical step from advanced engineering and development to full productization on a unified, embedded platform.

“Autonomous trucking is one of the most concrete applications for AI that can drive demonstrated revenue, business value, and industry transformation, and Torc is at the forefront of creating a safe autonomous solution with safety, scalability, and cost efficiency top of mind,” said CJ King, Torc’s Chief Technology Officer. “With our long-standing tenure in the autonomous space, this milestone reinforces Torc’s safety-focused commitment to driving the future of freight.”

 

Accelerating developments with simulation and AI

As CTO, King brings “a long history” of experience to Torc at automotive OEMs Toyota and Ford, at Amazon working on “pure technology,” and at Here Technologies, where he ran autonomous mapping and the foundational engineering organization. At Ford, he worked on the Argo AI team that built the SAE Level 4 passenger car system and virtual driver.

King provided more insight into company developments exclusively to Futurride at GTC.

Torc is building a hub in Dallas-Fort Worth in support of its operations there. In addition, it just opened an office in Ann Arbor, MI, to grow and centralize its engineering. According to King, the company wanted to consolidate all its operations in a time zone closer to the launch market. It now has a couple of hundred engineers working on “productionization.”

He says that Torc has been working with Nvidia and Flex for over a year, with the company displaying at GTC the Flex Jupiter compute design platform that holds Nvidia’s Drive Orin SoC and DriveOS. That autonomy system is fitted to the full-autonomous-ready Freightliner trucks delivered by Daimler, and Torc makes use of “all their expertise to help us get a fully ready validated system,” said King.

He said one of the biggest challenges “was when we transitioned at the beginning of last year into our full embedded stack.” That was followed by the driver-out demonstration on a close course up to 65 mph, and now Torc is back to running on its launch lane in Texas.

“Now we’re working on the edge cases to make sure we’re solving them and being able to have feature-complete next year,” he said.

It’s now time for his team “to deliver and execute.”

“We’re focused on that iteration and building up the efficiency at the edges of our platform, and that’s why we’re already running our all the roads, and we’re able to cover the core use cases,” he said. “But we know that’s not perfect, so we want to extend the coverage to all the other edge cases.”

Simulation is now playing a larger role in accelerating development.

“Our major investment is around simulation,” he said. “I think that’s a huge requirement. You can do a lot with driving vehicles on the road, but you can’t see everything, and you don’t want to see everything. We use our simulation environment and specifically add on our generative AI synthetic data. That allows us to take the 11 million miles we already have and go to hundreds of millions of miles and scenarios.”

That will be welcome news for the company’s target customers, the large freight operators. Torc has publicly announced Schneider and C.R. England as customers but promises more announcements soon.