Torc’s Approach to AI

Torc’s Approach to AI

by | Apr 30, 2025 | Uncategorized

The self-driving semi-trucks of the future have arrived – and they’re driving on highways now. At Torc, our autonomous trucks are beyond the research and development of prototypes and moving into our commercialization phase. The trucks that will solve customer needs tomorrow are online and on the road today.

Torc’s self-driving product is powered by both our cutting-edge technology and our rapidly scalable business model. Underpinning it all is responsible and expert use of Artificial Intelligence.

Artificial Intelligence (AI) is a subset of computer science, including deep learning and reinforcement learning, where software is taught to perform tasks for many different applications that would normally require a human touch or direction, like pattern recognition, computer vision, and other forms of decision making. Using AI is especially suited to designing and testing autonomous vehicles, as cutting-edge AI approaches have been shown to deliver the highest accuracy and performance for a vehicle to see, think, and act on the road.

"Training the systems to make decisions through reinforcement learning achieves optimal outcomes.”

– Felix Heide, Head of AI at Torc

AI unlocks the ability for our autonomous driving product to learn to accurately perceive and understand its surroundings, determine what others on the road may do, and safely determine the right actions to take. It ultimately makes our product more reliable and scalable on different road networks in different conditions.

Additionally, our approach to software testing also uses AI, specifically generative AI and neural rendering combined with physics models and techniques. With generative AI, we drastically increase and improve our software’s ability to drive high-volume freight routes, as well as deal with rarely encountered edge cases, allowing our autonomous driving software to experience billions of test miles in a fraction of the time and cost

The Power of Torc’s Virtual Driver

Driving is a skill that people never stop developing. Every time we put the key in the ignition, we learn from our roads and our fellow drivers. We’re constantly tweaking and refining our own driving behaviors with time and experience – and so should our machines.

Enter AI. AI is what powers our software to pick up on changes in the world around it and adjust accordingly, in real-time. For example, whether it’s the ability to pause at a newly installed stop sign or notice construction cones modifying where it is safe to drive, AI powers how the scene is seen and understood, and how the truck should behave, to be the safest vehicle on the road. Its use in self-driving applications cannot be understated. AI can both teach and learn driving skills in less time – specifically what computers were invented to do.

We refer to our AI software suite as Torc’s Virtual Driver: our advanced approach on seeing, thinking, and acting for self-driving trucks. Torc’s Virtual Driver combines cutting edge end-to-end learning and verifiable AI with algorithmic redundancy, allowing Torc to quickly evolve and scale to interpret the world around it, as well as adapt and absorb new sensor technologies and customer routes.

You Might Be Interested In: Daimler Truck’s Autonomous-Ready Fifth Generation Freightliner Cascadia Hits Texas Roads With Torc

Achieving Safe and Accurate Perception, Planning, and Prediction

To train and verify the models within the Virtual Driver, and make our technology as robust as possible, Torc uses novel generative AI approaches for our software testing. It assesses the Virtual Driver by driving billions of miles in every conceivable circumstance (more miles than a human would ever be able to drive in their lifetime) in simulation before it hits the road, and making changes as needed in real-time. AI isn’t just saving us time – it’s saving lives by helping us create situations most humans would never encounter, such as various weather conditions, unexpected pedestrian or vehicle behaviors, and many other edge cases to deeply test that the Torc Virtual Driver works correctly in all conditions and scenarios.

“Overnight, we can create a massive number of heavily optimized scenarios with parallel conditions that we can use for training very easily. This also allows us to scale the data we need much more easily for verification and validation of the software,” says Felix Heide, Head of AI at Torc. “It is possible to train software models and perform object detection without any training data on completely unseen, unannotated sequences of frames. We can use any real-world videos from a vehicle dash cam or test vehicle and use AI to match objects in an image frame, create our own settings, and finally derive a myriad of scenarios in a novel way to get away from the heavy reliance on limited training data sets. Training the systems to make decisions through reinforcement learning achieves optimal outcomes.”

Together, the modular AI powered Virtual Driver, taught and tested by next-level generative world simulation AI and further validated in the real-world, elevates Torc’s AI approach to the next level.

 

Torc’s technology unlocks high-performance verifiable AI, what we call AV 3.0, the highest performance and safety measures in the industry. With real world plus generative AI data loops working together, Torc’s AV 3.0 approach on our production embedded hardware enables fast and predictable product scaling for our fleet customers.

Partnerships Make It Happen

Importantly, our technology is built on best-in-class industry partnerships. Our Virtual Driver runs on the Flex Jupiter high-performance embedded compute platform, powered by NVIDIA DRIVE AGX™ technology. It is factory-integrated within the industry’s first and only autonomous-ready Class 8 truck chassis, the 5.0 Freightliner Cascadia developed by Daimler Truck, providing highest reliability and volume for our fleet customers.

Taken individually, it’s impressive. Together, it’s unequaled. We’re offering our customers unparalleled levels of redundancy and high reliability. And with the Freightliner Cascadia already owning nearly 60% of the U.S. Class 8 long-haul trucking market share today, this product and this partnership cannot be matched.

Torc has the best power, performance, and cost advantages for freight companies moving to adopt self-driving trucking. Our strong collaborations ensure autonomous trucks can be quickly produced at scale to meet manufacturing and market demands by our customers.

You Might Be Interested In: Torc Collaborates with Flex on Physical AI Platform for Autonomous Trucks, Accelerated by NVIDIA

Our AI-powered simulation
make us fast, flexible, and efficient. 

AUTONOMY FORWARD

Torc’s differentiations are distinct and our customer engagements prove we’re on the right path to delivering the right self-driving product at the right time. We’re combining all the pieces in the right order:

  • Our deep integration with Daimler Truck, building the autonomous-ready Freightliner chassis for safer, more reliable, day-to-day autonomous freight operations
  • Cutting-edge AI – AV 3.0 – providing end-to-end self-driving capabilities that are safe, scalable, and adaptable for US road networks
  • The embedded automotive-grade hardware needed to run Torc’s autonomous software, allowing for reliable real-time operation in harsh environments

But that’s not the end of the story. We’re not just working on the software and hardware. Our commercialization team is working closely with fleets today to understanding their needs tomorrow, their existing network infrastructure, and their pain points, so that we have the right application and tooling when the autonomous trucks are on the road.

The road to 2027 is paved with groundbreaking advancements, and with the support of AI technology, market and industry leaders, and the best team in the business, we’re driving toward a future where autonomous trucking transforms the way goods move across the world. Torc is driving the future of freight.

Our use of AI is governed internally by a cross-functional committee. The Generative AI Committee is dedicated to achieving Torc’s business objectives while adhering to the company’s commitment to Do the Right Thing. The Generative AI Committee’s mission is to establish a program to ensure AI systems used across Torc business functions conform with enterprise values, policy compliance, regulatory standards, and industry best practices.

  1. From engines to algorithms: Gen AI in automotive software development, January 2025, https://www.mckinsey.com/features/mckinsey-center-for-future-mobility/our-insights/from-engines-to-algorithms-gen-ai-in-automotive-software-development. AI tools have revealed a “productivity improvement of 44 percent when using gen AI with quality assurance measures, such as creating and automating tests to then enhance efficiency and code reliability.”

Does Torc offer co-ops or internships?

by | Jun 6, 2024 | Frequently Asked Questions

Yes! Our co-op program is a paid program that allows you to gain hands-on experience as you contribute to our mission of making the world a safer place. You’ll collaborate, learn from and network with a multi-disciplinary team of world-class professionals.

Are Torc’s driverless trucks currently on the road today?

by | May 2, 2024 | Frequently Asked Questions

Vehicle testing is paramount to our success. Our routes include public roads in Virginia and New Mexico, chosen in part for their diverse terrain, traffic, and weather conditions. Our testing on closed-course tracks includes Virginia Tech Transportation Institute and a Daimler Trucks facility in Madras, Oregon. In addition to test tracks, Torc has tested numerous vehicle platforms on public roads in 20 states over the past 14 years.

imaginAviation 2024 Panel: How Ideas Become Innovation

imaginAviation 2024 Panel: How Ideas Become Innovation

by | Apr 4, 2024 | Uncategorized

Innovation. Resistance. Transformation. Collaboration. These ideas paved the way for a panel discussion at imaginAviation 2024, featuring guest John Marinaro, Torc’s Vice President of Fleet Operations, along with host Dr. John A. Cavolowsky, NASA’s Director, Transformative Aeronautics Concepts Program (TACP), and guest Sheilla Torres-Nieves, Associate Professor, Fluid Dynamics and Turbulence at the University of Puerto Rico Mayaguez.

Among the more notable topics were resistance to innovation, transformational innovation, and inspiration for innovation.

Resistance to Innovation

When asked to provide examples of how to overcome an unwillingness to adopt innovation or accept change, Marinaro recounted a statement from the Columbia Accident Investigation board that “NASA Safety wasn’t as credible or competent as it should be.” He then explained: “I spent the rest of my career engineering that out of ever being said in an accident investigation again.” As he led innovation of a safety training program, he encountered resistance from some of the senior SMEs that training could be delivered online using revolutionary lecture-capture technology. However, thanks to beta testing, the program had 250 graduates on day one of the safety training’s deployment and proved a successful innovation.

At Torc, one of the primary challenges is resistance to the idea of self-driving vehicles replacing truck drivers. However, a shortage of drivers at the tune of 60,000, Marinaro explained, is disrupting the supply chain and resulting economics. Torc is looking to fill that gap. Marinaro indicated that Torc’s goal is to create safer conditions through technology that produces real-time reactions through awareness of a 360-degree environment, coupled with the reality that the truck doesn’t “get tired.” He concludes, “At the end day, we’re not going to replace the drivers. We’re just gonna augment them and make it safer.”

Transformational Innovation

When Dr. Cavolowsky posed the question of how we apply transformational innovation, how we get there and what kind of innovation we need to bring, Assistant Professor Torres-Nieves answered, “When we hear transformational, we think about changing the way we live drastically…changing culturally…changing from the fundamentals.” Torres-Nieves mentioned the “Change the World” talent competition offered at her university that she and a peer had engaged. In it, the competition gave both training and funding on how to push the idea out, get support, and advertise it to introduce transformational innovation into industry.

Marinaro offered a story regarding the integration of the Cirrus aircraft parachute system into aircraft which has proven to be a successful transformation in flight safety and resulted from an accident where it was clear the life-saving system was needed for pilots.

Inspiration for Innovation

Cavolowsky asked: “Our world is filled with so many issues and problems. How does one go about finding purpose or fulfillment in solving them?” Torres-Nieves’ suggested, “Do what you love.” She recommended aligning purpose with what you do – not that it’s not frustrating or challenging, but that you persist in spite of the challenges. Meanwhile, Marinaro agreed and expressed that one should continue to learn, to press forward. He said, “80% on time is better than 100% late.” To remedy this, he posed that individuals strategize realistic goals and pursue them to the finish, not necessarily to perfection.

View the entire panel above or on YouTube here.

“At the end day, we’re not going to replace the drivers. We’re just gonna augment them and make it safer.”

John Marinaro, Director of Fleet Operations

Understanding the Levels of Autonomy: 0-1-2

by | Mar 16, 2024 | Uncategorized

Depending on how much time you spend watching the news, you may hear a lot about “self-driving vehicles,” and you may even hear speculation about how they can do everything from drive themselves to rescuing and altering the American economy. Those are lofty goals to set for any piece of technology, let alone one that is still in its infancy. Rather than focusing on the benefits of this technology, this article will define what we mean by “self-driving” or “autonomous” technology, as well as the several types of autonomous technology that are currently being developed. 

An autonomous driving system is any system that is partially or fully capable of driving a vehicle without the aid, assistance, or intervention of a human driver. SAE International created the SAE Levels of Driving Automation in 2014 as a classification system for autonomous vehicles with the goal of providing greater clarity and transparency on the subject of autonomous technology for the benefit of everyone. 

An Introduction to the Levels of Autonomy 

Autonomous Driving Systems at Level 0 Autonomy: Manual Driving 

At Level 0 Autonomy, the vehicle is incapable of autonomous driving and is solely controlled by a human driver. All driving decisions and movements are the responsibility of the driver. In the event of an emergency, a warning or help system may be employed to provide assistance or advice. Prior to the development and commercialization of advanced driver assistance features, all vehicles were Level 0 Autonomy. Even today, with technology pervading every aspect of our lives, the vast majority of vehicles are still Level 0 Autonomy. 

Autonomous Driving Systems at Level 1 Autonomy: Assisted Driving 

Level 1 Autonomy is characterized as follows: one aspect of the driving process is taken over in isolation, using data from sensors and cameras, but the driver retains entire control of the vehicle. When the driver activates the driving automation system, it can execute either longitudinal or lateral vehicle motion control, depending on the scenario. The driver must always keep an eye on the system and be prepared to take control of the vehicle if necessary. The development of Level 1 Autonomy and autonomous driving technology began in the late 1990s with the creation of the world’s first radar-controlled cruise control and the debut of lane-keep assist in 2008. These were the initial stages in releasing commercially viable autonomous technology for the masses. They were also the first steps to placing less on the driver and developing vehicles with enhanced active safety features. 

Autonomous Driving Systems at Level 2 Autonomy: Partially Automated Driving 

At this level, computers take on many of the driver’s responsibilities. Today, there are many commercially available vehicles capable of running at Level 2 Autonomy. Their technology makes them intelligent enough to combine speed and steering systems together by relying on multiple data sources, including cameras, radar, LiDAR, and GPS. However, a driver always must be present in the vehicle to take control of the vehicle in the event of an emergency, must keep a continual eye on the system, and best be ready to take over the delegated tasks if necessary. The driver simply must activate the autonomous driving system to use it.  

Under specific scenarios, the driving automation system supports the driver with a portion of the active driving duties by steering the vehicle laterally and longitudinally through traffic flow. This complex form of cruise control combines directional, throttle, and brake operations. It uses sophisticated sat-nav data to autonomously brake for curves ahead, maintain a fixed distance from the car in front, and resume driving when traffic jams clear, all while the driver remains sitting. When the driver makes requests, the system will immediately disengage and allow the driver to resume control of the vehicle. 

In the next part of this series, we will continue learning about the remaining Level 3 to Level 5 of autonomous driving. We will also discuss which level of autonomy Torc is pursuing and why. 

Learn More 

For more information, read Torc’s Vehicle Safety Self-Assessment (VSSA) Report. It gives an overview of our safety culture, our technology, and describes how we are collaborating with industry experts to assure the safe development and commercialization of our autonomous driving systems for autonomous freight trucks.