By: Dawn Zoldi
This year’s CES highlighted that AV truck autonomy has moved from concept to concrete business tool, and Torc Robotics aims to position itself at the center of that transition. Rooted in nearly two decades of trucking experience and backed by Daimler Truck, Torc is building an autonomous platform designed to scale from the first commercial routes in Texas to diesel fleets by 2027.
A CEO Who Has “Been Here Before”
Torc CEO Peter Vaughan Schmidt is not your typical startup experimenter. He has already changed trucking once…and now plans to do it again. Schmidt has spent almost 20 years in the trucking industry, including leading development of Daimler’s “world engine” that powers the Freightliner Cascadia and other flagship models, which helped Daimler gain close to 10% market share.
Now, after serving on Torc’s board and then taking the helm about three and a half years ago, he describes this phase of his career as “the second time” he gets to do something “truly epic” for trucking.
That perspective shapes how Schmidt talks about autonomy. Instead of leading with futuristic scenarios, he starts from the customer side: chronic driver shortage, stubborn safety challenges and the opportunity to redesign supply chains that today are constrained by hours-of-service limits. For him, autonomy provides the best way to move freight more safely, more predictably and, ultimately, more profitably across an industry he knows intimately.
Torc’s Origin Story and Daimler Backing
Torc’s roots reach back to the DARPA Grand Challenges, where founder Michael Fleming and a Virginia Tech team placed third in the final competition that seeded much of today’s autonomy ecosystem. Instead of chasing robo‑taxis, the company turned early to real revenue by making autonomous systems for military and mining applications. Those efforts supported 14–15 years of profitable growth without outside capital.
Everything changed in 2019, when Daimler Truck acquired a majority stake and Torc pivoted fully into on‑highway Class 8 trucking. That relationship enables what Schmidt calls an integrated solution “from the rubber to the roof,” where Torc specifies the autonomous hardware stack (think: redundant chassis systems, sensor suites and compute placement), while Daimler handles sourcing, purchasing and factory integration. The result is hardware designed from day one to be built at scale on the latest Freightliner Cascadia platform, rather than retrofitting a handful of prototypes.
On top of that platform sits Torc’s core product: the “virtual driver.” The team develops the complete software suite, starting with middleware and the operating system up through perception, prediction, planning, motion control and mission control. They intend the stack to be generalizable, so that the same architecture can extend across different truck classes, propulsion types, including electric, and even adjacent markets such as mining or defense when the time is right.
Safety, Scale and a Clear 2027 Target
Torc has anchored its path to commercialization with the specific goal of a joint commercial launch with Daimler Truck in 2027, with driver‑out operations on real freight lanes. According to Vaughan Schmidt, Torc has its first corridor in focus: I‑35 between Dallas–Fort Worth, where Torc runs its operations center, and Laredo at the southern end of the Texas freight spine.
Schmidt characterizes 2026 as the year of “getting all ducks in a row” for this launch: finishing the safety case, completing internal proofs, satisfying Daimler’s requirements and engaging regulators and the public. What is different now, he says, is that the big unknowns are gone. Torc believes it understands what is needed in perception, what evidence is required to remove the driver safely and how to structure the safety case. The remaining challenge is the sheer volume of work, rather than fundamental feasibility.
That focus on scale shows up everywhere. Torc is not racing to be first with a flashy demonstration on a single tightly constrained lane. Instead, the product they are building to scale from tens to thousands and then tens of thousands of units, both in terms of manufacturable hardware and updatable software, right from the start.
Inside the Tech: Long‑Range Perception, No HD Maps and Innoviz LiDAR
Torc highlighted several technical decisions that it believes will matter in the long run. One is an end‑to‑end AI system that still exposes interim results, giving engineers visibility into what the system is “thinking” and why. This hybrid approach aims to retain the efficiency of end‑to‑end training while providing the introspection needed for safety‑critical verification and regulatory confidence.
Another notable decision is to operate without high‑definition maps. Torc’s autonomy stack is designed to drive using perception and more generalized representations rather than relying on detailed HD maps that must be painstakingly created and constantly maintained. That approach should reduce operating costs and allow faster expansion into new routes, an important consideration for fleets that want to add depots and lanes without waiting for mapping teams to catch up.
Perception range is part of the story as well. Torc’s system looks out roughly 1,000 meters, giving the virtual driver a long planning horizon at highway speeds. To achieve this, Torc and Daimler Truck recently selected Innoviz Technologies as their short‑range LiDAR partner for series‑production Level 4 autonomous trucks, with InnovizTwo as part of the sensor suite on autonomous Freightliner Cascadia models. Innoviz’s sensors will work alongside radar and cameras as one of several key components enabling Torc’s Level 4 virtual driver, with a focus on consistent performance and durability across harsh commercial truck environments.
NVIDIA, Flex and the Rise of “Physical AI”
A pivotal piece of Torc’s roadmap is its collaboration with Flex and NVIDIA on what the company calls a scalable “physical AI” compute system for autonomous trucks. This platform combines NVIDIA DRIVE AGX, powered by the DRIVE Orin system‑on‑a‑chip and DriveOS, with Flex’s Jupiter compute design and manufacturing capabilities to create an automotive‑grade brain for Torc’s virtual driver.
In Torc’s framing, physical AI is the core of its software stack that enables trucks to perceive, understand and execute complex actions in the real world using inputs from LiDAR, radar and cameras. The system supports end‑to‑end, real‑time perception and navigation so the truck can make informed decisions about lane changes, braking and obstacle avoidance at highway speeds, and it has already been validated in driverless product acceptance testing on a closed course.
This NVIDIA‑ and Flex‑based compute platform is designed to meet Torc’s strict requirements on size, performance, cost and reliability while aligning with the total cost of ownership expectations of long‑haul fleet customers pursuing near‑continuous operation. It also underpins a true software‑defined vehicle approach: the same hardware and operating system can adapt to new routes, hubs, sensor configurations and operational rules without redesigning the entire system.
NVIDIA notes that DRIVE AGX has already been proven in volume automotive deployments, offering the high compute performance, low latency and multi‑sensor connectivity needed for sophisticated autonomous trucking software. Torc, in turn, emphasizes that the collaboration provides a “low‑risk, high‑confidence path to production,” with a roadmap toward future upgrades such as NVIDIA DRIVE Thor as compute needs and capabilities evolve. By marrying Daimler’s autonomous‑ready Cascadia chassis, Torc’s virtual driver, Innoviz’s LiDAR and the NVIDIA/Flex compute platform, the company is building a stack that is meant not just to work, but to scale.
Safety Research With Stanford: AI You Can Trust
As autonomy moves closer to large‑scale deployment, the safety of the underlying machine learning systems is coming under increasing scrutiny. Torc has taken a visible step here by joining the Stanford Center for AI Safety to conduct joint research on AI safety for Level 4 autonomous trucking.
Membership gives Torc the ability to sponsor and coauthor research, access symposiums and seminars, and apply leading academic work on robust and reliable machine learning directly into its virtual driver. Torc’s Chief Safety Officer Steve Kenner framed the collaboration as support for Torc’s mission to deliver safe, scalable and trustworthy autonomous solutions, particularly as the company prepares for market entry.
For a technology that will be operating 24/7 on public highways, this kind of partnership is about more than mere optics. Stanford’s research on risk mitigation, robustness and safety protocols feeds into how Torc designs, tests and validates AI behaviors across the operational design domain. The goal is not only to meet regulatory expectations, but to strengthen confidence among fleets, drivers, regulators and the public that the system behaves predictably under both normal and edge‑case conditions.
Customer Pull, Depot‑to‑Depot Vision and AV Readiness
On the commercial side, Vaughan Schmidt describes customer demand as “enormous.” Large fleets see autonomy as a way to tackle driver shortages and safety risks while reshaping network design. Because autonomous trucks are not constrained by human hours‑of‑service rules, fleets can reconsider where to place depots, how to balance regional versus long‑haul runs and how to smooth peak flows across their networks.
Torc’s technology is being developed for depot‑to‑depot operations, not just the traditional “hub‑to‑hub” model. That flexibility opens opportunities for both short runs of around 200 miles and longer multi‑state routes, across common freight types. Initially, the focus will likely be on the largest fleets that have dense, repeatable flows between key nodes and the operational sophistication to integrate autonomous assets. Over time, as the technology proves itself and costs fall, Torc expects capabilities to flow down, full-circle, to mid‑sized and smaller fleets and to spill back into adjacent markets such as mining and defense.
From a vehicle perspective, Torc and Daimler are designing the platform so it can span multiple truck classes and propulsion systems. While today’s test vehicles are built on diesel Freightliner Cascadia tractors, the same core autonomy stack is intended to support other AVs, as Daimler Truck and its customers expand AV offerings in North America, Europe and Asia. That “from the rubber to the roof” integration strategy is particularly relevant for AV truck autonomy, where packaging sensors, compute and high‑voltage systems inside an OEM‑engineered chassis matters for safety, thermal management and uptime.
For fleets watching AV truck autonomy mature, Torc’s story provides yet another proof point that the autonomous vehicle industry is moving from pilots to products, with 2027 as a credible horizon for scaled depot‑to‑depot operations on real freight lanes.