# Doosan wants its robots to graduate from programmable arms into AI-first industrial workers, and NVIDIA is supplying the nervous system Source: TechNewsList (https://technewslist.com) Canonical URL: https://technewslist.com/en/article/doosan-nvidia-agentic-robot-os-2026-07-03-night Section: Drones & Robots (https://technewslist.com/en/drones-robotics) Author: TechNewsList Language: en Published: 2026-07-03T20:51:02.154+00:00 Updated: 2026-07-03T20:51:02.31362+00:00 > Doosan's June 7 expansion with NVIDIA matters because it combines Isaac Sim, Isaac Lab, Cosmos, Newton, and Jetson Thor inside an Agentic Robot OS aimed at making industrial robots more adaptive, mobile, and economically scalable. ## TL;DR - NVIDIA and Doosan Group expanded their collaboration to cover robotics, AI-factory power systems, and advanced data-center materials. - For robotics, Doosan is integrating Isaac Sim, Isaac Lab, Cosmos, Newton, and Jetson Thor into an Agentic Robot OS. - The key shift is from fixed collaborative robot arms toward systems that can perceive, simulate, learn, and reason through more varied industrial tasks. ## Key points - Doosan says the stack will support simulation-to-real workflows, physics calibration, on-device inference, and AI reasoning. - The companies are targeting use cases such as depalletizing, sanding, dual-arm systems, and humanoid form factors. - This is part of a wider Doosan strategy that extends physical AI beyond robot arms into construction machinery and power equipment. - NVIDIA's role keeps growing from AI compute supplier into an operating substrate for industrial autonomy. - The practical test will be whether these systems can move from demos into repeatable factory deployments with measurable labor and throughput gains. # Doosan wants its robots to graduate from programmable arms into AI-first industrial workers, and NVIDIA is supplying the nervous system ## What happened NVIDIA said on June 7 that it is expanding its collaboration with Doosan Group across physical AI, robotics, and AI factory infrastructure. The robotics portion is the most interesting for this category: Doosan Robotics is integrating NVIDIA Isaac Sim, Isaac Lab, Cosmos world foundation models, the Newton physics engine, and Jetson Thor into what it calls an Agentic Robot OS. ![Contextual editorial image for Doosan wants its robots to graduate from programmable arms into AI-first industrial workers, and NVIDIA is supplying the nervous system Doosan Robotics NVIDIA Agentic Robot OS Isaac Sim Jetson Thor NVIDIA Blog NVIDIA Developer technology news](https://www.wisematic.com/wp-content/uploads/2022/10/M-SERIES_Familyshot_1-reduced-size-1280x884.png) *Contextual visual selected for this TechPulse story.* This is not just another partnership headline about robotics optimism. The stated goal is to help Doosan's machines perceive, reason, simulate, learn, and run inference in ways that make them more adaptable inside real industrial environments. NVIDIA and Doosan also say they are exploring specific industrial tasks such as depalletizing and sanding, as well as new form factors including dual-arm and humanoid platforms. In other words, the ambition is to move Doosan from being mainly a robot-arm provider toward becoming what it describes as a full-stack AI-first robotics company. ## Why it matters This matters because industrial robotics is running into the limits of classical automation. Many factories already automate narrow, highly structured tasks well. The harder opportunity is everything that changes too often, arrives imperfectly, or requires a machine to adapt to variation in materials, placement, or physical context. That is where physical AI becomes commercially interesting. The promise is not simply that robots become smarter in the abstract. It is that the cost of handling variable tasks falls enough for more factories, warehouses, and industrial sites to automate work that previously stayed manual. Doosan's announcement is important because it frames the problem correctly. Simulation, learning, physics, perception, and on-device inference all have to work together. A better robot arm alone is not the answer if the operating system behind it cannot generalize. ## Technical details The announced stack covers the full development loop. Isaac Sim provides the simulation environment, Isaac Lab supports robot learning workflows, Cosmos contributes foundation-model-style world understanding, Newton provides physics support, and Jetson Thor handles edge AI execution. Doosan's Agentic Robot OS is supposed to connect those pieces into one system spanning perception, reasoning, simulation, learning, and deployment. ![Contextual editorial image for Doosan wants its robots to graduate from programmable arms into AI-first industrial workers, and NVIDIA is supplying the nervous system Doosan Robotics NVIDIA Agentic Robot OS Isaac Sim Jetson Thor NVIDIA Blog NVIDIA Developer technology news](https://www.wisematic.com/wp-content/uploads/2022/10/A0509-reduced-size-1280x1086.png) *Contextual visual selected for this TechPulse story.* That matters because industrial robotics increasingly depends on simulation-to-real transfer. It is not efficient to hand-program every environmental variation on a real line. The better route is to train and validate in simulation, calibrate physics, and then move policies into real-world tasks with stronger perception and reasoning. The target tasks mentioned by the companies are also revealing. Depalletizing and sanding are not trivial showcase motions. They involve irregularity, contact, force, and repeated adjustment. Targeting those workflows suggests Doosan wants to prove practical industrial usefulness rather than just autonomous novelty. ## Market / industry impact For robotics, this is another sign that the competitive center is moving from hardware shape to software stack depth. The companies that win may be the ones that can unify data generation, simulation, learning, and deployment into repeatable industrial workflows. For NVIDIA, it reinforces a pattern already visible across the market: the company is embedding itself not only in AI data centers, but also in the operating substrate for physical AI. That makes its ecosystem harder to replace because customers adopt a development pathway, not just a processor. For Doosan, the upside is strategic differentiation. If it can successfully move from collaborative robot hardware into AI-first robotic systems, it becomes more than a component supplier in a crowded automation field. ## What to watch next Watch whether Doosan publishes concrete customer deployments, throughput data, or task-level reliability metrics. Industrial robotics stories only become convincing when performance claims turn into operational numbers. It is also worth watching whether the company's humanoid and dual-arm references evolve into real pilot programs or remain long-range ambition. Finally, watch how much of the value comes from simulation-to-real speed. If Doosan can meaningfully shorten deployment and retuning cycles, that may matter more than any single hardware breakthrough. ## Sources - [NVIDIA Blog: NVIDIA and Doosan Group Collaborate to Advance Physical AI and AI Factory Infrastructure](https://blogs.nvidia.com/blog/nvidia-and-doosan-group-physical-ai/) - [NVIDIA Developer: Isaac robotics platform](https://developer.nvidia.com/isaac) Mentions: Doosan Robotics, NVIDIA, Agentic Robot OS, Isaac Sim, Jetson Thor ## Sources - [NVIDIA Blog](https://blogs.nvidia.com/blog/nvidia-and-doosan-group-physical-ai/) - [NVIDIA Developer](https://developer.nvidia.com/isaac)