NVIDIA 擴展 Isaac 平台，加入 Agile ONE SN 參考人形機械人

NVIDIA announced an expansion of the NVIDIA Isaac open robotics development platform at GTC Berlin on October 21, 2026, adding a reference humanoid, new datasets and a runtime alongside updates to its robot learning framework.

NVIDIA said the new technologies together give developers an integrated path from data to deployment aimed at bringing general-purpose robots to the real world faster. “General-purpose robotics is advancing at extraordinary speed — the ChatGPT moment for robotics is fast approaching,” said Jensen Huang, founder and CEO of NVIDIA. “With NVIDIA Isaac, our full-stack open robotics platform, we are giving developers the models, data, simulation tools and computing systems to bring physical AI to the world’s largest industries.”

Agile ONE SN Reference Humanoid

Agile Robots introduced Agile ONE SN, a full-stack reference humanoid preconfigured with NVIDIA Isaac GR00T and powered by the NVIDIA Jetson Thor edge AI platform. NVIDIA said the system unifies the simulation, training and deployment tools needed to take robot skills from development into the physical world, cutting setup time from weeks to days.

The humanoid uses the Agile ONE S body, standing 1.40 meters tall and weighing 40 kilograms, with 31 degrees of freedom and a 3-kilogram payload per arm. Its dual WUJI Hand 2 five-finger hands each provide 20 degrees of freedom, 1kHz force control and tactile sensing for dexterous manipulation. Jetson Thor delivers 2,070 FP4 teraflops of compute and 128GB of unified memory for onboard reasoning and real-time control, and a multimodal sensing suite combines head and wrist cameras, lidar sensors and inertial measurement units.

The Isaac GR00T platform supplies the GR00T foundation model, a workflow running from data capture to deployment and whole-body control for coordinated movement. Agile Robots also plans to integrate on-robot safety protocols with NVIDIA Halos, NVIDIA’s full-stack safety system for robotics.

“The breakthrough for humanoids is usefulness at scale: one adaptable machine that learns many skills and works across environments built for people,” said Zhaopeng Chen, founder and CEO of Agile Robots. Chen described Agile ONE SN as a full-stack, open reference humanoid that gives developers a ready system to create, train and validate the capabilities needed to put humanoids to work.

Agile ONE SN is scheduled for availability in early 2027 and is designed for frontier humanoid research and development, with interest from institutions including Stanford University and ETH Zurich, according to NVIDIA.

Isaac Lab 3.0, Isaac OS and Open Datasets

NVIDIA Isaac Lab 3.0 is now generally available as an open source, agent-ready robot learning framework for training and refining robot policies from desktop to data center, using multi-GPU workflows and physics engines including Newton and NVIDIA PhysX. The new NVIDIA Isaac OS is an open source runtime framework built on NVIDIA Holoscan, connecting AI, sensors, controls and robot hardware for deterministic, real-time performance from simulation to physical robots.

The SimReady Central Library offers more than 2,000 simulation-ready robotics and factory assets, while the Human 1M Dataset covers diverse human activities with synchronized three-view video and 3D hand tracking. Both are available on Hugging Face, with the full Human 1M Dataset accessible to Cosmos Coalition partners. SimReady Central Library assets come from Comau, FANUC, ForgeCAD, Franka Robotics, LG Electronics, Palatial, Robotiq, Standard Bots, Universal Robots and Vertiv.

NVIDIA also released three open source reference workflows for DGX Station, covering refinement of GR00T 1.7 models with reinforcement learning, training of dexterous manipulation skills in Isaac Lab, and post-training of NVIDIA Cosmos world models for synthetic data generation using ChatGPT Astra. Solutions from 1X, Agile Robots, Humanoid, Maven Robotics, Mentee, MicroAGI, Rhoda, Sereact and Skild AI will support the workflows.

NVIDIA Halos, built on NVIDIA IGX Thor, is a full-stack safety platform for moving autonomous vehicles and robots from development to deployment. Agile Robots, Agility, Dexory, FORT Robotics and KION are integrating Halos within their safety architectures, including for warehouse robot safety, and SICK is evaluating Halos to advance dynamic safety for industrial robots operating beyond fixed zones. Advantech and NexCOBOT are developing NVIDIA IGX-based industrial computing platforms, while Reynolds & Moore, Accenture and TÜV SÜD are working to accelerate certification for the safe deployment of industrial physical AI.

European Robotics Partners Build on Isaac

ABB Robotics is integrating NVIDIA Omniverse libraries with RobotStudio HyperReality to simulate and virtually commission robotic systems before installation. KUKA Group is using Omniverse libraries in Visual Components FactoryLens to add real-time rendering for factory simulation workflows. Sereact is using Isaac GR00T with Cosmos Transfer to train and post-train its Cortex goods-handling platform, which runs on Universal Robots industrial robot arms.

Humanoid is using Jetson Thor, Holoscan Sensor Bridge, Halos and Isaac OS to develop and validate robot behaviors ahead of industrial pilots. Hexagon Robotics is building perception capabilities for its AEON robot, including teleoperation video, localization, mapping and object pose detection. Bear Robotics is accelerating perception and model execution for its KR1 robot and post-training Isaac GR00T through its Kino-2 system for warehouse, returns and industrial workflows.

Toyota Material Handling Europe and SoftServe are using Omniverse and the Isaac Sim framework to develop autonomous material-handling solutions, an approach NVIDIA said reduces deployment time by up to 50%. Idealworks is using NVIDIA open models and Jetson hardware to develop memory capabilities for autonomous mobile robots in factories, and Analog Devices is building autonomous mobile robots on Jetson AGX Orin with the cuVSLAM library.

Hugging Face and Pollen Robotics are using Isaac Lab, Isaac Sim and Newton to build Microduck, a small open source robot designed to help developers and students train, modify and share new physical behaviors. Franka Robotics is using NVIDIA NIM microservices and Isaac Sim to turn natural-language requests into actions on its Franka FR3 Duo robot system. Bellboy is using Jetson and GR00T to power Billie, its first robot teammate, which is already deployed at hotels operated by Marriott, Hilton, IHG and OYO.

Flexion is using Isaac Lab and Newton to scale learning for Reflect, a unified robot intelligence model combining reasoning and whole-body movement for complex manipulation across robot bodies. Schaeffler is contributing its actuator technology to Newton as an exclusive foundational component, and NEURA Robotics is using Isaac Sim and Isaac Lab to generate synthetic training data and train robot skills in simulation for Neuron AI, its multimodal foundation model.