NVIDIA Develops Robotics Technology To Move Humanoid Robots Beyond Simulation

Nvidia has expanded its robotics development platform with a new humanoid robot reference system, artificial intelligence training tools and software designed to help developers move robots from simulation into real-world environments. Announced at the company's GTC Berlin event on October 21, 2026, the updates to its Isaac robotics platform aim to simplify the process of building, training and deploying general-purpose robots. Nvidia CEO Jensen Huang said advances in robotics are accelerating rapidly, with the company seeking to make physical AI more practical across industries.

A key announcement was the Agile ONE SN, a humanoid robot reference platform developed by Germany-based Agile Robots using Nvidia's Isaac GR00T technology and Jetson Thor computing system. The robot stands 1.4 metres tall, weighs 40 kilograms and features 31 degrees of freedom, allowing it to perform a range of coordinated movements. Its hands are designed for dexterous manipulation, while cameras, lidar sensors and other equipment help it perceive and interact with its surroundings. The integrated platform is intended to reduce the time developers spend connecting separate tools for simulation, training and deployment, potentially cutting setup from weeks to days.

Nvidia also introduced Isaac Lab 3.0, an open-source framework that enables developers to train and refine robot behaviours using simulated environments. The software supports multi-GPU workflows and physics engines, including Nvidia PhysX and Newton, helping researchers test how robots respond to different conditions before deploying them on physical machines. Simulation can reduce the need for repeated physical trials, potentially saving time and resources while helping developers identify performance problems earlier in the development process.

Also Read: Deepinder Goyal Introduces Scientists Behind Temple’s BrainFlow Technology

Another addition is Isaac OS, an open-source runtime framework designed to connect AI models with sensors, controls and robot hardware. Nvidia has also introduced new reference workflows for its DGX Station systems, covering reinforcement learning, dexterous manipulation and training of world models. The company expanded its training resources with the SimReady Central Library, which contains more than 2,000 simulation-ready robotics and factory assets, and the Human 1M Dataset, featuring human activity videos and three-dimensional hand tracking. These resources are intended to help developers generate training data and build more capable robotic systems.

The expanded platform is aimed at applications ranging from industrial automation and warehouse operations to humanoid robots designed to work in environments built for people. Several robotics companies, including ABB Robotics, KUKA and NEURA Robotics, are using Nvidia technologies to develop or simulate robotic systems. Agile Robots plans to make the Agile ONE SN available in early 2027, with universities and research institutions among the potential users. However, widespread adoption will depend on factors such as reliability, safety, cost and the ability of robots to perform useful tasks consistently outside controlled environments.

Also Read: Chinese Companies Spend $35 Billion on Nvidia GPUs Despite Limits on Advanced AI Chips