NVIDIA Isaac Sim is a reference application built on the NVIDIA Omniverse platform that lets developers design, simulate, test, and train AI-driven robots in physically-based virtual environments. As of 2025 it is open-source under Apache 2.0 and available on GitHub. The simulator is built around Pixar’s OpenUSD scene description format and uses NVIDIA’s PhysX 5 engine for high-fidelity rigid-body and articulated dynamics, with real-time RTX ray tracing for photorealistic camera, depth, and LiDAR sensor simulation. Robots can be imported from URDF, MJCF, Onshape CAD, or USD. Tightly integrated companion modules include Omniverse Replicator for synthetic data generation with domain randomisation, Isaac Lab for GPU-accelerated reinforcement learning, and Isaac ROS for hardware-accelerated ROS 2 packages. A full Python scripting API enables headless or hardware-in-the-loop workflows. The platform is used by Boston Dynamics, Figure, Agility Robotics, 1X, and many academic labs to train policies in simulation before deploying to physical robots, making it the de-facto industrial-strength simulator for modern humanoid and manipulation development. Requires an NVIDIA RTX-class GPU.
Open-source GPU-accelerated robotics simulation framework built on NVIDIA Omniverse and OpenUSD. Provides photorealistic ray-traced environments, PhysX-based physics, multi-sensor simulation, synthetic data generation via Replicator, and Isaac Lab for reinforcement-learning-based robot training.
NVIDIA Isaac Sim is a reference application built on the NVIDIA Omniverse platform that lets developers design, simulate, test, and train AI-driven robots in physically-based virtual environments. As of 2025 it is open-source under Apache 2.0 and available on GitHub. The simulator is built around Pixar’s OpenUSD scene description format and uses NVIDIA’s PhysX 5 engine for high-fidelity rigid-body and articulated dynamics, with real-time RTX ray tracing for photorealistic camera, depth, and LiDAR sensor simulation. Robots can be imported from URDF, MJCF, Onshape CAD, or USD. Tightly integrated companion modules include Omniverse Replicator for synthetic data generation with domain randomisation, Isaac Lab for GPU-accelerated reinforcement learning, and Isaac ROS for hardware-accelerated ROS 2 packages. A full Python scripting API enables headless or hardware-in-the-loop workflows. The platform is used by Boston Dynamics, Figure, Agility Robotics, 1X, and many academic labs to train policies in simulation before deploying to physical robots, making it the de-facto industrial-strength simulator for modern humanoid and manipulation development. Requires an NVIDIA RTX-class GPU.