What Is Isaac Sim?#

Import robots and scenes from URDF, MJCF, Onshape CAD, or USD. Simulate with PhysX or Newton, add RTX and physics-based sensors, generate synthetic data, prepare robots for Isaac Lab, and validate robot stacks with ROS 2.

Getting Started#

Pick the setup that matches how you work. Most users should start with Quick Install. Choose Python or containers when you need pip, conda, CI, or remote workflows.

Quick Install
Fastest path to a working local setup
Quick Install
Workstation Setup
Install the full app and local dependencies
Workstation Installation
Container Setup
Run Isaac Sim in Docker for repeatable setups
Container Installation
Python Environment
Use pip or conda for Python-first workflows
Python Environment Installation

Tip

Running into issues? See Setup Tips for common fixes or the Troubleshooting page.


Tutorials#

Start with the topics users look for most: first simulation, robot import, sensors, ROS 2, synthetic data, and robot learning.

BeginnerLearn the app, scenes, and core robot workflows
Basic Usage Tutorial

First steps: navigate the UI, load a scene, and run your first simulation.

Isaac Sim Basic Usage Tutorial
Python Scripting Intro

Write your first standalone script to control robots and environments.

Python Scripting and Tutorials
Import Your First URDF

Bring a URDF robot into Isaac Sim, configure it, and simulate it on a stage.

Tutorial: Import URDF
IntermediateConnect ROS 2, control simulations, and build data generation workflows
ROS 2 TurtleBot Series

Follow the TurtleBot flow from import and driving to sensors, timing, and transforms.

URDF Import: Turtlebot
Synthetic Data with Replicator

Generate labeled training data from Isaac Sim scenes with Replicator.

SDG Workflows
ROS 2 Simulation Control

Use ROS 2 services and actions to load worlds, spawn entities, and step simulations.

ROS2 Simulation Control
AdvancedTrain policies, randomize scenes, and deploy results
Prep a Robot for Isaac Lab

Rig your robot and stage a scene in Isaac Sim so Isaac Lab can train policies on it.

Isaac Lab
AMR Navigation Synthetic Data

Drive an AMR through randomized warehouse scenes and capture stereo camera data when it nears objects of interest.

Randomization in Simulation – AMR Navigation
ROS 2 Policy Evaluation

Run a reinforcement learning policy through ROS 2 while Isaac Sim publishes observations and receives actions.

Running a Reinforcement Learning Policy through ROS 2 and Isaac Sim

Isaac Sim Workflow Overview#

Simulation Development Loop

Bring assets in, configure the robot and scene, simulate behavior, then connect external stacks.

Each stage stays reusable: asset prep, robot and scene configuration, simulation, and stack connection all operate on the shared Isaac Sim scene.

For SDG, label the scene, vary conditions, simulate behavior, and render sensor outputs for downstream datasets.

For SIL, configure robot physics, sensors, and communication graphs, then validate the external robot stack before hardware.

01 Import
Scenes, robots, sensors, assets
Scene assets Target objects Sensor assets Robot assets Robot descriptions CAD / USD / NuRec
02 Configure
Shared setup, then workflow-specific wiring
Materials Sensors Scenarios Semantics Randomization Tune robot physics Communication graph
03 Simulate
Run the world and capture evidence
Physics stepping Sensor output Randomization Annotation capture Rendered frames Control loop Stack behavior
04 Connect / Deploy
Send results to training or robot stacks
Dataset writers Training pipelines Model evaluation Failure-case tests Robot stack Pre-hardware tests
Shared Isaac Sim Scene
USD scene, physics state, sensors, semantics, and graphs in one runtime.
01 Import Bring robot, scene, sensor, CAD, DCC, and reconstructed assets into a shared USD workspace.
02 Configure Set materials, sensors, scenarios, semantics, robot physics, and communication graphs.
03 Simulate Run physics, sensor output, Replicator capture, and stack behavior on the assembled scene.
04 Connect / Deploy Export datasets to training pipelines or connect external robot stacks for pre-hardware validation.

Robotics Ecosystem#

Understanding the components of the NVIDIA robotics ecosystem and where Isaac Sim fits among them.

Each row shows the workflow step and the NVIDIA component that supports it.

  1. 01
    Build the scene & rig the robot
    Isaac Sim Robotics simulator
    Open docs

    Assemble USD scenes, run physics and sensors, connect external robot stacks.

  2. 02
    Train an RL or IL policy Optional
    Isaac Lab RL / IL framework
    Open docs

    Train RL and imitation-learning policies with parallel environments.

  3. 03
    Evaluate policy at scale Optional
    Lab - Arena Policy benchmark
    Open docs

    Benchmark and compare trained policies across many scenes and seeds.

  4. 04
    Run the integrated SIL test
    Isaac Sim SIL test stack
    Open docs

    Run software-in-the-loop tests with your ROS 2 or Isaac ROS robot stack.

Each row shows the workflow step and the NVIDIA component that supports it.

  1. 01
    Bring in real-world environments Optional
    NuRec Reconstructed scenes
    Open docs

    Gaussian-splat reconstructions of real environments, loaded as USD assets.

  2. 02
    Assemble & configure the scene
    Isaac Sim Robotics simulator
    Open docs

    Assemble USD scenes, configure robots and sensors, run physics and rendering.

  3. 03
    Define variation, simulate, & write annotations
    Replicator SDG framework in Isaac Sim
    Open docs

    Script randomization, capture sensor outputs, and write labeled datasets.

  4. 04
    Photoreal augmentation Optional
    Cosmos Transfer Photoreal augmentation
    Open docs

    Convert rendered RGB plus a text prompt into varied photoreal images, offline.


Open Source & Community#

Isaac Sim is open source and built to fit into existing robotics stacks. Use the shipped tools, read the code, or extend the simulator with Python and Kit.

Open-source platform
Read the code, extend the simulator, and fit it into your stack.
Start with the built-in tools, then automate with Python, build custom Kit apps, or integrate Isaac Sim into your own ROS 2 and ML workflows.
Apache 2.0 Open-source licensing for the simulator stack.
USD-native One scene representation from asset import to deployment.
PhysX + Newton Switch between supported physics backends in one simulator.
RTX + Physics Sensors Use rendering and physics-based sensor models in one place.