Unitree Go2 twin in Isaac Sim 5.x — studio HDRI lighting, RT2 quality render.

Welcome to the Unitree Go2/G1 Digital Twins Project!

We are thrilled to announce that the Unitree Go2/G1 robot has now been integrated with the Nvidia Isaac Sim (Orbit), marking a major step forward in robotics research and development. The combination of these two cutting-edge technologies opens up a world of possibilities for creating and testing algorithms in a variety of simulated environments.

Get ready to take your research to the next level with this powerful new resource at your fingertips!

What this fork adds

This fork of abizovnuralem/go2_omniverse adds a real-robot digital twin mode and an Isaac Sim 5.x / ROS 2 Jazzy port, plus higher-fidelity default rendering.

Real-robot digital twin (./run_twinbot.sh)

twinbot mode drives the simulated Go2's joints directly from the physical robot's /lowstate stream (forwarded off the Jetson by scripts/twinbot_bridge.py), so the sim mirrors the real dog 1:1 via IMU-aligned kinematic playback. Details in docs/TWINBOT_IMU_FIX.md.

Scope, stated honestly:

• Real → Sim (physical robot state into the twin): implemented and IMU-aligned.
• Sim → Real (commands back to the robot): the standard Unitree SDK control path, not a closed, learned loop. We do not claim more than that.

Rendering

• The default scene is lit with Isaac's bundled studio HDRI (image-based lighting), so the robot's shells catch real reflections instead of reading as a flat gray dome.

• ./run_twinbot.sh launches the quality RT2 render preset (--rendering_mode quality).

• Hero stills are produced headlessly with --capture, which renders the Go2 from several angles via an isaaclab Camera render product (no on-screen window needed):

  ./run_sim.sh --capture 80 --headless --enable_cameras --rendering_mode quality \
      --capture_dir ./media
  

• For cinematic demo footage, record the joint trajectory live, then re-render it offline in RTX path-tracing mode (live closed-loop favours the real-time preset).

 

Isaac Sim 5.x boot fixes

• ROS 2 bridge libraries are resolved from isaacsim.ros2.core (Isaac 5.x relocated them out of isaacsim.ros2.bridge), with a fallback to the old path.
• Articulation buffers are converted from warp arrays to NumPy before indexing (IsaacLab 4.5 / Isaac Sim 6.0 compatibility).
• See JAZZY.md for the venv setup and what is / is not validated.

Real time Go2 Balancing:

Real time G1 Balancing:

Go2 Ros2 Camera stream:

URDF real-time joints sync:

Foot force data stream:

Real-time RTX lidar stream:

Custom envs (Office):

Custom envs (Warehouse):

VR support:

Project RoadMap:

1. PPO balancing algorithm ✅
2. Keyboard real time control ✅
3. Camera stream to ROS2 ✅
4. RTX Lidar stream to ROS2 ✅
5. IMU data stream to ROS2 ✅
6. URDF real-time joints sync ✅
7. Foot force data stream ✅
8. Real-time control from ROS2 ✅
9. Nav2 with Slam_toolbox ✅
10. Bunch of RL-envs for custom dog training ✅
11. Custom numbers of robots ✅
12. Added G1 Unitree support ✅

Your feedback and support mean the world to us.

If you're as enthusiastic about this project as we are, please consider giving it a ⭐ star on our GitHub repository.

Your encouragement fuels our passion and helps us develop our RoadMap further. We welcome any help or suggestions you can offer!

Together, let's push the boundaries of what's possible with the Unitree Go2/G1 and ROS2!

Cloud Services from @robolaunch

If your system is not sufficient for Isaac Sim or Isaac Lab etc., this is where Robolaunch offers you a new service. You can do real-time work by simply connecting to the computer on the remote server via a web browser.

If you would like to get more detailed information about Robolaunch web platform payments or anything else, contact us!

System requirements and installation

There are now two supported tracks:

Track A — original, Humble-based (Ubuntu 22.04, Isaac Sim 2023.1.1, Orbit 0.3.0)

This is the path the repo was originally designed for. Instructions below are unchanged.

Track B — experimental, Jazzy-based (Ubuntu 24.04, Isaac Sim 5.0, IsaacLab 0.54.3)

A pip-into-venv setup that uses the ROS 2 Jazzy runtime bundled inside Isaac Sim (resolved from the isaacsim.ros2.core extension on Isaac 5.x, or the older isaacsim.ros2.bridge as a fallback), rather than a system /opt/ros/jazzy install. See JAZZY.md for the full rationale (including why system Jazzy's Python 3.12 is not compatible with Isaac Sim 5.0's Python 3.11), setup, and what is / is not validated. ./run_sim.sh and ./run_sim_g1.sh are the launchers for this track; they no longer require conda activate orbit.

Track A requirements

1. Ubuntu 22.04
2. Nvidia Isaac Sim 2023.1.1
3. Nvidia Orbit 0.3.0
4. Ros2 Humble

Full instruction:

After installation of Nvidia Isaac Sim 2023.1.1 and Ros2 Humble:

1. Clone this specific IsaacLab repo version: https://github.com/isaac-sim/IsaacLab/releases/tag/v0.3.1
2. Execute in ubuntu terminal:
export ISAACSIM_PATH="${HOME}/.local/share/ov/pkg/isaac-sim-2023.1.1"
export ISAACSIM_PYTHON_EXE="${ISAACSIM_PATH}/python.sh"
and also put it inside .bashrc file
3. Inside the root folder of Orbit repo (https://github.com/isaac-sim/IsaacLab/releases/tag/v0.3.1) execute ln -s ${ISAACSIM_PATH} _isaac_sim
4. Execute ./orbit.sh --conda
5. Execute conda activate orbit
6. Execute sudo apt install cmake build-essential
7. Execute ./orbit.sh --install
8. Execute ./orbit.sh --extra rsl_rl
9. Verify the installation using "python source/standalone/tutorials/00_sim/create_empty.py" You should be inside conda env.
10. You need to check that you have "Isaac Sim Python 2023.1.1 - New Stage*" on the top of the window.
11. Clone this repo with git clone https://github.com/abizovnuralem/go2_omniverse/ --recurse-submodules -j8 --depth=1
12. Copy the config file Unitree_L1.json (located in go2_omniverse/Isaac_sim/Unitree/Unitree_L1.json) for the Unitree L1 LiDAR to the folder IsaacLab-0.3.1/source/data/sensors/lidar/Unitree_L1.json (if the path doesnt exists, create it)
13. Copy all material files in the isaac-sim-2023.1.1/data/material_files folder to IsaacLab_v0.3.1/source/data/material_files (if the path doesnt exists, create it)
14. Execute ./run_sim.sh (without activated conda orbit env)

Some suggestions:

1. You need to check nvidia-smi, it should work, before installing Isaac Sim
2. You need to install Miniconda and execute: conda config --set auto_activate_base false
3. Install Omniverse launcher and then install Isaac Sim.
4. You need to install ROS2 on your system and configure it:

https://docs.omniverse.nvidia.com/isaacsim/latest/installation/install_ros.html#isaac-sim-app-install-ros

Usage

The current project was tested on Ubuntu 22.04, IsaacSim 2023.1.1 with Orbit 0.3.0 and Nvidia Driver Version: 545. To start the project with Unitree GO2, execute:

./run_sim.sh

To start the project with Unitree G1, execute:

./run_sim_g1.sh

You can control the dog using "WASD" keyboard commands

ROS2 SDK

You can use https://github.com/abizovnuralem/go2_ros2_sdk as a basement for your ROS2 setup.

Select custom env

To use predifined custom envs, you need to download files from https://drive.google.com/drive/folders/1vVGuO1KIX1K6mD6mBHDZGm9nk2vaRyj3?usp=sharing and place them to /envs folder. Then you can execute it modifying run_sim.sh script with --custom_env=office and --terrain flat commands. If you are doing it first time, it will take 2-3 minutes to configure the env. Please, wait.

Development

To contribute or modify the project, refer to these resources for implementing additional features or improving the existing codebase. PRs are welcome!

VR support

To enable VR support on linux will take some time, but it works! I have tested it on:

1. Ubuntu 22.04
2. Nvidia drivers are 545.29.06
3. SteamVR 2.4.4 (IMPORTANT! It should be 2.4.4) and you need to go to Compatibility tab (Inside Steam app) and "Force the use of a specific Steam Play compatibility tool" and switch to "Steam-Play-None", additional info you can find in ALVR github issues tab.
4. ALVR streamer 20.8.1 + Oculus Quest 2 (client ALVR you can install via SideQuest app) (How to install it: https://github.com/alvr-org/ALVR)
5. Execute IsaacSim, Go to Window -> Extensions, find STEAMVR INPUT/OUTPUT then enable it and enable AutoLoad. Reopen IsaacSim. Use OpenXR mode.
6. Enjoy Omniverse in VR mode!

Troubleshooting:

The version of Isaac Simulator 2023.1.1 is no longer available on Omniverse Launcher, what should I do?

1. Go to Container Installation and follow all the steps in the Container Deployment section up to step 5).
2. Modify the command in step 6) with:

docker pull nvcr.io/nvidia/isaac-sim:2023.1.1

3. Now run this command:

docker run --name isaac-sim --entrypoint bash -it --runtime=nvidia --gpus all -e "ACCEPT_EULA=Y" --rm --network=host \
    -e "PRIVACY_CONSENT=Y" \
    -v ~/docker/isaac-sim/cache/kit:/isaac-sim/kit/cache:rw \
    -v ~/docker/isaac-sim/cache/ov:/root/.cache/ov:rw \
    -v ~/docker/isaac-sim/cache/pip:/root/.cache/pip:rw \
    -v ~/docker/isaac-sim/cache/glcache:/root/.cache/nvidia/GLCache:rw \
    -v ~/docker/isaac-sim/cache/computecache:/root/.nv/ComputeCache:rw \
    -v ~/docker/isaac-sim/logs:/root/.nvidia-omniverse/logs:rw \
    -v ~/docker/isaac-sim/data:/root/.local/share/ov/data:rw \
    -v ~/docker/isaac-sim/documents:/root/Documents:rw \
    nvcr.io/nvidia/isaac-sim:2023.1.1

4. Open another terminal and run the command bash docker ps to find the container's ID.
5. From the terminal inside the container, run this command to copy the entire Isaac 2023.1.1 folder to a folder on your computer (I suggest naming it isaac-sim-2023.1.1 so that the setup commands above work):

docker cp <id_container>:isaac-sim/. <your_local_machine_folder> # absolute path

6. On your PC, you will have a folder named isaac-sim-2023.1.1. Move it to the following path:

.local/share/ov/pkg/

Could not import 'rosidl_typesupport_c' for package 'go2_interfaces'

If running ./run_sim.sh results in the error "Could not import 'rosidl_typesupport_c' for package", refer to this thread.

Thanks

Special thanks to

1. Leul Tesfaye for his expertise in Orbit lidars;
2. Tamas @tfoldi for contributing to this project;
3. @ShaoshuSu for his hardwork in investigating the speed issues with new version of Orbit (IsaacLab 4.0 and 4.1);

License

This project is licensed under the BSD 2-clause License - see the LICENSE file for details.