Isaac Sim Local and Cloud Configuration Tutorial

After completing this chapter, you can determine whether your computer or cloud server is suitable for running Isaac Sim, and set up three common configurations: installation on a Windows local workstation, installation on a Linux workstation or cloud server, and installation of a Python environment using micromamba/venv. This article will also explain where Isaac Lab should be used and when it is necessary to enter GR00T for reproduction.

Here is the key judgment: If you only want to learn about Isaac Sim’s GUI, robot import, sensors, and basic scenarios, a Windows local workstation can serve as an excellent starting environment. However, if you plan to conduct reinforcement learning training with Isaac Lab, use Docker, enable remote desktop, or perform reproduction by multiple people, a Linux workstation and a cloud GPU server are more suitable. The commands for cloud servers and local workstations are not very different; the main difference is that cloud servers require additional handling for remote display, ports, security groups, and data persistence.

1. First, determine which installation path to choose

Isaac Sim is a high-fidelity simulation platform, not an ordinary Python library. It relies on NVIDIA RTX GPU, drivers, Omniverse Kit, PhysX, USD assets, graphics rendering, and numerous extensions. When installing, don’t just focus on whether pip install is installed successfully; more importantly, ensure that the GPU driver, memory, graphical interface, and extension cache all function properly.

2. Download Entry Summary Table

To enable everyone to reproduce the process step by step, we have listed here all the download links that will be used in this chapter. The Isaac Sim version is updated frequently. If you use version 5.2 or later in the future, please go to the official download page first to confirm the file name and version number, and then replace the URL in the command.

The most stable starting path is to download the Workstation package first and run the Isaac Sim GUI; if you later want to use Isaac Lab, switch to the pip environment. For the cloud server approach, configure the NVIDIA driver and Docker GPU runtime first, and then pull nvcr.io/nvidia/isaac-sim:5.1.0.

3. Hardware and System Requirements

According to the official requirements of NVIDIA Isaac Sim 5.1, the x86_64 platform supports Ubuntu 22.04/24.04 and Windows 10/11. The minimum configuration listed in the official table indicates 32GB RAM and at least 50GB SSD storage. The graphics card should be a NVIDIA RTX GPU with RT Core. The official also notes that complex scenarios, sensors, and large-scale training will further increase the demand for RAM and VRAM. GPUs with less than 16GB of video memory may not be able to handle complex rendering scenarios.

This means that machines like the RTX 3060 Laptop 6GB can be used for light learning and compatibility testing, but they are not suitable as primary environments for complex Isaac Lab training, high-definition sensors, multi-robot scenarios, or reproduction of GR00T large models. If learners do not want to rent servers, they can first complete the GUI and basic scenarios on their own machine; when training, synthetic data generation, or large model inference is required, they can switch to a cloud server.

4. Pre-installation Checks

In Windows PowerShell, first verify the NVIDIA driver, Python, and disk space. The pip installation for Isaac Sim 5.x requires Python 3.11; if the default Python on your machine is 3.12 or 3.10, create a separate environment instead of installing it directly into the system Python.

powershellnvidia-smi
python --version
python -m pip --version
Get-PSDrive C

If you plan to use the pip method, you can first check whether the Isaac Sim package is available in the PyPI and NVIDIA PyPI repositories. This command only queries the package index and does not download the complete installation package.

powershellpython -m pip index versions isaacsim --index-url https://pypi.org/simple --extra-index-url https://pypi.nvidia.com

On Linux, check GLIBC additionally. The Isaac Sim 5.x pip package requires GLIBC 2.35 or higher. Ubuntu 22.04 usually meets this requirement, while Ubuntu 20.04 defaults to GLIBC 2.31. It is not recommended to use pip directly.

bashnvidia-smi
python3 --version
ldd --version
df -h

Checkpoint 1: If the nvidia-smi does not detect the NVIDIA GPU, do not proceed with installing Isaac Sim. Usually, the issue is related to the graphics card driver, cloud server GPU instance specifications, Docker GPU runtime, or remote desktop environment, rather than Isaac Sim itself.

5. Installation on Windows Local Workstation

For beginners using Windows, the Workstation installation package is highly recommended, as it is more similar to the usage of ordinary desktop software. You can download it via a browser or directly in PowerShell. The installation package is about several dozen GB in size. It is recommended to check the disk space first, and then extract it to short paths like C:\isaacsim to avoid installation and cache issues caused by long Windows paths.

Step 1: Download the Windows standalone package:

powershellNew-Item -ItemType Directory -Force C:\isaacsim-downloads
cd C:\isaacsim-downloads
Invoke-WebRequest `
  -Uri "https://downloads.isaacsim.nvidia.com/isaac-sim-standalone-5.1.0-windows-x86_64.zip" `
  -OutFile "isaac-sim-standalone-5.1.0-windows-x86_64.zip"

If Invoke-WebRequest downloads slowly, you can copy the above URL into a browser, IDM, aria2, or school network downloader to download it, and keep the file name unchanged.

Step 2: Unzip and execute the post-installation processing:

powershellmkdir C:\isaacsim
cd C:\isaacsim-downloads
tar -xvzf "isaac-sim-standalone-5.1.0-windows-x86_64.zip" -C C:\isaacsim
cd C:\isaacsim
.\post_install.bat
.\isaac-sim.selector.bat

After starting App Selector, select Isaac Sim Full and click START. The first launch will preheat the shader cache and load extensions. It is normal for a blank window to persist for a few minutes. After entering the main interface, it is recommended to perform a minimum graphics verification: select Create > Environment > Simple Room, then Create > Robots > Franka Emika Panda Arm, and finally click the play button on the left. If the scene appears and runs, it indicates that the local GUI, rendering, basic assets, and physical simulation chain are connected.

Checkpoint 2: This verification only proves that Isaac Sim can be launched and run with basic scenarios on the local machine. It does not indicate that there is sufficient VRAM for large-scale training, nor does it mean that ROS2, Isaac Lab, or GR00T have been properly configured.

6. Installation via Windows micromamba / venv

If you want to integrate Isaac Lab later, or prefer to use Isaac Sim as a Python package manager, you can use micromamba or venv. It is recommended to create a separate env_isaacsim instead of installing it into the environment already used for LeRobot, OpenVLA, and PyTorch experiments.

powershellmicromamba create -n env_isaacsim python=3.11 -y
micromamba activate env_isaacsim
python -m pip install --upgrade pip
python -m pip install "isaacsim[all,extscache]==5.1.0" --extra-index-url https://pypi.nvidia.com

Isaac Sim requires acceptance of the NVIDIA Omniverse EULA for the first run. Environment variables can be temporarily set in PowerShell:

powershell$env:OMNI_KIT_ACCEPT_EULA="YES"
isaacsim isaacsim.exp.compatibility_check
isaacsim

If you only want to check whether the machine meets the requirements, you can first install a smaller compatibility check bundle:

powershellpython -m pip install "isaacsim[compatibility-check]==5.1.0" --extra-index-url https://pypi.nvidia.com
$env:OMNI_KIT_ACCEPT_EULA="YES"
isaacsim isaacsim.exp.compatibility_check

Checkpoint 3: isaacsim isaacsim.exp.compatibility_check will launch the Isaac Sim compatibility check application. It is more suitable as a pre-installation inspection than the full GUI, but it cannot replace the subsequent full scenario verification.

7. Installation of Linux workstations and cloud servers

The Linux local workstation can directly use the Workstation installation package. Step 1: Download the installation package:

bashmkdir -p ~/isaacsim-downloads
cd ~/isaacsim-downloads
wget -c https://downloads.isaacsim.nvidia.com/isaac-sim-standalone-5.1.0-linux-x86_64.zip

Step 2: Unzip and start App Selector:

bashmkdir -p ~/isaacsim
cd ~/isaacsim-downloads
unzip "isaac-sim-standalone-5.1.0-linux-x86_64.zip" -d ~/isaacsim
cd ~/isaacsim
./post_install.sh
./isaac-sim.selector.sh

If it is a cloud server, there is usually no physical monitor. The Workstation GUI requires VNC, NICE DCV, X11 forwarding, or Isaac Sim Livestream for support. In this case, it is recommended to consider “cloud server” and “local Linux workstation” as the same Isaac installation setup, with only additional remote display and port exposure issues for cloud servers. Do not open ports directly to the public network; instead, use an SSH tunnel or the security group settings of the cloud provider to restrict source IPs.

If using the pip method, the Linux command is as follows:

bashmicromamba create -n env_isaacsim python=3.11 -y
micromamba activate env_isaacsim
python -m pip install --upgrade pip
python -m pip install "isaacsim[all,extscache]==5.1.0" --extra-index-url https://pypi.nvidia.com
export OMNI_KIT_ACCEPT_EULA=YES
isaacsim isaacsim.exp.compatibility_check
isaacsim

If ldd --version indicates that GLIBC is below 2.35, do not install the pip version rigidly on old systems. A safer approach is to switch to Ubuntu 22.04/24.04, or use the official binary installation package and container method.

8. Linux Docker Installation

Docker is more suitable for Linux cloud servers, multi-person reproduction, and environment isolation. NVIDIA officially states that the Isaac Sim container only supports Linux. If you want to use the Isaac Sim GUI on Windows local, prefer using Workstation or pip; it is not recommended to use Windows Docker as the preferred teaching method.

Step 1: Install Docker. If the cloud server image already includes Docker, you can skip this step:

bashsudo apt update
sudo apt install -y docker.io
sudo systemctl enable docker
sudo systemctl start docker
docker --version

Step 2: Install the NVIDIA Container Toolkit. The following is the official installation method for Ubuntu. For other distributions, refer to the official documentation of the NVIDIA Container Toolkit:

bashcurl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey \
  | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg

curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list \
  | sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' \
  | sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list

sudo apt update
sudo apt install -y nvidia-container-toolkit
sudo nvidia-ctk runtime configure --runtime=docker
sudo systemctl restart docker

Step 3: Verify that a GPU can be seen inside the container:

bashdocker run --rm --runtime=nvidia --gpus all ubuntu nvidia-smi

Step 4: Pull the Isaac Sim 5.1 image. The image comes from NVIDIA NGC. Logging in to NGC may be required for the first pull. If authentication fails, register and generate an API Key in NGC, and then execute docker login nvcr.io.

bashdocker login nvcr.io
docker pull nvcr.io/nvidia/isaac-sim:5.1.0

It is recommended to create separate directories for Isaac Sim cache, logs, and user data, so that the cache is not lost after the container is deleted:

bashmkdir -p ~/docker/isaac-sim/cache/main
mkdir -p ~/docker/isaac-sim/cache/computecache
mkdir -p ~/docker/isaac-sim/config
mkdir -p ~/docker/isaac-sim/data
mkdir -p ~/docker/isaac-sim/logs
mkdir -p ~/docker/isaac-sim/pkg
sudo chown -R 1234:1234 ~/docker/isaac-sim

Start the interactive container:

bashdocker run --name isaac-sim --entrypoint bash -it --gpus all --rm --network=host \
  -e "ACCEPT_EULA=Y" \
  -e "PRIVACY_CONSENT=Y" \
  -v ~/docker/isaac-sim/cache/main:/isaac-sim/.cache:rw \
  -v ~/docker/isaac-sim/cache/computecache:/isaac-sim/.nv/ComputeCache:rw \
  -v ~/docker/isaac-sim/logs:/isaac-sim/.nvidia-omniverse/logs:rw \
  -v ~/docker/isaac-sim/config:/isaac-sim/.nvidia-omniverse/config:rw \
  -v ~/docker/isaac-sim/data:/isaac-sim/.local/share/ov/data:rw \
  -v ~/docker/isaac-sim/pkg:/isaac-sim/.local/share/ov/pkg:rw \
  -u 1234:1234 \
  nvcr.io/nvidia/isaac-sim:5.1.0

Run the compatibility check first after entering the container:

bash./isaac-sim.compatibility_check.sh --/app/quitAfter=10 --no-window

Checkpoint 4: If the system check passes here, it means that Docker, NVIDIA Container Toolkit, GPU driver, and Isaac Sim container can at least work in headless mode. However, it is not equivalent to a successfully configured remote GUI or Livestream.

9. How to open the screen on a remote cloud server

The Isaac Sim instance for cloud servers and local workstations is a single entity, but there is the issue of “how to view the screen” in the cloud environment. The simplest teaching solution is VNC or NICE DCV; if Isaac Sim provides a live stream, you can download the WebRTC Streaming Client.

Download this client when connecting Isaac Sim to the cloud from a Windows local computer:

powershellNew-Item -ItemType Directory -Force C:\isaacsim-downloads
Invoke-WebRequest `
  -Uri "https://downloads.isaacsim.nvidia.com/isaacsim-webrtc-streaming-client-1.1.5-windows-x64.exe" `
  -OutFile "C:\isaacsim-downloads\isaacsim-webrtc-streaming-client-1.1.5-windows-x64.exe"

Download this AppImage from the Linux client:

bashmkdir -p ~/isaacsim-downloads
cd ~/isaacsim-downloads
wget -c https://downloads.isaacsim.nvidia.com/isaacsim-webrtc-streaming-client-1.1.5-linux-x64.AppImage
chmod +x isaacsim-webrtc-streaming-client-1.1.5-linux-x64.AppImage

The configuration for remote display is closely related to the security groups of cloud providers. In principle, do not expose VNC or livestream ports directly to the public network. Instead, use an SSH tunnel or allow only your own fixed IP address to access them.

10. Isaac Lab Configuration

Isaac Lab is a robot learning framework built on top of Isaac Sim, suitable for reinforcement learning, imitation learning, task construction, and parallel simulation. The recommended process is to ensure Isaac Sim can start before installing Isaac Lab. Do not install Isaac Lab directly when Isaac Sim cannot be opened, as this will make it difficult to identify subsequent errors.

The following commands are based on Isaac Lab 2.3.x and Isaac Sim 5.1. The Isaac Lab source code can be downloaded from GitHub, and you can switch to the corresponding release version afterward. It is recommended to visit the Isaac Lab official release page to confirm the Isaac Sim version corresponding to the current version, before deciding whether to upgrade.

bashgit clone https://github.com/isaac-sim/IsaacLab.git
cd IsaacLab
git checkout v2.3.0

Install the basic compilation dependencies on Linux, and then install the Isaac Lab extension:

bashsudo apt update
sudo apt install -y cmake build-essential
./isaaclab.sh --install
./isaaclab.sh -p scripts/tutorials/00_sim/create_empty.py

Execute on Windows in an activated Isaac Python environment:

powershellisaaclab.bat --install
isaaclab.bat -p scripts\tutorials\00_sim\create_empty.py

The official verification script launches a blank scene with a black viewport. This result seems simple, but it proves that Isaac Lab can invoke the current Python environment, locate Isaac Sim, load basic extensions, and open the simulation application.

11. How to Choose ROS2 Integration

Isaac Sim 5.1 supports ROS2 Humble and Jazzy. For Ubuntu 22.04, Humble is recommended; for Ubuntu 24.04, Jazzy is recommended. In Windows 10/11 scenarios, the official documentation suggests using ROS2 Humble via WSL.

Beginners are advised not to install ROS2 and Isaac Sim together on the first day. A more stable order is to start with opening Isaac Sim, then installing ROS2, and finally enabling the ROS2 Bridge. Default message types such as std_msgs, geometry_msgs, and nav_msgs usually run smoothly; if you need to customize the ROS2 interface, you need to consider the compatibility between Isaac Sim Python 3.11 and the system ROS Python version.

12. When to start the GR00T tutorial

GR00T is not an introductory configuration for Isaac Sim, but rather a later-stage practice for robot base model development. The tutorial on阿里云 deploying Isaac Lab + GR00T in the repository has already set up a reproducible version: Isaac Sim 4.2.0, Isaac Lab v1.4.1, GR00T N1.6, PyTorch 2.5.1, and CUDA 12.1. This combination differs from the main line of Isaac Sim 5.1 used in this article.

If you only want to study Isaac Sim locally, there is no need to install GR00T immediately. If your goal is to reproduce GR00T, it is recommended to use a separate cloud server or isolated environment and follow the original tutorial version. Do not install GR00T as a dependency in env_isaacsim. This approach can prevent PyTorch, CUDA, transformers, flash-attn, and the Isaac extension from interfering with each other.

13. Common Questions

When the Isaac Sim installation fails, first troubleshoot by layer. The first layer is hardware and drivers: whether nvidia-smi is functioning properly, whether the drivers are sufficiently updated, whether the GPU has an RT Core, and whether the memory is too low. The second layer is the system: whether Windows supports long path support, whether Linux GLIBC meets pip package requirements, and whether at least several tens of GB are reserved on the disk. The third layer is Python: whether Python 3.11 is used, whether Isaac is installed in a clean environment, and whether the NVIDIA PyPI source is used.

If the first startup is slow, do not close the window immediately. The first launch of Isaac Sim will download extensions, compile the shader cache, and initialize user settings. It is normal for it to take 5 to 10 minutes. Only when the log clearly indicates a driver error, extension download failure, unacceptance of the EULA, or process crash, should you address it according to the error message.

If the pip installation on Windows reports a too-long path, enable long-path support first, or use a shorter environment path. For workstation installations, it is also recommended to use short paths such as C:\isaacsim, rather than deeply nested personal document directories.

If the GUI on the cloud server cannot be opened, first determine whether Isaac can start headless. If the compatibility check passes, the issue is mostly with VNC, WebRTC, DISPLAY, Xauthority, or security group ports, rather than the Isaac Sim本体. In this case, you can first verify using a container headless command, and then address remote display issues.

14. Light verification records of this tutorial

This article did not fully download the 40GB Isaac Sim installation package on this repository machine, as it would incur significant time and disk costs. The completed lightweight verification includes: the local nvidia-smi can recognize the NVIDIA GeForce RTX 3060 Laptop GPU, with a driver version of 581.95 and 6144 MiB of video memory; the current Python is 3.11.15; the micromamba version is 2.3.0; and through PyPI and NVIDIA PyPI indexes, the isaacsim package versions 5.1.0.0 and 5.0.0.0 can be queried.

This set of verifications proves that the local machine has the basic requirements for NVIDIA drivers, Python 3.11, and package index access. However, it does not prove that a 6GB GPU memory is sufficient to run complex Isaac scenarios. When actually delivered to learners, it is still recommended to confirm step by step using the Checkpoint in this article.

References

• NVIDIA Isaac Sim 5.1 Installation
• NVIDIA Isaac Sim 5.1 Downloads
• NVIDIA Isaac Sim 5.1 Requirements
• NVIDIA Isaac Sim 5.1 Workstation Installation
• NVIDIA Isaac Sim 5.1 Python Environment Installation
• NVIDIA Isaac Sim 5.1 Container Installation
• NVIDIA Isaac Sim 5.1 ROS 2 Installation
• NVIDIA Container Toolkit Installation
• Isaac Lab 2.3 Installation using Isaac Sim Pip Package