Skip to main content

Install the Isaac Lab

This wiki provides a step-by-step guide to install Isaac Lab. Isaac Lab requires Isaac Sim. This tutorial first installs Isaac Sim via pip, then installs Isaac Lab from source code.

tip

  • This tutorial is only applicable to Ubuntu 20.04 and 22.04 systems and does not currently support Windows. Please ensure your computer has NVIDIA graphics drivers and CUDA 12+ installed based on your GPU.

  • Check the official Isaac Sim link to verify if your hardware meets the requirements.

  • Miniconda is recommended and must be pre-installed.

Creating a Virtual Environment with Miniconda​

We recommend creating a virtual environment first. Ensure the Python version in the virtual environment is Python 3.10. 

conda create -n env_isaaclab python=3.10  
conda activate env_isaaclab

Installing PyTorch and torchvision​

Next, install PyTorch and Torchvision according to your CUDA version.

pip install torch==2.5.1 torchvision==0.20.1 --index-url https://download.pytorch.org/whl/cu118

Installing Isaac Sim​

pip install --upgrade pip  
pip install 'isaacsim[all,extscache]==4.5.0' --extra-index-url https://pypi.nvidia.com

Verifying Isaac Sim Installation​

isaacsim
note

When running Isaac Sim for the first time, all required extensions will be pulled from the registry. This process may take over 10 minutes and is necessary for the initial run of each experience file. Once the extensions are downloaded, subsequent runs with the same experience file will use cached extensions.

Installing Isaac Lab​

Clone the Isaac Lab repository 

git clone https://github.com/isaac-sim/IsaacLab.git

Install dependencies for Isaac Lab 

sudo apt install cmake build-essential

Install Isaac Lab's reinforcement learning libraries 

./isaaclab.sh --install  # or "./isaaclab.sh -i"

You can also install a specific RL library individually: 

./isaaclab.sh --install rl_games  # or "rsl_rl, sb3, skrl, robomimic"

Verify Isaac Lab installation
Navigate to the cloned Isaac Lab directory.

Option 1: Launch via shell script 

./isaaclab.sh -p scripts/tutorials/00_sim/create_empty.py

Option 2: Launch via Python 

python scripts/tutorials/00_sim/create_empty.py
note

The above commands should launch the simulator and display a window with a black viewport, as shown below. You can exit the script by pressing Ctrl+C in the terminal. On Windows, use Ctrl+Break or Ctrl+fn+B in the Command Prompt to terminate the process.

Training a Simple Robot​

You can train a group of spiders using the provided example script: 

./isaaclab.sh -p scripts/reinforcement_learning/rsl_rl/train.py --task=Isaac-Ant-v0 --headless

Or train a single dog: 

./isaaclab.sh -p scripts/reinforcement_learning/rsl_rl/train.py --task=Isaac-Velocity-Rough-Anymal-C-v0 --headless

Installing Isaac Gym (Optional)​

The previous installation steps are sufficient, but if you only want to experience the reinforcement learning part with Isaac Gym, you can install Isaac Gym separately.

Download and extract Isaac Gym code

Extract it to your home directory, then create a Conda environment and install dependencies: 

conda create --name isaac python=3.8  
conda activate isaac  
conda install pytorch torchvision torchaudio pytorch-cuda=12.4 -c pytorch -c nvidia  
conda install numpy=1.23

Install Isaac Gym 

cd <path_to_isaacgym>/IsaacGym_Preview_4_Package/isaacgym/python  
pip install -e .

Verify Isaac Gym installation 

cd <path_to_isaacgym>/IsaacGym_Preview_4_Package/isaacgym/python/examples  
python 1080_balls_of_solitude.py
danger

Possible error & solution
When running the Python script, you may encounter the following error: 

ImportError: libpython3.8.so.1.0: cannot open shared object file: No such file or directory

Fix it with this command (replace the path with your own): 

export LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/home/{Your Username}/anaconda3/envs/pi/lib
Loading Comments...