13. Manual Installation Process

November 2, 2023 ยท View on GitHub

This page documents the dependencies of the ALR SimFramework.

13.1. Minimal dependencies

These are the minimal dependencies needed to execute a pybullet simulation.

  • pybullet

  • pyyaml

  • scipy

  • opencv

  • gin_config

  • pinocchio. Important: Do not try to install via pip, as this is a different package. Either install through conda -c conda-forge pinocchio or systemwide through sudo apt

  • pre-commit to enforce style guides for developing

  • matplotlib (for plotting)

  • gym (for RL gyms)

  • for point cloud visualization, best install in this order:

    1. scikit-learn
    2. addict
    3. pandas
    4. plyfile
    5. tqdm
    6. open3d

13.2. Additional Mujoco Dependencies (v2.1)

Mujoco requires additional dependencies, which might need to be installed systemwide, such as

sudo apt-get libosmesa6-dev

Additionally, you might need to install

  • mesalib
  • glfw not required anymore with Mujoco 2.1 as far as we know
  • glew
  • patchelf

These libraries can be easily installed via conda.

Finally install the mujoco-py bindings by executing:

pip install mujoco-py

Afterwards, you must set two environmental variables:

LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$HOME/.mujoco/mujoco210/bin:/usr/lib/nvidia
LD_PRELOAD=$LD_PRELOAD:$CONDA_PREFIX/lib/libGLEW.so

This can be done by either editing your .bashrc file to include the lines:

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$HOME/.mujoco/mujoco210/bin:/usr/lib/nvidia
export LD_PRELOAD=$LD_PRELOAD:$CONDA_PREFIX/lib/libGLEW.so

or setting the environmental variables in your conda env only:

conda env config vars set LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$HOME/.mujoco/mujoco210/bin:/usr/lib/nvidia
conda env config vars set LD_PRELOAD=$LD_PRELOAD:$CONDA_PREFIX/lib/libGLEW.so

Installing the ALR Simframework

Right now the SimFramework must be installed with the -e option:

pip install -e .

This is because otherwise the assets, such as XML models etc., cannot be found.