Isaaclab Experiments

This repository has many research projects with RL on Isaaclab. It is designed to be modular, so each folder inside isaaclab_experiments has its own files for running the project.

Getting Started

After cloning this repository, run this command:

git submodule init

To install Isaac Lab, follow the pip installation from the original repository: https://isaac-sim.github.io.

During the official instalation, you are prompted for cloning Isaac Sim, so just use the folder IsaacLab in this repository instead.

After you are done with this setup, install this repository with the following command:

pip install -e .

To run an experiment for training:

python scripts/train.py --headless --task Go1-Navigation-v0

To view training logs

tensorboard --logdir logs

How the code is organized

All experiments are inside the folder isaaclab_experiments, take a look at the folder go1_low_level with an example of the possible parts.

Justfile

We included a Justfile for running the train and play commands for the available environments.

Install justfile utility: https://github.com/casey/just

Setup autocomplete:

echo 'source <(just --completions bash)' >> ~/.bashrc
source ~/.bashrc

Now you can enjoy autocompletions:

just train<tab> 

If you add a new environment, regenerate the justfile:

just generate_envs