Isaac Lab – SO‑ARM100 / SO‑ARM101 Project

This repository implements tasks for the SO‑ARM100 and SO‑ARM101 robots using Isaac Lab. It serves as the foundation for several tutorials in the LycheeAI Hub series Project: SO‑ARM101 × Isaac Sim × Isaac Lab.

📰 News featuring this repository:

• 10 June 2025: 🎥 LycheeAI Channel Premiere: SO-ARM101 tutorial series announcement! 🔗 Watch on YouTube
• 23 April 2025: 🤖 NVIDIA Omniverse Livestream: Training a Robot from Scratch in Simulation (URDF → OpenUSD). 🔗 Watch on YouTube
• 19 April 2025: 🎥 LycheeAI Tutorial: How to Create External Projects in Isaac Lab. 🔗 Watch on YouTube

🎬 Watch the Lift Task in action

🛠️ Installation

1. Install Isaac Lab by following the official installation guide (using conda).

2. Clone this repository outside the IsaacLab directory.

3. Install the package:

   python -m pip install -e source/SO_100

🚀 Quickstart

To list all available environments:

python scripts/list_envs.py

🐞 Debugging Tasks

Two scripts can help verify your setup:

Zero Agent

Sends zero commands to all robots, confirming that the environment loads correctly:

python scripts/zero_agent.py --task SO-ARM100-Reach-Play-v0

Random Agent

Sends random commands to all robots, confirming proper actuation:

python scripts/random_agent.py --task SO-ARM100-Reach-Play-v0

🏋️‍♂️ Training and Playback

You can train a policy for SO‑ARM100 / SO‑ARM101 tasks (for example, the Reach task, which is a basic RL-based IK) with the rsl_rl and/or skrl library:

python scripts/rsl_rl/train.py --task SO-ARM100-Reach-v0 --headless

After training, validate the learned policy:

python scripts/rsl_rl/play.py --task SO-ARM100-Reach-Play-v0

This ensures that your policy performs as expected in Isaac Lab before attempting real‑world transfer.

🧩 Sim2Sim Transfer

Work in progress.

🛠️ Sim2Real Transfer

Work in progress.

🤝 Contributing

We welcome contributions of all kinds! Please read our Contributing Guide to learn how to set up your environment, follow our coding style, and submit pull requests.

📄 License

This project is licensed under the BSD 3-Clause License. See the LICENSE file for details.

🙏 Acknowledgements

This project builds upon the excellent work of several open-source projects and communities:

• Isaac Lab - The foundational robotics simulation framework that powers this project
• NVIDIA Isaac Sim - The underlying physics simulation platform
• RSL-RL - Reinforcement learning library used for training policies
• SKRL - Alternative RL library integration
• SO-ARM100/SO-ARM101 Robot - The hardware platform that inspired this simulation environment

Special thanks to:

• The Isaac Lab development team at NVIDIA for providing the simulation framework
• Hugging Face and The Robot Studio for the SO‑ARM robot series
• The LycheeAI Hub community for tutorials and support

📚 Citation

If you use this work, please cite it as:

@software{Louis_Isaac_Lab_2025,
   author = {Louis, Le Lay and Muammer, Bay},
   doi = {https://doi.org/10.5281/zenodo.16794229},
   license = {BSD-3-Clause},
   month = apr,
   title = {Isaac Lab – SO‑ARM100 / SO‑ARM101 Project},
   url = {https://github.com/MuammerBay/isaac_so_arm101},
   version = {1.1.0},
   year = {2025}
}