ARC Humanoid Robot Training using Isaac Lab

Overview

This repository contains a project developed for training humanoid robots from the KIST Advanced Robot Control Lab. using Isaac Lab. It follows the structure of the Isaac Lab extension template.

Humanoid Robots:

• New KIST Humanoid Robot: Coming soon!
• MAHRU: MAHRU is a wheel-legged humanoid robot developed by KIST. This robot combines wheels and legs for enhanced mobility!
• G1: 29-DoF G1 robot from Unitree Robotics.

Also contains the 'Anymal-D' robot as an example!

Keywords: humanoid, reinforcement learning, isaaclab, sim2sim, sim2real
Maintainer, Author: Sol Choi

Installation

Step 1. Clone this git repository

git clone --recursive https://github.com/S-CHOI-S/Humarconoid.git

Step 2. (Optional) Rename all occurrences of humarconoid (in files/directories) to your_fancy_extension_name

python scripts/rename_template.py your_fancy_extension_name

Step 3. Install Isaac Lab, see the installation guide

Step 4. Using a python interpreter that has Isaac Lab installed, install the library

• humarcscripts library

  python -m pip install -e .
  

• humarconoid library

  cd source/humarconoid
  python -m pip install -e .
  

Usage

Train

Step 1. Check out the environments available in the Humarconoid project.

python scripts/list_envs.py

Step 2. Start training with humarconoid TASK!

python scripts/rsl_rl/train.py --task TASK --headless

Tip

Humarconoid is compatible with rsl_rl, Stable Baselines3, as well as custom reinforcement learning algorithms.

Play

Step 1. Choose your TASK and LOGDIR from your log.

python scripts/rsl_rl/play.py --task TASK --log_dir LOGDIR --num_envs NUM_ENVS

Monitor

You can monitor real-time training logs via TensorBoard!

tensorboard --logdir logs/rsl_rl/TASK/

Code formatting

We have a pre-commit template to automatically format your code. To install pre-commit:

pip install pre-commit

Then you can run pre-commit with:

pre-commit run --all-files