UPSTREAM RIGHTS (NO LICENSE IDENTIFIED)
The previewed robot description is assets/urdf/v6biped_urdf_v4_aug29/v6biped_urdf_v4_squarefoot_aug29.urdf and its STL meshes in generalroboticslab/legged_env, commit 8b21398ffaa2abc8f4d2c599f2c0c11122dfe5a1. That commit is the exact legged_env gitlink of generalroboticslab/DukeHumanoidv1 commit 0700cfbfe4e68b0db8c12a351f4c7dc449524356, and envs/cfg/task/Biped.yaml names this URDF as the Biped task asset. The legged_env repository at that commit contains no LICENSE, COPYING or NOTICE file and its README states no license. The parent DukeHumanoidv1 repository LICENSE is the MIT License, "Copyright (c) 2024 General Robotics Lab - Duke University", and its README describes the Duke Humanoid as "an open-source 10-degrees-of-freedom humanoid" and legged_env as containing "the simulation and RL code". Open Source Robotics does not assume that the parent MIT license covers the separate legged_env repository.
Under the user decision of October 4, 2026 for missing or ambiguous licenses, Open Source Robotics hosts the unchanged description folder (URDF, config.yaml, graph image and all 23 STL meshes), the legged_env README and Biped.yaml, and the parent LICENSE, README and .gitmodules (in a folder named for that repository and revision) in a ZIP bundle, and a triangulated preview of the URDF at zero joint positions, with attribution to the General Robotics Lab, Duke University. It is a simulation description, not the machined-part manufacturing CAD. No reuse, manufacturing, redistribution or other clearance is claimed.
--- UPSTREAM DukeHumanoidv1-0700cfbfe4e68b0db8c12a351f4c7dc449524356/LICENSE ---
MIT License
Copyright (c) 2024 General Robotics Lab - Duke University
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
--- UPSTREAM README.md ---
# Isaac Gym RL template environment for legged robots
This repository contains an Isaac Gym template environment that can be used to train any legged robot using [rl_games](https://github.com/Denys88/rl_games). This repository is deployed with zero-shot sim-to-real transfer in the following projects:
- [Text2Robot](https://github.com/generalroboticslab/Text2Robot): a framework that converts user text specifications and performance preferences into physical quadrupedal robots.
- [DukeHumanoidv1](https://github.com/generalroboticslab/DukeHumanoidv1): an open-source 10-degrees-of-freedom child size humanoid for locomotion research.
## Table of content
- [project structure](#project-structure)
- [Setting up](#setting-up)
* [Tested environment](#tested-environment)
* [Setup python virtual environment](#setup-python-virtual-environment)
* [Setup vscode](#setup-vscode)
* [to start conda env](#to-start-conda-env)
* [to train a policy](#to-train-a-policy)
* [to run example checkpoint:](#to-run-example-checkpoint)
## project structure
```
.
├── assets
│ ├── checkpoints # saved checkpoint
│ │ ├── A1Terrain.pth
│ │ └── ...
│ ├── joint_monkey.py # script to check the urdf
│ └── urdf
│ ├── a1
│ ├── anymal_c
│ └──...
├── envs
│ ├── cfg # stores all configurations
│ │ ├── config.yaml
│ │ ├── pbt
│ │ ├── task
│ │ └── train
│ ├── common # common scripts
│ │ ├── publisher.py
│ │ ├── terrain.py
│ │ └── utils.py
│ ├── exp.sh # experiment scripts
│ ├── __init__.py
│ ├── plot_juggler # plot_juggler configurations
│ │ └── robotdog7kg_play_debug.xml
│ ├── run.sh # run script
│ ├── setup # conda env setup script
│ │ ├── conda_env.yaml
│ │ └── ...
│ ├── tasks #
│ │ ├── __init__.py
│ │ └── legged_terrain.py
│ └── train.py
├── outputs # contains trained results
└── README.md
```
## Setting up
### Tested environment
- Operating system: Ubuntu 22.04 with CUDA 12.3
- Development environment: Visual Studio Code (VSCode)
- Python environment management: micromanba
### Setup python virtual environment
first [install Micromamba](https://mamba.readthedocs.io/en/latest/installation/micromamba-installation.html) if you have not done so. Recommended to install at `~/repo/micromamba`
```bash
"${SHELL}" <(curl -L micro.mamba.pm/install.sh)
```
Setup a python virtual envirnpoment named `py38` with conda yaml file `setup/conda_env.yaml`
```bash
alias conda="micromamba"
# Create environment
conda env create --file setup/conda_env_py38.yaml -y
# Activate the environment
conda activate py38
# Export library path
export LD_LIBRARY_PATH=${CONDA_PREFIX}/lib
```
Install addional dependencies:
```bash
conda activate py38 && export LD_LIBRARY_PATH=${CONDA_PREFIX}/lib
# install Isaacgym
cd ~/repo
git clone https://github.com/boxiXia/isaacgym.git
cd ~/repo/isaacgym/python
python -m pip install -e .
## verify
cd ~/repo/isaacgym/python/examples
python interop_torch.py
# if error: crypt.h: No such file or directory, do this:
# cp /usr/include/crypt.h /home/grl/repo/micromamba/envs/py38/include/python3.8/
# install rl_games
cd ~/repo
git clone https://github.com/Denys88/rl_games.git
cd ~/repo/rl_games/
# Optionally modify setup.py to remove the version control parts
python -m pip install -e .
# install IsaacGymEnvs
# https://github.com/NVIDIA-Omniverse/IsaacGymEnvs/tree/main
cd ~/repo
git clone https://github.com/NVIDIA-Omniverse/IsaacGymEnvs.git
cd ~/repo/IsaacGymEnvs
# recommended to modify setup.py to remove the version control parts
python -m pip install -e .
```
### Setup vscode
Install VSCode: [Download and install](https://code.visualstudio.com/download) vscode if you have not done so
Install vscode Extensions:
- Python: ms-python.python
Optionally Install other extensions such as Git based on your needs.
To configure the project using vscode:
- Open the project folder.
- Select the correct Python virtual environment.
### to start conda env
Ensure you're operating within the Python virtual environment. For convenience, **we'll assume all commands in this guide are executed within this activated environment**:
```
conda activate py38 && export LD_LIBRARY_PATH=${CONDA_PREFIX}/lib
```
### to train a policy
```bash
cd envs
# duke humanoid v1 baseline
bash run.sh dukehumanoid_baseline
# duke humanoid v1 passive policy
bash run.sh dukehumanoid_passive
# robot dog7kg
bash run.sh dog
# a1
bash run.sh a1Terrain
# anymal
bash run.sh anymalTerrain
```
### to run example checkpoint:
```bash
cd envs
# duke humanoid v1 baseline
bash run.sh dukehumanoid_baseline -p
bash run.sh dukehumanoid_baseline -pk # with keyboard wasd
# duke humanoid v1 passive policy
bash run.sh dukehumanoid_passive -p
# robot dog7kg
bash run.sh dog -p
# a1
bash run.sh a1Terrain -p
# anymal
bash run.sh anymalTerrain -p
# dry run (print the command only)
bash run.sh anymalTerrain -d
```
### Debugging
Use `bash run.sh [THE_TASK_TO_RUN] -d` to start a remote attache python debuging using debugpy with vscode, for example
```bash
bash run.sh dukehumanoid_baseline -d # train mode + debug
bash run.sh dukehumanoid_baseline -pd # play mode + debug
bash run.sh a1Terrain -pkd # play mode + keyboard + debug
```
Then go to the vscode debug panel (keyboard shortcut: `ctrl+shift+D`), in the `RUN AND DEBUG`, select option `Python Debugger: Remote Attach` to start debugging.