ROS 2 Essentials¶
A repo containing essential ROS2 Humble features for controlling Autonomous Mobile Robots (AMRs) and robotic arm manipulators. Please setup an Ubuntu environment before using this repo.
The goal of this repo is to allow seamless robot policy reuse between simulation and reality powered by Omniverse Isaac Sim and Isaac ROS. In general, the amd64 images support both simulation and real robot control, while the arm64 images only supports real robot control.
Please note that this repo is under rapid development. The code is not guaranteed to be stable, and breaking changes may occur.
The documentation is hosted on https://j3soon.github.io/ros2-essentials/.
Cloning the Repository¶
Pre-built Workspaces¶
Pre-built Docker images for each workspace can be pulled by running docker compose pull in the corresponding workspace directory. Pulling these images bypasses the time-consuming build process (for both Docker Compose and Dev Containers).
The docker image of the template workspace is share by most of the workspace, allowing saving spaces by sharing common packages. Click on the following workspaces to navigate to their respective documentation.
| Workspace | amd64 | arm64 | Notes | Maintainer |
|---|---|---|---|---|
| Template | ✔️ | ✔️ | Yu-Zhong Chen, Johnson | |
| ORB-SLAM3 | ✔️ | ❌ | Assume Zhan | |
| RTAB-Map | ✔️ | ❌ | Assume Zhan | |
| ROS1 Bridge | ✔️ | ✔️ | Skip linting | Yu-Zhong Chen |
| Cartographer | ✔️ | ✔️ | Assume Zhan | |
| Clearpath Husky | ✔️ | ✔️ | Real-world support | Yu-Zhong Chen, Johnson |
| Yujin Robot Kobuki | ✔️ | ✔️ | Real-world support | Yu-Zhong Chen |
| Velodyne VLP-16 | ✔️ | ✔️ | Real-world support | Assume Zhan |
| Gazebo World | ✔️ | ❌️ | Yu-Zhong Chen | |
| ALOHA | ✔️ | ✔️ | Simulation only | Johnson |
If you have trouble using a workspace, please open an issue and tag the current maintainers mentioned above.
System Requirements¶
| Use Case | Platform | Hardware | Software | Notes |
|---|---|---|---|---|
| Simulation/Deployment | x86_64 | RTX GPU, 500GB+ SSD | Ubuntu 22.04, NVIDIA Driver, Docker, NVIDIA Container Toolkit | See this page for more details. |
| Deployment-Only | Jetson | Jetson Orin, 500GB+ SSD | JetPack 6.0 | See this page for more details. |
Some functionalities may still work on lower-spec systems, such as those without GPUs or on operating systems other than Ubuntu 22.04. However, these configurations are not officially supported and may require manual adjustments. Use them with caution.
Building Documentation¶
# (Optional) Create and activate python virtual environment
virtualenv venv -p python3
source venv/bin/activate
# Install dependencies and start serving
cd docs
pip install -r requirements.txt
mkdocs serve
# Go to https://127.0.0.1:8000 to view the site.
Link Documentation to ROS2 Workspaces¶
Reusing Docker Build Cache¶
The default (docker) build driver does not support pulling pre-built build cache from Docker Hub. Changing to the docker-container build driver allow reusing build cache from Docker Hub, however may introduce a few minute overhead after building any image (for sending tarballs). This is a tradeoff you should consider.
The default docker build driver cannot pull pre-built caches from Docker Hub. Using the docker-container driver enables cache reuse but may add a few minutes of overhead for sending tarballs after the build. You should consider this tradeoff if you choose to switch the build driver.
Switching to docker-container Build Driver¶
# Install buildx as `docker build` alias
docker buildx install
# Create and use new builder
docker buildx create --name docker-container --driver docker-container --driver-opt default-load=true --use
After setting this, using docker compose build will use the build cache from Docker Hub. If you want to switch back to the default docker build driver, follow the instructions below.
Switching back to docker Build Driver¶
# Uninstall buildx as `docker build` alias
docker buildx uninstall
# Unuse the created builder and remove it
docker buildx use default
docker buildx stop docker-container
docker buildx rm -f --all-inactive
Acknowledgement¶
The code is mainly contributed by Johnson, Yu-Zhong Chen, Assume Zhan, Lam Chon Hang, and others. For a full list of contributors, please refer to the contribution list.
We extend our gratitude to ElsaLab and NVIDIA AI Technology Center (NVAITC) for their support in making this project possible.
Disclaimer: this is not an official NVIDIA product.
License¶
All modifications are licensed under Apache License 2.0.
However, this repository includes many dependencies released under different licenses. For information on these licenses, please check the commit history. Make sure to review the license of each dependency before using this repository.
The licenses for dependencies will be clearly documented in the workspace README in the future.
Supplementary¶
Installing Docker¶
Follow this post for the installation instructions.
Installing NVIDIA Container Toolkit¶
Follow this post for the installation instructions.