Skip to content

v0.1.0 - Nav2 GPU MPPI plugin, Python package, Colab & Docker

Latest
Latest

Choose a tag to compare

View all tags
@rsasaki0109 rsasaki0109 released this 11 Jun 03:36
· 14 commits to master since this release
v0.1.0
51b663f
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

v0.1.0 - Nav2 GPU MPPI plugin, Python package, Colab & Docker

First public release of CudaRobotics as a usable GPU robotics toolkit, not only
a demo gallery.

Highlights

  • Nav2 cuda_mppi_controller::CudaMppiController: GPU MPPI controller plugin
    with DiffDrive, Ackermann, and Omni motion-model support.
  • Python package cudarobotics with GPU MPPI and point-cloud registration
    bindings.
  • CUDA DLPack costmap input for Python MPPI, so PyTorch/CuPy costmaps can stay
    on the GPU.
  • Colab quickstart notebook for browser-based MPPI and registration demos.
  • GHCR Docker demo image for the Nav2 GPU MPPI controller.
  • manylinux Python wheels for CPython 3.10 and 3.12 on x86_64 Linux.
  • External registration baseline results against probreg and Open3D CPU
    baselines.
  • Nav2 parameter validation for fail-fast controller configuration and live
    parameter updates.

Try It

Colab:

https://colab.research.google.com/github/rsasaki0109/CudaRobotics/blob/master/examples/colab/cudarobotics_quickstart.ipynb

Docker demo:

docker run --rm --gpus all ghcr.io/rsasaki0109/cuda-mppi-controller-demo:v0.1.0

Python from release assets:

python -m pip install ./cudarobotics-0.1.0-*.whl
python -c "import cudarobotics as cr; print(cr.__version__)"

Source build:

git clone https://github.com/rsasaki0109/CudaRobotics.git
cd CudaRobotics
pip install -e python/
python examples/python/mppi_quickstart.py
python examples/python/registration_quickstart.py

Nav2 plugin:

cd ros2_ws
colcon build --packages-select cuda_mppi_controller
source install/setup.bash
ros2 run cuda_mppi_controller plugin_load_test

Artifacts

  • Source distribution: cudarobotics-0.1.0.tar.gz
  • CPython 3.10 manylinux x86_64 wheel
  • CPython 3.12 manylinux x86_64 wheel

The release intentionally does not publish to PyPI yet.

Notes

  • Wheels require Linux x86_64 and a compatible NVIDIA driver at runtime.
  • The Docker image is built from the v0.1.0 tag and pushed to GHCR by GitHub
    Actions.
  • Jetson/aarch64 support is not included in this release.
Assets 5
Loading