CSDecomp: Configuration Space Decomposition Toolbox

CSDecomp is a Python package that implements a simple GPU-accelerated collision checker and GPU-accelerated algorithms for computing approximate convex decompositions of robot configuration spaces. The package provides implementations of Dynamic Roadmaps (DRMs) and the Edge Inflation Zero-Order (EI-ZO) algorithm in cuda/cpp as described in our paper "Superfast Configuration-Space Convex Set Computation on GPUs for Online Motion Planning".

Since the publication I have continued adding useful functionality to this codebase for motion planning and have made it pip installable.

Contributions are welcome!

Installation

From PyPI (recommended)

pip install csdecomp

Requires an NVIDIA GPU with a compatible driver. The wheel bundles the CUDA runtime.

From source

1. Install the CUDA toolchain (12.x recommended): https://developer.nvidia.com/cuda-toolkit-archive

   (Your display driver must be compatible with the installed CUDA version.)

2. Install prereqs:

   Install bazel via bazelisk: bazelisk instructions

   Other prereqs: sudo bash setup.sh

3. Build and test:

   bazel test //...

   (If you are getting cudaMalloc errors, make sure there aren't any big applications running in the background.)

Usage

import csdecomp as csd

The build also outputs a pip-installable wheel at bazel-bin/csdecomp/src/pybind/csdecomp/.

To change Python version, edit tools/my_python_version.bzl and MODULE.bazel.

The unit tests demonstrate how the code should be used. The Python bindings closely follow the C++ syntax. There is experimental documentation that can be built with doxygen: cd csdecomp/docs/ && doxygen Doxyfile.

Running the Examples

The examples require uv for environment management.

As a user (using the PyPI package)

cd examples
uv sync
uv run python minimal_test.py

For notebooks:

uv run jupyter notebook

Select the .venv kernel in your editor.

As a developer (using a locally built wheel)

cd examples
bash dev_install.sh          # builds wheel from source and installs it
uv run python minimal_test.py
uv run python test_eizo.py
uv run python test_drake_bridge.py

The example tests are also Bazel targets, so bazel test //... runs them alongside all other tests.

For notebooks, select the .venv kernel in your editor.

Developing

Run the full test suite (requires GPU):

bazel test //...

CI runs lint and build checks only (no GPU required). Developers with a GPU should run the full suite locally before pushing.

For interactive debugging of C++ code with plotting, use the cc_test_with_system_python targets (tagged manual) in the test BUILD files. These require system Python with matplotlib and dev headers installed.

Citation

If you find this code useful, please consider citing our paper:

@article{werner2024superfast,
  title={Superfast Configuration-Space Convex Set Computation on GPUs for Online Motion Planning},
  author={Werner, Peter and Cheng, Richard and Stewart, Tom and Tedrake, Russ and Rus, Daniela},
  journal={arXiv preprint arXiv:2504.10783},
  year={2025}
}

Useful Commands

bazel build //...                                              # build everything
bazel build //csdecomp/src/pybind/csdecomp:csdecomp_wheel      # build pip wheel
bazel test //...                                               # run all tests (requires GPU)
bazel test //csdecomp/tests:csdecomp_test                      # run Python integration test

If Drake is slow to launch (LCM error):

sudo ifconfig lo multicast && sudo route add -net 224.0.0.0 netmask 240.0.0.0 dev lo