Getting Started¶
Prerequisites¶
- Linux (x86_64, glibc 2.28+) on a CPU with AVX2 (Intel Haswell / AMD Zen or newer)
- Python 3.10–3.14
Installation¶
Pre-built wheels are available for Python 3.10–3.14 on Linux x86_64. No local compilation required:
Verify installation¶
from nav_planning.locobot import HOME_JOINTS
from nav_planning.planning import create_planner
planner = create_planner("locobot")
goal = planner.sample_valid()
result = planner.plan(HOME_JOINTS.copy(), goal)
print(f"Planning {'succeeded' if result.success else 'failed'}")
Building Wheels from Source¶
Wheels are built with cibuildwheel inside
the manylinux Docker images (configured under [tool.cibuildwheel] in
pyproject.toml). To build one locally:
The output goes to wheelhouse/. Requirements: Docker must be installed and running.
Releasing¶
Pushing a v* tag runs .github/workflows/release.yml, which builds
manylinux_2_34 and manylinux_2_28 wheels for every supported Python and
publishes them to PyPI via Trusted Publishing (project nav-planning,
GitHub environment pypi):
Development Setup¶
For contributing or rebuilding the robot models, use pixi:
git clone --recursive https://github.com/H-tr/Nav-Planning.git
cd Nav-Planning
bash scripts/setup.sh
Or manually:
git clone --recursive https://github.com/H-tr/Nav-Planning.git
cd Nav-Planning
pixi install # also builds nav_planning into the environment
pixi run test
Rebuilding the robot models¶
The collision models ship with the package. To rebuild them from the official descriptions (see Robots):
bash scripts/download_assets.sh # official descriptions, pinned commits
pixi run cricket-build
pixi run generate-fk # build-robot -> spherize-robot -> generate-fk
pixi run python scripts/sphere_coverage.py nav_planning/resources/robot locobot
generate-fk depends on spherize-robot, which depends on
build-robot, so one call runs the whole pipeline. Rebuild the package
afterwards (pixi install).