- CPU: 4+ cores recommended for scene generation; paper measurements used Intel Haswell-class CPU
- RAM: 8 GB minimum; 16+ GB free recommended for
main_measurements.pkl.gz; 32 GB for full reproduction runs - Disk: 10 GB free space (Docker image, optional Zenodo download)
- Display: headless Docker + web browser
- Docker Engine 24+ and Docker Compose v2
- Web browser for the Streamlit UI at http://localhost:8501
- Optional (non-Docker): Python 3.12, see package.json for pinned dependencies
- Browser uploads: max 300 MB per file (Streamlit
server.maxUploadSize)
Python dependencies are pinned in package.json (pythonDependencies.common). Install locally with: python scripts/install_dependencies.py
- PYTHONPATH=/app/src
- ARTIFACT_DATA_DIR=/data
- ARTIFACT_OUTPUT_DIR=/output
- ENABLE_RRT_ANIMATION=false (headless RRT in artifact mode)
- MPLBACKEND=Agg
- Full evaluation data is published separately under CC-BY 4.0
- Dataset DOI:
10.5281/zenodo.20792733(concept DOI; resolves to latest version; hardcoded insrc/utils/artifact_config.py) - Quick demo (plotting & visualization): load
msr_measurements_for_full_coverage.pkl.gz(~26 MB compressed; 6-vessel cases, not for trajectories) - Full paper archive:
main_measurements.pkl.gz(~241 MB compressed; 16+ GB free RAM recommended)