Data

All files are in the repository under data/ and results/, produced by scripts that are also in the repository. Formats are versioned and each has a JSON Schema; lunarsim validate <file> checks any of them.

Datasets (`data/`)

File	Contents
`parameters.csv`	Every model parameter with the value used, its basis (source) and a confidence grade.
`mission_manifest.csv`	Landed payload mass at the lunar south pole by year, 2026–2032, low / base / high, with the base-case content.
`cargo_classes.csv`, `lander_types.csv`, `vehicle_classes.csv`	The demand and fleet models exactly as the code uses them.
`site_graph.json`	The nine-site base graph of the v2 study (coordinates in km, edges with slope class and path length).
`scenarios/*.json`	Named scenarios in the `lunarsim-scenario` v1 format: the three demand phases and the Phase-2 fleet-size variants. Load with the command line or import them on the Simulator page.
`synthetic_manifests/P*_30_years.csv`	30 seeded years of landings and cargo items per phase, one row per item.
`sample_traces/*.json`	One replayable simulated year per phase (marketplace, seed 3) in the `lunarsim-trace` format. The World page plays them.
`terrain/connecting_ridge/`	The terrain pack: 16 × 16 km of the Connecting Ridge from the LOLA 5 m DEM as 16-bit height maps (5 m and 20 m), slope, hillshade, metadata with citation, and `sites.json` with the nine sites and the landing-pad sets at 0.5, 1, 2, 3 and 5 km from the base.
`driving/driving-v1.json`	Per-leg driving tables on that terrain: time, energy and interventions for every ordered pair of sites, by autonomy level, control latency and lighting.
`policies/`	Trained landing policies (`lunarsim-policy` v1, plain JSON weights that run in Python and in the browser) with their training records and Monte Carlo outcome tables.
`capacity_ladder.csv`, `fragmentation_law_cells.csv`, `anticipation_paired.csv`, `verification.csv`, `cost_context.csv`	The derived tables behind the v2 paper's figures.

Results (`results/`)

Raw per-seed experiment outputs, one row per simulated year, one CSV per suite (E1–E14 from the v2 study; E15–E20 from the current phase), and the summary tables table_*.csv used by the paper and by this site. Column names are documented in results/README.md.

Formats

lunarsim-scenario v1: a full form (every scenario field) and a compact form (a demand phase plus the page's controls). Both load in Python and in the browser.
lunarsim-trace v1/v2: one simulated year: vehicles, trips with timed segments, cargo rows, nights, landings; v2 adds landing outcomes, plume keep-out windows and shelters.
lunarsim-terrain v1, lunarsim-driving v1, lunarsim-policy v1: the terrain pack, the driving tables and the policy weights described above.

Schemas live in lunarsim/schemas/.

Reproducing

pip install -e ".[full,terrain,ml]"
python -m pytest tests -q
python experiments/run_all.py            # v2 suites E1–E14
python experiments/phase3/run_phase3.py  # current suites E15–E20
python experiments/export_datasets.py
python experiments/make_figures.py

Seeds are fixed and documented in every results file; runs are deterministic. The World and Simulator pages state which engine produced what you see.

Citing

See the Paper page and CITATION.cff. Data produced by the project is released under CC BY 4.0; NASA data keep their own terms (public, credit NASA/GSFC).