Ariadne

Python-native cislunar mission design + TNO discovery — validated, tutorial-driven, source-available. Free for noncommercial use; commercial use requires a license.

The thread through the interplanetary labyrinth: a working, validated astrodynamics toolkit that fills the gap between heavyweight platforms (GMAT, Monte, Copernicus) and the existing Python libraries (poliastro, Tudat, Heyoka) for the things that fall between them, combined in one public source-available, Python-native package — CR3BP-aware cislunar mission design, invariant-manifold transport routing, and end-to-end TNO discovery from real survey astrometry.

What you can do in 30 seconds

import ariadne

# Real CR3BP system constants (Earth-Moon, Sun-Earth, all 4 Jovian moons, Sun-Mars, ...)
em = ariadne.system("EARTH_MOON")

# Build the L1 Lyapunov orbit family in one line
family = ariadne.lyapunov_family("L1", n=30)

# Construct NASA's Gateway-class 9:2 Near-Rectilinear Halo Orbit
nrho = ariadne.gateway_nrho()
print(f"period {nrho.period * em.T_star / 86400:.2f} d, periodic to {nrho.residual:.1e}")
# -> period 6.56 d, periodic to 2.4e-14

# Fit a Trans-Neptunian Object's orbit from its real MPC astrometry
fit = ariadne.discover_tno("90377")     # Sedna
print(f"RMS = {fit['rms_arcsec']:.2f}\"  IOD r = {fit['iod']['r_au']:.0f} AU")
# -> RMS = 3.95"  IOD r = 78 AU

Install

pip install -e ".[dev]"          # editable install + test deps
pytest -m "not slow" -q          # fast tests

Requires Python ≥ 3.10. NASA SPICE kernels (DE440) auto-download on first use.

Runnable tutorials

Seven tight, runnable scripts in examples/ — each ends with a PNG you can look at and a few lines of summary you can read (plus operational scripts 08-10 for the discovery pipeline).

Tutorial	What it demonstrates	Output
01_lyapunov_family.py	Build the Earth-Moon L1 Lyapunov family by amplitude continuation; plot family colored by Jacobi constant	01_lyapunov_family.png
02_gateway_nrho.py	Construct NASA's Gateway 9:2 NRHO via pseudo-arclength continuation; verify period 6.56 d, perilune 3,238 km, apolune 71,198 km, Floquet 2.18 (843× more stable than a deep L1 Lyapunov)	02_gateway_nrho.png
03_manifold_transport.py	Cislunar transport graph: L1↔L2 halo + NRHO↔L2 halo Poincaré patches. Velocity-mismatch alone: 112 / 119 m/s. Honest total with position-gap correction over a 1-day window: 162 / 646 m/s (see benchmarks/heteroclinic_honest_dv.py)	03_manifold_transport.png
04_tno_orbit_fit.py	Discovery-engine filter: pull real MPC astrometry for Sedna / Eris / Quaoar, fit orbits to a few arcseconds RMS, recover (a, e, i) within a few percent	04_tno_orbit_fit.png
05_helmholtz_hjb.py	Coherence-HJB: sampled-graph Helmholtz value function on full 6D CR3BP — 84% greedy reach to lunar goal in sub-second compute, no grid (sidesteps the grid-based curse of dimensionality)	05_helmholtz_hjb.png
06_veega_jupiter.py	Galileo-class Venus-Earth-Earth gravity assist to Jupiter on real DE440 ephemeris; C3 = 16.79 km²/s² (4.6× energy reduction vs direct transfer)	06_veega_jupiter.png
07_lambert_porkchop.py	Earth→Mars Lambert porkchop over a 1.5-yr launch window × 120-360 d TOF grid; finds the Oct 2026 launch at C3 = 9.27 km²/s² (near Hohmann theoretical)	07_lambert_porkchop.png

Gallery

L1 Lyapunov family (Tutorial 01)	Gateway NRHO 3D + Moon-distance (Tutorial 02)

Cislunar manifold transport graph (Tutorial 03)	TNO orbit recovery (Tutorial 04)

Coherence-HJB value field + Δv distribution (Tutorial 05)	Galileo VEEGA Jupiter (Tutorial 06)

Earth → Mars Lambert porkchop (Tutorial 07)

Command-line interface

After pip install -e . you get an ariadne command:

$ ariadne info                       # version + 7 CR3BP systems + capability summary
$ ariadne systems                    # list the systems with mu / L* / T* / V*
$ ariadne lyapunov --point L1 --n 30 # build an L1 Lyapunov family
$ ariadne nrho                       # construct Gateway NRHO + report period/peri/apo/Floquet
$ ariadne discover 90377             # fit Sedna's orbit from real MPC astrometry
$ ariadne benchmark                  # run the reference benchmark suite
$ ariadne tutorial 5                 # run example #5 (Coherence-HJB) end-to-end

Where Ariadne sits

Tool	Domain	Style	Availability
GMAT	Full ops mission design	C++/Java, GUI/script	yes (NASA)
Monte	Operational trajectory	closed	no (JPL)
STK / Ansys	Commercial	GUI	no, $$$$
poliastro	2-body + perturbations	Python	yes
Tudat	Academic OCP	C++/Python	yes
Heyoka	High-precision propagation	C++/Python	yes
Ariadne	CR3BP + cislunar + TNO discovery	Python-native, tutorial-driven	source-available, noncommercial

Capabilities at a glance

Subsystem	Capability	Validated against
CR3BP core	Synodic dynamics, pseudo-potential, Jacobi constant, STM	Conservation to 1e-12
Libration points	L1–L5 to ~1e-11 nondim, 7 systems	Earth-Moon ↔ Sun-Earth ↔ Jovian moons ↔ Sun-Mars
Lyapunov / halo / NRHO	Continuation families, periodicity to 1e-12	NASA Gateway 9:2 NRHO (6.56 d, 3.2 / 71 Mm)
Manifold tubes	Eigenvector-seeded, STM-transported, branched	Jacobi-conserved to 1e-12
Heteroclinic	Energy-consistent Poincaré-section crossings	0.0 m/s same-energy ballistic
Transport graph	Planar (y, vy) + 3D (y, z, vy, vz) + NRHO via y=0	L1↔L2 162 m/s, NRHO↔L2 646 m/s (honest total, 1-day correction window — see benchmarks/heteroclinic_honest_dv.py)
Interplanetary	Lambert porkchop, VEEGA + per-leg DSMs (1-DOF + 4-DOF)	Galileo C3 16.8, Earth→Mars 5.63 km/s
Coherence-HJB	Sampled-graph Helmholtz value function	6D CR3BP: ~84% greedy reach in sub-second
TNO discovery	HelioLinC linker + (r, rdot)-hypothesis IOD + LM	Sedna/Eris/Makemake/Quaoar/2001 FP185 fit 1.4–8.7″
Discovery inference	Calibrated sparse-evidence posterior + benchmark proof harness	Labelled known-object proxies, ZTF/LSST-like alerts, adversarial false positives, reliability curves, precision/recall, ablations
Real-data	MPC ITF (135 MB, 2.6 M tracklets)	515/863 known re-links + 0 new (the honest verdict)
Cross-validation	DE440 ephemeris, GMAT, REBOUND	149 m vs GMAT over 3 days
Proof-carrying routes	CR3BP → BCR4BP → DE440 promotion certificates	(see ariadne.certification)

Cross-validation — the honesty firewall

Every claim is checked against an independent tool:

Validation	Result
Lagrange points	All 5, all 7 systems, residual < 1e-11
Jacobi constant	Conserved to 1e-12 over many periods
Ephemeris (DE440)	spiceypy vs jplephem agree to 6 mm
Integrator	DOP853 vs Radau agree to 0.26 m
Earth–Moon transfer	Ariadne vs NASA GMAT agree to 148.6 m / 0.893 mm·s⁻¹ over 3 days (live re-run 2026-05-30, GMAT R2026a)
Lambert (Curtis Ex 5.2)	Ariadne vs hapsira agree to 0.002 mm·s⁻¹ on every velocity component (independent libraries, same answer to bit level)
Two-body conservation	Energy + ang-momentum drift over 10 elliptic periods: < 1e-12 relative
Circular-orbit closure	After 1 period, position returns to < 1 mm, velocity to < 1 μm/s
TNO orbit fit	Sedna a-error 2.0%, RMS 3.95″ on real MPC data
6D HJB	84% greedy reach to lunar goal, sub-second compute
Synthetic-injection (closed-loop)	100% pure Sedna recovery from a 200-interloper haystack, fit a=527 AU (JPL 506) — a self-consistency check, not an independent cross-validation

Honest scope — what this is, and isn't

It is: a source-available, high-fidelity, validated CR3BP + cislunar mission-design + TNO-discovery engine built on standard gravity and real ephemerides, cross-validated against NASA GMAT.

It is not:

• New physics. The dynamics are standard n-body gravity (CR3BP, BCR4BP, full DE440 ephemeris).
• An operational mission-planning platform. For that, use GMAT, Monte, or Copernicus.
• A discovery service. It's a toolkit you run on data; we're not an MPC alert stream.

The ITF discovery run found 0 new objects in 348 unmatched candidates — the correct, scientifically defensible outcome on a public archive that the MPC's own (excellent) linker has already processed. A genuine new TNO find would require Rubin/LSST-class data deeper than the public MPC bar.

Read more

• docs/ — documentation index (user guides, validation results, dev notes)
• scripts/ — index of the runnable tools and reproducible pipelines
• docs/WHITE_PAPER.md — capstone write-up: methods, full validation table, headline results, honest limitations.
• docs/REVIEWER_GUIDE.md — claim-to-evidence map for skeptical scientific review.
• notebooks/06_reviewer_replication.ipynb — notebook walk-through of the frozen reviewer evidence manifest and corpus.
• MASTER_PLAN.md — full vision, prior art, CR3BP foundations, validation roadmap (45 stages, all gates green).
• docs/PUBLIC_LAUNCH_AUDIT.md — exact public-release checks, package/doc/test evidence, and non-gates to avoid overclaiming.

License & contributing

PolyForm Noncommercial License 1.0.0. Free for any noncommercial use — personal, academic, research, teaching, non-profit, and government. No fee, no sign-up: use it, fork it, build on it, publish from it; just keep the notice intact.

Commercial use requires a paid license (use in a product/service you sell, internal for-profit business use including the for-profit arm of a non-profit, paid consulting, or building/training commercial AI/ML systems). See COMMERCIAL-LICENSE.md or email Josh@coherenceenergylabs.com.

AI/ML training: using this code or its outputs to train or develop AI/ML systems for commercial purposes requires a commercial license; non-commercial ML research is fine. See AI_POLICY.md.

Issues + PRs welcome on GitHub. Console commands after install: ariadne-atlas (build the open-release atlas bundle) and ariadne-figures (regenerate all figures).

Architecture in one diagram

                discovery/        validate/          examples/
                  ▲    ▲              ▲                 ▲
                  │    │              │                 │
                  │    └──────────────┼─────────────────┤
   astroquery → linkage   iod        gates           tutorials
                  │        │           │                 │
                  └────┬───┘           │                 │
                       │               │                 │
                       └──────────┐    │    ┌────────────┘
                                  ▼    ▼    ▼
                          orbits/   dynamics/   manifolds/
                              │  (CR3BP, eom)       │
                              └────────┬────────────┘
                                       │
                          connections/ + transport_graph/
                              │ (Poincaré 2D & 3D)
                              ├─────────────────────────────┐
                              ▼                             ▼
                       interplanetary/                   optimize/
                        (Lambert, VEEGA,            (autodiff, LM, Lambert,
                         per-leg DSMs)               coherence_hjb, Helmholtz)
                              │
                              ▼
                     io/  +  certification/
                  (GMAT export, atlas, proof-carrying routes)