io.github.ashfordeOU/kshana-mcp

It quantifies, in hard and reproducible numbers, what quantum clocks, quantum inertial sensors, and optical time-transfer buy a navigation system over classical PNT — scored against the operational figures of merit that matter for resilient navigation. Every result is reproducible from scenario + seed + engine version, and every sensor parameter is traceable to a published source — consolidated in one citable table in docs/PROVENANCE.md.

Validated against external oracles — every row CI-gated

Each row is checked against an independent external oracle (real dataset, independent reference implementation, or published reference vectors) and re-checked in CI. Full 91-row matrix →

Free and open source under the GNU AGPL-3.0 — with a commercial licence available from Ashforde OÜ for proprietary/closed integration (see LICENSING.md). Professionally developed and maintained by Ashforde OÜ; commercial support, integration, and proprietary extensions available.

Status: v0.23.0 · a validated, reproducible simulation substrate for PNT resilience. A fully reproducible engine spanning the PNT stack — orbit geometry and constellation design, a numerical (Cowell) propagator with a seven-perturbation force model, maneuver and trajectory design, time systems, inertial navigation (incl. map-aided and gravity-map-matching alt-PNT), GNSS/INS fusion (loose, tight, UKF, coupled clock+position, 17-state), orbit determination, ARAIM integrity, clocks, advanced time-and-frequency transfer, the GNSS measurement domain, resilience (jamming + multi-layer spoofing), and an open deep-space / Mars radiometric navigation engine (light-time + Shapiro, CCSDS-TDM, reduced-dynamic SRIF, one-/two-way fusion); plus first-order mission-analysis budgets (launch / re-entry / EO-coverage / pointing / ground-station passes / link), a space-weather environment model, an AI/ML RF-impairment evaluation testbed, and the versioned Kshana Interchange Format (KIF). Honest by design: every figure of merit is labelled validated or modelled, and optical-clock figures are space goals on ground hardware (no strontium optical clock has flown).

Validation ladder (maturity is not uniform across domains — and saying so is the point):

Domain	Tier
Earth PNT (orbit, frames, time, clocks, IMU, integrity)	Real-data validated — ESA SP3 (Galileo 0.13 m / 8 h · 0.61 m / 24 h, Swarm-A 0.10 m), NIST SP1065, SOFA/ERFA, heritage vectors
Deep-space / Mars navigation	Simulation-validated — synthetic closed-loop OD + analytic self-consistency; Sun-central dynamics cross-checked vs JPL DE440 (137 m @ 1-day arc)
Real-mission deep-space OD	Roadmap — pending real DSN/ESTRACK tracking-data validation

Deep-space figures (Mars-LMO OD ≈ 0.2 m; relay-PNT orbiter 0.4 m / rover 5.1 m) are simulation / covariance figures of merit, not real-mission results. See Capabilities for what it does, What it is / is not for scope, and docs/CAPABILITY.md / docs/VALIDATION.md for per-capability maturity. The overclaim closure ledger docs/CLAIMS-VS-REALITY.md tracks every historical overclaim, how it was resolved, and a CI guard (tests/no_overclaims.rs) that keeps it resolved.

Try it in your browser: the playground runs the engine client-side as WebAssembly — pick a scenario, edit the parameters, and see the result, with nothing uploaded. Build it locally with ./web/build.sh (see web/README.md), or publish it to GitHub Pages via the pages workflow.

New to this? In plain terms: GPS-style satellite signals tell things where they are and what time it is. When those signals are lost (jammed, blocked, or out of view in space), a system has to keep going on its own onboard clock and motion sensors — and they slowly drift. "Quantum" clocks and sensors drift far more slowly. Kshana measures, in honest numbers, how much longer a quantum-equipped system can coast before it exceeds its accuracy limits. New readers should start with the plain-language primer and the glossary.

Contents

• Why · What it is / is not · Capabilities · Results
• Install & build · Usage (Python, WebAssembly)
• Scenario format · Output · Architecture
• Repository layout · Validation & honesty
• Documentation · FAQ · Troubleshooting
• Roadmap · Contributing · Citing · Versioning & releases · License
• Support & professional services · References

Why

Resilient PNT depends on holding position and time when GNSS is denied or jammed. Quantum sensors promise far slower drift during those outages. There is no good open tool to quantify that advantage honestly and reproducibly — so primes, agencies, and labs each rebuild private one-offs. Kshana aims to be the neutral, citable reference for exactly this question.

The engine knows nothing about "quantum" vs "classical": each sensor is an error model plugged into a common pipeline, so a quantum and a classical device are compared apples-to-apples on the same scenario, with independent noise realizations.

What it is / is not

It is: a deterministic, dependency-light engine spanning the PNT stack — orbit geometry, inertial navigation, GNSS/INS fusion, integrity, clocks, and timing. It runs a scenario (often a GNSS outage), evolves calibrated sensor error models through the appropriate estimator, and scores the result against the operational figures of merit — emitting a reproducible JSON result and an SVG chart, from a Rust library, a CLI, a Python extension, an in-browser WebAssembly module, a Model Context Protocol (MCP) server for AI agents, or a JetBrains IDE plugin.

It is not: flight hardware, a quantum-payload design, a full GNSS signal receiver, or a certified avionics product. Quantum-hardware fidelity comes from published error models, not from this tool. The granular maturity of each capability is documented in docs/CAPABILITY.md.

It is not (yet): a full atom-interferometry physics engine (most quantum sensors consume published Allan/noise-budget coefficients; the CAI accelerometer has a first-principles layer — Mach–Zehnder phase, projection noise, contrast decay, and vibration coupling — but Coriolis and light-shift systematics remain a P2 roadmap layer, see ROADMAP.md and docs/QUANTUM-MODELS.md); a full GNSS signal-acquisition receiver (it now solves a single-point PVT position fix from real RINEX code observations — validated on real IGS data — but does not acquire or track raw signal); or a full mission-design suite (it has Lambert / porkchop / maneuver / orbit-determination building blocks, but is the performance-simulation layer above GMAT/Orekit, not a replacement). Owning this scope is deliberate. If you need first-principles cold-atom interferometer error budgets (e.g. CARIOQA-PMP-grade or X-37B-style validation), see the P2 roadmap and get in touch to collaborate.

Capabilities

One engine spans the whole PNT stack — and its maturity is honest per domain: Timing, Orbits and GNSS geometry are heavily externally validated; Lunar and several quantum/resilience domains are deliberately Modelled until real tracking data exists.

The full domain-by-domain detail follows; for a per-capability maturity ledger see docs/CAPABILITY.md and docs/VALIDATION.md.

Each capability is reachable as a Rust API, a runnable scenario kind, or both. Maturity per capability — validated, runnable, or library — is tracked in docs/CAPABILITY.md. A machine-checked verification matrix (src/verification.rs) renders the requirement → module → test → oracle → status cross-reference, with unit-tested honesty invariants that permit a validated label only where an independent external oracle backs it — and that record the hardware/PA capabilities Kshana deliberately does not provide.

Results

Each scenario compares a quantum sensor against its classical counterpart through a ~1.8 h GNSS outage. Numbers are reproducible (scenario + seed + version).

The advantage is outage- and vibration-dependent, with an explicit break-even where classical wins — shown honestly across the technology-readiness ladder (optical-clock figures are ground-demonstrator targets; no strontium optical clock has flown):

The capstone shows the fusion thesis: optical inter-satellite time-transfer keeps even a classical clock locked, isolating the inertial sensor as the classical suite's weak link — i.e. quantum inertial + optical timing together.

A further scenario, orbit-gnss-challenged.toml, derives GNSS availability from orbital geometry rather than hand-authored windows: a spacecraft inside the GNSS shell is propagated against a GPS-like Walker constellation, and the visible-satellite count (line-of-sight, Earth-occultation, elevation mask) sets the fix state at each step. Over a day the user is in fix only ~59% of the time; the quantum clock holds a 5 ns timing solution through every gap (availability 1.0), the chip-scale clock only ~0.83.

The constellation can also be given as real two-line element sets. A full TLE (line 1 + line 2) is propagated with the full SGP4/SDP4 model — including atmospheric drag and the deep-space lunar-solar and 12 h / 24 h resonance terms that matter for ~12 h GNSS orbits — validated against the official AIAA 2006-6753 vectors to a worst-case ≈ 4 mm. scenarios/orbit-sgp4-gps.toml ships a real Celestrak gps-ops snapshot of the operational GPS constellation (2021-07-28, 30 satellites) and requires valid TLE checksums — two-line element sets are open data from the US Space Force / 18th Space Defense Squadron catalogue, redistributed by Celestrak (Dr T. S. Kelso, celestrak.org); refresh with scripts/fetch_tles.sh. A line-2-only block keeps the analytic two-body propagation (scenarios/orbit-real-tle.toml); the two forms can be mixed in one constellation. A constellation can equally be built from a block of RINEX-3 GPS broadcast-ephemeris records — the format a receiver decodes — propagated by the IS-GPS-200 user algorithm and fed through the same geometry (scenarios/orbit-rinex.toml).

Install & build

Requires a Rust toolchain (≥ 1.75; developed on 1.93).

git clone https://github.com/ashfordeOU/kshana
cd kshana
cargo build --release
cargo test          # all tests pass

Usage

Run any scenario; the CLI dispatches on the scenario's kind field and writes a <scenario>.result.json and a <scenario>.chart.svg next to it:

cargo run -- scenarios/clock-holdover.toml
cargo run -- scenarios/imu-deadreckoning.toml
cargo run -- scenarios/timetransfer.toml
cargo run -- scenarios/hybrid-pnt.toml
cargo run -- scenarios/orbit-gnss-challenged.toml
cargo run -- scenarios/orbit-sgp4-gps.toml
cargo run -- scenarios/orbit-rinex.toml
cargo run -- scenarios/integrity-raim.toml

# Export a propagated constellation to an SP3-c precise-ephemeris file:
cargo run -- scenarios/orbit-sgp4-gps.toml --export-sp3 gps.sp3

# Export the constellation's mean elements to a CCSDS OMM catalogue (one OMM
# message per TLE-defined satellite, with its real NORAD id / COSPAR designator):
cargo run -- scenarios/orbit-sgp4-gps.toml --export-omm gps.omm

# Export the velocity-carrying state to a CCSDS OEM 2.0 ephemeris (GMAT/Orekit/STK):
cargo run -- scenarios/orbit-sgp4-gps.toml --export-oem gps.oem

Other CLI modes — lint a scenario, feed real Earth-orientation data, or run a whole suite:

# Lint a scenario without running it (checks the kind + required fields):
cargo run -- --validate scenarios/integrity-raim.toml

# Feed a real IERS Earth-orientation file (finals2000A) for frame precision:
cargo run -- scenarios/orbit-sgp4-gps.toml --eop tests/fixtures/agency/eop/finals2000A_2022001.txt

# Run a SUITE of scenarios into one aggregated, stamped study artifact
# (writes <suite>.study.json + <suite>.study.html next to the manifest):
cargo run -- --study scenarios/quantum-pnt-demonstrator.suite.toml --study-name "Quantum-Enabled PNT demonstrator"

A suite manifest is a small TOML — a title and a scenarios = [ … ] array of scenario paths — that the engine runs in turn, folding every result (with its MODELLED / VALIDATED labels) into one self-describing study artifact. See scenarios/quantum-pnt-demonstrator.suite.toml.

Interoperability role. Kshana is the performance-simulation layer that sits alongside the post-processing toolchain, not a replacement for it: feed its RINEX output into RTKLIB or gLAB for a position solution, and use its SP3 output as a precise-orbit product for tools like Ginan — Kshana answers what resilience a given PNT architecture buys before you have real signals, in formats those tools already ingest (--export-sp3, or export_sp3 = true in an orbit scenario, writes <scenario>.sp3). The same orbit can be published as standards-track CCSDS OMM mean elements (--export-omm, or export_omm = true, writes <scenario>.omm) — one OMM 502.0 KVN message per TLE-defined satellite, carrying each object's real NORAD catalogue number, COSPAR international designator, and epoch, for any OMM-aware consumer instead of a bespoke two-line element set.

Example output (clock holdover — note the Integrity and Security figures of merit):

scenario c827e5d40d25 | quantum holdover 6600s p95 0.0ns integrity 1.000 security 0.997 | classical holdover 2610s p95 19.7ns integrity 1.000 security 0.000
wrote scenarios/clock-holdover.result.json and scenarios/clock-holdover.chart.svg

The optical clock's tight detection floor keeps security 0.997; the chip-scale clock's own noise over the monitoring window exceeds the 20 ns spec, so it has no spoof-detection margin (security 0.000). The orbit scenario additionally reports a geometry block — fraction of samples with a fix, and best/median PDOP and position accuracy — alongside the clock result.

Read these two numbers carefully. security is an analytic spoof-detectability bound derived from each clock's stability — it is meaningful only against a configured spoofing scenario and is not a multi-satellite RAIM detector. integrity here is the filter's self-consistency (fraction of outage samples inside its own k-sigma bound), not an aviation HPL/VPL integrity figure. See docs/INTEGRITY.md.

For genuine receiver-autonomous integrity, the integrity scenario kind (scenarios/integrity-raim.toml) runs real snapshot and solution-separation (ARAIM-style) RAIM over the propagated constellation geometry: it computes horizontal/vertical protection levels (HPL/VPL) per epoch and reports the fraction of epochs that meet the configured alert limits, with a Stanford integrity diagram for error-vs-PL classification.

Reproducible study artifacts

Four open studies each regenerate a byte-deterministic artifact (fixed seed) from one command — the numbers behind the quantum-vs-classical crossover, RF-impairment optimism-gap, PNT-resilience-scoring, and timing-protection-level studies:

# Quantum-vs-classical resilience crossover map (writes paper/crossover/*.json):
cargo run --release --bin crossover_study -- paper/crossover

# RF-impairment optimism-gap study (13-detector panel, scaling laws, LOO predictor):
cargo run --release --example optimism_study -- paper-artifacts/optimism-study.json

# Framework-aligned PNT-resilience scoring + decision-instability study:
cargo run --release --example resilience_report -- paper-artifacts/resilience-study.json

# Conditional Timing Protection Level, calibrated on a real recorded spoof:
cargo run --release --example tpl_jammertest

Each artifact records its engine version, seeds, and a config hash and carries an honest MODELLED/VALIDATED label. The real-data probes (*_probe) run the same pipeline over recordings you supply locally; no datasets are shipped in the repo. The RF-impairment optimism-gap study is written up in the preprint arXiv:2606.22054, and the conditional timing protection level (tpl_jammertest above) in the preprint arXiv:2606.24210 (see Citing).

Python

An optional Python extension (PyO3, abi3) wraps the same engine. Build and install it with maturin:

pip install maturin
maturin develop --features python   # or: maturin build --features python

import json, kshana

result = json.loads(kshana.run(open("scenarios/clock-holdover.toml").read()))
print(result["quantum"]["fom"]["integrity"])

# json, svg, and a one-line summary at once:
result_json, chart_svg, summary = kshana.run_full(open("scenarios/orbit-gnss-challenged.toml").read())
print(kshana.version(), summary)

Beyond run / run_full / version, the module exposes run_typed (a structured result object), validate_toml (lint → list of error strings), list_kinds / scenario_kinds (the dispatchable kinds), and error_kind (the KshanaError tag for a rejected scenario) — see docs/PYTHON_API.md.

Wheels are built for Linux, macOS, and Windows by the wheels workflow on each release tag.

WebAssembly

The engine also runs in the browser via wasm-pack:

wasm-pack build --target web -- --features wasm

import init, { run, chart_svg, version } from "./pkg/kshana.js";
await init();
const result = JSON.parse(run(tomlText));
console.log(version(), result.classical.fom.timing_p95_ns);

The module also exports summary (the one-line result string), list_kinds / error_kind (introspection), and encode_permalink / decode_permalink — the shareable-URL codec the playground uses to round-trip a whole scenario through the address-bar fragment.

AI agents (MCP)

Kshana ships an MCP server, kshana-mcp, so AI assistants and agents can run the validated engine instead of guessing the math — usable from Cursor, JetBrains AI Assistant / Junie, and any MCP-compatible assistant or agent. It exposes run_scenario, list_scenario_kinds, validate_scenario, export_sp3, and export_omm (each a thin wrapper over kshana::api).

cargo install kshana-mcp                          # crates.io
docker run --rm -i ghcr.io/ashfordeou/kshana-mcp  # or OCI, no Rust toolchain

Then register kshana-mcp in your client's mcpServers config. In Claude Code it's one command — claude mcp add kshana -- kshana-mcp — or install the plugin: /plugin marketplace add ashfordeOU/kshana then /plugin install kshana@ashforde. Copy-paste config for Claude Code, Claude Desktop, Codex, Cursor, VS Code, Windsurf and JetBrains is in docs/integrations.md (per-client snippets also in mcp/kshana-mcp/README.md). The server is a standalone, workspace-excluded crate (the rmcp SDK is edition 2024), so it never affects the lean published kshana crate or its build.

In a JetBrains IDE you can also install the Kshana — PNT simulator plugin from the JetBrains Marketplace (or Settings → Plugins → Marketplace → search "Kshana") to run scenarios from a right-click — see ide/jetbrains/.

Scenario format

Scenarios are declarative TOML. A top-level kind selects the pack — forty-four in all (clock is the default if omitted): inertial, timetransfer, hybrid, hybrid-ukf, fusion, gnss-ins, orbit, ephemeris, gnss-sim, integrity, lunar-integrity, lunar-time-offset, spoof, spoof-detect, jamming, sweep, sweep-nd, gravity-map, terrain-nav, terrain-slam, combined-altpnt, pvt, mars-pnt, impairment-eval (AI/ML RF-impairment detection evaluation testbed — labelled synthetic corpus + detector-agnostic ROC/AUC harness + in/out-of-distribution optimism gap), quantum-trade (quantum-vs-classical PNT trade with measured-ADEV ingestion + GNSS-denied resilience envelope; MODELLED), space-weather (solar/geomagnetic indices + Jacchia-71 exospheric temperature + activity-driven thermospheric density over the static atmosphere; MODELLED), oem-interop (CCSDS OEM import/round-trip bridge for GMAT/Orekit/STK ephemerides; MODELLED), the mission-analysis trio launch-window (two-body launch azimuth / plane-change / opportunities), reentry (Allen-Eggers ballistic re-entry corridor), eo-coverage (EO swath / GSD / access / revisit geometry), space-packet (CCSDS 133.0 TM/TC Space Packet framing — exact bit layout, round-trip verified), and attitude-budget (3-DOF gravity-gradient torque + RSS pointing error budget), passes (ground-station rise/set pass prediction — AOS/TCA/LOS, max elevation, access), and link-budget (one-way CCSDS/DSN link equation — FSPL / Eb·N₀ / margin / closure); the lunar-PNT suite lunar-vlbi, lunar-joint-od-clock, lunar-frame-realisation, moonlight-service-volume, lunar-differential-pnt, lunar-interop-export; and the Quantum-Enabled PNT demonstrator quantum-time-transfer, quantum-gnss-free-nav, quantum-anomaly-detect — the mission-analysis trio and these later kinds all MODELLED. Common fields: seed, a [time] grid, a [gnss] availability timeline (the outage driver), and per-sensor blocks with provenance strings citing the source of every figure. Example (clock):

seed = 42
threshold_ns = 20.0
[time]
step_s = 10.0
duration_s = 7200.0
[gnss]
windows = [
  { t0 = 0.0,   t1 = 600.0,  state = "nominal" },  # 10 min GNSS sync
  { t0 = 600.0, t1 = 7200.0, state = "denied"  },  # ~1.8 h outage
]
[clock_quantum]
id = "optical-sr-lattice"
provenance = "Strontium optical lattice clock, space-oriented goal sigma_y(1s)=1e-15 (arXiv:1503.08457)"
y0 = 5.0e-17
q_wf = 1.0e-30   # white FM:  q_wf = sigma_y(1s)^2
q_rw = 0.0       # random-walk FM
drift = 0.0      # linear aging (per second)
[clock_classical]
id = "csac-sa45s"
provenance = "Microchip SA65 / SA.45s CSAC datasheet sigma_y(1s)=3e-10"
y0 = 5.0e-10
q_wf = 9.0e-20
q_rw = 0.0
drift = 0.0

Optional fields (off when absent): a clock may add flicker_floor (1/f FM Allan floor); an inertial sensor may add gyro_bias and q_arw (gyro bias and angular random walk), and bias_instability and q_aa (the Allan bias-instability floor and acceleration random walk) — together a single-axis (1-DOF) accelerometer error budget (VRW/ARW and bias-instability). This is the error budget the shipped inertial scenario pack runs. Separately, the library now carries a verified 3-axis strapdown navigator (src/inertial/{attitude,mechanization,imu_errors}.rs): quaternion attitude with coning/sculling compensation, a full NED mechanization (Earth-rate and transport-rate terms, WGS-84 Somigliana gravity), and a deterministic IMU error model in which scale-factor, misalignment, g-sensitivity, quantization, and rate-ramp are modelled (IEEE Std 952-1997 §A.2; Groves 2013 §4.3). That 3-axis path is now wired into a runnable loosely-coupled GNSS/INS pack (kind = "gnss-ins"): a 15-state error-state EKF disciplines the strapdown solution against noisy fixes while GNSS is up, then coasts through the outage, reporting the fused horizontal error against the open-loop free-INS coast. A tightly-coupled pseudorange update is also available (it forms the innovation in the range domain, so it keeps correcting with fewer than four satellites). A clock-holdover scenario may add runs (> 1) to run a Monte Carlo ensemble — each figure of merit is then reported as a mean with a 5th–95th-percentile spread and the chart shades the error confidence band (see scenarios/clock-ensemble.toml).

A fusion scenario (same blocks as hybrid) runs two independent Kalman estimators — one for the clock state, one for the position state — disciplined by GNSS and aided by optical time transfer, and reports a combined holdover FoM. The two blocks share no cross-covariance: this is a stacked pair of error budgets, not a true coupled clock+position joint filter (cross-block covariance is a roadmap item). See scenarios/fusion-pnt.toml.

A spoof scenario injects a time-spoof — one of four [attack.shape] kinds (linear_ramp, step_jump, meaconing, replay; a bare rate_ns_per_s is still accepted as a linear ramp) — and runs each clock's spoof detector. The detector is a two-sided χ²₁ energy / Neyman–Pearson test on the clock-aided monitor statistic: the threshold is set from a target false-alarm budget target_pfa, and the missed-detection probability P_md is reported both closed-form and by Monte-Carlo (mc_runs trials per hypothesis — the two agree to a few ×1/√N). The Security figure of merit is 1 − P_md at the operationally-harmful (spec) magnitude, so a quiet clock that catches a spec-sized spoof scores ≈ 1 and a noisy one that often misses it scores lower (see scenarios/spoof-attack.toml, scenarios/spoof-meaconing.toml).

A gnss-sim scenario is a measurement-domain simulation: for each visible satellite it synthesises the pseudorange ρ = geometric range + c·δt_rx − c·δt_sv + I + T + noise + multipath and the L1 Doppler, with the Klobuchar single-frequency ionosphere ([iono], IS-GPS-200 §20.3.3.5.2.5) and the Saastamoinen zenith troposphere projected by the Niell (1996) mapping function ([tropo]). The residuals feed snapshot RAIM for per-epoch HPL/VPL, and every satellite's pseudorange, Doppler, C/N₀, and iono/tropo corrections are emitted in the JSON gnss_measurements array. It is a forward simulator (it generates measurements from a known truth), not a receiver/solver — a zero-noise run reproduces geometry plus the corrections to sub-millimetre (see scenarios/gnss-sim-raim.toml).

A jamming scenario models RF interference as a link budget: a [jammer] (ECEF position, transmit power_dbw, type) raises the jammer-to-signal ratio at a [receiver] watching a Walker [constellation]. From the geometry (free-space path loss and the per-direction receive-antenna gain) it computes each satellite's J/S, the effective C/N₀ via the standard anti-jam equation (despreading processing gain × the spectral-separation factor Q; Kaplan & Hegarty §9.4), and flags loss of lock below a configurable tracking threshold — reporting an availability_under_jamming figure of merit. A 10 W broadband jammer at 1 km denies the receiver entirely (J/S ≈ 72 dB); the same jammer at 100 km only degrades the links (see scenarios/jamming-demo.toml).

A sweep scenario runs a trade study: it varies one parameter (threshold_ns, duration_s, quantum_q_wf, or classical_q_wf) from start to stop over steps points on a lin or log scale, records a metric (e.g. holdover_s) for both clocks, and charts the two curves. The base scenario goes under [base] (see scenarios/sweep-clock-stability.toml).

A sweep-nd scenario generalises this to any pack and any number of axes: it varies dotted TOML keys of a [base] scenario (of any kind) over the Cartesian product of [[axes]], re-runs each grid node, and records metrics given as dotted JSON paths into the result (e.g. classical.fom.holdover_s). It works for every pack because it operates at the TOML/result boundary; native runs evaluate the grid in parallel (no extra dependency, wasm falls back to sequential) and the output is deterministic and row-major (see scenarios/sweep-nd-inertial.toml).

An orbit scenario derives the [gnss] timeline from geometry instead of authoring it — give a [user] orbit, a [constellation], an elevation mask_deg, and the two clock blocks. It also reports position accuracy from the satellite geometry; the optional sigma_uere_m (1-sigma user-equivalent range error, default 1 m) scales the position dilution of precision into a position sigma. The user orbit may be made eccentric with eccentricity and argp_deg, and j2 = true adds Earth-oblateness secular drift (see scenarios/orbit-molniya.toml). The constellation can instead be a real one: give [constellation] a tle block of two-line element sets and the satellites are parsed from it (see scenarios/orbit-real-tle.toml). Add one or more [[constellations]] blocks for multi-GNSS (e.g. GPS + Galileo; see scenarios/orbit-multignss.toml):

kind = "orbit"
seed = 7
threshold_ns = 5.0
mask_deg = 10.0
sigma_uere_m = 1.0           # optional; position sigma = position-DOP * this
[time]
step_s = 60.0
duration_s = 86400.0
[user]                       # spacecraft (altitude in km, angles in deg)
altitude_km = 8000.0
inclination_deg = 0.0
[constellation]              # Walker-delta GNSS (GPS-like)
altitude_km = 20180.0
inclination_deg = 55.0
planes = 6
sats_per_plane = 4
phasing_f = 1.0
[clock_quantum]  # ... as above
[clock_classical]  # ... as above

The GPS-denied alt-PNT kinds navigate with no GNSS at all, matching a measured field sequence against a map through a particle filter. A gravity-map scenario flies a track through a spherical-harmonic gravity-anomaly field and recovers it from a cold-atom gravimeter's reading (scenarios/gps-denied-gravity-nav.toml); a terrain-nav scenario does the same against an SRTM elevation DEM (TERCOM/SITAN, scenarios/terrain-nav.toml); and a combined-altpnt scenario fuses gravity + IGRF magnetic + terrain in one filter (scenarios/combined-altpnt.toml).

A lunar-integrity scenario evaluates cislunar PNT: it runs a lunar south-pole ARAIM protection-level pass against a LunaNet/LNIS relay set and honestly reports the integrity gap — a ~30 m lunar σ_URE drives the protection level well above a 50 m alert limit, so the service is unavailable under aviation-style integrity rules (scenarios/lunanet-araim.toml).

A lunar-time-offset scenario reports the relativistic Earth–Moon clock rate — the basis of a Lunar Coordinate Time scale (LTC/TCL). A first-principles post-Newtonian identity sums the self-potential difference (IAU L_G geoid potential minus the Moon's surface self-potential) and the Moon's kinetic (second-order Doppler) term to a secular rate of ≈ 57 µs/day, reported with the published 56–59 µs/day band; it also gives the accumulated LTC−TT offset over a horizon and an inverse-variance ensemble (a lunar paper-clock). MODELLED — the headline figure is reference-dependent (Earth geoid vs lunar selenoid, averaging window), which is why a band, not a single certified number, is reported (scenarios/lunar-time-offset.toml).

See scenarios/ for at least one worked example of every kind (44 kinds, 56 scenario .toml files + 1 suite manifest — several kinds ship more than one example). A few kinds have an example file whose name differs from the kind: lunar-integrity → scenarios/lunanet-araim.toml, gravity-map → scenarios/gps-denied-gravity-nav.toml. List the dispatchable kinds at any time with cargo run -- --validate <file> errors, the Python list_kinds(), or the MCP list_scenario_kinds tool.

Output

The result artifact is versioned, self-describing JSON: per-step time series, the scored figures of merit, the active model specs (with provenance), the seed, a scenario hash — so any chart can be reproduced from the file — and, for each clock, an adev_curve ([{tau_s, adev, n_samples, noise, edf, ci_lo, ci_hi}]): the overlapping Allan deviation across octave-spaced averaging times — the standard way to read a clock's stability — now with a noise-type-specific 95% confidence band per point (the record's power-law type is identified from its modified-Allan slope, and the χ² interval uses the matching NIST SP 1065 effective degrees of freedom). The browser playground renders it as a log-log "Clock stability (ADEV)" chart. (MDEV, TDEV, and HDEV are available as library estimators; the exported result curve is the overlapping ADEV.) Every field, with units and a source pointer, is documented in docs/SCHEMA.md.

Every chart is self-describing. The browser playground, the CLI's *.chart.svg export, and the HTML scorecard all stamp each chart image with a footer reading Kshana v<version> · scenario <hash> · kshana.dev. The scenario <hash> is the first 12 hex characters of the run's scenario hash — a SHA-256 over the canonical scenario definition (seed, thresholds, model parameters, GNSS windows, …); the integrity and lunar reports, which carry no hash of their own, fall back to a SHA-256 of the scenario source. It is the same fingerprint shown in the one-line summary and the result JSON, so a saved or pasted chart always carries its version, the exact scenario that produced it (for bit-for-bit reproduction), and the source — change any input and the hash changes.

The figures of merit follow the standard operational PNT figures of merit:

New to these terms? Each is defined in plain language in the glossary.

Architecture

One engine, many front doors. A single Rust core (kshana) runs every scenario, reached through a CLI, a Python extension, an in-browser WebAssembly module, an MCP server for AI agents, and a JetBrains IDE plugin — all converging on one api::run_toml dispatch. Inside, the sensor packs plug into a common error-model interface; alongside them sit a reference-frame layer (IAU 2006/2000A precession–nutation and the CIO-based GCRS↔ITRS reduction), an astrodynamics/numerical layer (analytic SGP4/SDP4 and a numerical Cowell propagator with its EGM2008/perturbation force model, maneuver design, and orbit determination), an integrity/GNSS layer (RAIM/ARAIM, SBAS, the measurement domain, jamming, cislunar), a fusion / alt-PNT layer (the GNSS/INS estimators and the gravity/terrain/magnetic map-matchers), a deep-space & lunar layer (radiometric Mars-PNT and the MODELLED lunar PNT suite — LTC time, VLBI, joint OD+clock, frame realisation, service-volume, differential PNT, interop), a mission-analysis layer (launch / re-entry / coverage / pointing / pass / link budgets and the space-weather environment), and the open resilience & AI/ML study layer (RPCF resilience scoring, the RF-impairment optimism gap, and the quantum-enabled PNT demonstrator) whose reproducible artifacts ride the validated kernels.

Two standalone, workspace-excluded crates sit beside the core — mcp/kshana-mcp (the MCP server, built on the edition-2024 rmcp SDK) and xval/anise-frames (the ANISE/SPICE frame cross-check, which pulls MPL-2.0 deps) — kept out of the published crate's dependency graph, Cargo.lock, license gate, and MSRV build by the root Cargo.toml exclude list. The JetBrains plugin (ide/jetbrains) is a separate Kotlin project. See docs/ARCHITECTURE.md for the full set of diagrams.

flowchart LR
    SCN["Scenario (.toml)<br/>seed · GNSS timeline · sensor params"] --> ENG
    subgraph ENG["Engine (per step)"]
      direction TB
      M["Error model<br/>step(): evolve noise state"] --> E["Estimator<br/>GNSS-disciplined holdover"]
      E --> F["FoM scoring<br/>vs the 6 figures of merit"]
    end
    ENG --> OUT["result.json + chart.svg<br/>(reproducible: scenario+seed+version)"]

flowchart TD
    cli["CLI · Python · WebAssembly<br/>MCP server · JetBrains plugin"] --> api["api — run_toml<br/>typed dispatch over 44 kinds"]
    subgraph shared["Shared core"]
      types["types · scenario<br/>GNSS timeline"]
      allan["allan — ADEV/MDEV/TDEV/HDEV"]
    end
    subgraph frames["Time and reference frames"]
      ts["timescales · jd2<br/>UTC/TAI/TT/UT1"]
      cio["precession · nutation · cio<br/>GCRS to ITRS, SOFA-anchored"]
    end
    subgraph packs["Sensor packs"]
      p1["clock — models · estimator<br/>kalman · security"]
      p2["inertial — strapdown INS<br/>quantum-CAI"]
      p3["timetransfer — optical/RF<br/>TWSTFT/PPP"]
      p4["hybrid — fused PNT suite"]
    end
    subgraph astro["Astrodynamics and numerical"]
      orbit["orbit · walker · sgp4 · tle<br/>geometry to GNSS and DOP"]
      prop["propagator · forces<br/>gravity_sh · integrator"]
      odm["orbit_determination · maneuver<br/>precise_od — full-force POD"]
    end
    subgraph intg["Integrity and GNSS"]
      raim["raim · sbas — RAIM/ARAIM<br/>HPL/VPL · DO-229E"]
      gsim["gnss_sim · ionex · pvt<br/>measurements + SPP fix"]
      jam["jamming · navsignal<br/>J/S to C/N0 · anti-jam Q"]
    end
    subgraph spf["Spoof detection"]
      spoof["spoof — time-spoof attack"]
      spm["spoof_monitors — AGC power · SQM"]
      det["detection — test-stat theory"]
      spd["spoof_detect — runnable scenario"]
    end
    subgraph fnav["Fusion and alt-PNT"]
      fus["fusion — EKF · UKF<br/>17-state · coupled"]
      alt["gravimeter · mapmatch<br/>particle_filter · altpnt · igrf"]
    end
    subgraph deep["Deep-space · Mars · Lunar"]
      dsr["radiometric · ccsds_tdm<br/>deepspace_od · mars_pnt"]
      lun["lunar suite — cislunar ARAIM<br/>LTC time · VLBI · interop"]
    end
    subgraph resil["Resilience studies and AI/ML"]
      tpl["tpl · resilience<br/>conditional TPL + RPCF"]
      opt["impairment_* · eval_stats<br/>sdr · realdata · quantum_*"]
    end
    VER["verification<br/>machine-checked matrix<br/>SINGLE SOURCE OF TRUTH"]
    api --> packs
    api --> astro
    api --> intg
    api --> spf
    api --> fnav
    api --> deep
    api --> resil
    packs --> shared
    astro --> frames
    odm --> prop
    spoof --> p1
    spm --> det
    spd --> spm
    fus --> p2
    alt --> p2
    gsim -. uses .-> raim
    VER -. cross-refs .-> packs
    VER -. cross-refs .-> intg
    VER -. cross-refs .-> spf
    VER -. cross-refs .-> astro

Components & distribution. The core crate ships through the Rust, Python, and JavaScript ecosystems; the MCP server and IDE plugin reach AI agents and JetBrains IDEs. Each vX.Y.Z tag republishes every channel automatically (see Versioning & releases).

flowchart LR
    subgraph repo["One repository"]
      core["kshana core<br/>library and CLI"]
      mcp["mcp/kshana-mcp<br/>MCP server (excluded crate)"]
      ide["ide/jetbrains<br/>Kotlin IDE plugin"]
      subgraph xval["xval cross-checks (excluded)"]
        anise["anise-frames · lunar-od<br/>mars-od · service-geometry<br/>Rust ANISE / SPICE DE440"]
        orekit["orekit-passes<br/>Java Orekit"]
      end
    end
    core --> crates["crates.io"]
    core --> pypi["PyPI — wheels"]
    core --> npm["npm — WebAssembly"]
    core --> rel["GitHub Releases<br/>binaries · SBOM · SLSA<br/>validation summary"]
    core --> pages["kshana.dev<br/>GitHub Pages playground"]
    core -. archived .-> zen["Zenodo DOI"]
    mcp --> crates
    mcp --> ghcr["ghcr.io — OCI image"]
    mcp --> reg["official MCP registry"]
    ide --> jb["JetBrains Marketplace"]
    anise -. validates .-> core
    orekit -. validates .-> core

Repository layout

kshana/
├── src/                                       # the kshana core crate (library + CLI)
│   ├── api.rs · main.rs · lib.rs              # typed dispatch (44 kinds) + CLI + crate root
│   ├── python.rs · wasm.rs                    # optional PyO3 / wasm-bindgen bindings
│   ├── types.rs · scenario.rs · allan.rs      # shared core (time grid, GNSS timeline, Allan)
│   │
│   ├── models.rs · estimator.rs · kalman.rs   # Pack 1 — clock holdover + integrity
│   ├── security.rs · detection.rs · spoof.rs · spoof_monitors.rs  # spoof detection
│   ├── filter_health.rs · fom.rs · fom_label.rs · report.rs · chart.rs · run.rs  # health · FoM scoring + labelling · output
│   ├── suite.rs · study.rs                     # scenario suites + aggregated multi-scenario study artifacts (`--study`)
│   ├── inertial/                              # Pack 2 — strapdown INS (attitude · mechanization · imu_errors · quantum_imu)
│   ├── timetransfer.rs · timetransfer_adv.rs · timegeo.rs  # Pack 3 — TWSTFT/CV/PPP/optical, Sagnac
│   ├── hybrid.rs · ensemble.rs · sweep.rs     # Pack 4 — fused PNT, Monte-Carlo, trade sweeps
│   │
│   ├── timescales.rs · jd2.rs · ephem.rs      # time systems, two-part JD, Sun/Moon ephemeris
│   ├── precession.rs · nutation.rs · cio.rs   # IAU 2006/2000A precession-nutation + CIO GCRS↔ITRS
│   ├── frames.rs · *_data.rs                  # TEME↔ECEF + generated nutation/CIO/EGM2008/IGRF tables
│   │
│   ├── orbit.rs · sgp4.rs · tle.rs · walker.rs   # geometry, SGP4/SDP4, TLE, Walker design
│   ├── propagator.rs · forces.rs · gravity_sh.rs · integrator.rs  # Cowell + perturbations (EGM2008 d/o70, GR) + RK4/DOPRI
│   ├── maneuver.rs · batch_ls.rs · orbit_determination.rs  # burns/Lambert/porkchop, Gauss-Newton, OD
│   ├── cr3bp.rs · lunar.rs · lunar_frame.rs · lunar_od.rs  # Earth–Moon CR3BP + halo/NRHO STM corrector, cislunar/LunaNet ARAIM, MCI↔MCMF, lunar OD
│   ├── lunar_time.rs · lunar_vlbi.rs · lunar_combination.rs · lunar_frame_realise.rs · lunar_service.rs · lunar_dpnt.rs · lunar_interop.rs  # MODELLED lunar PNT suite — LTC time · geodetic VLBI · joint OD+clock · frame realisation · Moonlight service-volume · differential PNT · LunaNet/IOAG interop export
│   ├── body.rs · mars_frame.rs · ephem_provider.rs · radiometric.rs · ccsds_tdm.rs  # deep-space: multi-body · Mars frame · ephemeris seam · radiometric obs + CCSDS-TDM
│   ├── deepspace_od.rs · clock_state.rs · mars_atmos.rs · mars_pnt.rs · linkbudget.rs · gse_sim.rs  # SRIF OD · onboard clock · Mars drag · relay-PNT · link budget · GSE sim
│   │
│   ├── fusion/                                # GNSS/INS — EKF · UKF · tightly_coupled(17) · coupled · closed_loop
│   ├── raim.rs · sbas.rs                      # RAIM/ARAIM HPL/VPL, SBAS DO-229E PLs + L1/L5 iono-free
│   ├── gnss_sim.rs · ionex.rs · pvt.rs · jamming.rs  # measurement domain · ionosphere maps · single-point positioning · jamming
│   ├── navsignal.rs                            # nav-signal PSD (BPSK-R/BOC) · spectral-separation → anti-jam Q · DLL code-tracking jitter · multipath envelope
│   ├── gravimeter.rs · igrf.rs · mapmatch.rs · particle_filter.rs · altpnt/  # gravity/magnetic/terrain alt-PNT
│   ├── rinex.rs · rinex_obs.rs · glonass.rs · sp3.rs · oem.rs · omm.rs · permalink.rs  # interop formats
│   ├── launch.rs · reentry.rs · eo_payload.rs · attitude_budget.rs · passes.rs · space_packet.rs  # mission-analysis budgets + CCSDS Space Packet
│   ├── space_weather.rs · holdover.rs · tpl.rs   # space-weather environment · GNSS-denied clock-holdover calculator · conditional Timing Protection Level (under spoofing)
│   ├── resilience/                              # framework-aligned PNT-resilience scoring + decision-instability study (RPCF · Dirichlet · Kendall-τ · diversity collapse · assurance report)
│   ├── impairment_eval.rs · impairment_study.rs · impairment_ml.rs · eval_stats.rs  # AI/ML RF-impairment eval testbed · optimism-gap study · LR/MLP detectors · bootstrap/DeLong/Spearman stats
│   ├── sdr.rs · realdata/                       # software-defined-receiver front end (IQ/IF → E/P/L taps → SQM) + real-data ingest adapters (RINEX · UBX · GnssLogger · JammerTest · Yunnan · SatGrid)
│   ├── crossover.rs · quantum_trade.rs · frugal.rs · integrity_impact.rs  # quantum-vs-classical crossover map · PNT trade · cost-per-coverage ROI · integrity impact
│   ├── quantum_devices.rs · quantum_faults.rs · quantum_nav_od.rs · qtrade.rs · timetransfer_chain.rs · representativeness.rs  # Quantum-Enabled PNT demonstrator — device error models · fault catalogue · GNSS-free quantum OD · unified trade harness · quantum time-transfer chain · representativeness / gaps-to-flight ledger
│   ├── interchange.rs · verification.rs          # KIF artifact envelope · machine-checked verification matrix
│   └── bin/crossover_study.rs · bin/validation_report.rs  # crossover-study artifact generator · release validation-summary HTML
│
├── mcp/kshana-mcp/        # standalone, workspace-EXCLUDED crate — the MCP server (+ Dockerfile, server.json)
├── ide/jetbrains/         # standalone Kotlin/Gradle IntelliJ-Platform plugin
├── xval/                  # standalone, workspace-EXCLUDED external cross-checks: anise-{frames,lunar-od,mars-od,service-geometry} (Rust ANISE/SPICE DE440) + orekit-passes (Java Orekit)
│
├── examples/            # reproducible study generators: tpl_jammertest · resilience_report · optimism_study + real-data probes (jammertest_probe · yunnan_probe · satgrid_probe · texbat_probe · ingest_realdata)
├── paper-artifacts/     # byte-deterministic study artifacts, regenerable from examples/ (optimism-study.json · resilience-study.json); raw datasets stay out
├── scenarios/            # one cited .toml per kind + geometry-driven + GPS-denied
├── scripts/              # reproducibility + repo-hygiene + SBOM guards
├── docs/                 # CONCEPTS, ARCHITECTURE, CAPABILITY, VALIDATION, PROVENANCE, GLOSSARY, …
├── web/                  # the WebAssembly playground + kshana.dev site
├── tools/                # table generators (EGM2008 · IGRF · nutation · CIO) + fetch_tles.sh
├── .github/workflows/    # ci · release · publish · wheels · pages · mcp-publish · jetbrains-plugin · frame-xval
├── pyproject.toml        # Python packaging (maturin)
├── CHANGELOG.md          # Keep a Changelog + SemVer
└── CITATION.cff · ROADMAP.md · CONTRIBUTING.md · SECURITY.md

Documentation

Validation, reproducibility & honesty

• Every noise term is calibrated to a published, cited figure and validated against the standard relation (Allan deviation for clocks; Groves' dead-reckoning error growth for inertial; the timing→ranging conversion for time transfer). Status per term is tracked in docs/VALIDATION.md as validated or not modeled — nothing is presented as validated that is not.
• Reproducible by construction: scenario + seed + engine version → identical bits. scripts/check-reproducible.sh enforces it; quantum and classical runs use independent seeds so their noise is uncorrelated.
• Maturity is stated honestly: optical-clock and optical-link figures are targets / ground-demonstrator results, not flown.

Validation at a glance

Every row is enforced by a named test in CI. This table is a curated highlight; the full machine-checked matrix is 102 rows — 51 VALIDATED, 47 MODELLED, 4 PARTNER (src/verification.rs), with the complete evidence (and what is honestly not yet validated) in docs/VALIDATION.md and the per-release kshana-validation-summary.html artifact (generated by cargo run --bin validation_report, SLSA-attested).

The Status column states the kind of evidence, matching the validation ladder above: VALIDATED = checked against an independent external oracle (real data, an independent library, or published reference vectors); MODELLED = checked against analytic truth or simulation self-consistency (no independent external dataset). VALIDATED describes the method of checking, not a pass/fail — an honest miss against real data (the LRO row) is still VALIDATED. CI rows are process guards, not figures of merit. A few real-data islands (the measured caesium clock, Stable32 PHASE.DAT, and the OPS-SAT/ICGEM checks where the raw inputs carry no redistribution licence) are data-gated: the test prints a skip notice and stays green when the input is absent, and the public reference numbers are committed under tests/fixtures/. Reproduce the raw inputs with the matching scripts/fetch_*.sh.