OSOIX

WP-003 · Draft research architecture v0.1

A digital twin must expose what it knows—and what it does not.

OSOIX Twin is proposed as a provenance-aware representation of orbital assets, mission context and predicted behavior. It separates observation, estimation and simulation so uncertainty is visible instead of hidden behind a single dashboard value.

State architecture

Observed state

Authenticated measurements with source identity, timestamp, calibration status, quality indicators and uncertainty.

Estimated state

A reconciled representation derived from declared filters, assumptions and model versions rather than an unexplained single truth value.

Predicted state

Time-bounded forecasts with confidence intervals, validity horizons and sensitivity to uncertain inputs.

Asset model

Geometry, mass properties, propulsion, power, thermal, communications, payload, degradation and maintenance assumptions.

Environment model

Orbital conditions, conjunction context, radiation, illumination, atmospheric effects and external constraints relevant to the use case.

Scenario branch

A non-authoritative simulation branch used to compare maneuvers, failures, delays, degraded sensors or alternative mission sequences.

Trust controls

01Every state value carries provenance and uncertainty.
02Observed, estimated and predicted values remain distinguishable.
03Model, rule and dataset versions are recorded.
04Prediction validity expires rather than remaining silently trusted.
05Divergence between predicted and observed state creates an explicit event.
06Historical replay uses the information available at that time, not future knowledge.
07Simulation results cannot directly authorize safety-critical execution.

Divergence

Prediction error becomes evidence.

When observations leave the permitted uncertainty envelope, the twin should create a signed divergence event describing the prior prediction, actual observation, model version, elapsed horizon and operational impact.

Replay

Reconstruct decisions fairly.

Mission replay should preserve what operators and systems knew at each decision point. Later observations must not be silently inserted into earlier states, preventing hindsight from corrupting audit.

Explicit non-claims

No live orbital twin or certified simulator is claimed.

This publication does not claim a production orbital-data service, validated high-fidelity dynamics engine, flight-qualified model, customer fleet integration, certified conjunction product or operational mission simulator. The next evidence step is a simulation-only reference twin with synthetic telemetry, model lineage and measurable divergence tests.