Navigation Evidence Guide¶
Navigation evidence is claim-specific. Decoder acceptance, orbit accuracy, correction behavior, estimator convergence, integrity, PPP lifecycle, and RTK ambiguity resolution are different claims with different proof obligations.
Route The Claim¶
flowchart TD
claim["navigation claim"]
kind{"scientific family"}
format["format and product"]
state["orbit, clock, and time"]
correction["correction or model"]
estimate["position or filter"]
integrity["integrity, PPP, or RTK"]
evidence["family-specific evidence"]
scope["bounded scope statement"]
claim --> kind
kind --> format --> evidence
kind --> state --> evidence
kind --> correction --> evidence
kind --> estimate --> evidence
kind --> integrity --> evidence
evidence --> scope
| claim | required evidence | claim boundary |
|---|---|---|
| External data is interpreted correctly | realistic accepted data, malformed cases, time context, missing fields, and semantic round trip | one format revision and constellation at a time |
| Satellite state or clock meets an accuracy budget | independent broadcast or precise reference, frame, epoch, coverage edges, and uncertainty | named product family and interval |
| A correction has the expected sign and magnitude | independent formula or dataset, units, nominal cases, domain bounds, and refusal | named model and required inputs |
| A position or filter result is valid | truth, residuals, covariance, geometry, convergence, and impossible-case refusal | named estimator and scenario |
| Integrity, PPP, or RTK supports a claim | lifecycle, quality, downgrade, ambiguity or protection evidence, and refusal | named support level, products, and fixtures |
Evidence Must Be Independent¶
flowchart LR
implementation["implementation under test"]
independent["external reference,<br/>published formula, or invariant"]
observed["observed result"]
expected["expected result"]
comparison["comparison with<br/>derived tolerance"]
decision["supported or refused claim"]
implementation --> observed
independent --> expected
observed --> comparison
expected --> comparison
comparison --> decision
Generating expected values with the same parser, propagation routine, correction, or estimator only confirms self-consistency.
Use test strategy to locate evidence, invariants to state physical and result assumptions, and change validation to match proof to the change.
Bound Public-Data Claims¶
Public station and precise-product fixtures demonstrate comparability for selected epochs, geometries, products, atmospheric conditions, and models. They are not universal certification. Preserve provenance and name those bounds in every reader-facing claim.
Passing a parser test does not prove solution accuracy. Passing a standalone position test does not prove PPP, RTK, integrity, or another constellation. Run the estimator family that consumes the changed product or model.
Failure Signals¶
- A tolerance has no physical, reference, or numerical rationale.
- Missing precise data becomes a zero correction.
- A malformed product and an unsupported claim share one generic outcome.
- Convergence is asserted without covariance, residual, or quality evidence.
- One constellation’s reference test is cited for another constellation.
- A public-data fixture lacks provenance or reference-frame context.
- A long-run test replaces focused proof for the changed scientific family.
Use known limitations, the risk register, review checklist, and definition of done before broadening a claim.
Evidence Sources¶
The navigation test guide maps the major proof families. Representative evidence includes broadcast orbit reference, SP3 reference accuracy, position refusal, public PPP convergence, and RTK ambiguity fixing.