Platform Architecture

Six layers. One chain of meaning.

Each layer hands the next one its data together with its quality, so meaning is never added without a trace.

Evidence levelE1Architecture Defined

Six layers, one direction of travel.

Data moves from the physical world to governed outputs. At every step it keeps its time, its quality and its source.

Data flow: Data table
Stage	Value
1	Sources
2	Model
3	Policy
4	Interface
5	Consumer
6	Audit

Data flow CH13Data status: IllustrativeEvidence level: E0 · ConceptQuestion answered: How does data flow through the architecture?A pipeline of six stages, from sources to audit, shows how data moves through the model, the policy and the interface to the consumer, with a record kept at the end.Limitation: A conceptual flow. It does not describe a deployed system.

Eleven stages in four bands.

Each stage does one job and says what it passes on. Nothing is merged with its neighbour.

Observe
1. Physical worldAircraft, drones, birds and weather produce signals.
2. ObservationsEach source records what it saw, not what it concluded.
3. Time alignment and qualityClocks are aligned and every observation gets a quality mark.
Understand
1. CorrelationCandidate links between observations, proposed and not assumed.
2. FusionLinked observations become one supported track.
3. Unified stateOne state per object, with the unknown kept unknown.
4. Classification and contextA class with its own confidence, in its surroundings.
5. Confidence and evidenceHow sure, and what it rests on, kept separate.
Preserve
1. Digital TwinThe time-aware state, with its history kept.Read more — Digital Twin
Deliver
1. Governed outputsAlerts, views, an API and a replay, each with its limits.
2. Site-to-region scaleThe same model joins many sites.

Things the architecture keeps apart.

Most errors in airspace pictures come from merging two ideas that only look alike. These are kept apart on purpose.

Observation, correlation, fusion, stateFour steps with four meanings. A link is not yet a fusion, and a fusion is not yet certainty.
Confidence, evidence, healthHow sure, what it rests on, and whether the source works are three different answers.
Freshness and confidenceA recent value can be weak and an old value can be strong. Each carries its own mark.
A response and a factA successful API call says the request worked, not that the information is valid.

Health runs through every layer.

Each layer reports whether it is available, delayed or silent, and the layers above weaken their claims to match.

Normal
Degraded
Unavailable

System health over time: Data table
Time (relative)	Component 1	Component 2	Component 3
1	Normal	Normal	Normal
2	Normal	Normal	Normal
3	Normal	Normal	Normal
4	Normal	Normal	Unavailable
5	Normal	Normal	Unavailable
6	Normal	Degraded	Normal
7	Normal	Degraded	Normal
8	Degraded	Normal	Normal
9	Degraded	Normal	Degraded
10	Normal	Normal	Normal
11	Normal	Normal	Normal
12	Normal	Normal	Normal

System health over time CH07Data status: Synthetic dataEvidence level: E0 · ConceptQuestion answered: How has system health changed?Three bands, one for each of three example components, show normal, degraded and unavailable periods as different fills and patterns, so state never depends on colour alone.Limitation: Periods are invented and say nothing about any real component.

What this page leaves out.

This is a public description of the architecture. It shows no network topology, no endpoint, no site geometry and no frequency, and the diagrams are drawn for explanation only.

Limits of this description.

Read the architecture as a design, not as a record of performance.

A design, not a measurementThe stages describe how the platform is designed to work. They state no speed, range or accuracy.
Not every source is presentA site may run with fewer sources, and the state then says so.
Context is not completeTerrain and weather improve a reading but never turn a guess into a fact.