Multi-Sensor Fusion Engine

1. Four source traces arrive
2. They converge on one candidate
3. One source goes stale
4. The state persists with less confidence
5. The contribution record is replayed

Averaging sources hides conflict and invents confidence. Two sources that disagree do not make one source that is twice as sure; they make a question that anyone reviewing the state must be able to see.

• Correlated candidatesObservations already tied together by correlation, each with its source, time and quality.
• Per-source confidenceHow much each source supports a candidate, kept separately and never averaged away.
• Source availabilityWhich sources are present, delayed, absent or in disagreement at this moment.

1. Tests the relationship between observations before combining them.
2. Fuses compatible information, keeping each source's contribution and the provenance of every field.
3. Propagates uncertainty into the state and recomputes it when a source drops out or disagrees.

• A conflict never raises confidence.
• Every fused line can be reopened to see which sources made it.
• Disagreement stays visible as a mark on the state for as long as it lasts.

Fusion sits between correlation and the unified state. It takes the candidates that correlation has tied together, combines what is compatible and returns the result to the core platform with the contribution of each source still readable.

Read the architecture

Data status: Illustrative

Evidence ladder

1. E0 · Concept reached
2. E1 not reached
3. E2 not reached
4. E3 not reached
5. E4 not reached
6. E5 not reached

Project assurance panel

Mission and context

Combine independent sources into one candidate state without losing track of where each part came from or where the sources disagree.

Primary contribution

Fusion that records each source's contribution, the provenance of each field and a lasting mark wherever sources conflict.

Decision value

Shows a reviewer whether a confident state is confident because sources agree or because one source dominates. The judgement stays with the reviewer.

Evidence status

Evidence level: E0 · Concept Concept and research direction only. No specification, prototype or test result is claimed.

Next validation step

Write the arbitration rules first, then check them in simulation on synthetic conflicts, including the rule that a conflict never raises confidence.

Behaviour when degraded

When a source drops, the fused state is recomputed from the rest, the band widens and the claim softens. When two sources conflict, the disagreement mark stays until one of them changes.

Role in integration

Receives correlated candidates from the core platform and returns fused states with their contribution records. It does not task any sensor.

Human review

A person reviews any fused state that leads to an action and can reopen the contribution of each source.

Public-safe limitation

The board on this page uses invented sources and invented values to show a rule. It is not a measurement of any system.

Open unknowns

Which arbitration rules suit real sensors, how often sources will truly conflict and how reviewers will want conflicts presented are not yet known.

When a source drops, the fused state is recomputed from the rest, the band widens and the claim softens. When two sources conflict, the disagreement mark stays until one of them changes.

• A low-altitude object seen by two sources that disagree

  Why it matters here

  One source places it low and slow, another reads it differently.

  What it contributes

  A state that keeps both readings and shows the disagreement between them.

  What it does not replace

  A decision about what the object is.

• A source that drops out during an event

  Why it matters here

  Losing a source is common and should not make the scene look clean.

  What it contributes

  A recomputed state with a wider band and a softened claim.

  What it does not replace

  Bringing the source back, which is a maintenance matter.

Receives correlated candidates from the core platform and returns fused states with their contribution records. It does not task any sensor.

Write the arbitration rules first, then check them in simulation on synthetic conflicts, including the rule that a conflict never raises confidence.

• Fusion can only combine what the sources supply. It cannot recover information that none of them observed.
• Arbitration rules at this stage are concepts. Real sensors may conflict in ways that the illustrated rules do not cover.
• A fused state is not a verified state. Combining sources reduces some uncertainty and leaves the rest visible.
• The values shown are invented to demonstrate a rule. They say nothing about the accuracy of any sensor.

• Airspace Intelligence Core PlatformEvidence level: E1

  The core platform holds the unified state that fusion contributes to.

• Passive RF Sensing NodeEvidence level: E1

  A passive RF node supplies timestamped observations that fusion can combine.

• Passive Coherent Location (PCL) RadarEvidence level: E1

  Passive coherent location supplies spatial observations whose confidence depends on geometry.

• Electro-Optical AI Detection & ClassificationEvidence level: E0

  Electro-optical candidates arrive with separate detection and classification confidence.

• Airspace Data & Intelligence APIEvidence level: E0

  The data interface carries the fused state out with each source's contribution intact.