Causal Evidence Graph
Links observations, track associations, confidence transformations, decisions and replay runs into an inspectable evidence lineage.
COUNTER-UAS · VALIDATION / FORENSICS LAYER
Turn system outcomes
into inspectable evidence.
PAMIR-CUAS is a vendor-neutral, post-test validation and incident-reconstruction layer designed to explain why a detection, classification or track outcome became wrong.
SAFETY BOUNDARY No target selection · no weapon control · no engagement decisions · no interceptor guidance
INCIDENT RECONSTRUCTION / ABSTRACTED VIEWDETERMINISTIC
RADARobservation
EO/IRobservation
RFstale evidence
FUSIONdecision
CAUSAL PATHupstream evidence → temporal anomaly → decision impact → offline replayTRACEABLE
01 / MISSION
Not another sensor.
Not another effector.
PAMIR-CUAS sits above heterogeneous sensor and C2 evidence as an independent validation layer. Its job is to reconstruct evidence lineage, expose timing and disagreement failures, compare outcomes with independent ground truth and test causal hypotheses through safe offline replay.
Public material intentionally abstracts proprietary algorithms, sensitive integration details and operational parameters.
02 / SYSTEM ARCHITECTURE
From observations to defensible reconstruction.
01NORMALIZED EVIDENCESensor / C2 observations
Ground truth
Timing + confidence
Ground truth
Timing + confidence
→
02CAUSAL EVIDENCE GRAPHProvenance
Association
Contribution lineage
Association
Contribution lineage
→
03TEMPORAL INTEGRITYClock offset
Stale evidence
Out-of-order events
Stale evidence
Out-of-order events
→
04COUNTERFACTUAL REPLAYOffline exclusion / correction
Outcome comparison
Causal test
Outcome comparison
Causal test
03 / CORE METHODS
Temporal Integrity
Surfaces clock offset, stale evidence and out-of-order observations that can contaminate downstream fusion outcomes.
Counterfactual Replay
Tests whether removing or correcting suspect evidence changes the reconstructed outcome — offline, deterministically and without engagement functions.
ARC Approach
Attribution · Reconstruction · Counterfactual. A disciplined path from anomaly to traceable evidence, reconstructed decision context and causal replay.
Public-facing framework description; implementation remains proprietary.04 / VALIDATED PUBLIC SCOPE
Claims follow evidence.
The current CUAS branch is explicitly marked as a research prototype. Public claims below are limited to behavior exercised by the repository's current automated tests and prototype specification.
✓Deterministic incident analysisPrototype test suite currently passes on the CUAS branch.
✓Stale-evidence causal caseSynthetic CUAS-001 reconstructs a false positive, detects stale RF evidence and verifies an outcome-changing exclusion replay.
✓Sensor-disagreement causal caseSynthetic CUAS-002 detects disagreement and verifies an outcome-changing exclusion replay.
✓Evidence graph contribution pathTested output includes contributing-evidence relationships.
05 / INTEGRATION VALUE
Designed for the questions that appear after the test.
Why did this false positive occur?
Which evidence materially changed the outcome?
Was the anomaly causal or merely correlated?
Did timing integrity alter the decision?
Can the result be reproduced and inspected?
TECHNICAL DIALOGUE