Link survival across surveyed furniture arrangements
Current assessment (2026-09-15)
Session 01M14FXTAQ9QA13GCDCN5DEA65 ran 12 simulations across three surveyed arrangements and four paired device-jitter seeds. Link losses were 0/36, 3/36 and 5/36. Surviving-link open-transfer minus tables amplitude was −4.530 dB, 95% CI [−10.92, +0.06]. One real receiver outage matched the signature, but furniture causation and the real amplitude arm remain unresolved.
The original 'not run' statement and near-flat prediction are superseded by the session. The arrangements describe different scanned zones, not a chronological furniture sequence. Tables below the antenna plane do not rule out all reflected-path effects. The excluded catalogue arrangement used default heights, and lost-link sentinels must not be averaged as physical amplitudes.
Next evidence needed: Link 01M14FXTAQ9QA13GCDCN5DEA65 to the comparison and preserve its unresolved real-data criteria. Separate acquisition failure from propagation change before using link loss to trigger recalibration.
Basis: Thesis Narrative, the September 10 synthesis, and the September 15 register review. See the review record for evidence sources and outstanding checks.
Earlier protocol and dated results
The methods and observations below retain their original scope. Read their readiness statements and proposed next steps as historical. The current assessment above records what has changed; frozen thresholds have not been retrospectively revised.
Original hypothesis (before this review): The synthetic drift curve built on a toy corridor makes a checkable prediction about the FIIT library: because every scanned table (59–75 cm) sits below every node antenna (90 cm), furniture displacement cannot reach the losin deep-blockage regime that was the only arm to move occupancy error. The sim run on the real floor should therefore predict a NEAR-FLAT drift curve. If EXP-F2's measured month shows drift anyway, the cause is not the layout, and the recalibration argument must move from geometry to hardware.
The prediction the geometry already forces
The newest scan gives sixteen tables at measured heights 59–75 cm. Every node antenna
is at 90 cm. So no table in this room reaches the sensing plane, and the losin arm —
the only arm in csi-layout-drift whose occupancy error moved — is geometrically
unreachable by tables here.
The sim should therefore predict a near-flat drift curve on this floor. That is a falsifiable statement made in advance, and it is the opposite of the self-confirming pattern the 2026-06-24 rigor audit found across the sim corpus.
What a null means
If the simulated curve is flat and EXP-F2's measured month is flat, layout is not
the drift axis in this room and ble-periodic-calibration needs a different mechanism.
If the simulated curve is flat and the measurement drifts, the drift is not geometric
— the standing alternative is thermal, and fleet_ops joins node temperature to CSI on
(host, ts).
Either outcome is reportable. The outcome that would falsify the instrument rather than the hypothesis is a simulated curve that drifts steeply on furniture that cannot reach the antennas, which would say the scene staging is placing scatterers wrongly.
Explicit non-goals
- This does not answer RQ4 on its own. It pairs a prediction with a measurement.
- It does not use the impairment profile as validated. csi-impairment-fit runs first, or the sim's amplitude scale is a literature prior rather than this NIC's.