Through-wall, sparse-LOS CSI — locating the blockage-dominant regime

Question

When the link is wall-obstructed (Tx and Rx in different rooms), does mean attenuation (L4.6) become the leading occupancy feature — the blockage-dominant regime the in-room study never reached?

csi-static-occlusion found variance dominant across the whole range (crossover N≈0) because its corridor LOS was rarely blocked. Depatla's school works in exactly the opposite geometry — sparse-LOS, through-wall — where blockage dips dominate. This campaign places that geometry on a real apartment and tests whether the crossover moves to N*>0.

What we already know

• The scene bridge resolves each wall to an ITU-R P.2040 RadioMaterial, so through-wall attenuation is physical (not a fudge); max_depth ≥ 4 lets enough reflections reach a shadowed Rx.
• resplan-12439-floor-0 has well-separated rooms (bedroom-0/2/3) for graded wall counts.
• The per-body loss is now wired (occluded-link excess attenuation), so blockage on a shadowed link compounds the wall loss — the regime where the scalar should matter most.

Prerequisites

1. Multi-room link layout csi-link-resplan-12439-multiroom: Tx in bedroom-0; Rx in bedroom-0 (in-room control), bedroom-2 (1 wall), bedroom-3 (≥2 walls). Place via gis tools.
2. Scene-stage check (expect many walls; confirm the three Rx resolve).

What the supervisor does

One sweep (8 occupancy × 2 bands = 16 runs); the analysis splits links.parquet by link class (wall count between that link's Tx and Rx) and computes per-class mean-attenuation slope + crossover. Analysis-writer reports the N=0 wall-count path-loss ladder and the per-class crossover.

Figure render request

csi_through_wall — mean |H| & CV vs occupancy, one panel per link class; plus a floor map coloured by per-link attenuation showing the shadowed rooms.

Out of scope

• Cross-floor generalization (c-csi-cross-geometry-resplan); temporal motion (c-csi-crowd-temporal); real-data calibration.

Expected interpretation

1. Blockage regime found (through-wall N*>0, steeper mean slope) → "the L4.5↔L4.6 crossover is a geometry continuum: variance-dominant in-room (csi-static-occlusion) → blockage-dominant through-wall (here). The two schools are reconciled by link geometry." A direct chapter result.
2. Still variance-dominant → "even through-wall, diffuse multipath dominates at these occupancies in residential geometry" — a null worth stating (bounds Depatla's regime).
3. Criterion 1/2 fails → ITU material / scene-bridge finding.

Notes

Follow-on: a Tx-outside / Rx-inside variant (true through-wall sensing from a corridor) once a multi-floor or exterior-anchored layout exists.