Campaign session

01KVWN98XPHR2KVCTCXVDG1KEM

finished 2026-06-24 11:12:14.262397+00:00 → 2026-06-24 11:20:50.379667+00:00 · 1 run · supervisor: react-agent

“Blocked: furniture-scatterer injection is not in the deployed sionna-csi-runner image, so no furnished/displaced states can be staged; base chain validated, drift curve + T*(eps) unresolved pending IP-101 Phase 2.”

Archive snapshot, as of 9 h ago — the run corpus is rebuilt once a day, so this page is not a live reading. The fleet panel is the live one; it refreshes every 30 s.

Success criteria

Criterion	Resolved
Stage >=4 progressively-displaced furnished floor states and run exp-csi-static on each (scalar floor + links.parquet, no NaN/Inf).	no
Notebook emits drift_curve.parquet carrying both analytic estimators + relative schema.	no
Occupancy error grows monotonically with delta_layout; curve yields a T*(epsilon) cadence.	no
Synthesis records the curve as a campaign-design input that does NOT answer RQ4.	no

Synthesis

Verdict

Blocked on a simulator capability, not the science. The layout-drift study requires staging an ordered sequence of furnished floor states with progressively displaced furniture injected into each run's scene.json. The deployed sionna-csi-runner image (digest sha256:14a6e852…, source SHA cd1b428) does not support furniture scatterers, and the configured floor test-lab-synth-floor-0 carries no occupiables for scenegeom to emit. No (T*, ε) drift curve could be produced in silico in this session.

What we tested

One feasibility probe was run and gate-passed: exp-csi-static, test-lab-synth-floor-0 / csi-synth-corridor-link, n_agents=6, n_placements=4, phase=layout-drift, seed 0 (run_id 01KVWNC2HZ306KQH6CG3NEBXNB, 88.5 s wall-clock, 7 artefacts, no NaN/Inf). It confirms the base static CSI chain runs end-to-end locally on this floor: domain_metrics report mean_atten_db_per_person=2.18, k_vs_occupancy_slope=-1.32 dB/person, variance_monotonicity=-1.0 over 3 links × 52 subcarriers — the qualitative occupancy invariants hold.

Why it is blocked

The generated inputs/scene.json carries only walls, materials, body_material, transmitters, and receivers — there is no furniture / obstacles field. This matches the curated _knowledge/simulators/sionna-csi-runner note's stated known limitation ("Obstacles … are not yet inserted as scatterers — only walls + bodies"). The phase: layout-drift parameter was accepted by the config resolver but is inert: it injects no furniture and the scene is identical to a plain static run. The IP-101 _write_box_ply furniture-insertion path is therefore not present in the deployed digest. Per the brief's operator note, the study is gated on IP-092 furnished floors / occupiables landing in the runner; that gate is not yet met.

Criteria status

Unresolved — no furnished/displaced states could be staged (furniture unsupported).
Unresolved — csi_layout_drift.py reduction not run (no states to reduce).
Unresolved — no drift curve, hence no T*(ε) cadence.
Unresolved (honestly recorded) — there is no curve to record; this synthesis instead documents the platform dependency that blocks it.

What unblocks it

Wire furniture scatterers into sionna-csi-runner (IP-101 Phase 2: _write_box_ply + a furniture field in scene.json, with scenegeom emitting occupiables for a furnished floor such as a fiit/ResPlan layout, and a displacement-schedule param). Re-run this campaign once that lands. The base chain and the probe metrics confirm everything upstream of furniture is ready. recalibration-trigger-from-drift stays plausible; no in-silico evidence was added this session.

Attached runs

Run	Gate	Purpose	Replay
G3NEBXNB	✓	—	—