Asking the fleet what it is doing…
monad-knowledge Wi-Fi sensing lab · FIIT STU

Where it happens

The floor the fleet stands on.

The FIIT ground floor, as surveyed. 10 radios in 3 rooms — one transmits, the rest listen. Concrete is drawn heavy and drywall light, because that is the difference the signal feels.

Click a radio. These 10 have sealed 1,946 capture sessions between them, 3,110 node-hours and 6,715 million channel estimates. Every pin carries its own share of that, and what it is doing this minute.

Open reading roomSilent roomBack roommonad01 — Not reportingmonad01monad02 — Not reportingmonad02monad03 — Not reportingmonad03monad04 — Not reportingmonad04monad05 — Not reportingmonad05monad06 — Not reportingmonad06monad07 — Not reportingmonad07monad08 — Not reportingmonad08monad09 — Not reportingmonad09monad10 — Not reportingmonad10
Fig. 1 10 surveyed node positions across three rooms. Live state, layers, zoom and the inspector need JavaScript; what each node is doing and what it has recorded are in the tables below either way.

Node positions are carried over from the deployment survey and are being re-measured against this plan — read them as within a metre or so, not to the centimetre.

Right now

Every radio on this floor.

the full fleet table

Reading the fleet…

Since each was commissioned

What each radio has recorded.

One row per radio, summed from the sealed sessions the archive holds for it. A session appears here once it has sealed and been projected, so a capture running this minute is in the table above and not in this one.

Node Sessions Node-hours Estimates Useful yield Recording since Declared to
monad01 237 359 603,429,432 72.7% /172 2026-08-10 – 2026-09-10 EXP-007, EXP-009, EXP-011, EXP-C1, EXP-F1, EXP-F2, EXP-LIB-01, EXP-P1
monad02 207 346 683,899,207 94.5% /158 2026-07-27 – 2026-09-10 EXP-007, EXP-009, EXP-011, EXP-C1, EXP-F1, EXP-F2, EXP-LIB-01, EXP-P1, EXP-P2
monad03 171 314 693,739,515 97.1% /136 2026-08-11 – 2026-09-09 EXP-007, EXP-009, EXP-011, EXP-C1, EXP-F1, EXP-F2, EXP-LIB-01, EXP-P1
monad04 185 309 679,607,115 94.7% /150 2026-08-12 – 2026-09-10 EXP-007, EXP-009, EXP-011, EXP-C1, EXP-F1, EXP-F2, EXP-P1
monad05 180 313 707,219,937 94.0% /145 2026-08-12 – 2026-09-10 EXP-007, EXP-009, EXP-011, EXP-C1, EXP-F1, EXP-F2, EXP-P1
monad06 182 311 670,087,588 95.4% /150 2026-08-12 – 2026-09-10 EXP-007, EXP-009, EXP-011, EXP-C1, EXP-F1, EXP-F2, EXP-P1
monad07 188 319 709,199,101 96.9% /179 2026-08-18 – 2026-09-10 EXP-007, EXP-009, EXP-011, EXP-C1, EXP-F1, EXP-F2, EXP-LIB-01, EXP-P1
monad08 150 285 687,014,953 94.0% /139 2026-08-18 – 2026-09-10 EXP-007, EXP-011, EXP-C1, EXP-F1, EXP-F2, EXP-P1
monad09 282 248 583,910,854 94.6% /228 2026-08-19 – 2026-09-10 EXP-007, EXP-011, EXP-C1, EXP-F1, EXP-F2, EXP-P1
monad10 164 306 697,101,677 96.9% /158 2026-08-18 – 2026-09-10 EXP-007, EXP-011, EXP-C1, EXP-F1, EXP-F2, EXP-LIB-01, EXP-P1

Useful yield is the share of frames a radio received that produced a non-zero channel estimate. A low one means the traffic in the air was DSSS or CCK, which carries no usable estimate — not that the channel was quiet. 137 further sessions are withheld from every figure above: their own numbers disagree with each other, and pooling those into a total is the failure with no outward symptom. Ownership is declared on the experiment card, never inferred from a capture's profile name.

Booked

What this floor measures next.

the full schedule
StartsArmHoursRadiosWhy
lib-drift-weekend 6.0 monad02, monad03, monad04, monad05, monad06, monad07, monad08, monad09, monad10 Weekend channel drift — the same paced measurement on a building with a thinner crowd and a quieter network.
lib-ambient-weekend 6.0 monad01, monad02, monad03, monad04, monad05, monad06, monad07, monad08, monad09, monad10 Weekend ambient monitoring — the same passive capture, on the thinner weekend crowd.
lib-drift-weekend 5.3 monad02, monad03, monad04, monad05, monad06, monad07, monad08, monad09, monad10 Weekend channel drift — the same paced measurement on a building with a thinner crowd and a quieter network.
lib-empty-night 5.0 monad02, monad03, monad04, monad05, monad06, monad07, monad08, monad09, monad10 Empty-room reference — how the Wi-Fi channel drifts on its own, with nobody in the building.
lib-drift-weekend 6.0 monad02, monad03, monad04, monad05, monad06, monad07, monad08, monad09, monad10 Weekend channel drift — the same paced measurement on a building with a thinner crowd and a quieter network.
lib-ambient-weekend 6.0 monad01, monad02, monad03, monad04, monad05, monad06, monad07, monad08, monad09, monad10 Weekend ambient monitoring — the same passive capture, on the thinner weekend crowd.
lib-drift-weekend 5.3 monad02, monad03, monad04, monad05, monad06, monad07, monad08, monad09, monad10 Weekend channel drift — the same paced measurement on a building with a thinner crowd and a quieter network.
lib-empty-night 5.0 monad02, monad03, monad04, monad05, monad06, monad07, monad08, monad09, monad10 Empty-room reference — how the Wi-Fi channel drifts on its own, with nobody in the building.

Merging the plan is the whole act — the control host reads these committed files and arms the radios itself. Nothing on this page arms anything.

What this floor is evidence for

Experiments recorded here.

the register
CardQuestionStatus
EXP-007 The detectability of walking-speed motion in CSI collapses once the packet reception rate (PRR) falls below the Nyquist bound for that motion's Doppler shift; the empirical breakpoint in a downstream motion/occupancy-change metric matches the closed-form ceiling v_max = λ·PRR/4 (max Doppler f_D = 2… measured
EXP-009 Running the BLE reference scanner on a sensing node degrades that node's own 2.4 GHz CSI capture, because the Pi's Broadcom BLE controller and the AX210 are two uncoordinated 2.4 GHz radios centimetres apart with no coexistence interface between them. proposed
EXP-011 With a monitor-mode injector as the sounding source, the two bands differ measurably in delivered rate and inter-arrival regularity at the same commanded pace, same subcarrier basis and same geometry; and the injector itself delivers a quantitative Doppler axis (inter-arrival CV < 0.5) that hostapd… analysed
EXP-C1 IP-106 capture design — how many people, how long, and what to skip planned
EXP-F1 Counting unique BLE advertisements from the gamification mobile app, observed via [[_knowledge/hardware/esp32-c6|ESP32-C6]] listeners, matches a prescribed walk-task schedule and a spot-check manual count within ± 1 person across 90 %+ of 1-minute windows — establishing the BLE-anchored ground-trut… planned
EXP-F2 (a) A CSI crowd-counting model frozen at Day 0 measurably degrades over 30 days as the environment changes, and (b) short trigger-based BLE calibration campaigns (10–30 min, mobile-app-driven) restore the model to within 5 % of Day 0 accuracy. running
EXP-LIB-01 Over twelve hours in an unoccupied room the channel's own wander — measured as the two-chain CSI ratio, the only stable observable on this hardware — stays far enough below the amplitude perturbation a walking person causes that uncorrected drift is not what limits overnight occupancy sensing. analysed
EXP-P1 The lab inventory (10× ESP32-C5, 10× ESP32-C6, 5× ESP32-S3, 14× Pi 5 + M.2 HAT + AX210, mobile app) can be brought up as a single distributed sensing plane — synchronised to sub-100 ms across all nodes, end-to-end ground-truth-counting working via the mobile-app BLE adv chain, FeitCSI streams reach… planned
EXP-P2 A hostapd-based AX210 illuminator emitting paced non-HT OFDM broadcast gives the observer node a controlled, single-class CSI sounding source whose delivered rate matches the commanded pace. analysed

A card is here because it claimed sessions this floor's radios recorded. A card that merely writes about the library is not — the join runs on declared ownership, never on prose.

People on this floor

The recorded walks.

replay them

On 2026-08-28 one person walked this floor for 25 minutes while the radios listened for their phone. From 155,034 Bluetooth readings the fleet placed them 2.48 m from where they actually were. That gap is why this lab measures the channel instead.