Live status · Read the latest
monad-knowledge Wi-Fi sensing lab · FIIT STU
designed EXP-013

Illumination efficiency and wideband probe qualification

In plain words

Which illuminating packets give reliable CSI for the least radio airtime?

Why it matters
More transmitted bytes need not mean more useful channel measurements.
How it is done
Repeat controlled changes of frame size, bitrate, channel width and packet cadence on the same links.
Where it stands
Authored and held for deployment; no measurements or improved sensing performance claimed.

Asks, formallyAt fixed geometry and receiver tune, a small paced probe can deliver useful CSI at no worse than 95% of the 200-byte 6 Mb/s control's usable cadence while reducing nominal transmit airtime; VHT width and cadence gains must survive receiver-side PHY, timing and validity checks. These are engineering screening hypotheses, not established optima or occupancy claims.

StandingDesigned, not yet run

RoomPi 5 + AX210 fleet, fixed FIIT ground-floor geometry; monad01 ch1 reference, monad05 ch44 comparison; unattended instrument pilot

Setup

Fixed FIIT geometry. Monad01 supplies the existing ch1 reference to monad02/09. Monad05 supplies the ch44 comparison to monad03/04/06/07/08/10; at least five common receivers are required. Comparison receivers remain at 80 MHz with interval_us=0; BLE continues unchanged.

The protocol, rationale, statistical limits and deployment procedure are in ILLUMINATION-STUDY.md. The matrix manifest owns visit order and identity; the inventory owns radio parameters. This card owns evidence through requests and planned run IDs, not a broad profile claim.

Procedure

  • CR-102: 64/200/800/1400-byte frames at 100 Hz and 6 Mb/s.
  • CR-103: 6/12/24/54 Mb/s legacy OFDM, 200 bytes, 100 Hz.
  • CR-104: VHT20/40/80 at MCS0/NSS1, 100 Hz, 200 bytes; legacy controls bracket it.
  • CR-105: 25/100/250/500 Hz at VHT80, after VHT80 qualification.

Each request budgets three two-hour blocks, preferably on separate nights. Settings recur in reverse order; a thirty-minute closing control tests carry-over. Exclude the first thirty seconds of each visit. Endpoints are on-air PHY, source attribution, usable CSI/s, gap tails, misses and storage cost. Wrong-tune, wrong-PHY and unidentifiable-source visits remain recorded instrument failures, never silently relabelled treatments.

Results

Not measured. These are candidate capabilities, not new production defaults.

Limits

One geometry and fixed 80 MHz receivers. The ch1 cohort is a health/drift covariate, not a matched treatment control. Only VHT20/40/80 holds PHY family and transmit antenna selection fixed; legacy versus VHT changes both. Three blocks are a descriptive pilot, not statistical noninferiority evidence. Packets and receivers within a block are correlated. No occupancy accuracy, global optimum or improved motion recovery is established by this pilot.

Training-symbol rationale: zhang2022_5822 ↗ and bocus2022_ce7f ↗. Bandwidth/antenna/spectral-resolution distinction: cominelli2023_e6ee. The reviewed corpus establishes no optimum for this fleet.

Next steps

Deploy profiles and the runner while idle; validate and release the held requests after the existing authored horizon. CR-105 additionally requires receiver-verified VHT80 on this deployment. Inspect separate-night replication before consuming repeats. Claim both successful and failed captures by planned run ID, then audit. Size later confirmation from pilot variance and acquire independently labelled motion before promoting a sensing default.

Provenance

—
not recorded