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

Recorded walks

2026-08-28 · 25 minutes on the fiit floor.

Press play. One person walked 302 m while 9 radios listened for their phone. From 155,034 Bluetooth readings the fleet placed them 2.48 m from where they actually were.

Fig. 1 The recorded path in ink, every second radio fix in cyan. Replaying it with a clock, and the distance between the two as it runs, needs JavaScript. The numbers below are the same reading either way.

The recorded path is the phone's own tracking, fitted onto this plan on the surveyed radios the walker stopped at. It is a fit with a measured residual, not ground truth, and the residual is printed with every walk.

This walk

What it measured, and against what.

every recorded walk
Recorded2026-08-28, 25 minutes, 302 m walked
Path fitted onto the plan on 9 surveyed radios the walker stopped at, off by 0.26 m on them (RMS)
Radios that heard the phone9, over 155,034 Bluetooth readings
Positions computed758, one every 2.0 seconds
Typical miss2.48 m — 9 in 10 within 7.96 m
Without a propagation model4.47 m — a signal-weighted average of the radios that heard it
How this room carries signal falls off as d1.215, shadowing 2.86 dB, fitted on 5,468 readings from this session

The propagation model was fitted on this session and scored on it, so these metres are the best this room allows and not what the method would do on a floor it had never seen. Read the gap, not the decimal: a few metres is what beacons give you, and it is why this project measures what the room does to a Wi-Fi signal instead of how loud a beacon sounds.