A twelve-week track on locating people and devices indoors from radio signals — RSSI, ranging, fingerprinting, and filtering — built around the lab's own thesis direction: calibration as a periodic campaign, not a one-time fit. The throughline: every position estimate comes with an honest error distribution.
The 12-week arc
2 of 12 lectures published — each a self-contained read with a slide deck and runnable workbook. Start at the first published week and work forward.
| Wk | Theme & focus | Read |
|---|---|---|
| 01 | RF ranging fundamentals start here RSSI, ToF, AoA — the three ways a radio measures distance/angle, and what each costs | read → slides · workbook |
| 02 | Log-normal path loss The model everyone uses, its parameters, and why it drifts indoors | read → slides · workbook |
| 03 | Trilateration & DOP From ranges to a position fix; the geometry that makes or breaks accuracy | planned |
| 04 | Fingerprinting k-NN / probabilistic / learned radio maps, and the survey cost behind them | planned |
| 05 | Filtering I: smoothing & outlier rejection Taming RSSI noise before any tracking | planned |
| 06 | Filtering II: the Kalman filter State-space tracking — the workhorse, derived and implemented | planned |
| 07 | Particle filters & map constraints Non-Gaussian tracking constrained by the floorplan | planned |
| 08 | BLE specifics & beacon density Advertising, channels, and how many beacons a room really needs | planned |
| 09 | Periodic BLE calibration campaigns Keeping a radio map alive via scheduled recalibration (thesis core). Quote the POWERED numbers, not the pilot: cross-modal disagreement tracks error at ρ = 0.68 (pilot 0.93 did not replicate; TOST vs oracle FAILED), CSI-only self-monitoring false-fires on 17 % of error-irrelevant shifts, and the trigger's honest payoff is calendar-parity at ~¼ the calibration spend — condition-based maintenance, never "event-driven beats fixed" | planned |
| 10 | Sensor fusion (BLE + CSI + IMU) Combining modalities as one probabilistic estimate | planned |
| 11 | Ground truth & error CDFs Why the error CDF beats RMSE, and how to collect honest ground truth | planned |
| 12 | Capstone A calibrated multi-room BLE localisation pipeline, end-to-end | planned |
Capstone
Week 12: build a multi-room BLE localisation pipeline with a documented calibration procedure, and report its accuracy as an error CDF across rooms — including where and why it fails.