Domestic Network Analysis WEB EDITION

💡 The big idea in plain words

Wi-Fi signal fades with distance and walls. The model is a simple rule of thumb: every metre of air costs 3 dBm, every wall costs 7 dBm. Count the distance and the walls from any spot to the router and you get a prediction. The page then searches the house for the router position that covers the rooms best — and the subdivision slider decides how finely the heatmap samples the floor plan.

🎓 The Project

Domestic Network Analysis (SCON 2020, UPM) — a 5 GHz Wi-Fi scan of a home plus a simulation of router placement. The notebook models received signal strength with an empirical path-loss model with wall attenuation and searches for the room that maximizes coverage. This page reproduces the model and makes it interactive, with the real house rebuilt as a perfectly tiled plan plus three designed flats.

📡 Path-loss model

Predicted received power at distance \(d\) metres from the router, through \(W\) walls:

\[ P(d) = P_{\max} - 3\,d - 7\,W \]

i.e. 3 dB/m free-space-ish attenuation plus 7 dB per wall — the notebook's calibrated model (\(P_{\max} = -39.2\) dBm from the measurements). Each room's score is the mean predicted signal over its cells, and the optimal room is the maximizer of that score (verified: Pasillo-2, score 19.78 — matching the notebook bit-for-bit).

🔲 Discrete subdivision model

Instead of one value per room, the plan is rasterized into a uniform grid of cells (16–48 per side, adjustable). Every cell computes its own prediction — distance from the router to the cell centre plus the wall count between their zones — so the heatmap shows smooth signal gradients within rooms and sharp drops across walls. Walls are drawn on shared cell edges, so the floor plan tiles perfectly.

🔁 Repeater mode

A second source (repeater) is modelled identically; each cell takes the maximum of the two predictions, simulating the best-case coverage of a mesh.