rail-coverage-pl

How good is mobile internet on Polish trains, kilometre by kilometre?

This project measures it from the passenger seat: a laptop on the phone's hotspot sends 10 pings every 30 seconds, and each measurement is placed on the route from the phone's GPS track or, without GPS, from the train timetable. Base stations near the line come from the official UKE radio permit lists, so you can see whether the dead zones line up with gaps in the network.

Map: https://kprzybys.github.io/rail-coverage-pl/ · First trip report (PL): IC 141 Białystok → Kraków

Who is this for

• Passengers who work on the train. Know where the 2-second lag and the dead zones are before you schedule a call or start a big upload.
• People comparing operators. Each trip records the network; with more trips the same route can be compared across Orange, T-Mobile, Plus and Play.
• Rail operators and mobile networks. A public, reproducible record of where coverage along the busiest lines fails, with the nearest permitted base station for each gap.
• Journalists and researchers. Raw CSV data, a documented method, and code that rebuilds every chart.

How it works

Step	Tool	Output
Log ping every 30 s	railcov/logger.sh	ping.csv
Describe the trip (train, timetable, delays, network)	edit trip.json	trip.json, stations.json
Import the phone's GPS track, clipped station to station	python -m railcov.gpx	track.gpx
Base stations within 8 km of the route (UKE)	python -m railcov.bts	bts.json
Charts and report	python -m railcov.report	figures/*.png, report.md
Route strip and replay GIF for sharing	python -m railcov.promo	figures/strip.png, figures/replay.gif
Website: replay, chart, strip, longest gaps	python -m railcov.site	site/ (GitHub Pages)
All of the above after a trip	./finish.sh trips/<trip-id> [recording.gpx]

pip install -r requirements.txt
railcov/logger.sh trips/<trip-id>/ping.csv          # during the trip; Ctrl+C at the end
python -m railcov.gpx    trips/<trip-id> ~/Downloads/recording.gpx
python -m railcov.bts    trips/<trip-id>
python -m railcov.report trips/<trip-id>
python -m railcov.site

Quality grades (thresholds for interactive use such as SSH and calls):

Grade	Rule
good	no loss, average < 100 ms, max ≤ 200 ms
weak	some loss, or average ≥ 100 ms, or a spike over 200 ms
bad	loss ≥ 30% or average ≥ 300 ms
dead	100% loss

Details and limits: docs/methodology.md · data format: docs/data-format.md.

Contribute a trip

The most useful contribution is another trip: a different line, a different operator, the same line at a different time. See CONTRIBUTING.md. Android phones can also log signal strength and cell IDs, which would turn "where is it bad" into "why".

License

Code: MIT. Measurements: CC BY 4.0. Station coordinates: © OpenStreetMap contributors (ODbL). Base station locations: UKE public register. See LICENSE and DATA-LICENSE.md.

---

Po polsku

Jak działa mobilny internet w polskich pociągach, kilometr po kilometrze?

Laptop podłączony do hotspotu w telefonie co 30 sekund wysyła 10 pingów. Każdy pomiar trafia na trasę według śladu GPS z telefonu albo, bez GPS, według rozkładu jazdy. Do tego dochodzą stacje bazowe z wykazu pozwoleń UKE, więc widać, czy martwe strefy pokrywają się z dziurami w sieci.

Pierwszy przejazd: IC 141 Hańcza, Białystok → Kraków Główny, 25.10.2026, sieć Orange, raport.

Komu to się przyda: pasażerom pracującym w pociągu (gdzie nie planować rozmowy), osobom wybierającym operatora, przewoźnikom i operatorom komórkowym (publiczna mapa problemów na najważniejszych liniach) oraz dziennikarzom i badaczom (surowe dane i kod).

Jak pomóc: nagraj własny przejazd, inną linią albo w innej sieci, i wyślij pull request. Instrukcja jest w CONTRIBUTING.md. Szczególnie cenne są pomiary z telefonów z Androidem, które zapisują siłę sygnału i identyfikator komórki.