Hodos maps potholes from the bumps your Android phone already feels

The pothole supplies the input. Your phone supplies the complaint. Hodos takes the jolt that usually produces a small personal monologue about road maintenance and turns it into a candidate mark on a shared map.

The project’s README describes accelerometer and gyroscope readings collected during normal trips. Android activity recognition arms sampling during vehicular travel. Nobody needs to play pothole stenographer while driving (please don’t).

A single bump doesn’t earn a permanent accusation against the pavement. Independent trips have to corroborate a defect. That distinction matters because a phone feels everything between the road and its mounting position, including the vehicle’s suspension. A map that believed every bounce would quickly develop a very personal relationship with your cupholder.

The maker explicitly warns that results vary with the phone, suspension, mounting and road conditions. That makes this an interesting neighborhood experiment rather than a reason to stare at the screen and assume an unmarked lane is smooth. The pavement did not issue a warranty, and an empty map could simply mean nobody has contributed observations there yet.

Hodos also distinguishes the fleeting sensor stream from what gets shared. Its README says raw motion stays in RAM and continuous location trails aren’t logged. Compact hazard observations go through server aggregation. That is a narrower, useful claim; it doesn’t mean a crowdsourced location service somehow operates without locations.

For a reader considering a trial, the important comparison is with a manual report. A manual report can describe the broken asphalt; this approach can collect repeat observations without asking a driver to pull over and fill in a tiny civic questionnaire. Confirmation trades speed for confidence. How well that trade works on your own route remains something to establish, not something the word automatic settles.

Offline support is similarly specific. The app caches a local hazard database and a Wi-Fi-downloaded map area with a 20 km radius. That can keep a familiar area visible when reception disappears; it cannot conjure fresh community observations from a disconnected phone. Coverage and connectivity are different problems, even when both leave you looking at nothing.

The licensing deserves a stop before any fleet experiment. The application code is MIT-licensed, while the road dataset has noncommercial terms and a separate commercial licensing route. A company can like the software and still need permission for the data it wants to route vehicles around. The code can be free while the dataset still needs a commercial agreement for the routing a company intends to do.

If repeated trips produce a useful neighborhood map, the appeal is wonderfully ordinary: an existing phone helps identify an existing nuisance. The unresolved part is whether enough independent trips pass your particular broken stretch to make its dots worth trusting.

Road rage, now with coordinates.