Wheels4Wheels

A small pilot, a useful result

The Tallinn pilot started with 173 scooter rides. After confidence filtering and human review, 124 road sections without surface-quality data were added to OpenStreetMap, covering 13.2 km.

Those are modest numbers by fleet standards. Yet the result is already public and available to map developers. Useful coverage can start long before every street or vehicle is connected.

The useful part can become invisible

The pilot used a focused collection campaign. At scale, the same work does not need a new survey. Rental scooters already move through cities and carry the sensors needed to understand the ground beneath them. Their routine rides can reveal where a surface is smooth, rough, or painful for a wheelchair user to cross.

Wheels4Wheels turns that existing stream into surface-quality information. High-confidence results can be reviewed and added to OpenStreetMap, where navigation apps can use them. Collection can happen in the background during trips that were going to happen anyway.

Built for an agent to pick up

The public tooling repository includes an agent skill, sample data, and a tested CLI. An agent can inspect the required telemetry shape, adapt a fleet export, validate whether the signals are usable, and run the analysis without reverse-engineering the project first.

1. A scooter fleet accumulates accelerometer and location data during routine rides.
2. The operator gives an agent the Wheels4Wheels tooling repository and a fleet export.
3. The agent reads the supplied skill and identifies the required fields and safeguards.
4. It transforms the export and checks whether the telemetry is usable.
5. The CLI analyzes the rides and prepares OpenStreetMap review candidates, maps, CSVs, and GeoJSON.
6. A researcher reviews every candidate, then uploads manually or applies for an automated process under OpenStreetMap rules.
7. After approval, the analysis can run inside fleet or municipal infrastructure so useful map coverage keeps growing.

One signal among many

Rental scooters are one example. Buses, street cleaners, delivery fleets, and public-service vehicles already cross the same streets every day. With the right privacy, accuracy, and human review safeguards, their existing signals can help cities notice problems and improve infrastructure without launching a new manual survey for every question.

Our role

1ar labs handled data analysis, engineering, visualization, and web development. Plan.Net and Serviceplan Group created the web design. Bolt provided the fleet platform and data samples. The method, tools, and project documentation are open-source.