Abstract

How to translate e-scooter telemetry into wheelchair accessibility data.

The Core Concept

Each e-scooter ride generates telemetry data that may reveal surface quality, slope, and the existence of sidewalks and ramps. The Wheels4Wheels project explores whether analyzing accelerometer readings from regular rides can be used to estimate wheelchair accessibility at the street level and upload this information to OpenStreetMap. The following modules summarize the data sources, methodology, and workflow used throughout the project.

Finding the Data Source

E-scooters generate high-frequency telemetry data (30 Hz accelerometer readings, GPS coordinates, and speed), which can be repurposed to measure surface conditions. In particular, the Z-axis accelerometer data is processed to estimate surface roughness, slopes, and inclines – a quantitative signal for wheelchair accessibility.

Project Workflow
Raw E-Scooter Data Processing
Initial cleaning and preparation of e-scooter telemetry
Surface Roughness Analysis
Calculation of Z-axis standard deviation to assess pavement smoothness
Confidence Scoring
Assign reliability scores based on speed, coverage, and data quality
OSM-Ready Proposals
Generate tags and structured outputs compatible with OpenStreetMap

The Process in Detail

A step-by-step breakdown of the methodology, showing how e-scooter telemetry is transformed into actionable accessibility data. Each of the four steps feeds the next, ending in proposals that are ready for review.

1. Data Collection

Telemetry data is collected from e-scooter rides. These measurements provide the foundation for assessing surface conditions. The table below summarizes the data types and their purposes, where only data relevant to accessibility analysis is extracted.

Data Sources
Data SourceDescriptionWhat It’s Used For
30Hz Accelerometer DataX, Y, Z axes, sampled at 30 times per secondSudden spikes in the Z-axis values indicate a bumpy surface
GPS CoordinatesCollected every second (latitude, longitude), with timestampsMapping location of surface measurements and proposals
Speed ReadingsRaw telemetry in 1/10 km/h (converted by dividing by 10)Ensuring speed is within the wheelchair assessment range and supporting data filtering

2. Surface Analysis

Z-axis acceleration data is analyzed over 1-second intervals, smoothing out transient noise and producing a single value per second. This value correlates with the roughness of the surface, providing a standardized measure for wheelchair accessibility. The table below translates these standard deviation ranges into descriptive surface categories.

Surface Thresholds
ClassificationZ-axis Standard Deviation (mg)Description
Excellent0–60Smooth asphalt or concrete
Good60–110Minor imperfections
Intermediate110–180Noticeable bumps and cracks
Bad180–250Rough surface, significant obstacles
Very Bad>250Extremely rough, impassable

3. Wheelchair Calibration

To confirm that surface roughness measurements align with real-world conditions, the thresholds are systematically validated against existing OpenStreetMap tags. Streets that are already well mapped serve as the reference for calibrating the classification.

Calibration Steps
Overlapping Segments: Identify rides on roads already mapped in OSM
Z-Value Comparison: Measured values vs existing smoothness tags
Threshold Adjustment: Calibrate to match OSM’s classification system

4. Confidence Scoring

Each measurement receives a wheelchair-focused confidence score between 0 and 100%. The score expresses how much trust a reading deserves before it becomes a tagging proposal, and it is computed from three metrics.

Confidence Metrics
MetricDescription
Classification ClarityShows how confidently a reading fits within calibrated thresholds
Speed AppropriatenessChecks whether measurements fall within the 2–6 km/h assessment range
Movement ContinuityPenalizes readings taken during stops to maintain data reliability

Key Insight

Even a single ride can provide valuable data if its coverage and point agreement are high enough. The proposal generator can surface a way-level candidate when more than 30% of the segment is sampled. Applying that proposal with the current tooling requires at least 70% coverage and 70% point agreement.

Benefiting Parties

The Wheels4Wheels dataset is useful to more groups than wheelchair users. The following sections outline what each group can take from the project’s findings.

Wheelchair Users
FocusBenefit
Improved NavigationGet more accurate and up-to-date wheelchair accessibility information
SafetyAvoid dangerous roads and areas
InclusionFeeling of support from the community
OSM Contributors
FocusBenefit
AccuracyContribute to more accurate and up-to-date wheelchair accessibility data
InclusionContribute to a more inclusive and accessible world
SustainabilityHelp keep accessibility data maintained over time
Cities
FocusBenefit
Identify Problem AreasSystematic mapping of accessibility barriers
Track ImprovementsMonitor surface quality changes over time
Data-driven PlanningPrioritize infrastructure investments based on actual usage

Output

The exploration wraps up into a set of concrete artifacts that let others pick up the work. Everything is designed to be handed over – documented methods, working scoring code, reviewable maps, and proposals formatted for direct OSM contribution.

Deliverables
Documentation: Technical documentation ready for implementation
Confidence Score: Confidence scoring algorithms to adopt or improve
Interactive Maps: Visual review of data
CSV Proposals: CSV files with OSM-ready proposals

Behind the deliverables sits one reproducible pipeline. The stages below show where each artifact comes from, from telemetry ingestion through classification and scoring to the proposals that are finally exported.

Technical Workflow
StageDescription
InputAnonymized e-scooter telemetry in JSON format
ProcessingPython pipeline and OpenStreetMap CLI for Wheels4Wheels surface quality proposals
OutputCSV, GeoJSON, and HTML maps ready for review and integration
IntegrationCompatible with standard OpenStreetMap editing workflows
VisualizationsInteractive charts illustrating key analysis and metrics

Anyone Can Help

Anyone can contribute to improving accessibility maps, from individuals reviewing a single ride to organizations processing city-scale datasets. The two paths differ mainly in volume, not in kind, and both rely on the same tooling and the same review standards.

Individual Contributors

Contributors can analyze their own ride data, where even a single high-quality ride improves local accessibility mapping. They can review proposed accessibility tags through CSV exports and interactive maps, and focus on targeted contributions to fill gaps in wheelchair-accessible routes.

Organizations

Organizations can process large volumes of telemetry data and reach city-wide coverage. The confidence-scoring system helps prioritize high-impact contributions, while outputs are formatted for OpenStreetMap and other accessibility-focused platforms.