Surface Analysis

How to quantify street surface conditions from sensor data.

Overview

E-scooter telemetry is analyzed to estimate surface quality and map the results to OpenStreetMap road segments. The pipeline produces georeferenced, confidence-scored proposals for review. Once added to OpenStreetMap, these smoothness tags can support wheelchair-routing engines that use OSM accessibility data.

Classification Methodology

Surface quality is assessed using the Z-axis acceleration standard deviation, which strongly correlates with the roughness experienced by wheelchair users. The workflow below anchors that raw signal to OSM's established smoothness scale before any threshold is trusted.

Calibration Workflow
Find Overlapping Segments
Identify rides that traveled roads with existing OSM smoothness=* tags
Compare Z-Values
Analyze z_std distributions for segments with known classifications
Adjust Thresholds
Calibrate our classification system to match established OSM standards
Validate Accuracy
Ensure our proposals align with ground truth data
Classification 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

Surface Classification Timeline

The visualization below depicts how these classes unfold along a ride. Each point is color-coded by smoothness class, with a confidence overlay indicating the certainty of wheelchair accessibility for each segment.

Loading chart data...

Single Ride Capability

Even a single high-quality ride can provide valuable wheelchair accessibility data. What counts as high quality is defined by a small set of requirements covering sampling coverage, riding speed, and sensor integrity.

Ride Requirements
RequirementDescription
Proposal GenerationMore than 30% of the segment must be sampled to surface a candidate for review or map display
Optimal SpeedRide speed within the 2–6 km/h assessment range
Data QualityStable GPS accuracy and complete sensor readings
Apply FiltersAt least 70% coverage and 70% point agreement

Large Dataset Aggregation

A multi-ride workflow can aggregate rides on the same street segments to reduce the influence of anomalies, building on the same classification and confidence scoring that single rides already use. The parameters below define how such an aggregation weighs and combines rides.

Proposed Algorithm
MetricDescription
Spatial ClusteringGroup rides within a 10 m tolerance (allows for GPS error)
Temporal WeightingRecent rides receive higher weight
Consensus ScoringProposed minimum of 3+ rides for high-confidence proposals
Coverage CalculationProposed minimum of more than 30% segment coverage
Agreement ThresholdProposed 70% agreement threshold; 75% proposed for automatic handling
Quality Indicators
IndicatorDescription
Ride CountMore rides increase confidence in results
Coverage PercentageProportion of the segment that was sampled
Temporal SpreadData collected from different times and days
Speed ConsistencyMeasurements taken under similar conditions

Validation Results

Validation assesses how well single-ride sensor-derived data matches real-world conditions, using streets that already carry OSM smoothness tags as the reference. Multi-ride aggregation is a natural extension of the same approach and is described separately below.

Single Ride Performance

For single rides, the measurements showed 78% agreement with existing OSM tags, effective for segments longer than 50 m, and a false positive rate below 15% for high-confidence proposals. Agreement was measured against segments that already carry smoothness tags in OSM.

Multi-Ride Enhancement

Multi-ride aggregation can improve consistency, expand coverage systematically across the city, and support monitoring of surface quality changes over time. As more rides accumulate on the same segments, individual anomalies matter less and longer-term trends become visible.

Practical Application

The methodology tells contributors where collected data helps most: use it to improve local accessibility mapping in the areas where it is most needed. The guidance differs by audience, but the underlying data and thresholds are the same.

For Contributors

Individual contributors can start with small rides to fill gaps, focusing on areas lacking surface quality data. Each ride contributes georeferenced measurements that can be aggregated with other rides to strengthen accessibility proposals.

For Cities

Cities benefit from systematic mapping, monitoring improvements, and prioritizing interventions where accessibility barriers are most critical. Aggregating e-scooter and manual data can help urban planners identify high-impact areas and track changes in surface quality over time.