OSM Integration

Overview

Procedures for responsibly integrating accessibility data into OSM, combining automated and manual validation with best practices for contributors and cities. Coverage and accuracy metrics track the quality of contributions.

Core Principle

The system never overrides existing OSM tags. Its design focuses on identifying gaps in wheelchair accessibility data while respecting the work of current contributors. Gap-filling and overriding are treated as fundamentally different actions, and everything downstream is built around that distinction.

Gap Detection Strategy

Before filling gaps in accessibility data, the system first identifies which segments already carry OSM tags, so only missing or outdated information is targeted. The check runs before any proposal is generated, so existing contributor work is recognized up front rather than filtered out late.

def check_existing_tags(way_id, osm_api):
    """Check if way already has smoothness or wheelchair tags."""
    tags = fetch_way_tags(way_id)

    existing_tags = {
        'smoothness': tags.get('smoothness'),
        'wheelchair': tags.get('wheelchair'),
        'surface': tags.get('surface')
    }

    return any(existing_tags.values())

When the check finds existing values, the segment is not simply skipped. Each way is sorted into one of four gap types, which determines whether the system may propose a tag, flag data for review, or leave the segment untouched.

Proposal Generation Rules

After gaps are identified, proposals are generated following clear rules that determine when and how new accessibility tags should be applied.

def generate_proposal(way_data, existing_tags):
    """Generate proposal only for gaps, respecting existing data."""

    if existing_tags.get('smoothness'):
        return None  # Never override existing smoothness

    if existing_tags.get('wheelchair'):
        return None  # Respect existing wheelchair assessment

    # Only propose for genuine gaps
    return create_accessibility_proposal(way_data)

A proposal that clears the gap check still has to earn its way out of the pipeline. Three quality gates filter what reaches reviewers, so only measurements with enough confidence, coverage, and internal agreement turn into proposals.

Automatic multi-ride consensus is future work. A 75% consensus threshold remains a proposed target rather than an implemented trigger, so every proposal in the current workflow is reviewed and applied per ride rather than by automated agreement.

Manual Review Process

Automation only prepares the ground. Before anything reaches OpenStreetMap, each proposal is reviewed, and the process differs depending on whether it fills a gap or challenges a tag that already exists. Nothing in the workflow currently bypasses this step. Once the method is validated, these measures can be gradually relaxed.

Gap-Filling Proposals

Contributors follow a clear sequence that keeps updates consistent and verifiable, from confirming the affected road segment to documenting the evidence behind every edit before it is submitted to OpenStreetMap.

Override Decisions

Overrides are proposals that contradict a tag another mapper has already set. Even at high confidence they require careful scrutiny, so the evidence bar is deliberately higher than for filling an empty segment.

Following that sequence is easier with the right tooling. The project ships several aids that make review faster and more reliable, each aimed at a different part of the checklist – seeing proposals in context, verifying them visually, and tracking how segments change.

Quality Assurance

Quality assurance is proposed to combine automated checks and hands-on review to ensure the dataset is reliable and precise. Automated processes can flag potential issues, while manual validation can verify high-impact proposals. These two layers can catch different failure modes. Once the method is validated, these measures can be gradually relaxed.

Best Practices

These practices keep contributed data accurate, consistent, and useful. They outline steps for contributors to follow before moving on to city or organization-level workflows.

Success Metrics

Success metrics track improvements in coverage, data quality, and community adoption; the broader community benefits are assessed separately below.