Skip to content

WEP 1 Routing Service

Core routing engine selection and architecture for Alpine accessibility analysis.

Requirements

  • Support for transit data (GTFS) with time-dependent routing
  • Strong support for outdoor profiles (hiking, biking, ski touring)
  • Ability to integrate external data sources (e.g. Swisstopo hiking trails, ski routes)
  • Custom routing profiles and weighting (e.g. prefer ski routes, hiking difficulty)
  • Low memory consumption during graph build and runtime (< 4 GB target)
  • Isochrone generation, including transit isochrones
  • Matrix calculations (one-to-many, many-to-many)

Options

Valhalla

Overview Valhalla is a tile-based routing engine developed by Mapbox, with strong native support for multimodal and public transport routing.

Advantages

  • First-class GTFS transit routing and transit isochrones
  • Very efficient runtime performance for transit queries
  • Tile-based architecture allows partial loading at runtime
  • Good default support for car, bike, and pedestrian routing

Limitations

  • GTFS data is baked into transit tiles and must be rebuilt on update
  • Frequent GTFS updates are operationally expensive
  • Custom outdoor profiles (e.g. ski touring) require C++ costing model changes
  • Integration of external vector data is complex

Link https://github.com/valhalla/valhalla

GraphHopper

Overview GraphHopper is a flexible Java-based routing engine with strong support for custom profiles, external data, and precomputation workflows.

Advantages

  • Excellent support for custom profiles via Custom Models (YAML)

  • Clear separation between:

  • Hard profiles (encoded values, schema, external data – require rebuild)

  • Soft profiles (weighting and preferences – no rebuild required)
  • Can integrate external vector data (e.g. Swisstopo hiking and ski routes)
  • Supports isochrones and matrix routing (one-to-many, many-to-many)
  • GTFS transit data is preprocessed offline and loaded at startup
  • Allows frequent GTFS updates without rebuilding the road graph
  • Lower memory footprint than Valhalla for non-transit-heavy use cases
  • Well suited for static precomputation (hut isochrones, reachability)

Limitations

  • Transit isochrones are less mature and slower than Valhalla
  • GTFS updates still require preprocessing
  • New tags or schema changes require a graph rebuild

Link https://github.com/graphhopper/graphhopper

BRouter

Overview BRouter is a lightweight routing engine focused on biking and hiking, widely used in outdoor navigation tools.

Advantages

  • Very low memory usage
  • Highly flexible programmable profiles
  • Excellent routing quality for hiking and biking
  • Fast for single-route calculations

Limitations

  • No native support for transit
  • No isochrones or matrix calculations
  • Limited API and server-side capabilities

Link https://github.com/abrensch/brouter

Other Options (Brief)

pgRouting

  • PostgreSQL-based routing extension
  • Useful for custom graph experiments
  • Not suitable for large-scale or transit routing

Link: https://pgrouting.org/

OpenTripPlanner (OTP)

  • Strong public transport routing
  • Heavy memory requirements
  • Less flexible for custom outdoor profiles

Link: https://www.opentripplanner.org/

Preliminary Conclusion

  • GraphHopper is the preferred primary routing engine due to:

  • Flexible handling of outdoor and Alpine-specific profiles

  • Practical integration of external data sources (Swisstopo)
  • Clear distinction between build-time schema and runtime weighting
  • Easier and more frequent GTFS updates compared to Valhalla
  • BRouter can be optionally combined for high-quality hiking and biking time estimation.
  • Valhalla remains a strong alternative if public transport routing and transit isochrones become the dominant requirement, but at the cost of reduced flexibility and higher operational complexity.

Infrastructure Integration

Docker Deployment

  • Memory Target: < 4 GB RAM for routing service
  • Storage: PostGIS database for isochrones and results
  • Updates: Automated GTFS data refresh pipeline

Performance Targets

  • Single Route: < 500ms response time
  • Isochrone: < 2s for 4h radius
  • Matrix (1:100): < 10s for hut accessibility analysis