NOTE: There are several approaches to route calculation from a road network. You can use PostGIS’s pgRouting, a simple library like networkx, or even use Neo4j. See the example of pgRouting with a custom road network and a Flask web service https://github.com/voirinprof/gis_pgrouting_docker. If you want to see how to use networkx for route calculation, use this repo: https://github.com/voirinprof/gis_nx_docker.

Introduction to pgRouting

pgRouting is a PostGIS extension that adds route calculation and network analysis capabilities to a PostgreSQL/PostGIS database.

In geomatics, pgRouting allows:

  • Calculation of the shortest routes (fastest, shortest, multimodal)
  • Network analysis (accessibility, graphs, multiple routes)
  • Processing of large road networks directly in the database

NOTE: The official pgRouting documentation is very useful to explore all the available features.

Installing pgRouting

On a server with PostgreSQL and PostGIS, install pgRouting:

sudo apt install postgresql-15-pgrouting

Then enable the extension in your database:

CREATE EXTENSION pgrouting;

Tip: pgRouting works on graphs stored as edge tables.

Preparing a Road Network

Your road network must be structured as an edge table:

CREATE TABLE reseau_routier (
    id SERIAL PRIMARY KEY,
    source BIGINT,
    target BIGINT,
    geom GEOMETRY(LineString, 4326),
    cost DOUBLE PRECISION
);
  • source: ID of the start node
  • target: ID of the end node
  • cost: Cost to traverse the edge (distance, time, etc.)

You can use pgr_createTopology to automatically generate source and target from geom.

SELECT pgr_createTopology('reseau_routier', 0.0001, 'geom', 'id');

Connecting to pgRouting in Python

Install dependencies:

pip install psycopg sqlalchemy shapely

Simple connection:

import psycopg

conn = psycopg.connect("dbname=geomatique user=postgres password=password host=localhost")
cur = conn.cursor()

Calculating the Shortest Path

Assuming you know the source_id and target_id in your network:

start = 1001
end = 1050

cur.execute("""
    SELECT seq, id1 AS node, id2 AS edge, cost
    FROM pgr_dijkstra(
        'SELECT id, source, target, cost AS cost FROM reseau_routier',
        %s, %s,
        directed := true
    )
""", (start, end))

path = cur.fetchall()

for line in path:
    print(f"Step {line[0]}: pass through node {line[1]} via edge {line[2]} (cost {line[3]})")

pgr_dijkstra is the classic shortest path algorithm.

Other Useful pgRouting Features

  • Route to multiple destinations:
    pgr_dijkstraVia
  • Calculate cost matrices between multiple points:
    pgr_dijkstraCostMatrix
  • Isochrone search (reachable areas by distance or time):
    pgr_drivingDistance

Example of driving distance:

SELECT * FROM pgr_drivingDistance(
    'SELECT id, source, target, cost FROM reseau_routier',
    1001,
    2000  -- maximum distance
);

Best Practices

  • Create indexes on source and target to speed up calculations: 
    CREATE INDEX ON reseau_routier(source);
CREATE INDEX ON reseau_routier(target);
  • Optimize your network: limit unnecessary edges, check topology.
  • Use a metric projection system (e.g., EPSG:2154 for France) for accurate distances.
  • Separate directed and undirected networks according to your needs.

Use Cases in Geomatics

  • Vehicle route calculation (deliveries, buses, emergency services)
  • Accessibility analysis (distance to hospitals, schools)
  • Origin-Destination matrices for mobility flows

Useful Resources