Using PostgreSQL, PostGIS, and pgRouting for street sweeping

Community Chat: https://discord.gg/CVmyKb8

RSVP: https://postgresconf.org/conferences/postgres-webinar-series/tickets

Presented by James E. Marca

In responding to an RFP to optimize and improve street sweeping routes for a small city in California, we developed a solver based on Google's Operations Research Tools. While the guts of the solver uses OR-Tools, virtually everything else about the project is made possible by PostgreSQL, PostGIS, and pgRouting.

In this presentation, I talk about the various ways PostgreSQL and friends made our solver possible. I provide specific examples of the SQL used to:

• clean OpenStreetMap data using WITH RECURSIVE/Common Table Expressions (see https://www.postgresql.org/docs/12/queries-with.html)
• convert one-way and two-way streets into a one-way network of curbs
• convert the street network into its linegraph dual
• save the solver output to a new table using Python
• display the solver's routes using QGIS
• produce a smooth animation using PostGIS aggregate functions and windowing

This presentation is intended to be of "medium" difficulty, but I cover more advanced SQL techniques like WITH RECURSIVE as well as more advanced PostGIS queries.

This presentation will be of interest to people who have thought about using OpenStreetMap data in their projects, or who would like a practical example of using PostGIS and pgRouting in a real project. It may also be interesting to Data Science types—although there isn't any machine learning, there is plenty of practical advice about cleaning up data and presenting results.