Routing OpenStreetMap

Brief introduction to A* and routing on OpenStreetMap data.

Transcript

1. Routing OpenStreetMap Zero Cho
2. A quick recap on Graph
3. node edge 1 weight 5 1 10
4. node edge 1 weight 5 1 10 shortest path, weight = 7
5. Georgia Institute ofTechnology d a o R l l i t 4 1 t e e r t S a t t e i r a M W N t S h t 0 1 Tech Pk y N W W N e v A l l i h p m e H W N e u n e v A h t r o N FerstDrNW Fers tDr NW WilliamsSt SpringSt NW 10thStN WPeachtree St NW t S e e r t h c a
6. a Fers tDr NW
9. How is routing done?
10. many ways. We will talk about A-star algorithm today.
11. S 1 E 2 node 1/-10 5/6 1/1 10/10 7/5
12. S 1 E 2 node 1/-10 5/6 1/1 10/10 7/5
13. S 1 E 2 node 1/-10 5/6 1/1 10/10 7/5 1 2 0 + 5 0 + 6
14. S 1 E 2 node 1/-10 5/6 1/1 10/10 7/5 2 0 + 6 5 - 10 E 5 + 1
15. S 1 E 2 node 1/-10 5/6 1/1 10/10 7/5 E 5 + 1 -5 + 10
16. S 1 E 2 node 1/-10 5/6 1/1 10/10 7/5 E 5
17. which leads us to my first result… A* OpenStreetMap router written in Ruby https://github.com/itszero/osm-simple-router
18. which is great… except it’s…
19. so slow.
20. How do we improve it?
21. Moving to neo4j a Graph-based database.
22. Cypher Query Language MATCH (actor:Person)-[:ACTED_IN]->(movie:Movie) WHERE movie.title =~ "T.*" RETURN movie.title as title, collect(actor.name) as cast ORDER BY title ASC LIMIT 10;
23. Sure it helps having proper storing and querying engine, but the data is still too much. ~25x improves? 5 secs?
26. intersections
27. overlay link 3 5 8*
28. take Atlanta metro area for an example 2,567,916 nodes ➔ 452,313 (-83%) 10,669,432 relationships ➔ 1,843,440 (-83%) 4.65 secs. routing time ➔ 1.30 secs.
29. Now we have the algorithm, all we need to do is plug it into the database.
30. which leads us to my second result… A* OpenStreetMap router for Neo4j https://github.com/itszero/neo4j-astar-example (code is really ugly tho)
32. Questions?