A Fast and Customizable Route Planner By Peter: @timetabling - PowerPoint PPT Presentation

A Fast and Customizable Route Planner By Peter: @timetabling Berlin, WhereCamp, 2014-11-13 graphhopper.com

Overview of this Talk ● What is GraphHopper? ● Who uses GraphHopper? ● Numbers Find this presentation under graphhopper.com/public/slides/ 2

What is GraphHopper? A complete online maps application requires 1. Geodata: OpenStreetMap and SRTM (NASA) 2. Drawing: Display map from vector or raster data Like Mapnik, MapSurfer.NET, … 3. Geocoding: Search address, get GPS coordinate Like Photon, Gisgraphy 4.0, Pelias, ... 4. Routing : find best path(s) between coordinates → GraphHopper is all about routing! Still we offer routing and geocoding in our Directions API for business 3

GraphHopper Maps = Address Search + Tiles + GraphHopper 4 Try at graphhopper.com/maps

What is GraphHopper? ● Open Source, Apache License ● Fast routing library and server ● Handles OpenStreetMap data by default ● Ordnance Survey experimental; Navteq possible ● Very memory-efficient but still simple Java API ● Scales to your problems ● Runs world wide as well as indoor graphs ● Can utilize lots of RAM for maximum speed ● Still can run on Android with only 32MB 5

What is GraphHopper? ● Routing with optimality guarantees ● Heuristics not necessary ● Large test suite: unit, integration and load tests ● Commercial Support & Consulting & Matrix and Directions API → Business friendly ● Written in Java: Runs on the Serverside, Desktop, Rasperry Pi, Android, offline in the Browser and (drum roll) ... (github.com/konsoletyper/graphhopper) 6

GraphHopper for iOS! … and offline on iOS! ● graphhopper/graphhopper-ios ● Also available under Apache License ● Still experimental, same code base ● Fast: ~1sec for a route through Germany for iPhone 6 ● Big Thanks goes to Tobias and Calin! ● Works for iOS 7.0+ and OS X 10.10 ● might work for older versions ● Xcode 6.0+ ● including a sample project → 7

Demos GraphHopper Maps https://graphhopper.com/maps ● Online Maps powered by GraphHopper Directions API In-Browser Offline Routing thanks to TeaVM https://graphhopper.com/teavm ● Showcase usable in real world for hybrid solutions ● Combine with pure JS solutions like instantnavi/ffwdme 8

GraphHopper Use Cases ● Point to point routing for car, bicycle, mountainbike, motorbike, horse, foot, ambulance, wheelchair, ... ● Track vehicles via map matching: 'snap' real world GPS points to digital road network ● Urban planning ● Traffic simulation ● Isochrone calculation = reachability for cars or bikes ● For tourism e.g. find neighbouring POIs ● Graph traversal and statistics ● Education 9

GraphHopper Use Cases ● Indoor routing ● Warehouse optimizations ● Tradeshow planning and visitor routing ● Virtual reality games ala 'scotland yard' ● Outdoor routing like biking ● For logistics: Fast route matrix calculation 10 … and more!

Who uses GraphHopper? → For Outdoor and Logistics! ● GPSies: used a self-hosted GraphHopper and will switch to our Directions API ● Gnome: uses our API for the 'Maps' Desktop app ● Komoot: we made routing over 500x faster than home-grown routing engine ● Some IT companies for the taxi sector ● Several IT companies for logistics OpenDoorLogistics, Optitool, market-ip, … ● Rome2rio: multimodal routing ● University Clausthal: education and traffic simulation 11

Who uses GraphHopper? → For Outdoor! Komoot Route planning Route suggestions Komoot Wearable 12

Who uses GraphHopper? → For Logistics! A typical scenario: Get the optimal route to deliver 'goods' to many locations ● With GraphHopper or our Matrix Add-on: Calculate the distance matrix ● With external software: Optimize order of locations e.g. for one or more vehicle 13

Flexibility & Speedup Mode The flexibility mode ● allows customized queries on-the-fly changes possible: e.g. prefer/avoid silent roads, steps, mountains ● still fast: <1sec through entire Germany The speedup mode (Contraction Hierarchies) ● very fast queries: <50ms through Europe ● requires preparation ● smaller RAM usage per query ● more RAM for base graph 14

Numbers! For Worldwide Scenario For car: 130 mio edges, ~100 mio nodes ● Takes ~1h to import and requires 20GB RAM ● Using memory mapped configuration decreases RAM but increases query and import time ● To run this ~10GB are required Speed-up mode ● preparation takes ~2h (car) and requires 25GB ● to run this 17GB are required Comparable to the fastest commercial Maps APIs ● for embedded or LAN queries even ~5x faster 15

Numbers! Routing API, Europe 20 via-points for 1 route 16 Disclaimer: of course, firebug is no scientific performance analysis ;)

Numbers! Same 20 points but use Matrix API → results in 380 distances 17

Thanks! ● Join Mailing list! ● Get Started: graphhopper.com/#community ● Sources & Issues at github.com/graphhopper ● Twitter: @graphhopper 18

Backup slides 19

What is a Graph? ● In a graph we have nodes and edges ● In the real world we have junctions and streets ● Edges and nodes get properties like position Real world Graph 20

Dijkstra → Input : one start and one end node 1. nodeX := start node 2. Get all neighboring nodes of nodeX 3. Put distance of edges for those nodes into a priority queue later steps: add old distance 4. nodeX : getMin(priority queue) 5. Go to 1, break if nodeX == end node → Output : Smallest distance from start to end 21 Get final path via shortest path tree

Bidirectional Dijkstra ● Search from start and end at the same time 22

A* ● Goal directed approach: investigate more 'appropriate' nodes first ● Use linear distance to estimate which node is 'appropriate' 23