A fast and optimal routing service for cars, bikes, and pedestrians that is based on OpenStreetMap data. The project consists of a server and several programs for the preprocessing (parser, refine, partition, metric, kdtreebuilder). The service scales up to the whole planet meaning that the preprocessing takes a few hours and queries on >1000km routes take less than a second (assuming a modern computer).
Florian Benz, Steven Schäfer, Bernhard Schommer
- Go version 1.0 or above (tested with 1.0, 1.2, and 1.3)
- Metis: 'gpmetis' has to be available for the preprocessing
To build everything, use:
build.sh
To execute the preprocessing steps, use:
preprocess.sh pbf_file path
To start up the server, use:
bin/server -dir path -port port
The following parameters occur in the commands above:
- pbf_file: OSM PBF file
- path: absolut path to the graph dir
- port: port of the server
This project started during the Algorithm Engineering course at Saarland University in summer 2012 (https://www.mpi-inf.mpg.de/departments/d1/teaching/ss12/alg_eng/). In a team of three we successfully completed the course project where the task was to create a fast routing service based on OSM data. The main algorithm is based on ideas from the following paper:
D. Delling, A. V. Goldberg, T. Pajor, R. F. Werneck, Customizable route planning. SEA'11.