Datasets

In the following, I collect some interesting real-world datasets in spatial databases.

Road Networks

These real-world road networks were extracted from OpenStreetMap on Feb. 2021, based on the processing method in our VLDB’18 paper. Hence, these road networks are usually much larger than the public road networks that I have ever seen (e.g., the DIMACS ones). Moreover, I extract many Chinese provinces and cities, which are not included in the DIMACS datasets. If you want to use the road networks below, please consider citing this homepage and our VLDB’18 paper (BibTex). Briefly speaking, the datasets are obtained based on the following steps.

1. We download region/country .osm from Geofabrik, which maintains the OpenStreetMap data;
2. We extract the exact .osm based on the corresponding .poly via Osmconvert;
3. We extract the graph files based from the corresponding .osm via OsmToRoadGraph, where network type is car.
4. The speed to travel each edge is set to the 80% of the maximum speed of the corresponding path type.

Notice, in these graphs, the unit of distance is meter and the unit of travel time is second. If you want to do shortest distance/path queries on these graphs, I recommend one VLDB’18 paper named “An Experimental Study on Hub Labeling based Shortest Path Algorithms” from Ye Li et al. (see their paper for more details). I have revised their source code to fit the double-type of edge weights (the source code and license are here).

Large datasets (USA and Chinese provinces) can also be downloaded from my Tencent Cloud.

USA

China (Cities)

China (Provinces)

Urban Logistics

These datasets are under maintainence.