The 9-th DIMACS challenge focused on shortest path computations. During the challenge several benchmark instances were made available. These are graphs that represent road networks. This repository contains a conversion of these graphs to the PACE 2016 tree decomposition challenge format.
The USA benchmark instances are based upon a TIGER data set released by the US Census Bureau. The original road graphs were built for the DIMACS challenge and are available at the contests website. Since its appearance, the data has been used in numerous publications as benchmark data. I converted the graphs to the PACE 2016 format. In the resulting graphs, nodes correspond to geographic positions and edges connect two positions between which there is a road. The resulting graphs are guaranteed to have no multi edges, no reflexive loops, and are guaranteed to be connected. The node IDs of the converted graphs are identical to original DIMACS graphs. The edge IDs differ because I sorted them and removed multi-edges. Geographic positions of the nodes are available from the original contest website.
Unfortunately, the Europe graphs used in the DIMACS challenge are only available for research purposes. A few emails need to be sent and a form must signed to obtain the data. I therefore additionally extracted a graph and the associated geographic position from OpenStreetMap.
Please cite the following book if use the graphs in a publication:
- The Shortest Path Problem: Ninth DIMACS Implementation Challenge. Camil Demetrescu and Andrew V. Goldberg and David S. Johnson. 2009.