Open source map matching with Markov decision processes: A new method and a detailed benchmark with existing approaches
A. Wöltche, "Open source map matching with Markov decision processes: A new method and a detailed benchmark with existing approaches", Transactions in GIS, vol. 27, no. 7, pp. 1959–1991, Oct. 2023, doi: 10.1111/tgis.13107
Float Car Data (FCD) map matching dataset to download from:
https://doi.org/10.6084/m9.figshare.19785835
Place in this directory from the benchmarks.zip zip file
from the benchmarks/docker/initialize/data/ folder
the oberfranken-latest.osm.pbf OSM file, please see odbl-10.txt for licence information,
the points.csv csv file, and
the ground_truth.csv csv file.
Additionally, we provide the anonymized points_anonymized.csv raw FCD file with permission
from Map and Route GmbH & Co. KG for further testing purposes.
The points.csv file from the data set above is a subset for that we provide hand-corrected results with the
ground_truth.csv file.
You need the same OpenStreetMap data as above from the region of Oberfranken, Bavaria, Germany.
In case you want to use a more recent version than the one from above, you can download the needed extract for remplacement at Geofabrik / Oberfranken.
The following commands need to be run from the base map_matching_2,
the path to the map_matching_2 executable can be adjusted as needed, e.g.,
for GCC: run/gcc/release/bin/map_matching_2,
for Clang: run/clang/release/bin/map_matching_2,
for MSVC: run\msvc\release\bin\map_matching_2.exe.
# prepare network
./map_matching_2 --conf data/floating-car-data/conf/prepare.conf
# match tracks
./map_matching_2 --conf data/floating-car-data/conf/match.conf
# compare matches with ground truth
./map_matching_2 --conf data/floating-car-data/conf/compare.conf
The provided raw FCD tracks can be matched with the following command:
# match raw tracks
./map_matching_2 --conf data/floating-car-data/conf/match_raw.conf
Also given is an example on how to use custom tags for OpenStreetMap, for example for foot ways.
# prepare network using all roads, i.e., for foot
./map_matching_2 --conf data/floating-car-data/conf/prepare_all.conf
The results will be in the results folder that is automatically created.
Results on our test system (accuracy is the weighted mean correct fraction in percent):
| Mode | Time (s) | Max RAM (MiB) | Accuracy (%) |
|---|---|---|---|
| Prepare | 32.80 | 2,271 | N/A |
| Prepare * | 12.87 | 1,543 | N/A |
| Match | 0.66 | 422 | N/A |
| Compare | 0.59 | 38 | 99.58 |
| Match (RAW) | 5.56 | 1,620 | N/A |
| Prepare (All) | 37.38 | 2,602 | N/A |
| Prepare (All) * | 14.48 | 1,688 | N/A |
Max RAM contains shared memory (cached memory backed by disk, not actually used).
The prepare runs marked with * were run with --memory-mapped-preparation off.
In this case, only the final data is written into memory mapped files and during the preparation,
non-cached RAM (actually used by the process) is used.