Academia.eduAcademia.edu

Figure 4 – uploaded by Jody Sultan

See full PDFdownloadDownload figure
Figure 6. A single OSM road polyline (green line, ID=1) is split according to intersections with the result of the map-matching algorithm (red dashed polyline) into three segments (ID=11, 12, 13).

Figure 6 A single OSM road polyline (green line, ID=1) is split according to intersections with the result of the map-matching algorithm (red dashed polyline) into three segments (ID=11, 12, 13).

Related Figures (12)

Table 1. Tags used in OSM base map for roads describing where cycling is allowed — and where not "Unknown" refers to roads not defined in OSM, i.e., it is unknown whether cycling is allowed — or not. All road downloaded OSM segments and their attributed data are converted to ESRI's shapefile file format (“.shp”). The two study areas analyzed in this research are Amsterdam, The Netherlands, and Osnabriick, Germany; both cities are known for their massive cycle use and well-established and developed cycleway infrastructure. It is assumed that these cycle-friendly cities can serve as good case study for the paradigm presented here.
Table 1. Tags used in OSM base map for roads describing where cycling is allowed — and where not "Unknown" refers to roads not defined in OSM, i.e., it is unknown whether cycling is allowed — or not. All road downloaded OSM segments and their attributed data are converted to ESRI's shapefile file format (“.shp”). The two study areas analyzed in this research are Amsterdam, The Netherlands, and Osnabriick, Germany; both cities are known for their massive cycle use and well-established and developed cycleway infrastructure. It is assumed that these cycle-friendly cities can serve as good case study for the paradigm presented here.
Figure 1. GPSies trajectories (in black) and OSM roads (in grey) for Amsterdam.
Figure 1. GPSies trajectories (in black) and OSM roads (in grey) for Amsterdam.
Figure 2. GPSies trajectories (in black) and OSM roads (in grey) for Osnabriick. 3.2 Pre-processing of cycle trajectories
Figure 2. GPSies trajectories (in black) and OSM roads (in grey) for Osnabriick. 3.2 Pre-processing of cycle trajectories
Figure 5. Two cycling traces (light blue) matched to the OSM base map roads (black) and the final road match result (green). The algorithm handles the matching of a single trajectory even when the trajectory is deviated to an adjacent road, or when the trajectory does not go through a junction. Since GPS-data is accurate to 10 meters in position, it is obvious that the trajectories are not positioned precisely on the roads. To carry out the analysis, trajectories should be assigned to a road. To solve the matching task, this study adopts the HMM-based map-matching algorithm proposed by Haunert and Budig (2012), which proved to be a reliable method for the problem of matching a GPS trajectory with a road data set. We review the method in this section and discuss its suitability for our purpose. This algorithm can be classified as an offline map-matching global optimization algorithm, since it requires the entire trajectory dataset as input, and also as a topological map-matching algorithm, because it allows a user to constrain the output path to the given network. However, because the algorithm allows some GPS points to be left unmatched, it is relatively robust against missing roads or, more generally, incomplete or outdated network models. This is particularly important when matching trajectories of cyclists with a general-purpose network model, i.e., a model not focusing on a particular mode of transport. Furthermore, we think that it makes sense to tailor the probabilistic model underlying the algorithm to a certain mode of transport, for example, to define that candidates on a motorway are very unlikely if the GPS trajectory was recorded by a pedestrian. In the current implementation of the algorithm, however, the road type is not taken into account. The algorithm decides whether the existing class of a road edge should be kept or the class of a corresponding trajectory edge should be assigned to that road edge. Figure 5 depicts two examples for matching results of the algorithm.
Figure 5. Two cycling traces (light blue) matched to the OSM base map roads (black) and the final road match result (green). The algorithm handles the matching of a single trajectory even when the trajectory is deviated to an adjacent road, or when the trajectory does not go through a junction. Since GPS-data is accurate to 10 meters in position, it is obvious that the trajectories are not positioned precisely on the roads. To carry out the analysis, trajectories should be assigned to a road. To solve the matching task, this study adopts the HMM-based map-matching algorithm proposed by Haunert and Budig (2012), which proved to be a reliable method for the problem of matching a GPS trajectory with a road data set. We review the method in this section and discuss its suitability for our purpose. This algorithm can be classified as an offline map-matching global optimization algorithm, since it requires the entire trajectory dataset as input, and also as a topological map-matching algorithm, because it allows a user to constrain the output path to the given network. However, because the algorithm allows some GPS points to be left unmatched, it is relatively robust against missing roads or, more generally, incomplete or outdated network models. This is particularly important when matching trajectories of cyclists with a general-purpose network model, i.e., a model not focusing on a particular mode of transport. Furthermore, we think that it makes sense to tailor the probabilistic model underlying the algorithm to a certain mode of transport, for example, to define that candidates on a motorway are very unlikely if the GPS trajectory was recorded by a pedestrian. In the current implementation of the algorithm, however, the road type is not taken into account. The algorithm decides whether the existing class of a road edge should be kept or the class of a corresponding trajectory edge should be assigned to that road edge. Figure 5 depicts two examples for matching results of the algorithm.
Table 2. Trajectories obtained from GPSies for the study areas. 4.1 Highway-tag-based statistical use analysis
Table 2. Trajectories obtained from GPSies for the study areas. 4.1 Highway-tag-based statistical use analysis
Figure 7. OSM base map of Amsterdam showing roads allowed for cycling (in green) and prohibited (in red).
Figure 7. OSM base map of Amsterdam showing roads allowed for cycling (in green) and prohibited (in red).
Table 3. Road-type use in Amsterdam according to general classification: allowed (bold), prohibited and unclassified (grey shaded). Usage values below 1% are not presented. Table 3 shows that the total bicycle usage percentage of designated cycle roads in Amsterdam is 76.6%, whereas prohibited road usage is 8.5%; the latter is a relatively small number, but still one that should be addressed by planners (translates to a total length of close to 70 km of roads). The total cycle-use percentage of unclassified routes stands at 15.0%. In general, it can be concluded that with the data analyzed, cyclists prefer using cycle- designated and friendly roads. Table 4. Road-type use in Osnabriick according to general classification: allowed (bold), prohibited and unclassified (inderlined). [Jsase values helow 1% are not nresented.
Table 3. Road-type use in Amsterdam according to general classification: allowed (bold), prohibited and unclassified (grey shaded). Usage values below 1% are not presented. Table 3 shows that the total bicycle usage percentage of designated cycle roads in Amsterdam is 76.6%, whereas prohibited road usage is 8.5%; the latter is a relatively small number, but still one that should be addressed by planners (translates to a total length of close to 70 km of roads). The total cycle-use percentage of unclassified routes stands at 15.0%. In general, it can be concluded that with the data analyzed, cyclists prefer using cycle- designated and friendly roads. Table 4. Road-type use in Osnabriick according to general classification: allowed (bold), prohibited and unclassified (inderlined). [Jsase values helow 1% are not nresented.
Table 4 shows that the total bicycle usage percentage of designated cycle roads in Osnabriick stands at 74.7%, very similar to the value in Amsterdam, whereas prohibited road usage stands at 16.2% (double the value in Amsterdam). The total-cycle use percentage of unclassified roads stands at 9.1% (approximately 140 km). The highest rate in Amsterdam (50.0%) is associated with cycleways, whereas in Osnabriick the highest rate (36.9%) is associated with residential roads. Such
Table 4 shows that the total bicycle usage percentage of designated cycle roads in Osnabriick stands at 74.7%, very similar to the value in Amsterdam, whereas prohibited road usage stands at 16.2% (double the value in Amsterdam). The total-cycle use percentage of unclassified roads stands at 9.1% (approximately 140 km). The highest rate in Amsterdam (50.0%) is associated with cycleways, whereas in Osnabriick the highest rate (36.9%) is associated with residential roads. Such
Figure 9. Unclassified-type Amsterdam streets Keizersgracht, Runstraat and Huidenstraat, showing evidence of bicycle activity: Google maps (top-left), OSM (bottom-left), Google maps imagery (right). When considering the OSM base map, the results show that a relatively large number of roads is agged as unclassified. In Amsterdam this rate stands at 12.7% and in Osnabriick at 8.0%. It is inclear to which type of travel mode such roads are designed for, and whether cyclists are allowed o use these roads. In fact, in several cases, an analysis showed that cyclists, pedestrians and vehicles simultaneously use unclassified roads. Figure 9 shows unclassified roads in Amsterdam, in which cyclists and vehicles share the same space. For this reason, the unclassified roads were not aken into account in the analysis.
Figure 9. Unclassified-type Amsterdam streets Keizersgracht, Runstraat and Huidenstraat, showing evidence of bicycle activity: Google maps (top-left), OSM (bottom-left), Google maps imagery (right). When considering the OSM base map, the results show that a relatively large number of roads is agged as unclassified. In Amsterdam this rate stands at 12.7% and in Osnabriick at 8.0%. It is inclear to which type of travel mode such roads are designed for, and whether cyclists are allowed o use these roads. In fact, in several cases, an analysis showed that cyclists, pedestrians and vehicles simultaneously use unclassified roads. Figure 9 shows unclassified roads in Amsterdam, in which cyclists and vehicles share the same space. For this reason, the unclassified roads were not aken into account in the analysis.
Table 6. Road-type use in Osnabriick according to classification: designated cycleways, roads cyclists share with pedestrians; roads cyclists share with motorized vehicles; and roads cyclists are not allowed to use. One point should be mentioned, related to the issue of data- incompleteness and errors, which might stem from an incorrect classification (derived from ill-defined or erroneous matching, or ambiguous pre-processing), lack or wrong data contributed by users or update problems. On the other hand, results still prove that such a methodology can assist in decision making for future planning, due to the fact that volumes of data do exist since many cyclists use these routes. This information is accessible, free to use and can be used in many applications and planning, proving yet again the potential contributed geographic data holds.
Table 6. Road-type use in Osnabriick according to classification: designated cycleways, roads cyclists share with pedestrians; roads cyclists share with motorized vehicles; and roads cyclists are not allowed to use. One point should be mentioned, related to the issue of data- incompleteness and errors, which might stem from an incorrect classification (derived from ill-defined or erroneous matching, or ambiguous pre-processing), lack or wrong data contributed by users or update problems. On the other hand, results still prove that such a methodology can assist in decision making for future planning, due to the fact that volumes of data do exist since many cyclists use these routes. This information is accessible, free to use and can be used in many applications and planning, proving yet again the potential contributed geographic data holds.
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved