Vienna has one of the most advanced open data policies of any European city. The Austrian capital makes a vast array of rich data sets available, reaching far beyond the typical offerings found elsewhere. One of the reasons the Open Government Data Initiative (OGD) in Vienna is so successful is that the data and services which are used within the administrative processes are the same ones provided to the public. In this way Vienna’s public offering of open data incurs only minor extra costs, thereby doing away with the argument of other municipalities that providing open data to the public is too expensive.
OGD Vienna offers several accessibility-related data sets such as road and sidewalk surfaces, the position and height of dropped kerbs, acoustic and tactile signals, elevators in stations, handicapped parking, and a precise elevation model. In the course of the My.Accessible.EU project these data sets will come in handy in a variety of ways.
MyAccessible.EU partner, Fraunhofer IAIS, has started integrating Vienna’s dropped kerb data set into the mapping platform OpenStreetMap (OSM), making this information available to the entire OSM community. Every record of the dropped kerb data set provides two pieces of information: the position of a dropped kerb defined by its geographic coordinates and the height of the kerb at that point, measured in one of three values: 0, 3, or 5 cm. One might think it is simple enough to just import these records as new points into the OSM data base, and in this way the dropped kerbs could be shown on the map as points. But because the location of a dropped kerb in the context of wheelchair routing is only interesting in combination with the street that is accessible because of the dropped kerb, things get a little complicated.
This has to do with how streets are represented on the OSM. The complication that arises is illustrated in the figure above: On the aerial image there are several pins representing the dropped kerbs at a crossing. Looking at this photo, the locations of the pins make sense. They usually come “in pairs” facing each other at opposite sides of a street. Now have a look at the overlaid graphic showing the situation in OSM where streets and ways are most often modelled only as thin lines in the map’s internal database, shown by the blue lines in the graphic. The green stars in the graphic represent the dropped kerb points with the same coordinates as those in the aerial image. Now, forget what you know from the aerial image and try to match the green stars to the “their” skinny blue lines, or streets. That is not easy anymore! You will probably find several points that you cannot intuitively assign to one single street.
In order to resolve these ambiguities one has to come up with a set of assumptions based on what is known about the location of dropped kerbs on real streets. This set of assumptions then results in the most plausible overall assignment of dropped kerbs for the whole crossing on the OSM. This involves an automated process that incrementally reduces the possible solutions until the most plausible solution has been found. As this somewhat complex example shows, the assignment of dropped kerbs to streets on the OSM is quite a challenging task. Fortunately, the automated process doesn’t get confused easily and – according to on-site checks of its solutions at dozens of complex crossings – delivers almost 100% accurate results.
And why is this important? Once the positions of dropped kerbs and their locations in the OSM streets are known, we will be able to optimize the OSM-based wheelchair routing and navigation tools being developed within the MyAccessible.EU project. If we know where a street can be safely crossed by a wheelchair user we can plan ideal routes, and re-plan them when an existing dropped kerb is blocked or impassable. One such use for the kerb data is the University of Heidelberg’s brand new wheelchair profile on openrouteservice.org. A current beta version of this wheelchair routing tool is available online at: http://openrouteservice.org/wheelchair-2.1/.