The fact that mobile users of location-based services (LBS) have to be able to go anywhere on earth without changing the application brings up a major requirement: there has to be one global platform that provides a transborder and continuous access to the information necessary for this kind of information systems - and the most important information source for location-based services is geospatial data. However, the world of geospatial data is split into pieces, and the integration of those pieces is a difficult task since the data are highly heterogeneous: they have been acquired according to differing conceptual schemas (or application perspectives), they are available in different formats and scales, with different accuracies, etc. The main problem concerning spatial data integration results from the fact that the same real world objects are stored in multiple, inconsistent representations (MRep) in different spatial databases. Thus, in order to achieve a common view on the underlying spatial data within a global information platform for location-based services, mainly the inconsistencies between multiple representations have to be considered and dealt with adequately.

The problem can be depicted by means of a navigation service. Consider the task of finding the shortest car route between Stuttgart/Germany and Vienna/Austria. You will not have any problems to find a solution if you have one single, continuous street data set comprising the whole query area. However, you will encounter difficulties in case you have two or more separate, partly overlapping source data sets (or patches) that you have to assemble (like a mosaic) in order to form the query area. In this case your navigation application has to be able to merge (or conflate) the source data sets that contain multiple conflicting representations of one and the same real world (street) object in the overlapping areas to again create one consistent, single representation (SRep) street database on which the navigation algorithms can operate. The merge operation, though, is time consuming. This paper describes an approach to find an optimized solution for the navigation in MRep street networks by generating explicit relations between multiple representations and by exploiting them during shortest path search