рус eng укр deu
About me
Practice in Germany: About IPVS Target Setting About work
Efficient Indexing for Spatial Events Processing
Executor: Ignatova Kateryna
Supervisor: Tobias Drosdol
Target setting:
To manage the information about mobile objects positions (especially indoors) the support of symbolic coordinates is necessary. This means that descriptors like "Room 1.123" will be used to refer to certain geographical area. Unlike geometrical coordinates, which are delivered by GPS, symbolic coordinates do not make spatial relations like distance or inclusions obvious. For this reason we need the management of this metainformation in so-called location model. This model delivers all possible information about areas' spatial relations, which are described with symbolic coordinates.
To achieve the most efficient management of position information for symbolic coordinates different index structures, which take into account spatial relations, have been analyzed. In this work based on this analysis efficient index structures, which process simple spatial events, should be developed. Two different basic events should be supported:
1. OnEnter: A notification is sent immediately when a mobile object enters a certain area, described with symbolic coordinates.
2. OnLeave: A notification is sent immediately when a mobile object leaves a certain area, described with symbolic coordinates.
To realize this functionality by each position actualization it should be checked, whether a registered event took place. Moreover, the appropriate indexing of all registered events is needed. This indexing should take into account inclusion relations between symbolic coordinates. To make is possible to compare the efficiency of different approaches two indexing structures were developed and compared to each other in the context of this work:
Version 1: Each registered event is being referred by each leaf of spatial index that lies inside of given area. Then by position actualization the events, which are referred by the old and the new position have to be compared to each other to ascertain which events took place. By the bigger number of events, this comparison gets too expensive.
Version 2: Each registered event is being referred by corresponding node of spatial index, which represents the given area. Then by traversing of index structure for every position actualization we can find out which events took place.
The underlying location model defines a quantity of symbolic coordinates.