Mobile Intelligence

8.3 PREDICTIVE LOCATION INDEXING TECHNIQUES

In the previous section, we presented the dominant approaches for predictive location tracking for the case of a symbolic topology model. In many cases though, we are interested in tracking the location of mobiles at a finer granularity of space and time. Consider, for instance, our interest in tracking the trajectories of birds, airplanes, or satellites, which are considered as points, or our interest in tracking the movement of a tropical storm, of fires. This interests stems from our need to answer queries like, “When two satellites are going to meet?”, “Is the fire threatening village Thetidio?”, and so on.

This leads to the idea of storing in a database for each moving object not the current position but rather a motion vector, which amounts to describing the position as a function of time. That is, if we record for an object its position at time t₀ together with its speed and direction at that time, we can derive expected positions for all times after t₀. Of course, motion vectors also need to be updated from time to time, but much less frequently than positions. Hence, from the location management perspective, we are interested in maintaining dynamically the locations of a set of currently moving objects and in being able to ask queries about the current positions, the positions in the near future, or any relationships that may develop between the moving entities and static geometries over time.

To support such functionality, the ...