On optimal spatial probability density estimation of passive mobile positioning events

In passive mobile positioning, the cell-level measurements must be translated into a geographical location, which can be expressed as spatial probability density function (SPDF). In this paper, we present the results of a study where we compared different methods to estimate SPDF of passive mobile positioning. The mobile operators provide spatial data about network cells. It is called cellplan. Cellplan can be provided in various formats: location of cell towers and azimuth of antennas, or radio signal strength levels from radio propagation models, etc. Each such source requires specific processing to infer SPDF for passive mobile positioning. We investigated the probabilistic properties of different processing algorithms on different variants of input data. Some investigated methods apply Bayes rules to take into account the effects from overlapping neighbor cells. The results indicate that even on relatively small dataset one can clearly see different accuracy from different processing parameters. The accuracy of SPDF varies significantly depending on processing parameters. and sophisticated radio propagation models did not have significant advantages over simple procedures using Voronoi diagram.