Probabilistic Location Estimate of Passive Mobile Positioning Events
Name
Joosep Rõõmusaare
Abstract
Researchers, who are trying to understand human mobility patterns, collect data from cellular telephone networks. Mobiles are creating events every time they are used for calling, SMS, or the Internet. The events contain the information, in which network cell that mobile was at the moment of the event. Cell's coverage can be used for estimating the geographical location of the mobile. The estimated locations are not a point on the map, but the possible area, where the mobile may be when they are connected to that specific cell.
Mobiles connecting to cells are depending on multiple variables, meaning, that a mobile may not always connect to the cell with the strongest signal. That makes estimation of the mobile location more difficult, as the coverage areas may overlap with each other.
Cell plan is a description of cell coverage areas and there are multiple ways for defining cell coverage areas.
This thesis is about estimating mobile events positioning quality with spatial probability density functions. Different cell plan variants will be implemented and real ground truth location data is used to find the modification that maximizes the likelihood estimation.
Compared RSSI-based and Voronoi-based cell plans and their modifications and was found that Voronoi-based cell plans are better for location positioning than the RSSI-based cell plans.
Furthermore, Bayesian overlapping method was examined to see does applying it would improve location positioning accuracy. It was found that applying Bayesian overlapping methods improved the accuracy of the worse cell plans, but made accuracy worse for the better cell plans.
Mobiles connecting to cells are depending on multiple variables, meaning, that a mobile may not always connect to the cell with the strongest signal. That makes estimation of the mobile location more difficult, as the coverage areas may overlap with each other.
Cell plan is a description of cell coverage areas and there are multiple ways for defining cell coverage areas.
This thesis is about estimating mobile events positioning quality with spatial probability density functions. Different cell plan variants will be implemented and real ground truth location data is used to find the modification that maximizes the likelihood estimation.
Compared RSSI-based and Voronoi-based cell plans and their modifications and was found that Voronoi-based cell plans are better for location positioning than the RSSI-based cell plans.
Furthermore, Bayesian overlapping method was examined to see does applying it would improve location positioning accuracy. It was found that applying Bayesian overlapping methods improved the accuracy of the worse cell plans, but made accuracy worse for the better cell plans.
Graduation Thesis language
English
Graduation Thesis type
Master - Computer Science
Supervisor(s)
Toivo Vajakas
Defence year
2016