Communications and Signal Processing Lab (ComSP)

Title	Técnicas avanzadas de geolocalización en redes UMTS
State	Finished
Author	Juan Antonio García Fernández
Director/s	Mariano Fernández Navarro ,  Antonio Jurado Navas
University	Universidad de Málaga
Center	Escuela Técnica Superior de Ingeniería de Telecomunicación
Department	Ingeniería de Comunicaciones
Reading date	19/09/2016
File	PDF

Mobile communication networks have evolved exponentially over recent years. This

growing has highlighted the need of having tools that help us to do troubleshooting, optimization

and monitoring of the performance of these networks. In such a context, geolocated

maps play a fundamental role as they provide a simple, efficient and visual manner

to fulfill these tasks as well as offering a low cost alternative to drive tests. Under these

circumstances, and from call traces, comes the possibility of utilizing different geolocation

techniques to estimate the positions of the events reported by the User Equipments (UEs).

Naturally, the two most important issues for this approach are the accuracy point-to-point

and the visual quality of the maps.

The goal of this PhD is to analyze and develop several advanced geolocation algorithms

for UMTS networks, but most of them can be easily extrapolate to LTE networks. These

algorithms are implemented into stand-alone tools with the aim of further testing through

real data coming from real networks currently in operation. The use of real data is a key

factor, since provides high robustness and reliability to results and conclusions drawn in this

work.

First of all, a hybrid geolocation tool which combines various algorithms and source of

information is developed. Its main objective is to establish a trade-off between populated

and realistic geolocated maps, and a proper point-to-point accuracy. A set of sequential processes

are defined in order to gradually introduce and combine signal level measurements,

time measurements, parameters of the server and neighbors Node Bs, signal propagation

predictions, and even geographical data known as clutters. The final outcome is the capability

of generating maps, for instance, RSCP or best server maps, providing great visual

coherence and cohesion, in addition to an accuracy improvement due to the combination of

different strategies.

In the second instance, a new tool based on Observed Time Difference of Arrival (OTDOA)

technique is presented in order to jointly estimate both the position of the different

UEs and the Relative Time Differences (RTDs) between Node Bs in real UMTS networks.

For this purpose, and employing the parameter TM provided by the events Measurement Report,

this method results in a non-linear least squares estimation problem which is solved by

employing an iterative method. In particular, a comparative between the well-known Gauss-

Newton, Levenberg-Marquardt and a proposed modified version of Levenverg-Marquardt

algorithms is carried out. In addition, an optimal spatial geometry, the star-topology, for the

Node B stations involved in such an iterative method is analyzed for avoiding the appearance

of local minima. As a result, all modifications proposed increase the inherent accuracy

vii

while maintaining a fast convergence and a high robustness, apart from preventing the cost

of deploying external system to recover the synchronization, such as LMUs.

Finally, a technique for compressing and adjusting in a smart manner the position estimates

of any UE given by different geolocation/positioning methods in a mobile communication

network based on different strategies (OTDOA, Angle of Arrivals - AoA, Propagation

Delay - PD ...) is detailed. Accordingly, the basic idea is to move from at event level geolocation

to at call level geolocation, and distinguishing whether they are static or dynamic.

Concretely, the proposed method forces each position estimate towards a virtual anchor previously

calculated from the 95% confidence intervals of every located event. In turn, these

confidence regions or error estimates are obtained with simulators which depend on several

parameters of the mobile network and the events themselves. The main advantage of this

proposal is to improve the accuracy in a remarkable way, and to mitigate the adverse effect

of multipath and other sources of errors that induce to inaccuracy in the terminals position

estimates.