64 lines
6.7 KiB
TeX
64 lines
6.7 KiB
TeX
\section{Introduction (3/4 - 1 Seite)}
|
|
\label{sec:intro}
|
|
|
|
Many indoor positing system (IPS) rely on radio technologies for estimating a pedestrian's position inside a building \cite{Ebner-15}. Probably the most widespread approaches are those based on Received Signal Strength Indication (RSSI) provided by Wi-Fi, as many buildings nowadays offer good infrastructure and as the be located receiver, standard smartphones can be used. These two features make Wi-Fi particularly interesting for the commercial consumer market, e.g. navigating inside an airport or shopping mall.
|
|
|
|
%Probleme von RSSI
|
|
By measuring the RSSI emitted from Wi-Fi access points (AP) at known locations, the receiver and thus the pedestrian is able to get its current position using the principles of multilateration.
|
|
Here, at least three APs are needed for two-dimensional positioning and at least four APs are needed for three-dimensional positioning.
|
|
However, as a simple RSSI-based multilaterian is very prone to errors in real world scenarios, causing unacceptable inaccuracies, it is necessary to add more advanced methods to approach the positioning problem in a more accurate and stable way.
|
|
The two most popular are fingerprinting and signal strength prediction \cite{davidson2017survey, Afyouni2012}.
|
|
|
|
Fingerprinting is the process of taking RSSI measurements at known positions distributed throughout the building in an so called offline or recording phase.
|
|
These fingerprints can then be used in the online or localization phase to obtain the current position by finding the closest match, e.g. by using a nearest neighbour search.
|
|
Given the current RSSI value the most likely position is the one which is closest to other similar RSSI fingerprints.
|
|
This method includes the characteristics of the environment into the prerecorded fingerprints.
|
|
Thus, the structure of the building (e.g. walls and furniture) and the positions of the access points need not necessarily be known.
|
|
In principle, very small positioning errors in the lower single-digit metre range can be achieved by using this method.
|
|
However, the accuracy depends heavily on the density of the fingerprints and whether the smartphone used for the recordings will also be used later for the localization, as the RSSI characteristics can highly differ between single devices \cite{he2016wi}.
|
|
Recording the fingerprints is a time-consuming and tedious process for large buildings and needs to be redone whenever the environment changes significantly.
|
|
|
|
In contrast, signal strength prediction methods determine signal strengths for arbitrary locations by using an estimation model instead of real measurements.
|
|
This makes an offline phase unnecessary and only the positions of the access points have to be known as well as some other model parameters.
|
|
However, complex scenarios with many non-line-of-sight (NLOS) transmissions require an equally complex signal strength prediction model, taking into account the characteristics of the environment like the different attenuation coefficients of the building's walls \cite{PathLossPredictionModelsForIndoor, WithoutThePain}.
|
|
The determination of such parameters is often only possible with a great deal of effort and material knowledge, especially for older buildings \cite{Fetzer-18}.
|
|
Furthermore, to determine a more or less realistic signal prediction for a location in question, intersections checks of each obstacle within the line-of-sight to the AP have to be calculated, which is costly for larger buildings.
|
|
For this reason, simplified models are often used for the calculation, which do without intersection checks and precisely determined parameters.
|
|
The accuracies achieved in this case are often in a reasonable medium single-digit range of meters \cite{Ebner-17, farid2013recent}.
|
|
|
|
As both, fingerprinting and signal strength prediction, are based on RSSI readings, the accuracy is affected by similar factors.
|
|
For example, temporary and unpredictable occurrences such as large crowds are very difficult to compensate, as the signals are highly attenuated by the human body.
|
|
Without additional sources like cameras, there can be no indication where and how many people are in an area, what makes it hard to adapt the above discussed methods accordingly.
|
|
|
|
|
|
|
|
higher accuracy with more information about the environment.. walls etc
|
|
|
|
ap positions und physikalisches model wie log distance. modell relys on many parameters, hard to receive. however provides also good results with only a few parameters, not so accurate then fingerprinting but without large setup time..
|
|
|
|
-> großer auftritt FTM
|
|
|
|
Wir wollen in dem Paper zeigen wie sich FTM als Sensor für Indoor verhält und kein komplettes IPS vorstellen. Die Arbeit soll ein Gefühl vermitteln, wie sich FTM für IPS einsetzen lässt und wie es sich in unterschiedlichen estimation methoden verhält. wir nutzen ..., ... und particle filtering für die schätzung der aktuellen position des fußgängers. um die performance von ftm möglichst deutlich herrauszustellen wird in dieser arbeit ein sehr einfaches bewegungsmodell, ohne informationen der imu adaptiert.
|
|
|
|
dem gegenüber stellen wir die signal strength prediction rssi, da die beiden verfahren sich im grundkonzept sehr ähnlich sind. beide benötigen lediglich die position der APs sowie ein kleines set an parametern, welche auch nach best practices methoden ausgewählt werden können.
|
|
|
|
|
|
unter der (sehr realistischen) annahme, dass die von FTM gelieferten Distanzen nicht optimal sind, könnte man natürlich auch hier ein fingerprinting einsetzen. jedoch zielt diese arbeit drauf, die grundliegenden eigenschaften und fähigkeiten von FTM (as is) für die indoor lokalisierung zu erforschen und in \cite{} verspricht das gremium von IEEE nämlich ... und ... .
|
|
|
|
\begin{itemize}
|
|
\item New technology ftm -> rtt usw.. wird auch schon von ersten smartphones unterstützt gibt protokoll und api's
|
|
\item klassiches rssi ungenau wegen dämpfungsfaktren, wände usw.
|
|
\item ftm zum lokalisieren weil laufzeit stabiler
|
|
\item ziel dieser arbeit: vergleich von klassischem RSSI (Signal Prediction Log Distance) mit FTM im Anwendungsgebiet der Indoor Lokalisierung
|
|
\item Wir haben: Log-Distance based RSSI + Vorgeschalteten Kalman Filter und FTM + Kalman Filter
|
|
\item Zuerst untersuchen wir die Range-Performance
|
|
\subitem Wer kann die Distanz zum Access-Point besser schätzen)
|
|
\subitem Daraus resultiert auch, dass der Kalman Filter genutzt wird.
|
|
\subitem für Indoor Lokalisierung hat das aber grdz. noch keine große Aussage.
|
|
\item Danach untersuchen wir die Lokalisierungsperformance. (Jeweils Kalman an und aus)
|
|
\subitem Klassische Trilateration
|
|
\subitem Simple Probabilistischer Ansatz
|
|
\subitem Particle Filtering Ansatz (Mit ganz einfachem Bewegungsmodell)
|
|
\item Aufbau der Arbeit in 2 Sätzen
|
|
\end{itemize}
|