没有合适的资源?快使用搜索试试~ 我知道了~
首页wifi location tracking
wifi location tracking
需积分: 9 14 下载量 90 浏览量
更新于2023-03-03
收藏 3.22MB PDF 举报
wifi location tracking for indoor. beyesian algorithm application.
资源详情
资源推荐
Development of a
location system
based on Wi-Fi
signal strength
Master's thesis - rapport de T.F.E.
Internship at the LACIME laboratory of the
Ecole de Technologie Supérieure (University of Québec)
February 19, 2006 to August 19, 2006
Arnaud Bétrémieux, TP3 EA
September 2006
Development of a location system based
on Wi-Fi signal strength
Master’s thesis - rapport de T.F.E.
Arnaud B
´
etr
´
emieux
´
Ecole Sp
´
eciale des Travaux Publics, du b
ˆ
atiment et de l’industrie
TP3 EA
September 2006
Internship at the LACIME laboratory of the
´
Ecole de Technologie Sup
´
erieure
(University of Qu
´
ebec)
February 19, 2006 to August 19, 2006
Abstract
I did my final project for the
´
Ecole Sp
´
eciale des Travaux Publics, du b
ˆ
atiment et de l’industrie
(ESTP, Paris) from February 19 to August 19 2006, in Montr
´
eal, Qu
´
ebec, at the LACIME lab-
oratory of the
´
Ecole de Technologie Sup
´
erieure. The report of this project is in the following
pages.
My objective was the development of an indoor location system, based on Wi-Fi signal
strength, as part of the larger WINS (Wireless Indoor Navigation System) project.
There are numerous applications to indoor positioning, for example in the fields of security
or process management, but the existing techniques either require expensive equipment or are
of very limited efficiency. Wi-Fi is a proven and widespread technology, which has become
very affordable, and already equips most buildings. This makes it an interesting basis for the
development of an indoor positioning system.
When I started working on the projet, the groundworks for the development of the position-
ing system had been laid down: two methods, named “Perimeter” and “Signature”, had been
designed, for the location of a Wi-Fi device. Those methods where both based on a technique
called RF fingerprinting.
RF (Radio Frequency) fingerprinting consists in recording the characteristics (fingerprint)
of signals at various locations in the building in a calibration phase. The calibration database
can then be used to deduce the location corresponding to observed characteristics.
Based on Perimeter and Signature, I had to design as efficient and complete a positioning
system as possible. This involved developping or researching the necessary algorithms, then
designing the corresponding software.
After an extensive bibliographic research on Wi-Fi, RF fingerprinting and the related sub-
jects, I started by writing the programs that would generate fingerprints using both the Perime-
ter and Signature methods, and store those fingerprints in a database. By recording fingerprints
in many different locations and situations, I was able to get a relatively accurate view of the re-
lationship between fingerprint and location.
The complexity and unpredictability I discovered in this relationship, as well as my re-
search, led me to build a system based exclusively on probabilities. I chose the concept of
Hidden Markov Models.
In order to attain a higher precision, I wanted the system not only to consider the last fin-
gerprint, but to associate the set of fingerprints recorded until the time of computation to the
corresponding set of points (trajectory). This way, the errors introduced by measuring inaccu-
racy and the unpredictable signal strength variations would have less impact, considering that
the corresponding trajectory needs to be valid. Valid both in the absolute sense (no sudden
jumps ) and for the considered device (maximum possible speed, probability of moving, etc.)
The Viterbi algorithm seemed to me like the best solution.
To reduce the time needed for calibration, I then looked for the best interpolation method.
I chose a technology based on an artificial neural network, which allows the generation of a
dense set of reference fingerprints, with a scarce calibration.
I had the opportunity to validate my choices and the reliability of the proposed system
by doing a demonstration (cf. presentation slides in appendix A) for the Cari-all company,
the leading shopping cart producer in Qu
´
ebec. The goal of this demonstration was to prove
i
剩余100页未读,继续阅读
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功