MSC_GIMA_Justin_Stook_V1.6
Planning an indoor navigation
service for a smartphone with
Wi-Fi fingerprinting localization
Author:
E-mail:
Justin Stook
justinstook@gmail.com
Supervisor:
Professor:
Reviewer:
Ir. Edward Verbree (TUD)
Prof. dr. ir. Peter van Oosterom (TUD)
Dr. Corné van Elzakker (UT-ITC)
Version (Final):
1.6 – August 18th , 2011
MSc Thesis Indoor Navigation with Wi -Fi Fingerprinting |2
Dedicated to my father – may he navigate well in an unknown world to us
Selamat jalan
“The wisest men follow their own direction” – Euripides (484 BC – 406 BC)
MSc Thesis Indoor Navigation with Wi -Fi Fingerprinting |3
Prologue
Navigation has become booming business recently. Well-known navigation services, such as
TomTom, Garmin and Google Navigation, are dominating the market for outdoor, road based
navigation. The fact that such applications are able to determine a location on a platform has always
fascinated the author. Simple location determination is further utilized in specialized applications and
functions, such as SportyPal, which logs the movements and interpolates several units, or the
geotagging of photos. And so, this further piqued the interest to investigate matters in the indoor
environment, as one has to look beyond what one sees.
However, navigation as a word can also be used as a metaphor to describe one man’s life journey.
How can one navigate as a human being? On the paths of life, one has to navigate through a web of
complicated matters by fulfilling expectations and by making decisions that will influence your
future. Decisions might be based on past experience and/or knowledge, there is no ‘turn back’
option, and things might heavily alter the path of life in unforeseen circumstances, both positive and
negative. Navigation is then based on the abstract level.
The author, too, almost navigated towards another direction. Unforeseen circumstances in the
private life forced him to choose a different flow than intended. The result is this thesis project, in
which much time and effort has been put. The author hopes that the thesis is readable and
understandable. He hopes that every wise man should follow his own direction and that as such,
navigation applications will contribute in making those decisions.
Justin Stook
Acknowledgements
I would like to thank the following persons who helped me with my thesis research:






Geert Wirken for his advice and excellent help op Java programming – we have spent a lot of
time together tackling the problems!
Mehmet Yildirim from the TU Berlin, for providing me excellent help as well, with the location
determination scripts, as well as helping me out to get familiar with the Android environment.
Gavin Brown from the TU Berlin, for providing me new insights of this domain.
My supervisors Edward Verbree and Peter van Oosterom for their endless patience and who
always believed I could do it, and provided me extensively feedback I could use.
Joel Haasnoot for his advice and help on Java programming.
My family and friends who (mentally) supported me to accomplish this feature.
MSc Thesis Indoor Navigation with Wi -Fi Fingerprinting |4
Table of Contents
Summary ..................................................................................................................................... 9
List of abbreviations and acronyms ............................................................................................. 11
Terminology ............................................................................................................................... 12
Positioning vs. localization and accuracy vs. precision ..................................................................... 12
Building geometry ............................................................................................................................. 14
PART 0 - Introduction ................................................................................................................. 15
1. Introduction ........................................................................................................................... 16
1.1. The research topic ...................................................................................................................... 16
1.2. Contents of this thesis ................................................................................................................ 17
2. Goal, objectives and scope/limitations .................................................................................... 19
3. Research questions ................................................................................................................. 21
PART I - Theoretical framework .................................................................................................. 23
4. Sensor techniques .................................................................................................................. 24
4.1. Why GPS cannot be used ........................................................................................................... 24
4.2. Other sensor techniques compared ........................................................................................... 25
4.3. General background of Wi-Fi...................................................................................................... 28
4.3.1. Use of spread spectrum ....................................................................................................... 28
4.3.2. Frames / layers .................................................................................................................... 30
4.4. General strengths and weaknesses ............................................................................................ 31
5. Positioning techniques ............................................................................................................ 33
5.1. Five techniques........................................................................................................................... 33
5.2. Fingerprinting elaborated .......................................................................................................... 35
5.3. Building constructions influencing fingerprinting ...................................................................... 36
5.4. Using signal strengths to locate ................................................................................................. 37
6. Navigation .............................................................................................................................. 39
6.1. What is navigation? .................................................................................................................... 39
6.2. Navigation principles .................................................................................................................. 39
6.2.1. General principles ................................................................................................................ 39
6.2.2. Certainty and route confirmation ........................................................................................ 40
6.3. Navigation algorithms ................................................................................................................ 40
6.4. Navigation infrastructure ........................................................................................................... 40
6.5. Spatial referencing ..................................................................................................................... 41
7. Building an application using standards ................................................................................... 43
Conclusion Part I ........................................................................................................................ 44
PART II - Theory and application ................................................................................................. 45
8. The user involved ................................................................................................................... 46
8.1. User profiles ............................................................................................................................... 46
8.2. User needs in this project ........................................................................................................... 46
8.2.1. Personal information ........................................................................................................... 47
8.2.2. Human centred preferences (requirements) ....................................................................... 47
8.2.3. Service related information and preferences ...................................................................... 47
8.2.4. Device related information and preferences (requirements) .............................................. 47
9. User allocation: gateway services ............................................................................................ 49
9.1. Components of the gateway ...................................................................................................... 49
9.2. Site surveying and positioning.................................................................................................... 49
10. Directory and geocoding services .......................................................................................... 51
10.1. Directory services ..................................................................................................................... 51
10.2. Geocoding services ................................................................................................................... 52
MSc Thesis Indoor Navigation with Wi -Fi Fingerprinting |5
11. Modelling and attributes ...................................................................................................... 53
11.1. Modelling the indoor environment .......................................................................................... 53
11.2. Non-usage of geometry ............................................................................................................ 54
12. Visualization: presentation services....................................................................................... 55
12.1. Users and general constraints and limitations ......................................................................... 55
12.2. Presentation services by the OGC ............................................................................................ 56
12.3. Acknowledgement of 3D modelling and 3D visualization ........................................................ 56
12.4. Cartography of text .................................................................................................................. 56
12.5. Initiatives to a text based indoor navigation application ......................................................... 59
13. Navigation services ............................................................................................................... 60
13.1. Navigation elements of the OGC .............................................................................................. 60
13.2. Mandatory parameters ............................................................................................................ 60
13.3. Generating navigation guidance .............................................................................................. 61
14. Architecture: functional and technical design ........................................................................ 64
Conclusion Part II ....................................................................................................................... 67
PART III - Implementation results ................................................................................................ 68
15. Surveying and databases ....................................................................................................... 69
15.1. Site survey preparation ............................................................................................................ 69
15.1.1. Measure points .................................................................................................................. 69
15.1.2. Measuring and determining suitable fingerprints............................................................. 70
15.1.3. Fingerprinting methodology used in this project .............................................................. 73
15.2. Databases and information storage ......................................................................................... 75
15.2.1. Generalities........................................................................................................................ 75
15.2.2. The database in the application ........................................................................................ 76
16. Location determination algorithms ....................................................................................... 81
16.1. Two methods of location determination being used in this thesis ....................................... 81
16.2. Geocoding and reverse geocoding ....................................................................................... 91
16.3. Location determination and geocoding in Java.................................................................... 91
16.4. Navigation algorithm ........................................................................................................... 93
17. Layout and results ................................................................................................................ 96
17.1. Layout configuration ................................................................................................................ 96
17.2. Testing the application ........................................................................................................... 100
17.2.1. Application validation...................................................................................................... 100
17.2.2. Criteria of location check and test performances............................................................ 100
17.2.3. Indoor Wi-Fi Navigation testing ...................................................................................... 106
17.3. Remarks and conclusions on the application ......................................................................... 108
PART IV - Conclusions ................................................................................................................109
18. Final Conclusion ...................................................................................................................110
19. Reflection: discussion, limitations and recommendations .....................................................112
19.1. Limitations in general ............................................................................................................. 112
19.2. Regarding the application ...................................................................................................... 113
19.3. Regarding the localization methodologies ............................................................................. 113
20. Additional remarks on privacy and legal issues .....................................................................115
20.1. Personal information .............................................................................................................. 115
20.2. MAC as personal information................................................................................................. 115
20.3. Coverage ................................................................................................................................. 115
References ................................................................................................................................117
Annexes ....................................................................................................................................121
A. Tables and scripts ........................................................................................................................ 122
A.1. Table of overview possible sensor technologies for location based services ....................... 122
A.2. Navigation flows .................................................................................................................. 124
A.3. Database contents ............................................................................................................... 126
MSc Thesis Indoor Navigation with Wi -Fi Fingerprinting |6
A.4. Database samples ................................................................................................................ 128
A.5. Codes for location determination......................................................................................... 130
A.6. Codes for navigation algorithm (Dijkstra’s algorithm)......................................................... 134
B. Background literature ................................................................................................................. 138
B.1. Background information on pedestrian movements and navigation algorithms ................ 138
B.2. Mobile cartography and the user ......................................................................................... 140
B.3. Background information on cartographic rules ................................................................... 142
C. Results ......................................................................................................................................... 144
C.1. Scan results for location estimation ..................................................................................... 144
C.2. Scan results for route prediction .......................................................................................... 145
List of figures and tables
Main thesis
Figure 1. Accuracy and precision ........................................................................................................... 12
Figure 2. Interrelation of the chapters .................................................................................................. 18
Figure 3. GPS positioning on the road and in a maze to simulate indoor navigation ........................... 25
Figure 4. Normal versus Spread Spectrum ............................................................................................ 28
Figure 5. DSSS messaging ...................................................................................................................... 29
Figure 6. Channels of the 2.4 GHz for OFDM ........................................................................................ 30
Figure 7. Frames of a 802.11 signal ....................................................................................................... 30
Figure 8. 802.11 Frames and its coding ................................................................................................. 31
Figure 9. Allocation methods ................................................................................................................ 34
Figure 10. An example of fingerprinting................................................................................................ 35
Figure 11. Position (35.629535, 139.879818): Bayside Station, Tokyo, Japan ...................................... 42
Figure 12. User profiles defined by ETSI (2010) .................................................................................... 47
Figure 13. Nodes and edges representation (topological dual space) .................................................. 53
Figure 14. Fingerprinting objects and navigation .................................................................................. 54
Figure 15. Example text output using text visualization and variances ................................................ 59
Figure 16. Routing, guidance, orientation ............................................................................................. 62
Figure 17. Architecture of this indoor navigation application .............................................................. 64
Figure 18. Functional and technical design of the application.............................................................. 66
Figure 19. Screenshot of the Eclipse Environment ............................................................................... 69
Figure 20. Photo of wing 2 and wing 3 of the OTB building .................................................................. 70
Figure 21. Screenshot of reconfigured RSSI scan .................................................................................. 71
Figure 22. Graphical representation of indoor model (graph) with correct orientation ...................... 72
Figure 23. Graph layer on photo ........................................................................................................... 73
Figure 24. Time aspect of fingerprinting ............................................................................................... 73
Figure 25. Temporal aspect of fingerprinting (bar chart) for given location FID=3210 ........................ 74
Figure 26. Database relations ................................................................................................................ 75
Figure 27. Matching live fingerprints with recorded fingerprints using snapping tolerances .............. 83
Figure 28. Flow scheme of location determination for count method with search space ................... 86
Figure 29. Visual representation of count method upon walking around (desirable simulation) ........ 87
Figure 30. Exclusion of access points to narrow down search space .................................................... 90
Figure 31. Main Screen: input and AutoCompleteTextView ................................................................. 96
Figure 32. Settings screen ..................................................................................................................... 99
Figure 33. Navigation screen: output of a route ................................................................................... 99
Figure 34. New database method: using coverage areas instead of storing best signals................... 101
MSc Thesis Indoor Navigation with Wi -Fi Fingerprinting |7
Figure 35. First scan results - grouped by match, specified by location (fingerprint) node, 12 nodes,
wing 2+3. ............................................................................................................................................. 102
Figure 36. First scan results - grouped by match, specified by location - wing 3 nodes ..................... 103
Figure 37. Second scan results – grouped by match, specified by location - wing 3 nodes ............... 105
Table 1. Overview selection of possible techniques for localization .................................................... 27
Table 2. Properties of IEEE802.11 protocols ......................................................................................... 28
Table 3. SWOT analysis of Wi-Fi and Wi-Fi Fingerprinting .................................................................... 32
Table 4. Mobile phone constraints........................................................................................................ 55
Table 5. Parameters Navigation Services .............................................................................................. 61
Table 6. Temporal aspect of fingerprinting (table view) ....................................................................... 74
Table 7. Sample database on signal strengths on two different locations ........................................... 82
Table 8. Sample database on location information .............................................................................. 82
Table 9. Sample database on room information................................................................................... 82
Table 10. Example of received signal strengths .................................................................................... 83
Table 11. Example of determination of range signal strengths ............................................................ 84
Table 12. Example of comparing recorded and received signal strengths ........................................... 84
Table 13. Localization methods: count method versus sum of squares method ................................. 88
Table 14. Outcome of counts and sum of squares ................................................................................ 88
Table 15. Scenario for bad reception for SS method ............................................................................ 89
Table 16. Overview of test nodes for location checks ........................................................................ 101
Table 17. Scan results attempt 1 ......................................................................................................... 103
Table 18. Scan results attempt 1 - Wing 3 only ................................................................................... 104
Table 19. Scan results attempt 2 - Wing 3 only ................................................................................... 105
Table 20. Results of route confirmation results within wing 3 ........................................................... 107
Box 1. A XML Example for OGC Directory Services ............................................................................... 51
Box 2. Example of utilizing text for recognizable objects ...................................................................... 56
Box 3. Example of SQL statements in XML: creation and population of database table ...................... 76
Box 4. Snippet overview of DatabaseHelper.java: getting the contents from database tables from
XML files ................................................................................................................................................ 77
Box 5. Snippet overview of MainScreen.java: creation of entry fields and actions upon tapping
buttons .................................................................................................................................................. 78
Box 6. Snippet overview of RoomList.java: populating an entry field with a directory service ............ 79
Box 7. Snippet overview of queries for searching for nodes within the search space ......................... 92
Box 8. Snippet overview of RouteCalc.java ........................................................................................... 95
Box 9. XML file of the Main Screen ....................................................................................................... 98
Annexes
Figure AP. 1. Shortest paths in buildings ............................................................................................. 139
Figure AP. 2. Reichenbacher's Context Model for Mobile Cartography ............................................. 140
Figure AP. 3. Bertin's geographic variables ......................................................................................... 142
Table AP. 1. Extended overview of sensor techniques used for allocation of receiver. ..................... 123
Table AP. 2. OGC Navigation flows ...................................................................................................... 125
Table AP. 3. Database: fingerprint locations ....................................................................................... 126
Table AP. 4. Database: connectivity .................................................................................................... 126
Table AP. 5. Database: room information ........................................................................................... 126
Table AP. 6. Database: POI information .............................................................................................. 127
Table AP. 7. Relation of graphic variables and visual isolation ........................................................... 143
Table AP. 8. Information perception of cartographic information ..................................................... 143
Table AP. 9. Scan results for all locations - best 10 macs in vicinity ................................................... 144
MSc Thesis Indoor Navigation with Wi -Fi Fingerprinting |8
Table AP. 10. Scan results for wing 3 locations - only routers from same floor registered ................ 144
Table AP. 11. Scan results for 3 series of 5 different routes in wing 3 ................................................ 145
Box AP. 1. Sample database for table_location .................................................................................. 128
Box AP. 2. Sample database for table_mac ......................................................................................... 128
Box AP. 3. Sample database for table_connectivity ............................................................................ 129
Box AP. 4. Sample database for table_rooms ..................................................................................... 129
Box AP. 5. Sample database for table_poi (unused) ........................................................................... 129
Box AP. 6. Overview of code for location determination (MainScreen.java and Navigation.java) .... 133
Box AP. 7. Code for DijkstraAlgorithm.java ......................................................................................... 135
Box AP. 8. Code for Graph.java ........................................................................................................... 135
Box AP. 9. Code for Vertex.java ........................................................................................................... 136
Box AP. 10. Code for Edge.java ........................................................................................................... 137
MSc Thesis Indoor Navigation with Wi -Fi Fingerprinting |9
Summary
One possible way to determine a position in the indoor environment is by using (the reflection and
absorption of) signal strengths, also called multipath, from Wi-Fi routers. By recording the signal
strengths in dBm over time at a certain position, a fingerprint can be created. These fingerprints can
be unique in the sense that it is possible to distinguish positions. However, multipath are both a bless
and a burden: the set of signal strengths are made because of the reflection and absorption, but at
the same time, they are prone to changes in time, since at a position, signals can be received directly
and indirectly. In this sense, it is not possible to record unique fingerprints at each single spot
consistently, since the differences would not be significant to distinguish.
In this thesis, research has been done on how to make use of fingerprints in a smartphone based
indoor navigation application, which uses fingerprints as a positioning technique, and which does not
uses geometric maps or coordinates as a guideline. Since the fingerprints are directly translated to
relative locations, such as “Room 2.200”, the method of positioning has been changed to localization
instead of positioning. The focus is on the confirmation of the appropriate location regarding the
requested route in topological sense: a user traverses a route constructed by nodes and edges
without any explicit coordinates. The framework of the Open Geospatial Consortium’s (OGC) Open
Location Services (OpenLS) has been used a guideline to set up a prototype. Parameters cannot be
used, since those involve maps and WGS84 coordinates, while the project aims at the non-geometric
features.
Signal strengths from a set of Media Access Controls (MAC) which transmits the Wi-Fi signals, have
been recorded at 40 locations at the OTB research centre in Delft. In the application a scan will be
compared to the recorded signal strengths and if there is a match, the location will be returned. This
location is being matched with the requested route and it is possible to tell whether the user is at the
correct location or not. As with the nature of multipath, one has to take into account the matching
signal has to be within a range of the recorded signal: if the recorded signal was -65 dBm, sufficient
search space has to be created around this signal. In this project, a search space of +4 dBm and -4
dBm has been proved sufficient. This means that a live signal of -67 dBm would provide a matching
location, since this is within range of the original recorded signal (-69 < RSSI < -61).
The strength of this methodology is that it is easy to maintain a database with the recorded signals,
although the recording, or site surveying, is rather time consuming, and the MACs can be easily
replaced or removed. Multipath itself is inherent to changes in the time domain, which results in
delays upon live tracking. As such, the fingerprint database requires a frequent update. Another
strength is that there is no need for additional data processing: since there are no maps needed,
there is a fast processing, as a certain degree of map matching does not apply here.
From the results, it was visible that it is highly recommended to store only mac addresses with their
signal strengths, which are in the same physical space as the location fingerprint, as it improved the
returning results on location predictions. Moreover, upon surveying, it is recommended to use the
same class of receiver as the application is to be installed. Scan results obtained from high-end
receivers can be considered as incompatible with scan results obtained from a medium-end receiver
such as a smartphone. It is most likely the survey results from the latter will match the live results. As
a result, a disadvantage is that separate fingerprint databases needs to be recorded for specific
devices.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 10
Further improvements are suggested regarding the translation (geocoding) of the fingerprint
locations to absolute positions based on coordinates (allowing graphs or maps to be displayed), using
extensions of smart navigation (as now, neutral directions have been used) and by distinguishing
certain user properties (such as navigating through restricted areas, navigation for disabled users, or
time of arrival estimations). In the developed prototype, room has been reserved for extensions such
as these. The database for MAC and locations can also be placed on a server, rather than in an
inherent database included in the application, for up scaling towards a large coverage area.
Issues will remain regarding the privacy of MAC addresses, since it is not allowed to record MAC
addresses along with location information in several countries, as it is possible to trace identifiable
persons, which contradicts legalities.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 11
List of abbreviations and acronyms
3G
A-GPS
AOA
AP
API
BSSID
Cell-ID
dBm
DSSS
FHSS
GPS
IEEE
IEEE 802.11
IR
IRDA
LAN
LOS
MAC
MIMO
OFDM
OGC
OpenLS
POI
RFID
RSS(I)
SQL
SS
SSID
TOA
UMTS
UWB
Wi-Fi
WLAN
XML
Third generation mobile telecommunication
Assisted GPS (see GPS)
Angle of Arrival
Access Point
Application Programming Interface
Basic Service Set Identifier, same as MAC in value
Cell Identification
Decibel miliwatt
Direct Sequence Spread Spectrum
Frequency Hopping Spread Spectrum
Global Positioning System
Institute of Electrical and Electronics Engineers
See WLAN
Infrared
Infrared Data Association
Local Area Network
Line of Sight
Medium/Media Access Control, same as SSID in value
Multiple Input Multiple Output
Orthogonal Frequency Division Multiplexing
Open Geospatial Consortium
Open Location Services
Point Of Interest
Radio Frequency Identifier
Received Signal Strength (Indication)
Structured Query Language
Sum of Squares
Service Set Identifier
Time of Arrival
Universal Mobile Telecommunications System
Ultra Wide Band
Wireless Fidelity
Wireless LAN (see LAN)
eXtendible Markup Language
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 12
Terminology
Positioning vs. localization and accuracy vs. precision
A distinction can be made between positioning and localization. Haenselmann (2005) uses the
following definitions:


Positioning is the determination of a location with absolute (global) coordinates, such as
(35.629535, 139.879818).
Localization is the determination of a location with relative coordinates, such as “Room 5.12”,
and depends on the scale as such, e.g. a large room could be subdivided in smaller particles.
In this research, these two words are often mixed in use. When they are used in the context of
describing a sensor technique, both refer to the same style of location determination. However, in
the event these terms are used as a noun themselves, they are referring to their respective trait. In
this perspective, the use of accuracy and precision, in combination with these techniques has also a
distinction, as defined by the BIPM (2008):


“*Accuracy is the degree of] closeness of agreement between a measured quantity value and a
true quantity value *…+. “
“*Precision is the degree of] closeness of agreement between indications or measured quantity
values obtained by replicate measurements on the same or similar objects under specified
conditions.”
A high accuracy means that measurement points (in this case location points) are close to the aimed
target area. A high precision means that the measurement points together are close to each other
This is displayed in Figure 1, where the purple dots represents the allocation of the GPS receiver, and
the red dot represents the actual position.
This research is mainly focused on the accuracy being used, rather than the precision of the
measurements of the signals.
Figure 1. Accuracy and precision
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 13
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 14
Building geometry
In this thesis the combination of words building geometry is consistently being used. This can refer
to:


Creating geometry (building as constructing)
The geometric features of a construction (the dimensions of a construction)
In this thesis, the second definition is being used. Building geometry refers to the dimension of the
building as information and not to the development of geometric features itself.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 15
PART 0 - Introduction
Contents



1. Introduction
2. Goal, objectives and scope/limitations
3. Research questions
This thesis has been subdivided in five parts: part 0 with the general context of introduction, goal and
research questions; part I with the main theory, part II with application oriented theory, part III with
the application implementation and empirical research, and part IV in which the thesis will be
concluded.
In this part 0, this report starts with a topic introduction and the structure of the thesis in chapter 1.
Chapter 2 continues with a description of the goal, objectives and the scope (and limitations) of the
research. Chapter 3 follows with the research questions.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 16
1. Introduction
In this chapter the research topic (1.1) and the contents (1.2) of this thesis are described.
1.1. The research topic
Recent studies (Kolodziej & Hjelm, 2006; Zhang et al., 2010; Dovis et al., 2010) have shown the
market in outdoor location based services (LBS) has increased over the past decade. With the use of
a variety of sensor techniques, such as the Global Positioning System (GPS), BlueTooth, IEEE 802.11
Wireless Local Area Network (WLAN, also called Wi-Fi) and the Universal Mobile Telecommunications
System (UMTS) and its generations in data transfer (e.g. 3G & 4G), software equipped mobile devices
are capable to request and display vital on-the-fly information. GPS is the prime technology being
used for these applications. Examples are the car navigation, in the form of TomTom or Garmin
installed on independent devices and Google Navigation on cell phones, store locations requests
(Albert Heins Appie), or friend finders (Google Latitude).
LBS allows on-the-fly navigation and service information, on-the-spot advertisement and
notifications possibilities and other location related information, but are primarily outdoor, while the
indoor possibilities have been neglected somehow (Manodham et al., 2008, p. 296). One can think
of navigating through a large stadium, a convention centre, an office complex or a railway station.
But the application of indoor location information can also be used for robotics, for example. Indoor
LBS especially proves its worth in emergency rescue or incident management, where high position
accuracy is needed under difficult circumstances
However, GPS services have their limitations, which include a low accuracy in densely built
environments and no available services at all inside buildings. Indeed, there has been, and still is,
ongoing research about how these techniques can be improved (Kolodziej & Hjelm, 2006, Zhang et
al., 2010; Yim et al., 2010; Dawes & Chin, 2010; Woo et al., 2011, to name just a few). This research
also focuses on the indoor navigation.
Despite the constraints of the current techniques and applications, existing sensor technologies and
existing software can be enhanced to cater these difficulties. At the moment, Wi-Fi technology seems
to be the most promising technique to detect indoor locations, although it has some weaknesses
regarding its positioning accuracy. The question is whether it is possible to develop an indoor
navigation system, which makes use of existing Wi-Fi technology. Furthermore, one has to deal with
the target user (user profiles), to determine what support of applications and/or services is desirable,
as this affects the general design of a service. As with sensor technologies, one must also take the
objects and materials in account, as signal penetration behaves different for plywood than for
concrete bunkers.
This research concerns about 1) the Wi-Fi fingerprinting technique and 2) the indoor navigation itself.
From the latter, the focus is on how to use fingerprints to confirm where the user is, and whether the
user is still on track in the context of the requested route towards the target, as this confirmation is
part of navigation. Although various studies have dealt with Wi-Fi allocation processing as stated in
Ibrahim & Ibrahim (2010), the novelty of this research that it will be strictly focused on the aspect of
certainty and confirmation for civilian use.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 17
The end product is a prototype for indoor navigation to be used on a smartphone, in particularly on
an Android based device. The prototype can be used to navigate from one place in a building, to
another place in a building, without having knowledge of the complex.
1.2. Contents of this thesis
First the objectives will be explained in chapter 2. Following these objectives, the research questions
will be discussed in chapter 3. The content body of this thesis is divided into three parts.
In part I, the theoretical framework and background will be described in a full literature review. In
chapter 4, some explanation about possible sensor techniques is provided. It also elaborates more on
the principles of IEEE 802.11 (Wi-Fi) technology. Chapter 5 enhances this with positioning techniques,
including fingerprinting. Chapter 6 deals with the navigation context, which forms another
fundamental part of this research. Chapter 7 continues with the Open Location Standards (OpenLS)
of the Open Geospatial Consortium (OGC), which can be used as a framework to develop a
prototype.
Part II concerns with the application of the theoretical frame and is thus a mixture of both theory and
practical issues and relies largely on a literature review, but allows room for ideas of the developing
application. Chapter 8 deals with the application requirements, by dealing what is desirable: this is
the part where the user preferences or user profiles will be stressed. Chapter 9 continues with
connecting the sensor technique with an application, in which the OGC on OpenLS calls this gateway
servicing. Chapter 10 deals with the implementation of navigational services, including algorithms
and the use of OGC/OpenLS, in particular the directory and geocoding services. Chapter 11 follows
with the modelling of the indoor environment. Chapter 12 proceeds with the output presentation of
the chapters 8-11 and emphasizes on the visualization. Chapter 13 combines the previous chapters
on the question how the navigation application can be constructed. Chapter 14 described how the
architecture looks like.
Part III emphasizes on the results and provides detail on the prototype being exploited, which will be
discussed in the chapters 15 (surveying methodology and results as well as database issues), 16
(localization methodologies) and 17 (layout and results).
Finally, part IV closes the thesis with the final conclusion in chapter 18. A reflection in the form of
discussion on limitations and recommendations are to be found in chapter 19. In addition, remarks
on privacy and legal issues form an epilogue in chapter 20.
Figure 2 displays how the chapters interrelate to each other.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 18
Figure 2. Interrelation of the chapters
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 19
2. Goal, objectives and scope/limitations
The goal is to investigate whether it is possible to use Wi-Fi fingerprinting technology as a positioning
technique for indoor navigation, which relies on the framework of the Open Location Services
standards, and which does not necessarily utilizes geometric features. The focus is on assessing the
use of fingerprints as a confirmation tool, which will be incorporated in an application prototype.
The development of a prototype itself is subject to various factors which have to be taken into
account, such as building material which will affect the Wi-Fi signal strengths and thus the positioning
of the user. User profiles – described by ETSI (2010) as a “total set of user-related information,
preferences, rules and settings” influences the development of a service as well.
Thus, the following objectives have been set for this project:
Regarding the theoretical context, it must become clear…




Why Wi-Fi technology is being used, and why not other technologies.
What infrastructural requirements are needed to develop this.
What the target user and goal of the application should be.
How it can be possible to build a system.
Regarding the application development, an application must be build, which…



Its framework shall be based on the OGC Standards for Open Location Services (OpenLS).
Suits the user profile and goal of the application (including the end visualization and style of
interaction).
Includes a method to actively utilize the route confirmation aspect of navigation.
Regarding the service deployment


The application will have to be tested and demonstrated on a mobile device, to check if the
positioning accuracy is acceptable.
A site survey and positioning model will be carried out.
The project is carried out in the OTB building in Delft. In particularly, only the ground and first floor
will be used, to determine if it is possible to navigate in the vertical dimension, and to make sure the
prototype will work in a relative small area.
The data and software being used will be dependent on the target area. However, further research
will be conducted about the database, server and spatial referencing system at a later stage. Yet, the
proposed platform will be a HTC Desire Z Android 2.2.3 smartphone.
The scope of the project is limited to Wi-Fi localization technology and its application in an indoor
location based service. In particular, the focus is on the navigation side, where the question is how
you can navigate from A to B if A and B are named locations inside a building? This means that
anything else is beyond the scope and will not be discussed, which includes other possible sensor
techniques, economic viability and market deployment, privacy matters and an extensive positioning
accuracy and positioning precision estimation. However, for the latter, an evaluation is proposed, to
verify and demonstrate the usefulness of the created software. This means it is not the intention to
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 20
modify existing accuracy and precision algorithms. Finally, inverse indoor positioning, as in placing
Wi-Fi access points just outside the building to determine the position inside as described by de la
Roche et al. (2010) has been acknowledged as a possible allocation technique, but is regardless
beyond the score of this proposal as research in this field is relatively new and a significant
uncertainty exist in various aspects. More about Wi-Fi functionality is described in chapter 5.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 21
3. Research questions
From the previous chapter, the following main research question has been derived:
Is it possible to develop an indoor navigation service for a mobile platform, based on the use of only
Wi-Fi fingerprinting technology and the framework of the Open Location Services standards, but
without using building geometry?
From the main research question, the following sub research questions can be posed:




How can Wi-Fi be used in mobile indoor navigation?
o This research question addresses the motivation of the chosen localization
technique/approach. There are multiple techniques possible, but upon answering this
question, it will become clear why Wi-Fi technology has been chosen in favour of other
techniques, and how Wi-Fi fingerprinting localization actually works.
o Part of navigation is the regular check whether the user is still at the right place. This
assumes a route is present. Route confirmation is a fundamental aspect in this sense. The
question is how Wi-Fi Fingerprints can handle this.
How can an existing infrastructure be used and modified for indoor navigation?
o Creating a new infrastructure is in the first place considered extremely time consuming
as the thesis research is constraint by time limits it is not desirable to create it.
Therefore, minor modifications to the infrastructure are ought to be more realistic, e.g.
readjusting router access points placements. In the end, it will become clear what
infrastructure is needed, and how the different components are linked together.
o This also includes the use of some sort of reference system. Where traditional outdoor
absolute (x,y,z or φ,λ,h) reference systems are being used, the indoor environment
requires a coordinated based scale of detail, hence the outdoor referencing system is
considered unsuitable. However, a true reference system can be neglected if the
emphasis is on the fingerprints as a confirmation tool. Since each position has a unique
set of fingerprints, one only need to know if the receiver is at the right position as that
has been told.
How can the creation of an application be possible, without using building geometry?
o This question features the inclusion of two problems that in general mobile users will
face, and must be addressed in the creation of the application.
 First, the limited accuracy of the position. Even when the position deviates
significantly from the true position, the question is if this information still can be
used, and how. How can an application be created without the inclusion of
building geometrical features, but solely fingerprints?
 Second, on a mobile device, smart designing is a key factor as well, since the
device has limitations in data transfer and data visualization, due to device
limitations such as screen size and processing time.
What are the benefits of using the framework of Open Location Service standards in an indoor
navigation application?
o It is time consuming to implement an own set of standard procedures. Instead, existing
standards for open location services from the Open Geospatial Consortium will be used
upon designing the application, as it contains a large knowledge and expertise base.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 22

However, the main question implies that a possible prototype will deviate from using the
standards themselves. Therefore only the main aspects are probably a guideline upon
developing a prototype. How does a framework of OpenLS fits into a prototype which
does not follow the parameters themselves? How can these modified for indoor use?
Which types are there? Are there already existing datasets and if yes, how are they being
used.
How can Wi-Fi fingerprinting determine a location which does not utilize geometric features, and
how can it be implemented in an application as such?
o A final question deals with how Wi-Fi fingerprinting can be used as a localization
technique. Upon surveying, one ends up with a comprehensive database, but how
exactly can that information being used to reconvert it to location information?
o As there is no geometry, it is difficult to tell a user to turn left or right as such, since there
is no orientation aspect. Yet, a route which has been calculated from a node-and-edge
based network is still suitable for navigation, although the network does not contain any
coordinates. This means that the importance and usage of a topological network plays a
key role upon answering this research question.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 23
PART I - Theoretical
framework
Contents




4. Sensor techniques
5. Localization techniques and fingerprinting
6. Navigation
7. Open Location Services by the Open Geospatial Consortium
In this part, the pure theoretical contents are described. Chapter four describes several sensor
techniques which can be used. It elaborates on why Wi-Fi is being used, thus attempting to answer
the first research question how can Wi-Fi be used in mobile indoor navigation?
Chapter five continues with the localization techniques, with fingerprinting being the central method.
Next, in chapter six, the navigational issues are considered. General thoughts of navigational
algorithms, as well as the necessary equipment and reference systems are included here, which leads
to an answer of the second research question how can an existing infrastructure be used and
modified for indoor navigation?
Finally, in chapter seven, the use of standards is discussed, in particular the standards as set by the
Open Geospatial Consortium.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 24
4. Sensor techniques
This chapter deals with the sensor techniques. Section 4.1 describes briefly why GPS is not suitable
for the indoor environment. Section 4.2 continues with alternatives for GPS, while section 4.3 picks
up Wi-Fi as a technique to be used in this thesis. Section 4.4 closes with a general overview of Wi-Fi.
4.1. Why GPS cannot be used
At the moment, Global Positioning System (GPS, or more general: Global Navigation Satellite System,
GNSS) and the concept of cell identification in the more mobile telecommunication systems (such as
GSM and 4G) are the most dominant modes of techniques being used to determine the location of a
compatible mobile receiver in the outdoor world. With the use of those signals, services can be
deployed, such as retrieving information where a receiver is located. One of those services is
navigation.
GPS is suitable for the outdoor navigation, as the navigable spaces are wide enough. The accuracy
positioning of GPS is between 6 and 12 meters 95% of the time. The problem is that the reception of
the GPS signals is bad in densely build areas, including the indoor environment. GPS has been
enhanced in accuracy, but the accuracy positioning is still insufficient indoors, as it cannot solve the
multipath interference, which is the reflectance (scattering) of the signal via objects (Manodham et
al., 2008). This is caused by the requirement of the line-of-sight (LOS) of GPS in order to function.
When a user is spending time indoors, the GPS signals will not be able to penetrate through
obstacles. Depending on the material and volume of the object, signals can be completely absorbed,
reflected, or even when it does penetrate the obstacle, the signal possible come out weaker,
(Swangmuang & Krishnamurthy, 2008), since GPS requires line-of-sight. Another issue that has been
mentioned is the high power demand of GPS receivers, and that GPS often behaves unpredictable
regarding its propagation conditions, with the result that the true location of the user deviates
significantly than desired (Dovis et al., 2010). Finally, GPS performs poorly in the vertical position. As
stated above, accuracy between 6 and 12 metres is mostly reached for the outdoor environment,
which is sufficient for its location based services. Indoor, that same accuracy is not wishful, as 1-2
metres is needed, to prevent the user to be allocated on the wrong floor, or in the wrong room. The
3D environment is thus extremely important for indoor location based services.
If a closer look can be taken on the outdoor navigation services, one can see some discrepancies, too.
The user usually finds itself at some distance of the modelled environment. It might be possible the
entire road network is modelled as a large network of nodes and edges with geometry. The
navigation software allocates the user’s location somewhere on the network. The accuracy is
important to provide the user with a list of directions. An accuracy of 10 metres would be sufficient
in the context of car navigation. The allocation of the user’s position on the network meets the
requirement, as displayed in the left half of Figure 3. The GPS of the user will not be allocated
improperly on the road to the south. However in a maze with high walls, and narrow paths, a higher
accuracy than 6 meters is needed, in order to navigate properly. If the sensor and the service do not
meet the desired accuracy, both might not be usable at all, as displayed in the right half of Figure 3,
where the user will be prompted to turn right, while there is no turn at all. One could compare this
simulation with an indoor environment. The aim is to ensure at least an accuracy of 2 metres to
navigate the user.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 25
GPS is thus unsuitable for the indoor environment. In the next chapter, other sensor techniques are
being described, with the emphasis on Wi-Fi technology.
Figure 3. GPS positioning on the road and in a maze to simulate indoor navigation
4.2. Other sensor techniques compared
Indoor signal measurement is difficult, since the majority of the signals hardly (or cannot) penetrate
through objects, but are being reflected instead. Other factors that decrease the reliability and
accuracy includes the hardware of the transmitter, construction objects such as walls and doors,
including their material, variations in temperature and humidity that might disturb signal strength,
access constraints due to opening hours or security zones, and human activity (Mehmood et al.,
2010, p. 1220; Lim et al., 2010, p. 406).
Kolodziej & Hjelm (2006), Manodham et al. (2008), Lim et al. (2010) and Woo et al. (2011) outlines a
variety of sensor technologies that might be used for localization. Amongst these, Infrared,
Ultrasound, Radio Frequency Identification (RFID), IEEE 802.11 (Wi-Fi) and Ultra Wide Band (UWB)
are commonly named techniques. Infrared (IR; and the extended Infrared Data Association, IRDA)
technology allows easy positioning due to ID matching of devices, but due to its short range and lineof-sight requirement, it is not a suitable system for localization. Radio Frequency Identification (RFID)
makes use of tags and readers, which allow fast read/write data communication, but there is no real
localization and positioning involved. Ultrasound technology, based on emitting sound pulses, is
using the response time to determine a location, but it cannot countermeasure the problem of
multipath effects and its weakness to environmental noise. Bluetooth is forwarded as a solution,
since it is capable of high speed data transfer, within an acceptable range, but since the entire
network is portable, positioning can be difficult. Ultra Wide Band (UWB) offers fast, low-power and
large capacity data transfer, and has similar traits as both Bluetooth as Wi-Fi. The advantage is that it
is capable of very precise localization (near 15 cm) but it is rather expensive. Finally, The IEEE 802.11
(Wi-Fi) technology of data exchange proves to be the most balanced, as the signal range of around
30-40 meter indoor is acceptable and its economic viability is stable. The technology is still being
developed, with the current state of 802.11n reaching up to an indoor range of about 91 m, and an
outdoor range of 182m (Ibrahim & Ibrahim, 2010). However, the localization accuracy at its current
state is not sufficient and today there is still research to be conduct regarding which filters should be
applied. Another weakness is that in larger buildings a large amount of transmitters needs to be
installed, although the same applies to the other techniques.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 26
Despite of each technology’s shortcomings, Wi-Fi is considered the most attractive alternative to GPS
for indoor usage, as its range is at a fair level (≈ 32 m), and Wi-Fi is widely available in many places,
which saves significantly time and effort in arranging an infrastructure. Nevertheless, it is also
possible to combine the different technologies in one application. For instance, an application that is
capable of toggling between GPS, Bluetooth and Wi-Fi at will. However, it is not the goal of this thesis
doing so, although it might serve as a recommendation for future work.
A brief summary of the technologies and techniques is shown in Table 1. A full overview is found in
the annex, Table AP. 1. The allocation techniques, also known as the positioning techniques, are
further described in chapter 5.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 27
Technique
(A)GPS
Range / accuracy
Accuracy:
6.0 m - 10.0 m
GSM / UMTS
Range:
≈ 35.0 km
Range:
0.7 m – 2.5 m
IR
IRDA
Range:
≈ 2.5 m
RFID
Active
Passive
Ultrasound
Bluetooth
Ultra Wide Band (UWB)
IEEE 802.11 (Wi-Fi)
Range:
≈ 100 m
Range:
1.5 m – 2.0 m
Accuracy:
3.0 cm – 1.0 m
Range:
≈ 100 m
Accuracy:
10 m – 20 m
Range:
≈ 50 m (cap)
Accuracy:
15 cm – 4 m
Range:
≈ 32 m (indoor)
≈ 95 m (outdoor)
Accuracy:
1m–5m
Remarks
+ Low barrier entry
- Slow computation and processing time
- Very susceptible to reflectance and multi-paths
+ Globally available
- Cell-based accuracy
- Short range of detection limits infrastructure
- No penetration of materials / multipath
- Line of sight
- Signal can be disturbed easily
Same method as IR, yet with higher speed
Allocation meth.*
Trilateration
+ High-speed response time
+ Read/write capabilities
- No communication network
- No allocation information
Cell-ID
- No penetration of materials / multipath
- Extremely sensitive to environment
+ High speed data transfer
- Positioning via triangulation (no objects into account)
- Explicit links between devices required
- Its mobility also limits positioning and topology
+ Multipath immunity/high precision)
+ High speed data (nearly 10xWi-Fi)
- Not everywhere legal
- Economically expensive
+ Large scale available over the world
+ Economical viable
- High power consumption
- Slightly multipath susceptible
Trilateration
Table 1. Overview selection of possible techniques for localization
Table based on Manodham et al. (2008) and Kolodziej & Hjelm (2006) and Lim et al. (2010). See also Table AP. 1 for a full overview
*The allocation (positioning) methods are further explained in chapter 5.
Cell-ID
Signal Strength
Cell-ID
Cell-ID
Trilateration
Signal Strength
Trilateration
Signal Strength
Trilateration
Signal Strength
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 28
4.3. General background of Wi-Fi
Wi-Fi is the common name for the standard framework of a Wireless Local Area Network (WLAN), as
set up by the 802 committee of the Institute of Electrical and Electronics Engineers. In fact, the
official name is IEEE 802.11. The commonly standards that are being used are the 802.11b, 802.11g
and the 802.11n versions. The contents below are based on the IEEE 802.11-2007 document, which
can be retrieved at their website
http://standards.ieee.org/about/get/802/802.11.html.
4.3.1. Use of spread spectrum
IEEE 802.11 technology makes primarily use of modulation techniques in the 2.4 GHz industrial,
scientific and medical radio band (ISM), or in the lesser used 5.0 GHz band. The radio waves make
use of the spread spectrum modulation technique, which means that the generated carrier waves
are over the full frequency domain, thus wider than the information signals, as displayed in Figure 4.
There is a transmitter, usually a router, and a receiver, usually a computer or mobile device. The
modulation technique for the 802.11 depends on the version being used, as displayed in Table 2.
Figure 4. Normal versus Spread Spectrum
Image derived from Kessler (2006)
802.11 Frequency
(GHz)
a
b
g
n
2.4
3.7, 5.0
2.4
2.4
2.4, 5.0
Bandwidth
(MHz)
Data rate per stream (Mbit/s)
20
20
20
20
20
40
1, 2
6, 9, 12, 18, 24, 36, 48, 54
5.5, 11
6, 9, 12, 18, 24, 36, 48, 54
7.2, 14.4, 21.7, 28.9, 43.3, 57.8, 72.2
15, 30, 45, 60, 90, 120, 135, 150
Allowable
MIMO
streams
1
1
1
1
4
4
Modulation
technique
DSSS, FHSS
OFDM
DSSS
OFDM, DSSS
OFDM
Table 2. Properties of IEEE802.11 protocols
Source: IEEE (2011) and Wi-Fi Alliance (2009)
The original 802.11 and subsequent 802.11a are no longer in use today. The original 802.11 made
use of Infrared technology, the Frequency Hopping Spread Spectrum (FHSS) or the Direct Sequence
Spread Spectrum (DSSS), but was rather slow due to its low data rate. The 802.11a has become
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 29
obsolete as well, because the signals of the Orthogonal Frequency Division Multiplexing (OFDM)
modulation technique at the frequency of 3.7 GHz and 5.0 GHz could not penetrate well through
walls, despite their higher frequency and their higher data stream. The 802.11b and the 802.11g are
nowadays commonly used protocols, again operating at the 2.4 GHz frequency, knowing that
interference with appliances in the same frequency (such as microwaves and Bluetooth) still occurs.
Using the DSSS at this frequency ensured a reliable working. The main differences between the b and
the g protocols are the levels of data rates, and the modulation techniques of OFDM in addition (yet
instead of the 3.7 or 5.0 GHz level, at the 2.4 GHz level). Finally, the 802.11n was based on prior
protocols, and by allowing multiple input multiple-output antennas (MIMO) (Wi-Fi Alliance, 2009;
IEEE, 2011). This means that more than one transmitters and receivers are being used to increase
capacity.
The Frequency Hopping Spread Spectrum (FHSS) is the transmitting of radio wave signals that
constantly change between the frequencies. For the originally IEEE 802.11, this was less suitable,
even though the security is rather high, because it is hard to intercept these.
Both the 802.11b and 802.11g make use of the DSSS modulation technique, with 802.11g also
capable of using the OFDM technique. The 802.11n explicitly uses the OFDM, also at the 2.4 GHz.
In DSSS, waves are being send, a pseudo noise (PN) code symbol (a chip) is being transmitted, which
is shorter and faster than an information bit. The data transmissions are being multiplied with a noise
transmission, which is a sequence of 1 and -1 values, at a frequency higher than the original signal. A
receiver can use the same PN sequence to reconstruct the data signals. This is displayed in Figure 5.
Figure 5. DSSS messaging
Image derived from Kessler (2006)
OFDM for 802.11 divides the 2.4 GHz band in 14 channels, each with a width of about 20-22 MHz
wide, and 5 MHz apart. Not all channels are available (legal) in each country, such as the 14th
channel, which is only available in Japan. In Figure 6, an example is provided. Within that channel,
conventional data is being sent. The advantage of this technique is that it can switch to other
channels in case problems are detected that interferes with the signal, such as Bluetooth or
microwaves due signal loss despite being closely located to the access point. Most importantly it is
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 30
possible to use overlapping channels for signal transmittance, countering selective fading unlike the
case being for a single channel (Intini, 2000).
Figure 6. Channels of the 2.4 GHz for OFDM
Image derived from Intini (2000)
4.3.2. Frames / layers
The signals that 802.11 broadcast, contain frames, or also called layers. Each frame contains a
Multiple Access Control (MAC) address. In combination with the received signal strength, expressed
in dBm, it is possible to estimate how close an access point is. The IEEE 802.11 documents specifies in
great detail what the wide array of the exact contents. The Wi-Fi Virtual Laboratory led by Thomas
Sturgeon (2011) graphically explains briefly what a signal contains, as displayed in Figure 7; while
Intini (2000) explicitly visualize the coding in Figure 8.
For Wi-Fi Fingerprinting, the only information that is required is the information in the management
frames, since the signals are only being used to determine a location, and not to transfer data.
Fingerprinting is described in the next chapter, among other techniques.
Figure 7. Frames of a 802.11 signal
Screenshot taken from the Wi-Fi Virtual Laboratory (Sturgeon et al., 2011)
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 31
Figure 8. 802.11 Frames and its coding
Source: Intini (2000)
4.4. General strengths and weaknesses
Wi-Fi system is low-cost and low-entry: Wi-Fi access points are nowadays widely available. The main
strength is that it is able to partly penetrate through objects. However, some weaknesses are the
security (easy to intercept, unless further encrypted) and interference with other applications in the
same frequency, such as Bluetooth and microwaves, although this depends on the channel being
used in the case of OFDM. Limited data bandwidth is considered another limitation: Lim et al. (2010)
for example noted the decrease of data transfer speed and bandwidth during office hours. In Table 3,
a summation of the SWOT analysis can be found.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 32
SWOT
Strengths
Aspect
1. Low-cost and low-entry
2. Able to penetrate walls in where
GPS fails
Weaknesses
3. In available spaces, fairly good
available signal strengths
4. Specific location fingerprints
available.
5. Susceptible to variations in signal
strength due to time
6. Earthbound based, requires more
infrastructure
7. Multi-path influenced by present
objects
8. Might interfere with other
appliances in the 2.4 GHz ISM
9. Site surveying and registering
time consuming
10. MAC address related – prone to
changes
Opportunities
Threats
11. Security is weak
12. Speed decreases with traffic
13. Fingerprinting does not require
geometric surveys
14. Fingerprinting only necessary at
selected places
15. Bluetooth (BT) or Ultra Wideband
(UWB) might overtake Wi-Fi for
positioning
16. Privacy related elements might
block registration of MACs and
thus fingerprinting
Table 3. SWOT analysis of Wi-Fi and Wi-Fi Fingerprinting
Remark
1. Widely available at affordable
prices
2. Although effect is depleted
after thick layers, such as thick
concrete walls
3. Due to multi-path, good signal
differentiation; up to 100m
4. Coverage of entire building, if
access points are well placed
5. A recorded fingerprint cannot
exactly be reproduced
6. For fingerprinting, more access
points are needed, unlike
satellite based
7. The more objects there are, the
more differentiated the
fingerprints will be
8. For example Bluetooth and
microwaves
9. Must be repeated for interior
and movement changes, and for
each building
10. If system fails, or when access
point fails, MAC cannot be
used.
11. Only applicable to data transfer
12. Only applicable to data transfer
13. Time and effort can be saved on
mapping
14. It is not necessary to measure
every n meter; only at places
with important topological
meaning or at least where
fingerprints look different.
15. BT rather stable, UWB powerful
in better catering multi-path +
better range.
16. Issue of registration of private
MACs – limitation to public
buildings only
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 33
5. Positioning techniques
In the various literature studies, the word localization is being used to determine a position.
However, in this research the term allocation (method) is being preferred, since it does not explicitly
refer to an absolute or relative location, unless stated otherwise. See also the terminology at the
beginning of this thesis research.
5.1. Five techniques
Different methodologies based on signal metrics can be used to allocate devices. Although there are
more allocation techniques available, Woo et al. (2011), Zhang et al. (2010) and Kolodziej & Hjelm
(2006) distinguished five main methods, using: A) Cell Identification, B) Angle of Arrival
(triangulation), C) Time of Arrival (trilateration), D) Time Difference of Arrival and E) Received Signal
Strength Categories, also called Fingerprinting.
A) With Cell Identification (Cell-ID), transmitters (or: access points) are dividing an area in tiles, or
cells. Within those cells, the receiver can be detected. However, this method is imprecise,
because it is hard to determine where in the cell the receiver can be spotted.
B) The Angle of Arrival (AOA) determines the position of the user by measuring the angle towards
the receiver from the transmitter (Woo et al., 2011). The transmitters must be capable of
calculating such information. This can be done with directional antennae. Yet, the method is
unreliable, since it is prone to multi-paths, plus it requires the line of sight to detect the receiver,
which is hard to counter in the harsh indoor environment (Zhang et al., 2010; Kolodziej & Hjelm,
2006). In the extension of the AOA, triangulation can be done to map an entire region, based on
the angles.
C) Time of Arrival (TOA) measures the exact distance by using the travel time of the signal from the
transmitter to the receiver. Using the equation R = time x speed, where speed is a constant, only
time needs to be measured to determine the exact location R (Zhang et al., 2010). However, in
order to get more accurate results, synchronization of the receivers is needed.
D) TOA requires synchronization of both transmitters and receivers, whereas the Time Difference of
Arrival (TDOA) requires synchronization of the receivers (Woo et al., 2011). Both methods belong
to the trilateration group, as it involves the intersection of the radii of the transmitters (Zhang et
al., 2010).
E) Lastly, the Received Signal Strength (RSS), also called fingerprinting, where a radio map is being
created. There is an offline and an online phase, also called the training phase and tracking
phase. The offline phase of fingerprinting is determining the signal characteristics of a given point
and storing it in a database. The online phase is picking up the signal characteristics and
compares it with the database, so that the place can be determined. Ideally, a filter is needed to
increase the positional accuracy and estimations of the locations, even though the reflection and
absorbance of the signals have been taken into account (Kolodziej & Hjelm, 2006; Woo et al.,
2010).
Figure 9 presents an overview of the localization techniques mentioned. Figure 10 presents a more
comprehensive illustration of the offline phase of fingerprinting. In the online phase, the device
attempts to match the signal characteristics with its position and is in this way able to determine
where it is located. In the next sections, the emphasis is on fingerprinting as a technique.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 34
Figure 9. Allocation methods
Illustrations are based on Woo et al. (2011).
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 35
Figure 10. An example of fingerprinting
Based on Woo et al. (2011). This is a pure theoretical illustration; in reality the signal strength distribution heavily depends
on present objects, and overlapping coverage areas. In this figure, only the radio maps of access points A and D are
displayed; in reality all radio maps together form one overview. As can be seen, this placement of Wi-Fi access points might
not be the most efficient one.
5.2. Fingerprinting elaborated
Recent studies (Jan et al., 2010; Lim et al., 2010; Yim, 2008) have shown that fingerprinting is a
fundamental method of indoor allocation, even though the level of accuracy is not at its desired level
of approximately 2 metres, because of its relative robustness to handle multi-path effects. Still,
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 36
various challenges exist. As soon as objects have moved (which includes the presence of peoples),
the fingerprints might alter. In addition, the timestamps of the received signal strength (RSS) is not
constant over the day. In public buildings for instance, the RSS is about 5 dBm less during office hours
(from approximately 9 am to 6 pm), which could influence the fingerprint database, but not too
much (Lim et al., 2010), since the data layer of Wi-Fi is needed. In order to create a fingerprint
database, one must be able to log the signal strengths throughout the entire building, with the help
of data logging services (Dawes & Chin, 2011). That might cause problems in case of sealed off zones.
Various studies have introduced different filters for improving positioning on maps directly. Yim et al.
(2010) used an extended Kalman filter to improve the accuracy of the Wi-Fi positioning during the
offline fingerprinting phase. The Kalman filter basically reduces the noise in locational determination.
However, the use of Kalman filters is disputed by various authors. Chiou et al. (2009) developed an
alternative method for fingerprinting, which is called adaptive positioning but it still makes use of the
Kalman filter. Jan et al. (2010) suggested that a particle filter would be better instead of a Kalman
filter, since a Kalman filter is a linear model, suitable for linear environments. However, the real
world is usually not a linear build environment. Jan et al. (2010) and Dawes & Chin (2011) however
found out that Bayesian map matching algorithms have difficulties in environment with high variable
transmitters and fingerprinting (RSS) values. They also showed that in the same fingerprinting phase,
the accuracy is highest, when nearest neighbour filters are applied. The nearest neighbour assumes
that a specific point that has to be measured is most likely similar, or at least close in value, to the
most nearby points, depending on the weight (Roos et al., 2002).
An extensive offline/training phase will be carried out in the study area, and after that, the
online/tracking phase will be tested for its accuracy. This will be taken into account for the evaluation
of the application, as the user allocation will be used for the navigation. It is not the goal of the
research to fully assess the accuracy of Wi-Fi fingerprinting, however. As a final remark, it is possible
to determine signal strengths in the vertical position as well. If the router is attached to the ceiling,
the received signal strength is a bit weaker near the floor. However, since it is unlikely for persons to
keep their receiver at near 2 metres or at 0.1 metres above the floor, this research will only conduct
a survey at the height of near 1 metre, which should be the height of the users to keep their phone in
their hands. In addition, since it is not the aim of the study to enhance positioning accuracy (instead,
localization is considered), a real map matching does not take place.
Due to high variations in signal strengths caused by multipath, the fingerprint database needs to be
updated frequently, and it might be reproduced for a variety of devices, since a high-end Wi-Fi
receiver perceives signal strengths to be stronger than low-end Wi-Fi receivers.
5.3. Building constructions influencing fingerprinting
As already pointed out, GPS suffers significantly from radio wave propagation (Jan et al., 2010).
Signals are unable to penetrate buildings, and even if they do so, the signal strength significantly
decreases. Although Wi-Fi belongs to the same category of frequency, the UHF (Ultra High
Frequency, which is between 300MHz and 3 GHz), Wi-Fi is better able to cater the propagation as it is
in the higher, longer range of the frequency: Wi-Fi at 2.4 GHz (although 802.11n is capable of using
the 5.0 GHz band) and GPS L1 on 1.5-1.6 GHz and L2 on 1.2 GHz (Ibrahim & Ibrahim, 2010).
This does not mean it can perfectly handle this problem. Not only intensive use of the access point of
Wi-Fi influences the signal strength, and thus fingerprinting, but so does the building constructions.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 37
Klukas et al. (2004) studied the effects of building materials on the UHF signals and found evidence
aluminium tissues and concrete, cinder block objects interferes the signal strength: a loss of 10-15
dBm for aluminium objects and even 20-26 dBm for cinder block objects. This is in contrast of the
plywood and gyprock, where in both cases a power loss of between 0 and 2 dBm was found. A
combination of the latter two materials resulted in moderate power loss of 5-10 dBm. This temporal
aspect has been left out: in theory, an object can receive a signal via the line-of-sight (directly), but
also via reflectance of other objects (indirectly). In this case, the receiver did receive a signal, but it
was either delayed (indirect signal) or it was a weaker signal (direct signal).
In addition, Ibrahim & Ibrahim (2010) found out that the signal strength also sharply drops after a
distance of 20m. This means that, regardless of the building construction and materials used, the
placement of the Wi-Fi access points should be placed at a proper distance from each other, to
ensure the fingerprinting results will not be heavily influenced (prone to alterations) by
misplacements. Combined with the signal propagation as described by Klukas et al. (2004), great care
has to be taken as well for the placement next to specific types of walls or ceilings as such. In Figure
10 for example, the walls of the elevator shaft might contain a thicker coating than normal, resulting
in lower signal strength.
5.4. Using signal strengths to locate
The fingerprinting method itself is not new, as various (MSc) researchers have tried to use various
filters upon location detection directly projected on maps, and at each possible location. Mostly,
attempts have been made to calibrate the position in the indoor space. Lee (2007) addresses the
paradigm of fingerprinting by converting discrete signal strength values to continuous values and
implements various probabilistic filters to increase location determination for constant tracking at
each single space. De Koning (2010), albeit with GSM signals outdoors instead of Wi-Fi signals
indoors, first correlates signal space with Euclidean space and implemented variations on the
weighted nearest neighbour filter, to retrieve location information. Yim et al. (2010) proposes a
similar combination with weighted nearest neighbour filters and trilateration, but concludes that an
extended Kalman filter would provide better results.
Milioris et al. (2010) proposed an analysis in which the distances in terms of signal strength of the
training phase were compared with the tracking phase, and in addition, a multivariate Gaussian filter
has been used to do so. Although attempts have been made to calculate mean errors, the idea of
comparing them with each other itself is sound. Finally, Shum et al. (2011) have chosen a method to
determine positions without filters, also by comparing differences in signal strengths in the training
and tracking phases, but then apply a least sum of square methods, which has been regarded very
promising.
In this thesis project and prototype, a similar approach to Shum et al. (2011) has been chosen, as
with the lack of geometry, no direct map matching algorithms can be implemented, resulting in the
fact that no filters, such as the Kalman or Weighted Nearest Neighbour, can be used. In addition, it is
not the aim to pinpoint exact locations of the entire space, as it is assumed that locations which are
very close to each other have similar fingerprints. This is why locations have been chosen and
recorded, in which the fingerprints can be clearly distinguished from each other, as that was the
assumption of Shum et al. (2011) similarly. However, the text format outputs being provided, can be
geocoded into real world coordinates. To do so, two methods have been proposed: a count method,
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 38
which only counts the number of matches, and a least sum of squares method as proposed by Shum
et al. (2011). These methods have been chosen in the first place because there is no map matching
which implies there is no need in exact positioning, and in the second place to underline the
simplicity of the application. These two methods are described and compared in chapter 16.3.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 39
6. Navigation
In this chapter general theory about navigation and its required infrastructure are described, which
includes the general thoughts about navigation (6.1), as well as the importance of certainty and route
confirmation (6.2), algorithms (6.3), required infrastructure (6.4), and spatial referencing (6.5).
6.1. What is navigation?
“Navigation is the process of monitoring and controlling the movement of a craft or vehicle from one
place of another” – Bowditch.
This statement from Nathaniel Bowditch (2002) is still applicable. Navigation can also be described as
going from source A to destination B, while knowing where you are from, what to expect on your
way, while being in motion from source to destination (or differently said: while the source is moving
towards a fixed destination). It assumes that a start and destination is known, and that one has to go
from start to destination via a certain path. Present-day navigation is no longer solely based on the
position of human interference with the eye line-of-sight such as in watch towers, celestial processes
or dead reckoning. Today, navigation can be fully done without the use of any direct human
interference, such as radio, radar and satellite navigation. As with these modern navigation
techniques, going from A to B itself is no longer the main subject, but the question is rather how one
should going from A to B and how one would retrieve information along the way. The aspects of
monitoring and controlling is very important, with a rapid emerge of applications which utilizes
navigation for information, ranging from tracking and tracing parcels, to chip-tagged products leaving
unauthorized a shop (resulting in ringing alarms at the front gate of that same shop). In short,
navigation requires information on:



Where the user is (start location);
Where the user would like to go (destination), which can be queried from an address, coordinate,
points of interest, or a combination. This is strongly related to the use of user profiles to steer the
end user; and
How the user should go from start to destination: the information of the navigation itself, as well
as information along the way, such as the availability of an ATM, depending on the user
preferences and user profile. This is the path, and can be configured according to the wishes of
the end user, such as the shortest or fastest path.
6.2. Navigation principles
6.2.1. General principles
In order to perform a navigation service, a start point, an end point, the navigable space (the
network), (re)orientation and directives are required. All locational information between start and
end point can be used for this desire, such as attributes of the start and end point which can be
queried, fastest or shortest route (although pedestrian routes do not differ much regarding this
aspect, unlike motorized transport, since the velocity of the walking user is usually constant),
characteristics of the route in-between and its orientation. All of these sets need to be modelled
somehow. Kolodziej & Hjelm (2006), mention the fundamental inclusions of the following
procedures:

Input module for start and destination of user;
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 40



Route database with positions and connections (route network map);
o Topology and structure is coupled to this route database; it is not absolutely necessary to
have coordinates, however.
Route calculation module, using the route database and algorithms such as Dijkstra’s shortest
path algorithms;
Presentation of the route guidance (list of directives).
As soon as a route has been calculated, a list of directives and a map can be constructed. The route
will be dependent on what the profile and wishes of the users are. For instance, the user is a guest in
a large power plant, and cannot enter all zones due to security restrictions. Another example is
disabled person, who can only take the elevator. This will clearly influence the route. Either way, the
application has to keep track in which phase the user is currently at. Kolodziej & Hjelm (2006)
approach this in such way, that each directive in the list has a start and an end; as soon as the end
has reached, the directive will be removed from the list, and the next directive will start. This will
continuously loop, until the destination has been reached.
6.2.2. Certainty and route confirmation
Indoor, the user has a wide array of recognition points available. Since the indoor environment is
very detailed, the user easily passes his destination or important turns, the user cannot constantly
watch his mobile screen for confirmation. The user wants to have a level of certainty.
In navigation, one fundamental technique is to check whether the user is still at the assigned route.
The receiver will pass through certain points and if the position matches the predicted point,
confirmation is being sent to the application the user is on route. If this is not being the case, a
recalculation has to take place. In this thesis, the fingerprints can be considered as such check points.
6.3. Navigation algorithms
The navigation itself is calculated via navigation algorithms, with various optimizations, such as the
shortest, fastest or cheapest route. In order to implement a navigation algorithm, the navigable
space needs to be converted to a graph, compromised with nodes and edge, along with its topology.
In a study of shortest path algorithms, Lu & Lai (2006) confirmed the robustness of Dijkstra’s shortest
path algorithm, even though the shortest routing depends on the structure of the road network. The
paths that are possible in a building are not much different. Even though there are no large physical
entities hindering, such as rivers or mountains, the buildings structure influences the travel time. A
building with two dead ends like structure does not contribute to a short path availability. See also
annex B.1. for background on pedestrian movements.
6.4. Navigation infrastructure
For indoor navigation utilizing Wi-Fi fingerprinting as a positioning technique, it is required there are
multiple access points, or Wi-Fi routers, are available. In the study area (the OTB building in Delft), no
changes will be made to the placements of the routers, as in reality, not always precise coverage
occurs, despite it is desirable to do so. Each access point is designated with a Media Access Control
(MAC) Address, thus having unique transmitters. Those can be used to determine the position.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 41
6.5. Spatial referencing
A spatial referencing system is needed to properly maintain the locations of the various entities.
“Room 329” might be on the third floor in an educational institute, but in a large, single level hotel
complex this is not the case. Positioning can be done absolute (positioning of the user is relative to a
reference position) or relative (user depends his location on other objects nearby) (Kolodziej &
Hjelm, 2006).
In the outdoor navigation, a simple absolute referencing longitude-latitude is sufficient for the user,
although the user would prefer a relative referencing system, such as a street with a house number,
or a system which supports map backgrounds in which the user can recognize objects. Especially in
the indoor environment, the 3D aspects should not be neglected, which is why a position accuracy of
2 metres should is desirable. Multiple objects can be on the same horizontal dimension, yet on a
different vertical dimension. Proper geocoding (x,y,z) is thus necessary for each space. With this
context, cell ID positioning (absolute) as shown in Figure 9A (page 34), is most likely. A coordination
or reference system can be added to each access point, or to each floor, as access points on the same
floor usually have the same z-value. The only thing that is required is a switch between the reference
systems. However, cell ID positioning is rather inaccurate, since it does not tell where exactly the
user within the cell is located. Reference positioning in the form fingerprinting as described in the
previous chapters, is a welcome solution instead, since the accuracy is higher. Even though the
positioning algorithms are not yet at the desired shelf-like level of accuracy, the deviation of two
meters is still considered to be useful, as the user is still able to navigate properly.
Kolodziej & Hjelm (2006), mention that in reality, relative referencing is better understandable. Going
right or left is in a humans mind easy to understand, unlike going west or east, as one usually does
not have a clue what the orientation is. The same applies to the positioning: building A, wing B, room
2.10 (second floor) prevails over position (35.629535, 139.879818; Figure 11) in the human
perception. In this context, it is possible to make a set of geocodes from the components of the body,
so they can be distinguished from each other, but it is required to recode them again for the end
product. This means that spatial referencing will be mostly done relatively.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 42
Figure 11. Position (35.629535, 139.879818): Bayside Station, Tokyo, Japan
Source: Google Maps (2011)
The signal strengths of the fingerprinting can be used as a standalone, relative reference system.
Each fingerprint has its own location. A position can then be assigned to that location. A reference
location (0,0,0) can be assigned to one corner of the building, or perhaps even a little bit outside the
building. The exact locations can be recoded to relative positioning parameters, such as
<building>.<wing>.<floor>.<room>. Since the goal is to allocate a user using this method, we speak of
localization instead of positioning (see Terminology). More about not using an absolute reference
system (and geometry) is being discussed in chapter 12.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 43
7. Building an application using standards
An application will be made on a smartphone, which should be able to connect via 802.11/Wi-Fi in
the first place. The application shall be installed on a HTC Desire Z device, which uses the Android
2.2.1 Operating System. It is possible to build an application according to existing frameworks. In this
thesis research, the aim is to make use of the framework of the Open Location Services (OpenLS)
from the Open Geospatial Consortium (OGC).
The OGC has standards for mobile applications which uses a GeoMobility Server, which is a service
on a server, where applications can request various services and obtain responses which can be used
in applications. The GeoMobility Server is capable to produce maps and routes, as well as accessing
other databases via the Internet. GeoMobility makes use of the following five core services (OGC,
2008a):
1. Directory Services (what do you want to find, and where is it located?)
2. Gateway Services (the localization of the user on the service/network)
3. Location Utility/Geocoding Services (the transformation of a location in readable parameters,
such as from address to longitude/latitude in decimal degree and vice versa)
4. Presentation Services (the visualization of the results on the device)
5. Route/Navigation Services (the navigation from origin to destination)
The parameters and attributes of these services are described in the OGC document OpenGIS
Location Services (OpenLS): Core Services. However, most of these parameters require the presence
of geometric maps and (by default WGS84) coordinates. As maps are not present in the intended
application, only the framework of the GeoMobility Server OpenLS can be useful, in the sense they
provide a structure on which services should be present in an application, and how they are
interrelated: the same services can be utilized, but with different parameters.
For navigational purposes, the OGC (2010) recommends that at least three factors should be present
in the model, and therefore the application:
(1) localization technique and localization infrastructure, which has been discussed in the chapters 5
and 6,
(2) mode of transport (walking; elevators only etc.), by using user profiles (to be discussed in chapter
10) and
(3) navigation constraints, partly resulting from data and modelling issues, as well as spatial
referencing, which has been discussed in general in chapter 7, and will be further discussed in
chapter 14.
From these three components, the navigation engine can be carried out.
In the next part of this thesis research, the Open Location Services are being described in more
detail, along with how it shall be implemented in this project.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 44
Conclusion Part I
How can Wi-Fi be used in mobile indoor navigation?
The IEEE 802.11 technology is synonym to wireless internet, also called WLAN (Wireless Local Area
Network) or Wi-Fi (Wireless Fidelity). The technology makes use of the 2.4 GHz Industrial, Scientific
and Medical band, and uses spread spectrum modulation. In the transmittance of Wi-Fi, multiple
layers are being broadcast. For Wi-Fi fingerprinting, only the header of the power management layer
is necessary, which contains information about the header information, and the signal strength. Wi-Fi
is already widely available and installed in many public and private buildings.
Fingerprinting is the process of position determination, using the received signal strength at one
point. This can be stored for each point, at given heights, although in this research only the height of
about 1 metre above the floor is considered. Once collected in a database, the application can check
at which position the receiver is, and thus position the user. A major drawback of fingerprinting is
that it requires extensive surveying, and that multi-paths works too efficient, since human bodies and
closed doors are intervening the path of the signal strength characteristics.
An important aspect of navigation is the confirmation of the user being at the route. If the user gets
confirmation, the user can continue to follow his route. If not, an alternative route needs to be
calculated. The fingerprints can be considered as check points, in which this thesis will elaborate in
the next chapters.
How can an existing infrastructure be used and modified for indoor navigation?
Indoor navigation is different than outdoors, since pedestrians usually do not vary too much in
speed, making the shortest path also the fastest path. In addition, absolute positions such as (long,
lat) are unsuitable for indoor navigation, as the level of detail is different, although these details still
can be used for maps once the indoor locations are being geocoded. New reference systems with a
(0,0,0) can be used at the background, while they remain unknown to the user. Relative reference
systems are for better use, such as <building>.<wing/section>.<floor>.<room>, in a logical order and
sorted. These differences from outdoor navigation are the most important directives for indoor
navigation. We then speak of localization, as we use a relative position, instead of an absolute
position.
Applications can be built based on the Open Location Standards (OpenLS) set by the Open Geospatial
Consortium (OGC). A navigation service can be developed around directory services, gateway
services, geocoding services, presentation services and navigation services. In this research, the
gateway service is strongly connected with the Wi-Fi technology and fingerprinting positioning.
Information about what is available in the indoor environment can be found in the directory services,
and geospatial queries can be combined with it, as soon as that information is geocoded according to
the geocoding services. The navigation service can thus start if these are combined. The final
visualization or presentation is being described in the presentation services, in which another
important aspect needs to derive from outdoor navigation: indoor navigation requires much more
detail. Hence, a different approach is needed for the final visualization, especially since the
navigation also comes on a small screen, where normal cartographic rules differ from large screen
maps or paper maps.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 45
PART II - Theory and
application
Contents







8. The user involved – user profiles
9. User allocation: gateway services
10. Input: Directory services and geocoding services
11. Modelling and attributes – processing and modelling the indoor space
12. Processing visualization
13. Navigation services as output
14. Architecture
In this second part, theory and application are made concrete. This part focuses on the application of
theory within the to be developed prototype. With the Wi-Fi technology and fingerprinting methods,
a navigational application can be created, with the use of the OGC Framework for Location Services.
First, in chapter 9, one must know what the target user is, what one can do and what one can expect.
As soon as these user requirements are known, one can further develop the application. In chapter
10, the link between the sensor technology and the application is made. In terms of OGC Location
Services, this includes the gateway services. Thus in chapter 11, the directory and the geocoding
services are addressed, to make sure fundamental information can be queried and stored. It is
important where the points of interests are. The uniqueness of this project is that it does not use
building geometry for navigation and the indoor modelling. Chapter 12 emphasizes this aspect.
Chapter 13 continues with the presentation and visualization of the before mentioned contents.
Chapter 14 deals with the core service of navigation itself, and its extensions. Finally, chapter 15
altogether displays the architecture, which includes the functional and technical design, emphasizing
all chapters.
In the end, the research questions how can the creation of an application be possible, without using
geometry? And what are the benefits of using the framework of Open Location Service standards in
an indoor navigation application? can be answered.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 46
8. The user involved
Upon building an application or product, it is important to know what your target user is, and what
the user would like to do with it. For instance, it does not make sense to provide directions along
motor highways if your user is a pedestrian. Creating an application is not any different. This is why
there is a variety of location based services available on the market, which utilizes various sensor
technologies: from trackers to navigators, from compasses to tagging software. It allows applications
to become flexible, as it adapts itself to the user preferences.
8.1. User profiles
It is also possible to install a module in an application that automatically recognizes the user’s habits
and characteristics each time the application is being used. The aim of such automated modification
is to serve the user by supplying information that will most likely pique the interest. Examples are the
customized ads in Google’s Gmail, where advertisement is closely related to the user’s e-mail
contents. This is being described as user profiling. Wu et al. (2009) describes this as “the
characteristics and the relationships of the different users” being used, although the European
Telecommunication Standards Institute (ETSI, 2010) better describes user profiling as a module
“which formally captures the user requirements (preferences) *…+ and deploys those profiles *…+”,
covering explicitly users desires. They are going even further, by defining user profiles as a “total set
of user-related information, preferences, rules and settings, which affects the way in which a user
experiences terminals, devices and services”. User profiles are different from user requirements, as
user requirements involves the desire of user in technical fashion in the development of the product,
rather than the adaptability of user profiles in the actual usage of the product.
8.2. User needs in this project
In this project in particular, the goal is to guide a person in a public building with its mobile device
from the entrance to a given destination.
The aim in this project regarding user profiles, is to develop such module within the application at a
simple level, id est storing the information that is being inserted, so that in future events the
application is able to predict what settings the user most likely wants. From the final definition the
ETSI gave, (1) personal information, (2) human centred preferences, (3) service related information
and preferences (specific information-feed for the service) and (4) device related information and
preferences, all belong to the user profile. Ultimately, it is desirable to create an application that is
able to read device information, making the application suitable for a multitude of platforms, but in
this project this will be limited to the capabilities of the Android 2.2.3. HTC Desire Z device, since it is
the test bed. This results in the focus of user profiling in the user personal information, preferences
and service preferences.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 47
Personal Information
Resources
(describe
things that may
Human Centered
be referenced
Preferences
by rules or by
other settings
e.g. alerting
Service Related
Information and
Preferences
cadences, set
of members of
a group,
pointers to
Device Related
sound files)
Information and
Preferences
Figure 12. User profiles defined by ETSI (2010)
8.2.1. Personal information
In this project, the person is modelled with the following properties (see also Usher & Strawderman,
2010, for extensive research on characteristics on pedestrians):

The user has limited information about the indoor environment.
This means that the output of the application should compromise clear descriptions and overviews
about what to do.
8.2.2. Human centred preferences (requirements)
The following characteristics will co-determine the application regarding the preferences of the user



The user is provided with a simple overview of how the application should work.
The user wishes to get directions to walk towards the destination.
The user wishes to get confirmation whether the user is on route or not.
8.2.3. Service related information and preferences
Combined with the human centred preferences, the service related information and preferences
aims at the possibility to adapt the application to the user. This means that:

The application allows a customizable feature that influences the whole concept.
 Or: that the user can change settings of the application and adapt it to its situation.
8.2.4. Device related information and preferences (requirements)
Finally, the device related information and preferences are coping with hardware constraints. The
application needs to fit accordingly:
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 48




The user uses a HTC Desire Z smartphone model, with Android, API level 8 (2.2 or higher).
The application should run on a platform with fast processing (short processing time).
The application allows a touch screen interface.
The application utilizes Wi-Fi technology and must be enabled.
An example of using user profiles is illustrated by Foerster (2010) about map usage: a map of a
specific region is the same, but the contents being displayed for a geologist is different than for an
urban planner. The map is adapted according to the user. Some background information about maps
being adapted on mobile devices has been provided in Annex B.2.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 49
9. User allocation: gateway services
Differences between indoor and outdoor navigation includes the sensor technique (GPS versus nonGPS) and the scalability: the destination in the outdoor world are usually points of interests, while
indoor navigation could be seen as a subset or complement of the outdoor world, meaning that a
single point of interest (building) possibly contains other points of interests as well. These points of
interests need to be carefully maintained, since a large amount of attributes can be coupled to these
points of interests, which can help the user to navigate inside buildings.
The localization of the user in the indoor environment can be further used and integrated in location
based services, which is being described in this chapter. Further details about the building itself
needs to be mapped, and the rest depends on what the service provider would like to offer (Jan et
al., 2010; see also the remaining chapters of this part II).
9.1. Components of the gateway
The positioning and point placement of the user to the location service (the gateway service) is
perhaps the most challenging part of the project, as different algorithms applies to Wi-Fi positioning.
Kolodziej & Hjelm (2006), propose the following steps need to be done, in order to secure a decent
gateway service:







Site survey performance (signal strength model / fingerprinting database)
Creating the positioning model
Calibration
Access point placement and configuration
Tracking
Maintenance
Rights (due to changes over time)
Since it is the aim of the thesis to make a working prototype with existing access points, only the site
surveying is being discussed. It is not the intention to fine tune the access point placement, since the
main research question is whether it is possible to develop a Wi-Fi fingerprinting application without
using geometric properties.
9.2. Site surveying and positioning
Site surveying is the preparation phase of fingerprinting. A database will be constructed with the
specific fingerprints. It is possible to conduct this phase in the three dimensional space, by measuring
in vertical positions as well. However, since most people hold their phone around the 1 metre height,
only measurements at this corresponding height take place, at each floor.
The temporal aspect has to be acknowledged, since the fingerprint can change over time, because a
receiver can obtain signal strengths directly via a line-of-sight, but also indirectly, via reflectance (the
multipath). The process of measuring has thus to be repeated at various time slots.
The specific fingerprints will be assigned to a “label”, which actually includes other information about
what is present at that place, such as recognizable objects (plants, pillars, inboxes, and signs). This
process is similar as illustrated in Figure 10 (page 35). In this thesis, position is considered synonym
for a fingerprint, since the fingerprint can contain already information about its environment. Real
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 50
map matching and geometry are not required as such. It is therefore not possible to create a
positioning model. In chapter 11.2, more background is given on the non-usage of geometry.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 51
10. Directory and geocoding services
The reason why these two services are in one chapter is that they are closely related to each other, in
the sense they are both concern information about data: the directory services provides information
about a place or object, whereas geocoding services provides information about what the location is.
10.1. Directory services
The OGC (2008a) describes a directory service as [a] service [providing] subscribers with access to an
online directory to find the nearest or a specific place, product or service. This implies that a user can
use a search engine to find something with or without explicit location specifications. Users might
find a place, product or service by providing specific or explicit properties, which is called pinpointing.
Alternative, they can use spatial queries to find the nearest place, product or service, called
proximity. Both pinpointing as proximity require the following minimal information:


A property of the point of interest, such as a unique id, name, description, absolute location
The relative location information to the point of interest, such as nearest to me, within 500
metres.
The OGC (2008a) provides specific directory request parameters. The only mandatory parameters
include (although this depends on the nature of the query, in which other values can be stated as
mandatory):



POIproperties, in which a listing is provided, that can be used to do the search (keywords)
Name, in which characteristic values are stored, such as the reference object, the name and
the type.
Value, in which the information can be reflected to the user.
An example is illustrated in Box 1, in which the question is asked where the nearest copy machine is
being searched, while making use of a relative location statement.
Where is the nearest copy machine of where I am (Room 1.002)?
<DirectoryRequest>
<POILocation>
<Nearest>
<POI ID="1">
<POIAttributeList>
<POIInfolist>
<POIinfo name="POI Name" value="Room 1.002">
</POIInfolist>
</POIAttributeList>
</POI>
</Nearest>
</POILocation>
<POIProperties directoryType="Campus">
<POIProperty name="type" value="Machine">
<POIProperty name="subType" value="Copy">
</POIProperties>
</DirectoryRequest>
Box 1. A XML Example for OGC Directory Services
The provided output is list of points of interests with the distances.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 52
This assumes that upon developing the application, at least a database with retrievable information
for the interest of the user should be populated.
10.2. Geocoding services
In order to make locational queries possible, it is necessary the various points of interests needs to
have information on their position, also called geocoding. The other way around is called reverse
geocoding. The geocoding is usually carried out on the background, since the services need to know
exactly where the point of interest is located, and the user usually does not know on which absolute
position he is. As soon as the point of interest has been traced and located in the absolute reference
system, it will be translated back to the user in named locations, such as Aisle 1.
For geocoding services, the following parameters are mandatory:

Address, which can be inserted the returning value would be a position (X,Y,Z)
For reversed geocoding services, the following parameters are mandatory:

Position (X,Y,Z), which is the starting position of the user which can be inserted, in terms of
absolute positions.  the returning value would be an address (street, number)
In our iindoor environment, it is unlikely one can insert position information on its own, since a
fingerprint is being used. Unless the user knows exactly what fingerprint he or she is going to insert,
it is not possible.
The returnable responses for geocoding services that are possible include the following:


numberOfAddresses, so that the user might choose from matching addresses, if there are
multiple matches (e.g. if one inserts “Commission Room”, it is possible that there are 3
“Commission Rooms”)
Address, which matched the inserted address.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 53
11. Modelling and attributes
11.1. Modelling the indoor environment
The indoor environment can be modelled in many ways, dependent on what the service is aimed at.
The importance is that one must know the level of granularity should be. As it is the goal of this
project to let users navigate in buildings to a room, the accuracy level should be a room and not a
part of a room (shelf-like accuracy), unless the target is a very large space. A node and edge structure
like has been proposed by Kolodziej and Hjelm (2006) and the OGC (Open Geospatial Consortium;
2010) to model the environment for indoor navigation. Especially the OGC highlights the importance
of the relation between the nodes and edges, especially in the larger rooms and corridors. It is
therefore recommended to break larger rooms in smaller segments; the OGC calls this partitioning.
The vertical dimension can also be part of the model, as shown in Figure 13. In fact, this way of
modelling resembles of remodelling the topological dual space as the node-relation-structure (OGC,
2010), which can be used for the navigation service. The dual space can be described as a model of
the space in terms of linkages, which resembles a graph. The topological dual space explicitly
preserves the topology of the graph.
It is hard to tell whether rooms are neighbouring, since there is little information about the size and
distance of the environment, unless extra information can be added to the graph. Therefore,
attributes can be assigned to both the node as the edge. Those attributes can be used for locational
queries, in which the user can start its navigation in the first place. This also includes non-navigable
spaces for specific users, such as bumps for wheelchair users, which in turn can be used to determine
alternative routes.
Figure 13. Nodes and edges representation (topological dual space)
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 54
11.2. Non-usage of geometry
Unlike the outdoor navigation, data on indoor locations is not widely available. There are no vendors
for networks such as FalkPlan and TeleAtlas, and not all companies are willing to provide floor plans
for public use, or even if there is a provision, it might be possible that not everything is covered to
keep it secret. This might cause less accuracy, or missing data, although this might change in the
future.
In fact, the geometry does not necessarily have to be used in order to deploy the navigation service.
Indoor, a user only needs to know whether he is still on route, going from A to B. Going on route,
there are significant recognizable objects present, such as plants, pillars, postal boxes or paintings,
which provides help to confirm the user is in the right direction. For this, only fingerprints are an
absolute necessity. The requirement is that the fingerprints should be distinguishable from each
other and attached to POIs. Those fingerprints can be stored in a database and can be coupled with
attributes (recognizable objects) about what is present at that place and organized in the graph.
These attributes can be displayed on the screen in text or image. If the user can see those objects, as
the screen tells them, the user can be certain he is at the right place. If the user reaches a node
where a vertical movement can be made, information about possible stair climbing can be returned
as well. This whole process is illustrated in Figure 14.
Figure 14. Fingerprinting objects and navigation
The above implies that there is no geometry necessary, but this means that topology is
indispensable. As indoor pedestrian navigation (see chapter 7) has the opportunity to recognize the
space in multiple ways through present objects, it is not necessary to plot the entire floor plan: a
fingerprint is sufficient. The only information the pedestrian can have, is how to go from one end to
another. However, future extensions include the overview of the graph, so that the user can see at
least the route on screen. Another extension is the estimated time of arrival, which. However, the
travel time variables significantly, as this depends on the user profile. Some users need more effort
to open a door, while others need more effort upon climbing a staircase. This suggests that only
information about the edges is necessary, and not about the geometry surrounds it, as long as the
topology remains. The result is to be found in Figure 22, and is discussed in chapter 16.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 55
12. Visualization: presentation services
In this chapter, the visualization and representation aspects are discussed. As standard cartography
deviates on mobile platforms, first generalities on small screen cartography are discussed in 12.1. In
12.2 some technical information on visualization is related to the OGC standards on presentation
services. Finally, in 12.3, some comments are made about 3D visualization and explained why this is
not part of the thesis.
12.1. Users and general constraints and limitations
A decent visualization is necessary to communicate the proper information to the user. The user is
most likely not interested in the underlying principles, but in the end product. Small mobile screen
require a different visualization than paper maps or large computer screens. Nagi (2004)
distinguishes constraints on both the device-end as the network ends, as displayed in Table 4.
Category
Constraint
Comment
Display
Infrastructural
constraint
Device
Screen size
Usually very small
Display
Device
Resolution
Usually very limited
Display
Device
Colour range
Low range, although depends on device
Interaction
Device
Input opportunities
Interaction
Device
Output capabilities
Limitations of input fields, although depends on
screen & resolution
Usually restricted due to limitations in processing
Performance
Network
Connection
Performance
Network
Latency
Performance
Device &
Network
Device &
Network
Limited storage
Performance
Processing power
Slow in contrast to regular (wireless) internet,
unless device has a Wi-Fi data connection
Newer generation mobile phones allows larger
capacity with the expansion of (micro-)SD-cards
Restricted in the sense it does not equal normal
computer-like processing speed
Table 4. Mobile phone constraints
Table is based on Nagi (2004).
Although the information in Table 4 is dated from 2004, with performance and display related
constraints being partly solved or relieved. Still, the display and interaction categories remain
important limitations on the end visualization. The cartography largely depends on the displaying
capacities of the device: the larger the scale of a map, the more detail is needed. On a small screen,
this is considered challenging to overcome this problem. Yet, the better the resolution, the better the
application can overcome this challenge. A low resolution smartphone might display text labels or
symbol too large, for instance (Nagi, 2004). Newer generation smartphones are able to scale the
display; in that sense this barrier has overcome.
With the use of zooming options, one can toggle through different levels of details; however this
usually takes a lot of processing, when dynamic zooming is being used. This means that for each scale
a different data set and data representation is being visualized, which is a smooth zoom in fact. This
can also be done in steps (static stepped zooming) or static linear (no new data sets being visualized
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 56
at all). Either way, this demands a lot of processing time for both the device as the network (Nagi,
2004).
12.2. Presentation services by the OGC
The presentation service of the OGC (2008a) makes use of Web Mapping Services (WMS) and Web
Feature Services (WFS) to display on demand maps. However, since there is no geometry, only static
figures (images, topological network graph without coordinates, maps) can be visualized instead, as
there is no active and adaptive map viewing system. The parameters which belong to the
presentation services are all not applicable for the purpose of this thesis research.
12.3. Acknowledgement of 3D modelling and 3D visualization
Although developments of 3D modelling and 3D visualization are still ongoing, it is decided in this
research the visualization remains on the 2D or the 2.5D plane, as the focus is on the confirmation of
positions, and not the visual representation. However, the acknowledgement of 3D visualization is
important in future enhancements on the screen. Even though the presence of flyovers or
overpasses in the outdoor world is missing indoor, pedestrians still needs to deal with vertical
movements along staircases and elevators inside buildings. Döllner et al. (2007), suggests that 3D
visualization of objects or buildings certainly improve recognition patterns of users, which will further
improve the navigation process. It is not necessarily to make photo realistic representations, but 3D
sketch-based commands are already sufficient. In addition, utilizing the camera function of
smartphones, augmented reality (placing layers of information on top of the visible environment)
belongs to another possibility which can be incorporated in the application.
12.4. Cartography of text
The well-known outdoor large points of interests in the form of buildings do not apply to the indoor
context, however. Instead, smaller points of interests are present. Simple room destinations can be
translated to concrete goals and information. For example, room 1.001 can be an information point
for administration, and room 1.002 contains a copy machine. The visualization for the indoor
navigation follows thus the same rules and relations as described in the previous two sub sections.
However, as a result of not using the geometry for the modelling (chapter 11), no maps can be
displayed. It is possible to provide the user with a graph or with simple listed text to direct the user,
instead. The necessity of a map thus disappears, although this assumes there is a significant amount
of recognizable objects present. The presentation of the service could be as follows (Box 2):
1.
2.
3.
4.
Keep
Keep
Keep
Move
straight as you pass a copy machine
straight as you pass inboxes
straight as you pass room 1.011 to your left.
up one floor.
Box 2. Example of utilizing text for recognizable objects
The advantage of no cartography (maps) means that time and effort can be saved for deploying the
application on a larger scale, as one only needs to determine the fingerprints. As Hsu (2001) points
out, text usually takes longer to process for the user, which can be considered as a disadvantage.
Ideally, the text should be as short as possible, and supported by icons. The sub sections below will
discuss the underlying principles of text usage and text visualization as a whole.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 57
Text visualization
From the general cartographic rules, special attention is provided to the text visualization in this
paper first. Although Kraak & Ormeling (2003) only describe the rules for text usage on maps, the
same rules can be used for general text emphasis without maps. Text is able to show levels of
importance between words/names, by creating variation in:






Highlights, using boldness or italics;
Font face/style, using different fonts to distinguish texts from each other;
Font size, to emphasize the (non)-importance of words;
Font spacing, to emphasize the (non)-importance of words;
Font colour and/or font colour background;
Upper and lower case, by creating a contrast on what should be read.
Again, colour can be used to differentiate text on maps, but also on non-maps. The same limits on
colour perception apply: a maximum of approximately 7-8 colours can be used to ensure the reader
is able to select the most important information from a line of text.
Text attractiveness and uniform design
The screens on mobile phones thus has limited space, and although the scaling on screens have been
improved over the last years for smartphones, the emphasis on information visualization remains a
significant contribution to an application, aside from what the application does. Roto & Kaikkonen
(2003) found out that the topmost information on a page or interface got the highest priority, since it
is first item perceived. In a study on the user perception of web pages’ interface, Hsu (2011)
researched the attraction of elements on a webpage and listed them from best attractive element to
least attractive:
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
Title/Logo
Image
Information display/presentation
Willingness to read
Colours
Structure
Attraction
Layout
Usability
Hyperlink
Readability of texts
As can be seen, the importance of text is rated low. Suggestions on images, colours and layout are
better rated in contrast to texts (rank 10 and 11, aside from the aspect willingness to read). However,
when it comes to the functionality, texts were provided more important; better than the title or logo.
Text has to be readable to obtain knowledge as such, and Hsu (2011) found evidence that the texts
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 58
are in general better perceived when the layout of the text follows the webpage or application
standard design, which has to be uniform. This does not differ from location based applications.
Variation in text elements in their context
Michailidou et al. (2008) also underlined the aesthetics of text visualization in webpages and
applications. They acknowledge the importance of using clean, clear, and symmetrical design, in
which bullets and buttons can further enhance the readability of text. As with the general
cartographic rules, further refinements can be shown, using text variation in colour, size and style.
Wang & Chen (2003) provided evidence that important text can be distinguished within the main
text, by placing extra space on both sides, in which they call this jump length. A minimum distance of
0.35 cm to 0.70 cm has been considered effective – if the distance is too long, the text would appear
to be discontinued, in which users perceive as disturbing (too large holes in the text).
Text colour usage
Wang & Chen (2003) researched in particular the combination of colours on screens, for both texts as
backgrounds. They stated that colours have to be carefully chosen when they need to be combined.
Two extremes of the colour spectrum have to be avoided, such as red and blue, as they provide the
user from a “visual discomfort”. More important is the luminance contrast between background and
text colours, as they are perceived much better. For example, the combinations of black-on-white,
white-on-black and blue-on-yellow were better read than red-on-white, blue-on-white and green-onwhite, although the combinations are dependent on the condition of the screen, as well as the suns
illumination outdoors.
Moving text
Wang & Chen (2003) also investigated the effect of using motion and speed of text, and they found
that moving text in application has little impression on the user. In fact, users were tempted to
ignore the moving text, which has also been acknowledged in a study by Albrecht-Buehler et al.
(2005). Moving text in applications should not be too slow, but not too fast either, in order to
maintain the attention of the user. In combination with background and colours, one has to ensure
the background should not be disturbing to read the texts themselves. If the motion is too slow or
too fast, or if the background and colours makes it hard to read, change blindness (ignorance of
texts, thus ignorance of information and ignorance of the application) is being the case, which has to
be avoided.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 59
12.5. Initiatives to a text based indoor navigation application
Based on the above sections, an indoor navigation application, which is based on a small screen, and
does not utilize geometric maps, but solely text display for directions, a few recommendations can be
made regarding the output:







The interface must be adaptive to the user.
The interface itself has to be uniform in layout.
A contrast in colours (luminance) has to be guaranteed, in order to let the user perceive the
information correctly. Blue-on-yellow has been proposed.
To the top of the screen, the most important information has to be displayed.
o This coincides with the current location and the target destination.
When it comes to the output of text directions, orientation aspects have to be distinguished from
the body.
There should be a limit to the motion of text – text scrolling should be made possible otherwise.
Text size can be considered as well.
The proposed indoor navigation application is as follows:
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
The device automatically detects where the position is.
This position is made visible on the top of the screen.
The user has to insert a specific start or current location and destination.
The user has to navigate from one end to another.
The device calculates a route.
The user sees his current position on the top of the screen.
The user sees if he is on route/off route directly below that position information line.
The user sees his destination and the estimated time of arrival.
The user sees a list of directives below that on/off route information.
The first line of the directives is placed in a bigger text size.
The subsequent lines are in a different colour and size, to emphasize the importance of the first
direction output.
12. In any of the lines, information about directions can be made in a separate colour or style.
13. In any of the lines, information about orientation objects can be made in a separate colour or
style.
14. Bullets or numbers have to be used to indicate steps.
For example:
Current location: near room 5.99
ON ROUTE | Heading to: Exit
- Coming from the stairs, walk straight
towards Room 5.97 (to the left you see
glass works, to the right you see a coffee
machine)
- Turn right, you see an elevator
- Take the elevator to ground floor
- From the elevator turn left
- Exit the building
Figure 15. Example text output using text visualization and variances
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 60
13. Navigation services
The navigation engine is another vital part of navigation as a whole. The OGC (2008b) enhanced their
route services with the navigation services, with the assumption that the user is looking for more
than solely a route, such as the reconfirmation of their progress, and editable features.
A brief description of the Navigation Services from the OGC is in section 13.1. Section 13.2 deals with
the mandatory parameters to be used in a possible Navigation Service. The full overview can be
found in OGC (2008b). Section 13.3 concerns with the generation of route descriptions.
13.1. Navigation elements of the OGC
The OGC (2008b) listed 30 possibilities determining or influencing the course of the user, in which the
following are considered to be the most important. The full list can been found in Table AP. 2:


Core navigation:
o Specification of route preferences
o Specification of origin and destination point
o Determination of position
o Route calculation using user profile/preferences
o Presentation of route
o Route guidance
Extensions of rerouting:
o By detour, with traffic information
o By waypoints, as defined by the user
o By getting off route
o By change of route settings and preferences
In this thesis, getting off route piques the most interest. This process assumes that the user have to
be tracked, to determine whether the user is on or off route. Thus, the confirmation of the user’s
position plays a fundamental role in this aspect. In the case of fingerprinting as a positioning
technique: routing is made possible by linking the fingerprints with each other and the user has to
pass each point (fingerprint) on the route to stay on route. The navigation client should tell whether
the user is on or off route, by comparing those fingerprints with each other.
It is not the aim of this thesis to extensively describe all parameters and specifications of the
navigation application itself. The OGC (2008b) provided all exact inclusions in the document of
Navigation Services already. In the next section, only the mandatory parameters are briefly
discussed.
13.2. Mandatory parameters
The OGC (2008b) states that “map matching is the method of determining where the “mobile device”
has moved in the navigable network based on the device’s previous location(s) and data about the
device’s motion from external input (such as, but not limited to GPS)”. Map matching is strongly
related positioning, since a position is needed in order to determine where the user is. The OGC
(2008b) describes recommends hard lat-long coordinates in the WGS84 reference system, but this
cannot be implemented for the interior of a building. Instead, fingerprints in the context of a
topological graph are a possible solution.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 61
At specific points on the navigable network, attributes, conditions and relationships (i.e. prohibited
manoeuvres and directions) needs to be assigned (OGC 2008b), in order to provide the user
directions. This means that entries and exits are also part of this.
In Table 5 below, the most important parameters are displayed. A full overview can be found in the
document Navigation Services of the OGC (2008b).
Route requests
RoutePlanType
Parameter
RouteHandle
ExtendedRoutePlan
BoundingBox
RouteGuidanceRequest
FirstBucketSize
Priority
provideRouteHandle
DistanceUnit
Mandatory
Yes
Yes
No
No
No
Yes
No
No
Short description
Request information about the route
Specifies criteria route determination
Rectangular area of route
Return of turn-by-turn route instructions (text)
Stores directions in a list (“bucket”)
Priority of requests
Return of a route handle
Unit for measuring distance
WayPointList
AvoidList
Yes
No
List of waypoints along the route
List of areas, locations and features in which the route
should avoid passing through
The criteria upon which a route is determined
Cost Criteria
Ordered list of links and node travel costs for the path
Display of where the route navigation is now
Reference to the route stored at the Navigation Server
Describes overall characteristics of the route
ExtendedRouteControl
CostCriteria
RouteLinkAndCost
FirstBucket
RouteHandle
RouteSummary
Table 5. Parameters Navigation Services
RouteControlType*
Route Response*
Yes
Yes
Yes
No
No
No
Source: OGC (2008b). * More parameters are included, but these include vehicle information and actual traffic information,
which does not apply to indoor navigation in general.
As derived from Table 5, the control of the route, including the waypoints is fundamental for the
navigation. In order to navigate, one must include navigable points along the way, and assign costs to
it. These costs can be used to inform the user about time and distance estimations.
13.3. Generating navigation guidance
In this thesis research, a simple navigation algorithm, using Dijkstra’s shortest path route, can be
used for navigation instead. As the intention of the thesis is to use solely the framework, the
parameters in table 8 have to be transformed in Java (see chapter 16). However, before using the
parameters, thoughts on how the route descriptions (the guidance) have to be explained first.
When no maps are preserved, it is challenging to determine orientation aspects. Upon travelling
from node A to C via B, it is hard to tell from solely a graph whether one has to turn left, right or keep
straight, as one can flip the graph, for instance. It is then important to assign orientation aspects to a
connectivity database, when nodes (the fingerprints) and edges (connectivity information between
fingerprints) can be stored. It is then important that the application should take the current and
predicted positions into account.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 62
When the current and predicted (or the previous and current) position is known, the application can
then predict on how the orientation can be helpful. In Figure 16, a route has been calculated from
node 2004 to node 2012. Upon travelling from the first nodes, from node 2004 to node 2002 is
basically straightforward: it implies that going from node 2004 to node 2003 is straight on, and the
user has then turn left. However, there is no explicit implication that this graph is mirrored: perhaps
going from node 2003 to node 2002 is a turn to the right. When storing orientation information
between two links, one cannot add left or right, since it depends on where the user is coming from.
In the inverse route, the turn 2003-2004 is ‘right’, instead of ‘straight on’. It is useful to store
information of the current or previous location, the predicted location, or both. In a table, one could
store the route 2004-2003-2002, with 2004-2003 straight (previous location was 2005) and 20032002 as a turn to the left, while 2002-2003-2004 deals with a 2002-2003 straight (previous location
was 2007) and 2003-2004 as a turn to the right.
Figure 16. Routing, guidance, orientation
However, there are many routes possible, and the question is how to program the orientation
aspect. One possible solution, is by pre-programming every single possible list of route descriptions,
as described above. Another possibility is to store unique location, object information, without using
orientation terms, such as ‘turn towards the floor sign’, although it is difficult to state, when there is
no geometry involved. It is then important that the object information has to be stored along with
the fingerprint or with the connectivity information (or both). In the case from travelling from node
2004 to node 2012, the following can be constructed:







(2004) Pass through the doors ,
(2003) turn towards plant ,
(2002) walk towards sofa’s ,
(2007) pass the reception ,
(2009) pass Anneke van Cootenzaal ,
(2011) turn towards statue ,
(2012) destination reached.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 63
In this sense, no orientation (as in: head north/south, turn left/right), or rather, neutral directions are
being used. Only in the vertical direction, explicit orientation directions can be stated (go up/go
down).
Next, the application has to check the current position of the user constantly. If the user finds itself
on or near the provided nodes, then this must be confirmed. If not, re-routing has to be done, in
which the user should be warned. At the same time, the application can already predetermine which
nodes can be detected. It is therefore useful that the application should only get nearby nodes
(fingerprints). For instance, the application should at least eliminate outlying nodes: if one traverses
in the route from node 2004 to node 2012, then the application should not detect the user at node
2005, node 3005 or node 3205, after passing node 2009, for example.
In the next chapter, an architectural overview is being presented.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 64
14. Architecture: functional and technical design
With the use of the services and user information, the final architecture can be determined, which is
displayed in Figure 17 below. The described core services in the chapters 9, 10, 12 and 13 are
embedded in the centre of the architecture, based on the principles mentioned in chapter 11, as they
deal with the processing of the application. The application itself is constructed in Java with Android,
and is located at the heart of the architecture. The positioning feed is derived from the position
determination equipment; in this case, this is the Wi-Fi receiver of the Smartphone.
Figure 17. Architecture of this indoor navigation application
Derived and edited from OGC (2008b)
The location contents form the database of the application. This also includes the fingerprints. As the
application itself is standalone, there is no server needed. The application constantly links to the
inherent databases, which contains information on rooms, POIs, which are combined with
fingerprints and their connectivity. The application itself makes use of the Wi-Fi receiver of the
smartphone, which is processed by a Wi-Fi Extended library, also in the Java environment. The client
(the Graphical User Interface, the GUI) finds itself at the top of the architecture, as it can request the
information from the application.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 65
The functional design allows seeing what the user can do with the application. Based on the
documents of the OGC (2008a and 2008b) and the visualization described in the chapters 9-14, both
the functional as the technical design is displayed in Figure 18. Ideally, only a database with
directions and inherent information about the fingerprints themselves are necessary. The
fingerprints directly link to the locational information, and the route information.
The functional design is explained as follows (in the figure, in red denotes actions to be taken for the
technical properties of the prototype):
1. The user has to make sure Wi-Fi is enabled. In the case the inherited databases are placed on a
server, the user has to make sure a 3G OR Wi-Fi connection is enabled, so that the databases can
be downloaded. 3G is necessary if the user cannot establish a Wi-Fi data connection.
2. The route preferences have to be stated via the user profile. By default, this has been set to a
normal pedestrian, with a guest status. This implies that certain areas are restricted, unless
stated otherwise. The user can change this by clicking the Settings button.
3. The user can then define the departure. The current location is being displayed, which helps the
user orientate. The user have to define the departure location by typing (selecting) a predefined
location.
4. The user has then to define the destination. The same categories are possible as for the origin, as
assumed before, the user does not have knowledge about the indoor positioning, nor the
addresses (other than room numbers), which makes it impossible to pinpoint an exact location.
As the application also does not use maps, it is not possible to pinpoint it on a map.
5. Optionally, a via-point can be inserted. The same categories as the destination can be chosen,
with the addition that it can be set as an inclusion, or exclusion. In the case of an inclusion, the
route given will prompt the user to navigate to this first set point. In the case of exclusion, the
route will take into account it is not desirable to pass this point at all. An example is the exclusion
of a certain segment, which is now restricted due to refurbishments.
6. Now, the user can plan the route, by touching the plan button.
7. A list of directives will be provided. Each direction has information about what is visible at that
point, so the user can see himself whether he is at the right position or not. In addition, it will be
visible if the user’s location matches the predicted fingerprint.
8. In a possible extended application, the user is capable to modify the presentation, which
contains the text display, speech engine and most importantly, a confirmation setting. This can
be done by reserving a section on the screen which shows a green approval or red disapproval
sign or something similar. Alternatively, this can be spoken as well. In addition, a graph can be
displayed, but this depends on whether a graph will be included of the building. For now, the
focus is on text directives. The map can be used, for instance, to display the entire route.
At this stage, it will be possible to still adjust the route preferences, as well as the inclusion and/or
exclusion of certain areas.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 66
Figure 18. Functional and technical design of the application
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 67
Conclusion Part II
How can the creation of an application be possible, without using geometry?
The user would like to navigate from one source to destination. Indoor, the user has points of
recognition available, as the level of detail is higher than outdoor. The user can thus check whether a
navigation application provides correct information, if the application makes use of the surroundings.
As the focus is on utilizing fingerprints only, instead of using geometry, positioning and
representation are provided a different context. Fingerprints can be coupled to recognizable objects
at that place. A real (x,y,z) positioning (as stated in the terminology) is then not necessary, as user is
only interested if he is at the right place at that time, since there are no obvious addresses in the
vicinity, with exception of numbered rooms. In addition, pedestrians have specific characteristics,
and thus minute estimations are hard to tell, since a user might need more time to climb the stairs or
to open a door. This implies that geometry is not required, as long as information about connecting
fingerprints is preserved.
As with the level of detail surrounding a user, a simple text list is considered sufficient, as long as it
includes information about recognizable objects. However, a weakness is that there is no clear
direction or orientation (left/right) possible, since there is no geometry. Instead, neutral descriptions
in combination with recognizable objects have to be provided. This assumes that the user constantly
has to carefully watch the real environment because of the high detail. In the case graphs or floor
plans are provided, one must consider the constraints and limitations of the small screen and
processing speed.
How can the use of the framework of Open Location Service standards be implemented in the indoor
navigation application?
Each component of the Location Services can be used in the Technical design. The most important
feature is the gateway service, where the users location as picked up by the Wi-Fi receiver, has to be
compared to the fingerprints database. The user itself can define the constraints and preferences,
which determines the route handling. Upon entering information on origin or current location,
destination and the inclusion or exclusion of via points, information can be transformed via the
directory and geocoding services. This information can be further processed, with the navigation
service. However, since the aim of this thesis is not to use geometry at all, large parts of the
parameters of the OpenLS standards cannot be implemented. Therefore, only the framework will be
used instead of the parameters.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 68
PART III - Implementation
results
Contents



15. Surveying and databases
16. Location determination algorithms
17. Layout and results
In this third part, the final output is being discussed. First, the methodology of surveying as well as
the creation of the database is discussed in chapter 15. Two different location determination
algorithms based on fingerprinting are then assessed in chapter 16. Finally, in chapter 17, the test
results on location determination and navigation will be discussed. In the end, it is possible to answer
the question:
How can Wi-Fi fingerprinting determine a location which does not utilize geometric features, and how
can it be implemented in an application as such?
The conclusion is directly provided in part IV, as it further discusses the results related to it.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 69
15. Surveying and databases
Android applications are composed by using at least two languages: Extensible Markup Language
(XML) for the layout and interface, and Java for the functionality. On top of the Java, a separate
Android library is installed, resulting in an adapted version of Java to Android. Additional libraries,
such as an extended Wi-Fi library, can also be attached.
This application is no difference, and also utilizes the same style, as programmed in Eclipse (Figure
19). First, the interface was designed in XML. Screenshots are made via the emulator, which does not
support Wi-Fi detection however (only the real device does so). The XML files can be portrayed at the
client side of the application, with the server side being the inherited chain of Java files and
databases, as there is no remote server for this prototype.
Figure 19. Screenshot of the Eclipse Environment
Chapter 15.1 first discusses the methodology of surveying and chapter 15.2 continues with the
database implementation.
15.1. Site survey preparation
15.1.1. Measure points
The surveyed locations are parts of the ground, first and second floor of the OTB building at Delft
University of Technology. In particularly, the OTB building wings 2 and 3, as wing 1 is reserved for
another department of the university. As the office spaces in these wings can be locked, these are
the only restricted areas in the test bed. In wing 3, at the ground, first and second floor, twelve
points on each floor has been assigned a fingerprint, to ensure the fingerprints can be compared with
each other regarding signal strengths, which allows a proper snapping: if the fingerprints would be
too similar to each other, distinguishing the locations would be very hard (see also 16.2.3).
Differences are obtained because of receiving direct and indirect signals due time, which are being
caused by multipath effects. The points were measured in the aisles of the wings, at specific places,
such as in front of the elevator, near the toilets, and on junctions of connecting office rooms. Figure
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 70
20 shows a photo of the location; Figure 22 displays the graphical representation of the measured
points.
Figure 20. Photo of wing 2 and wing 3 of the OTB building
15.1.2. Measuring and determining suitable fingerprints
An ASUS K72J laptop, equipped with an Atheros AR9285 Wireless Network Adapter, has been used to
register the signal strength in the first test runs. Signal strengths are expressed in a negative value of
signal loss and noted with the unit dBm.
The signal strengths were measured with the inSSIDer 2.0 software, which could be downloaded at
http://www.metageek.net/products/inssider/. inSSIDer 2.0 is able to register all received signal
strengths, at a given interval, along with their unique MAC address. The uniqueness is necessary to
distinguish access points with the same SSID, as otherwise measuring using SSIDs will result in
collated signal strengths. At each location, a scan has been carried out for ten seconds, registering at
a 1 second interval. Those locations were then tagged with location information, and merged into
one file.
In the second test runs, new measurements have been performed, but instead of using a separate
device to measure the fingerprints, the same device as the application is being used: the smartphone
instead of the laptop has been used to do so. This is further elaborated in chapter 16.
The ten seconds scan, with a 1 second interval is necessary to determine stable, average signal
strength, since the signal strengths can fluctuate in a short time, as displayed below in Figure 21: for
instance at FID=201 (this has later been converted to FID=3201, to note the fingerprint is located in
wing 3), from the first record: the signal strengths at this point changes between -43 dBm and -49
dBm, caused by the reception of both direct and indirect signals due multipath effects. It also detects
the stability of access points: at a given location, if at least four out of ten are present, this can be
considered as relatively stable. This will eliminate unstable access points, which are usually outlying
routers. The information is first being stored in a SQL database from Microsoft Access. In Android,
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 71
each database table can be stored in a separate XML file, using SQL statements instead of database
tables. This conversion has been done as shown in annex A.4.
The next step is filtering the weak signals. It is possible that at a given location, a fingerprint consist
of 30 MAC-addresses, in which the majority are below a RSSI of -80 dBm. It might be the case they
will not appear in a scan at another given time. On the other hand, RSSIs higher than -70 dBm are
most likely to be constantly present. Even then, it is possible that at a measured location the signal
strength always remains relatively weak (for example, always lower than -70 dBm). In this process, it
has been chosen that the 10 strongest signals picked up, will determine the position, or to be more
precise: the fingerprint.
An unedited example is provided in Figure 21, ordered by location, then by average RSSI from highest
to lowest. As can be seen, there are duplicate SSIDs, but there are different MACs. Also the number
of detected signals within ten seconds changes. Do note that the location naming (FID) were
preliminary in this figure. The final FIDs have been assigned as visible in Figure 22 (full graph) and
Figure 23 (part of the graph on top of a photo).
Figure 21. Screenshot of reconfigured RSSI scan
In this figure, temporarily constructed Location and FID tags have been generated. The values RSSI1-RSSI10 represents
measurements within ten seconds, where the number of RSSI depicts the second.
As displayed in the Figure 21, within ten seconds, using an interval of one second, RSSIs from several
routers (MAC+SSID) at specific locations (Location/FID) are measured and recorded. These results in
a maximum of ten values (RSSI1-RSSI10), from these values an average can be calculated (RSSI_AVG).
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 72
Figure 22. Graphical representation of indoor model (graph) with correct orientation
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 73
Figure 23. Graph layer on photo
15.1.3. Fingerprinting methodology used in this project
Because the Wi-Fi signal carrier is not consistent itself, it is hard to identify a location without a fixed
signal strength. Signal strengths vary over time, caused by multipath, and these signal strengths
themselves are not consistent on top of that. However, by averaging the signal strength values, and
by creating a search space (a range) around this average, it is still possible to predict a location. This
is the methodology being used in this project. This is illustrated in Figure 24: a recorded fingerprint
deviates in signal strengths when it is compared at a later time stamp. At the x-axis, the signal
strengths are plotted for one fingerprint. Other fingerprints are being measured simultaneously,
which can be plotted at the z-axis in the figure below.
Figure 24. Time aspect of fingerprinting
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 74
At time stamp 1, at a specific location, a certain amount of signal strength is being received, which
has been obtained directly. At time stamp 2, at the same location, the amount has been changed,
because it received the signal strength delayed because of the multipath effect. For example, in
Figure 25/Table 6, at location 3210, there is a given signal from MAC 00:1A:A2:FA:0E:B0 of -72 dBm
at time stamp T1, which is received directly. At time stamp T3, it received a delayed signal strength
caused by multipath of -78 dBm, after a while, it receives another value of -74 dBm at T7. To counter
these differences, an average has been taken based on 10 time stamps.
MAC
SSID
FID
AVG
T1
T2
T3
T4
T5
T6
T7
T8
T9
T10
00:1A:A2:FA:0E:B0
eduroam
3210
-75,2
-72
-72
-78
-78
-78
-78
-74
-74
-74
-74
Table 6. Temporal aspect of fingerprinting (table view)
Figure 25. Temporal aspect of fingerprinting (bar chart) for given location FID=3210
Values in both Figure 25/Table 6 are in dBm.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 75
15.2. Databases and information storage
15.2.1. Generalities
In this prototype, the database is stored in the application, as Android allows doing so with SQLite
(Android Developers, 2011). If a large scale service deployment could be realized, a large database
can ideally be stored on a server, in which the user needs a connection to obtain information. Since
the database of the prototype is rather small, it is not necessary to work with a remote server. This
results in a standalone mobile application.
Originally, several database tables have been proposed, which included information about location
(fingerprints and general separated), connectivity, location related information (points of interests
and room-specific information), and personal information. However, personal information is heavily
restricted due to privacy and legal issues, in combination with locational information. In the final
prototype, the personal information has thus been left out. This issue is further discussed in chapter
18. Moreover, the table_poi has been merged with the table_rooms, as ideally, POIs have to be
assigned with the same location FIDs. This decreases probabilities in data redundancy as well.
Another prototype restriction is the availability of selected buildings. In a large scale service
deployment, the user is able to browse through various buildings, at given addresses and places.
Since the prototype only contains one building, the option to search for locations in other buildings
than the OTB has been disabled. More discussion is to be found in chapter 19 (limitations)
In Figure 26, an overview of these relations and database tables can be seen, along with their data
type written in capitals. The database fields are further explained in Annex A.3, with the exception of
the table for persons, since that has been omitted (see chapter 20 on discussion on privacy issues).
Table_mac has become an independent table, and only shows the MAC address along with their
RSSI, at given locations, which is joined by the original location table.
Figure 26. Database relations
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 76
The most important fields include the location ID from the table_location, which is used in any other
table. This coincides with the fingerprint, as the fingerprint is the node itself. The table_location
states the address, location, and type of location (whether this is an exit, a room or stairs as such).
The table_connectivity provides information on the connection between the fingerprints; each link or
edge has a unique ID (the connection ID, or the cid), and states whether it traverses over objects, as
well as which objects are to the left or to the right. In the overview above, the fields which are
underlined are truly used in the prototype. This effectively means that any double field in other
database tables can be removed, decreasing the data redundancy.
15.2.2. The database in the application
The database has been inserted as SQL statements in separate XML files, as databases in Android can
be configured as such. In Box 3 below, an example of a XML file is shown.
<sql>
<statement>
CREATE TABLE IF NOT EXISTS table_rooms (
rid INTEGER PRIMARY
KEY,
fid INTEGER,
room_name VARCHAR(50),
wing INTEGER,
floor INTEGER,
building VARCHAR(50),
department VARCHAR(50),
office VARCHAR(50),
room_type VARCHAR(50),
restricted BOOLEAN))
</statement>
<statement>INSERT INTO table_rooms VALUES(2002,2008,'Survey
Room',2,0,'OTB','','Room','Computer Room',1);</statement>
<statement>INSERT INTO table_rooms VALUES(2004,2008,'Grote
Vergaderzaal',2,0,'OTB','','Room','Commission',1);</statement>
<statement>INSERT INTO table_rooms VALUES(2006,2009,'Anneke van
Cootenzaal',2,0,'OTB','','Room','Commission',1);</statement>
<statement>INSERT INTO table_rooms VALUES(2007,2010,'0.070
(Secretary)',2,0,'OTB','','Work Office','Office',1);</statement>
Box 3. Example of SQL statements in XML: creation and population of database table
As such, an XML document can be filled with statements, in which those statements can be read in
the Java file again. Other examples are included in the annex A.4.
The database can be read with the use of a Java class “DatabaseHelper.java” which utilizes the
XML/SQL statements. It is up to the programmer to define which fields can be used for which
purpose. The fields that the user gets to see, are the room names, and the object information, which
is stored in the table_connectivity (object_left and object_right) and table_location (loc_prop). The
room name is coupled with a room id (rid), which is combined with a specific location (fid). The user
gets to see a reverse geocoded specific location: a specific fingerprint (mac+rssi from the table_mac),
can be found, which is then looked up in the database table, which room and location belongs to that
fingerprint.
In Box 4 below the database is called from the DatabaseHelper.java file. In this file, the
parseXmlFile(R.raw.roomsdata, db); states that the XML file, which contains SQL statements,
roomsdata is to be used as a database. This coincides with the contents of the table_rooms as shown
in Box 3. The statement NodeList statements = doc.getElementsByTagName("statement"); explicitly
tells that the SQL statements can be found in the XML document, which has the tags <statement>
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 77
public class DatabaseHelper extends SQLiteOpenHelper {
public static final String DATABASE_NAME = "rooms";
protected Context context;
public DatabaseHelper(Context context) {
super(context, DATABASE_NAME, null, 1);
this.context = context;
}
@Override
public void onCreate(SQLiteDatabase db) {
parseXmlFile(R.raw.roomsdata, db);
parseXmlFile(R.raw.connectiondata, db);
parseXmlFile(R.raw.macdata, db);
parseXmlFile(R.raw.poidata, db);
parseXmlFile(R.raw.locationdata, db);
}
private void parseXmlFile(int id, SQLiteDatabase db) {
String s;
Log.i("DbHelper", "Executing SQL statements");
try {
//
Toast.makeText(context, "1", 2000).show();
InputStream in = context.getResources().openRawResource(id);
DocumentBuilder builder =
DocumentBuilderFactory.newInstance().newDocumentBuilder();
Document doc = builder.parse(in, null);
NodeList statements = doc.getElementsByTagName("statement");
for (int i=0; i<statements.getLength(); i++) {
s = statements.item(i).getChildNodes().item(0).getNodeValue();
db.execSQL(s);
}
} catch (Throwable t) {
Toast.makeText(context, t.toString(), 50000).show();
Log.e("DbHelper", t.getMessage());
}
}
@Override
public void onUpgrade(SQLiteDatabase db, int oldVersion, int newVersion) {
db.execSQL("DROP TABLE IF EXISTS table_rooms");
onCreate(db);
}
}
Box 4. Snippet overview of DatabaseHelper.java: getting the contents from database tables from XML files
The database can be further used for input for the client. In the main screen, the user can select and
enter rooms and locations from the table_rooms to calculate a route. The MainScreen.java is shown
in Box 5. The setContentView(R.layout.main); refers to the fact that the main.xml document has to
be used as layout. Then, there are two fields which need to be connected to the database, in where
the client has to provide information on where the client wants to go, and what the client’s
departure place is. This can be done by using an AutoCompleteTextView, which is an extension of
Android, where the user have to type something, which will be automatically completed by a list, so
that the user can select an item from the list. This is being recalled with:
AutoCompleteTextView dep1 = (AutoCompleteTextView) findViewById(R.id.inputDeparture);
RoomList roomList = new RoomList(this);
ArrayList<String> items1 = roomList.getAllRooms();
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 78
However, it must explicitly state which fields are required to display. Then, a separate Java file needs
to be written, which is shown in Box 6.
Box 5. Snippet overview of MainScreen.java: creation of entry fields and actions upon tapping buttons
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 79
public class RoomList {
protected SQLiteDatabase db;
protected Cursor cursor;
public RoomList(Context c) {
Log.i("Roomlist", "Creating DB");
db = (new DatabaseHelper(c).getWritableDatabase());
}
public ArrayList<String> getAllRooms() {
// Log.i("Roomlist", "Query execute");
Cursor cursor = db.rawQuery("SELECT rid, room_name FROM table_rooms",
null);
ArrayList<String> rooms = new ArrayList<String>();
// Log.i("Roomlist", cursor.getCount()+"Amount found");
if (cursor.getCount() >= 1) {
while (cursor.moveToNext()) {
rooms.add(cursor.getString(1));
}
}
cursor.close();
return rooms;
}
}
Box 6. Snippet overview of RoomList.java: populating an entry field with a directory service
A query must be formed first, in order to populate the AutoCompleteTextView fields. First this field
(an Array List) must be formed, which is done by public ArrayList<String> getAllRooms(), next the
query will be summoned and filled in the Array List, with:
Cursor cursor = db.rawQuery("SELECT rid, room_name FROM table_rooms", null);
ArrayList<String> rooms = new ArrayList<String>();
Then this step will be repeated for the entire database table, which is done by:
if (cursor.getCount() >= 1) {
while (cursor.moveToNext()) {
rooms.add(cursor.getString(1));
}
}
cursor.close();
return rooms;
This way, the AutoCompleteTextView fields are fully populated. Both departure and destination can
use the same Array List. A message will be briefly shown, how many rooms are added:
Toast.makeText(this, "Amount of rooms loaded: " + items1.size(),
Toast.LENGTH_SHORT).show();
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 80
However, upon confirming the fields, one must be certain that there are no empty fields, or that the
same records are being used for departure as destination, as soon as the client clicked on the button
to calculate a route. This is solved with:
public void onClick(View v) {
if (v.getId() == R.id.buttonNavigationStart) {
TextView inputDeparture = (TextView) findViewById(R.id.inputDeparture);
TextView inputDestination = (TextView) findViewById(R.id.inputDestination);
if (inputDeparture.getText().length() == 0) {
Toast.makeText(this, "Please input departure",
Toast.LENGTH_SHORT).show();
return;
} else if (inputDestination.getText().length() == 0) {
Toast.makeText(this, "Please input destination",
Toast.LENGTH_SHORT).show();
return;
} else if (inputDeparture == inputDestination) {
Toast.makeText(this, "Departure and Destination is the same",
Toast.LENGTH_SHORT).show();
return;
}
if (false) {
}
Intent n = new Intent(this, Navigation.class);
startActivity(n);
} else if (v.getId() == R.id.buttonSettings) {
Intent s = new Intent(this, Settings.class);
startActivity(s);
As the following Java codes checks if the field is empty:
if (inputDeparture.getText().length() == 0) {
Toast.makeText(this, "Please input departure",
Toast.LENGTH_SHORT).show();
return;
In this prototype, the decision has been made not to combine the current location with an input field
for origin point, to allow the user to calculate a route in advance, without being on the present spot.
This might be useful if the user would like to prepare his route, rather than finding out on the fly.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 81
16. Location determination algorithms
The prototype resembles three major functionalities:
1. The determination of a location;
2. Routing between pre-determined locations or current location with destination;
3. Route confirmation during the navigation.
The routing itself is pre-determined in the sense that the user is able to type and select specific
locations, but the application only accepts predefined locations. This is comparable with planning
railway journeys, where there are only a limited number of stations to travel between. The user can
only use a railway station from a database. In the prototype, the user can only select rooms or POIs
and the user does not necessarily need the sensor to calculate a route. This can be useful if the user
is planning ahead. The route confirmation information is then not usable.
From here, it is desirable to combine two major functionalities in a fourth aim:
4. To navigate between locations, where the current location is constantly adapted to the users
situation, as the user sees the current location constantly at the top half of the screen.
Section 16.1 describes which types of location determination is being used in this thesis. Section 16.2
continues with the translation of the obtained locations, while section 16.3 continues with how the
previous sections were programmed. Finally, section 16.4 deals with the aspect of navigation and
route confirmation.
16.1. Two methods of location determination being used in this thesis
Method 1: count method with search space
One possible method to determine the location, is by picking up the signal strengths in negative dBm
per MAC address. These values have to be compared to a recorded database, in which location
information can be returned. This results in the following algorithm:
1. Pick up signal strength on the fly (within 1 second).
2. Snap signals within a range of n dBm.
3. Compare the live received signal strengths with the signal strengths ranges in the database
(which is a range query). Count the number of matches per FID.
4. Return the location ID of the highest match. Based on the location ID, return additional
information from other tables as well.
5. If there are multiple matches, take both.
6. To prepare and speed up the prediction of the next location, eliminate outlying locations (see
also chapter 14.3).
The databases that will be used are displayed below.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 82
First, the existing databases on signal strengths:
FID
3001
3001
3001
3001
3001
3001
3001
3001
3001
3001
MAC
00:1a:a2:c0:0e:85
00:1a:a2:c0:0e:82
00:1a:a2:fa:0f:a0
00:1a:a2:fa:0f:a2
00:1a:a2:fa:0f:a4
00:1a:a2:fa:0f:a5
00:1a:a2:fa:39:90
00:1a:a2:fa:39:92
00:1a:a2:fa:39:94
00:1a:a2:fa:39:95
RSSI
-74
-75
-59
-60
-60
-59
-41
-49
-49
-48
FID
3002
3002
3002
3002
3002
3002
3002
3002
3002
3002
MAC
00:1a:a2:fa:0f:a0
00:1a:a2:fa:0f:a2
00:1a:a2:fa:0f:a4
00:1a:a2:fa:0f:a5
00:1a:a2:fa:25:40
00:1a:a2:fa:25:45
00:1a:a2:fa:39:90
00:1a:a2:fa:39:92
00:1a:a2:fa:39:94
00:1a:a2:fa:39:95
RSSI
-63
-62
-64
-64
-72
-72
-42
-43
-43
-43
Table 7. Sample database on signal strengths on two
different locations
The following information is available about the location itself (the table has been truncated in
comparison to the actual table for this purpose):
FID
3001
3002
Building Address
OTB
Jaffalaan 9
OTB
Jaffalaan 9
Place
Delft
Delft
Floor
0
0
Wing
3
3
Access
0
0
Locprop_01
Emergency Exit
Showers
Table 8. Sample database on location information
The following information is available about the rooms (the table has been truncated in comparison
to the actual table for this purpose):
RID
FID
Room_name
3001 3001 Room 0.010 (Pieter
Groetelaerszaal)
3002 3001 Room 0.020
3003 3001 Room 0.030 (Shower)
3004 3002 Room 0.040
3005 3002 Room 0.050 (Shower)
3006 3002 Room 0.060
Wing Floor Office
3
0 Commission
Room
3
0 Work Office
3
0 Shower
3
0 Work Office
3
0 Shower
3
0 Work Office
Room_Type Restricted
Room
1
Office
Shower
Office
Shower
Office
1
0
1
0
1
Table 9. Sample database on room information
Note that one node (FID) contains several navigable spaces (RID), as the level of detail allows this. If it
is desirable to make the rooms separate navigable as well (for instance, two specific points of
interests are located inside a room), then separate fingerprints and thus nodes will need to be added
to the database.
Now, step by step the algorithm will be explained.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 83
Step 1: pick up signal strength
During the receiving phase, the following signals are retrieved, as displayed in Table 10:
MAC
00:1a:a2:c0:0e:85
00:1a:a2:c0:0e:82
00:1a:a2:fa:0f:a0
00:1a:a2:fa:0f:a2
00:1a:a2:fa:0f:a4
00:1a:a2:fa:0f:a5
00:1a:a2:fa:25:40
00:1a:a2:fa:25:45
00:1a:a2:fa:39:90
00:1a:a2:fa:39:92
00:1a:a2:fa:39:94
00:1a:a2:fa:39:95
RSSI
-78
-78
-61
-61
-61
-61
-75
-75
-43
-50
-50
-49
Table 10. Example of received signal strengths
Step 2: create range for the signal strength with ±n dBm
After filtering the weak and unstable RSSIs, an average RSSI can be calculated. However, since the
RSSIs fluctuate over time and space, a snapping tolerance is necessary to determine a position.
Therefore, a standard deviation has been calculated over all measurements, which was an average of
2.5 dBm (without filtering, this was 2.4 dBm). As a result, a rounding off to 3.0 dBm has been
accepted as the snapping tolerance. To make sure at least a signal was being recognized, an
additional 1.0 dBm has been added, resulting in a 4 dBm search space margin. If the received signal
strength in the online phase is within a range of 4 dBm, then the measured signal matches the
location recorded signal. This is illustrated in Figure 27.
Figure 27. Matching live fingerprints with recorded fingerprints using snapping tolerances
The values in this figure are purely indicative and do not represent actual values, nor snapping values
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 84
From the received signal strengths, n dBm has to be added and subtracted to get a range. For
example, if a snapping value of 3 dBm has been used, the result is the following Table 11:
MAC
00:1a:a2:c0:0e:85
00:1a:a2:c0:0e:82
00:1a:a2:fa:0f:a0
00:1a:a2:fa:0f:a2
00:1a:a2:fa:0f:a4
00:1a:a2:fa:0f:a5
00:1a:a2:fa:25:40
00:1a:a2:fa:25:45
00:1a:a2:fa:39:90
00:1a:a2:fa:39:92
00:1a:a2:fa:39:94
00:1a:a2:fa:39:95
Lower value
-75
-75
-58
-58
-58
-58
-72
-72
-40
-47
-47
-46
Upper value
-81
-81
-64
-64
-64
-64
-78
-78
-46
-53
-53
-52
Table 11. Example of determination of range signal strengths
Step 3: comparison and matching the database
This step exists of four sub steps:
A. From the existing database table_mac, compare if the recorded value is within the range of
received signal strengths table.
B. If this is being the case, accept the value. If this is not being the case, discard the value.
C. Count the number of matches per fingerprint FID; accept the FID from the recorded database.
D. If multiple IDs were matching, then probably the user finds itself at a location between two
recorded places. Both IDs can be returned. If no FIDs were matching, the statement “Location
not found” can be returned.
First, the received signal strengths are now being compared with the recorded signal strengths and
the signals are being rejected or accepted (step A and B, Table 12):
MAC
Received
00:1a:a2:c0:0e:85
-78
00:1a:a2:c0:0e:82
-78
00:1a:a2:fa:0f:a0
-61
00:1a:a2:fa:0f:a2
-61
00:1a:a2:fa:0f:a4
-61
00:1a:a2:fa:0f:a5
-61
00:1a:a2:fa:25:40
-75
00:1a:a2:fa:25:45
-75
00:1a:a2:fa:39:90
-43
00:1a:a2:fa:39:92
-50
00:1a:a2:fa:39:94
-50
00:1a:a2:fa:39:95
-49
Lower
-75
-75
-58
-58
-58
-58
-72
-72
-40
-47
-47
-46
Upper FID=3001 Accepted FID=3002 Accepted
-81
-74
NO
-81
-75
YES
-64
-59
YES
-63
YES
-64
-60
YES
-62
YES
-64
-60
YES
-64
YES
-64
-59
YES
-64
YES
-78
-72
YES
-78
-72
YES
-46
-41
YES
-42
YES
-53
-49
YES
-43
NO
-53
-49
YES
-43
NO
-52
-48
YES
-43
NO
Table 12. Example of comparing recorded and received signal strengths
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 85
Then, the number of matches needs to be counted. The result is 9 and 7 out of 10, respectively (step
C and D). FID=3001 is being accepted as the location, FID=3002 is being rejected. The user finds itself
near FID=3001.
Step 4: returning the information
The user has no information to extract from solely “FID=3001”. As soon as the matching took place,
information has to be returned. If the user is located near FID=3001, information from Table 8 and
Table 9 can be returned, such as: “you are near the emergency exit”, or “you are near Room 0.010,
Room 0.020 and Room 0.030”, at “Wing 3, Floor 0”.
The location information can also be used to monitor the actual location and from there recalculate a
route, if necessary.
The procedure steps 1-4 have been visualized in a flow scheme, as visible in Figure 28. Figure 29
shows a visualization of the process upon walking from one end to another: as the number of
matches are being counted.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 86
Figure 28. Flow scheme of location determination for count method with search space
Figure 29 is purely illustrative (it is not to scale and not linear) and simulates how the detection and
counting of correct fingerprints should take place upon walking around.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 87
Figure 29. Visual representation of count method upon walking around (desirable simulation)
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 88
Method 2: Least sum of square method for location determination
It is also possible to make use of a least sum of square (SS) method as proposed by Shum et al.
(2011). Instead of counting the number of matches, the differences between the actual and surveyed
are being used, similar to taking a physical distance into account. The lower the SS value, the most
likely it is the user finds itself at the predicted location. An example has been showed in Table 13
below – to the far right the outcome of the SS method can be seen.
FID
3001
3001
3001
3001
3001
3001
3001
3001
3001
3001
3002
3002
3002
3002
3002
3002
3002
3002
3002
3002
3003
3003
3003
3003
3003
3003
3003
3003
3003
3003
3004
3004
3004
3004
3004
3004
3004
3004
3004
3004
MAC
00:1A:A2:C0:0E:82
00:1A:A2:C0:0E:85
00:1A:A2:FA:0F:A0
00:1A:A2:FA:0F:A2
00:1A:A2:FA:0F:A4
00:1A:A2:FA:0F:A5
00:1A:A2:FA:39:90
00:1A:A2:FA:39:92
00:1A:A2:FA:39:94
00:1A:A2:FA:39:95
00:1A:A2:FA:0F:A0
00:1A:A2:FA:0F:A2
00:1A:A2:FA:0F:A4
00:1A:A2:FA:0F:A5
00:1A:A2:FA:25:40
00:1A:A2:FA:25:45
00:1A:A2:FA:39:90
00:1A:A2:FA:39:92
00:1A:A2:FA:39:94
00:1A:A2:FA:39:95
00:1A:A2:FA:0F:A0
00:1A:A2:FA:0F:A2
00:1A:A2:FA:0F:A4
00:1A:A2:FA:0F:A5
00:1A:A2:FA:25:42
00:1A:A2:FA:25:44
00:1A:A2:FA:39:90
00:1A:A2:FA:39:92
00:1A:A2:FA:39:94
00:1A:A2:FA:39:95
00:1A:A2:FA:0F:A0
00:1A:A2:FA:0F:A4
00:1A:A2:FA:0F:A5
00:1A:A2:FA:25:40
00:1A:A2:FA:25:42
00:1A:A2:FA:25:44
00:1A:A2:FA:25:45
00:1A:A2:FA:39:90
00:1A:A2:FA:39:92
00:1A:A2:FA:39:95
RSSI
O
M
-75
-78
-74
-78
-59
-63
-60
-62
-60
-62
-59
-63
-41
-45
-49
-45
-49
-46
-48
-47
-63
-63
-62
-62
-64
-62
-64
-63
-72
-77
-72
-75
-42
-45
-43
-45
-43
-46
-43
-47
-61
-63
-60
-62
-61
-62
-61
-63
-63
-62
-62
-62
-62
-45
-62
-45
-61
-46
-61
-47
-64
-63
-63
-62
-64
-63
-59
-77
-57
-62
-59
-62
-59
-75
-64
-45
-64
-45
-64
-47
L
-78
-77
-62
-63
-63
-62
-44
-52
-52
-51
-66
-65
-67
-67
-75
-75
-45
-46
-46
-46
-64
-63
-64
-64
-66
-65
-65
-65
-64
-64
-67
-66
-67
-62
-60
-62
-62
-67
-67
-67
Range 3
H
-72
-71
-56
-57
-57
-56
-38
-46
-46
-45
-60
-59
-61
-61
-69
-69
-39
-40
-40
-40
-58
-57
-58
-58
-60
-59
-59
-59
-58
-58
-61
-60
-61
-56
-54
-56
-56
-61
-61
-61
1
1
0
0
1
1
0
0
0
1
1
1
1
1
1
0
1
1
1
1
0
1
1
1
1
1
1
0
0
0
0
1
1
1
0
0
1
0
0
0
0
L
-80
-79
-64
-65
-65
-64
-46
-54
-54
-53
-68
-67
-69
-69
-77
-77
-47
-48
-48
-48
-66
-65
-66
-66
-68
-67
-67
-67
-66
-66
-69
-68
-69
-64
-62
-64
-64
-69
-69
-69
Range 5
H
-73
-72
-57
-58
-58
-57
-39
-47
-47
-46
-61
-60
-62
-62
-70
-70
-40
-41
-41
-41
-59
-58
-59
-59
-61
-60
-60
-60
-59
-59
-62
-61
-62
-57
-55
-57
-57
-59
-59
-59
1
1
1
1
1
1
1
1
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
1
1
1
0
1
1
0
0
0
0
Range 10
L
H
-85
-65
-84
-64
-69
-49
-70
-50
-70
-50
-69
-49
-51
-31
-59
-39
-59
-39
-58
-38
-73
-53
-72
-52
-74
-54
-74
-54
-82
-62
-82
-62
-52
-32
-53
-33
-53
-33
-53
-33
-71
-51
-70
-50
-71
-51
-71
-51
-73
-53
-72
-52
-72
-52
-72
-52
-71
-51
-71
-51
-74
-54
-73
-53
-74
-54
-69
-49
-67
-47
-69
-49
-69
-49
-74
-54
-74
-54
-74
-54
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
1
1
1
0
1
1
0
0
0
0
DIF
3
4
4
2
2
4
4
-4
-3
-1
0
0
-2
-1
5
3
3
2
3
4
2
2
1
2
-1
0
-17
-17
-15
-14
-1
-1
-1
18
5
3
16
-19
-19
-17
SS Method
DIF²
SS
9
107
16
16
4
4
16
16
16
9
1
0
77
0
4
1
25
9
9
4
9
16
4
1013
4
1
4
1
0
289
289
225
196
1
1628
1
1
324
25
9
256
361
361
289
Table 13. Localization methods: count method versus sum of squares method
O = Original, observed surveyed value, M = Measured value, L = Low (O-Range value), H = High (O+Range value), 1 = True, 0
2
= False, DIF = M-O, SS = ∑ DIF . Values are in dBm.
FID
3001
3002
3003
3004
Range 3
5
8
6
4
Number of TRUES
Range 5
Range 10
8
10
10
10
6
6
5
5
Table 14. Outcome of counts and sum of squares
SS
SS
107
77
1013
1628
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 89
At a specific location, a certain amount of signal strength is received from the various routers (M).
These are being compared with the original, surveyed signal strengths (O). In the count method, first
a range of search space is being created. The larger the search space is, the more matches can be
found, which leads to possible duplicate locations. In this example, if a range of ± 10 dBm would be
used, one cannot tell anymore if the person finds itself at location FID=3001 or FID=3002, as can be
seen in Table 14.
Count method and least SS method compared
In the least sum of squares method, large differences are contributing significantly to the elimination
of unlikely locations, since they add up to the total value due to their large value. For a lower ranging
value, there is no clear preference in which method can be used. However, for larger ranges, it
appears that the least sum of square methods prevails, since it is capable of distinguishing locations,
in the event of equal true counts. In this project, the SS method has been chosen to ensure a certain
degree of certainty.
Two major weaknesses can be noted, however. A first weakness of both methods is that it requires
all identifiable nodes to have an equal amount of access points, as both the count method as the sum
of squares method would be disproportional to each other. In this project for example, each
fingerprint must contain 10 MAC addresses. If there are only 4 MAC addresses in one fingerprint, the
maximum matching value would be 4, in which this can be easily overpassed. This requires smart
router placing to counter. One example includes the use of threshold and ratio counting. If, for
instance, from 6 detectable routers, 4 routers are within range, then a score of 3 out of 6 (50,0%)
would seem to be worse than 3 out of 5 (60,0%), despite there were more routers detectable.
A second weakness is the sensitivity of fingerprints with bad reception. If, for instance, recorded
signal strengths at a certain fingerprint are bad, by default, the difference to the weakest assumed
detectable signal strength (-100 dBm) is only 10 dBm. In Table 15, this is being shown: the algorithm
will detect the user then at FID=3001, because of the lower SS value, although in reality, the user
may find itself somewhere else.
RSSI
FID
3001
3001
3001
3001
3001
3001
3001
3001
3001
3001
3002
3002
3002
3002
3002
3002
3002
3002
3002
3002
MAC
00:1A:A2:C0:0E:82
00:1A:A2:C0:0E:85
00:1A:A2:FA:0F:A0
00:1A:A2:FA:0F:A2
00:1A:A2:FA:0F:A4
00:1A:A2:FA:0F:A5
00:1A:A2:FA:39:90
00:1A:A2:FA:39:92
00:1A:A2:FA:39:94
00:1A:A2:FA:39:95
00:1A:A2:FA:0F:A0
00:1A:A2:FA:0F:A2
00:1A:A2:FA:0F:A4
00:1A:A2:FA:0F:A5
00:1A:A2:FA:25:40
00:1A:A2:FA:25:45
00:1A:A2:FA:39:90
00:1A:A2:FA:39:92
00:1A:A2:FA:39:94
00:1A:A2:FA:39:95
O
-75
-74
-59
-60
-60
-59
-41
-49
-49
-48
-63
-62
-64
-64
-72
-72
-42
-43
-43
-43
M
-100
-100
-100
-100
-100
-100
-100
-100
-100
-100
-100
-100
-100
-100
-100
-100
-100
-100
-100
-100
DIF
-25
-26
-41
-40
-40
-41
-59
-51
-51
-52
-37
-38
-36
-36
-28
-28
-58
-57
-57
-57
Table 15. Scenario for bad reception for SS method
SS Method
DIF²
SS
625 19250
676
1681
1600
1600
1681
3481
2601
2601
2704
1369 20084
1444
1296
1296
784
78
3364
3249
3249
3249
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 90
Alternatively, if from a specific MAC no signal can be detected at all, it is possible to assign a blank
value, but that would result in least sum of squares based on unequal amounts of MACs. This
effectively means that unless it can be guaranteed all MACs per location are within reach.
A fifth step: sub selecting
It is therefore an additional step can then be followed, which deals with the prediction of future
locations. It is implied that users do not appreciate false location information. Providing false location
information when using fingerprinting is most likely to occur when unstable signal strengths are
being received. For instance, a user might be prompted to be located at a floor lower or higher than
he or she is, or that the device places the user at the end of the hallway, while the user has not
reached that location yet. There are possible ways to cater these, by doing the following:

Include only fingerprints of the locations along the determined route.
o The application can be programmed in such way, only location determination can take
place using the nodes (fingerprints) along the calculated route.
 For example, for the route 2004-2003-2002-2007-2009-2011-2012, the
programme only needs to check the access points coupled to these fingerprints.
In the Figure 30 below, from origin to destination, the user receives a list of
direction for a route. This route does not contain any signal strengths from
routers B and D, so they should be eliminated.
 However, this leads to solely filtering of routers: in the event that a location’s
fingerprint contains signal strengths from distant routers, the signal strengths
from the excluded router will never be picked up, decreasing the probability in a
correct location.
Figure 30. Exclusion of access points to narrow down search space
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 91


Exclude fingerprints and location information that are beyond the user’s current location.
o If the fingerprints are named in such way it is recognizable which fingerprint belongs to
which section or floor, it is possible to remove any access points for an area.
 For example, upon surveying, it has been predetermined nodes are numbered in
the convention <1-digit wing><1-digit floor><2-digit location>. For the route
2004-2003-2002-2007-2009-2011-2012 again, the nodes tells that these nodes
are located in Wing 2, at the ground floor (0). By excluding all location
information below the numbers 2000 and higher than 2100, it can be certain the
user will not be promptly allocated on a wrong floor.
If, despite the above actions have been undertaken, complete different sets are obtained than
predicted as derived from the calculated route, the application has to say the user is off route, in
which re-calculation of the route is proposed.
However, due to time restrictions and capability constraints, this fifth step has not been
implemented in the prototype. It is therefore highly recommended that a further narrowing down of
the location detection can take place. This includes one (or a combination of) the methods above
described.
16.2. Geocoding and reverse geocoding
When the signal strengths and fingerprints are determined, the device can look up the ID of the
fingerprint, which is coupled to both a general, navigable database (the table table_location), and a
rooms database (table_rooms). The inverse occurs upon entering a room – when a user selects a
room, this room is being geocoded in a fingerprint ID, which is being used to calculate a route. In this
project, one room can only have one fingerprint ID, whereas one fingerprint ID can have multiple
rooms. In the example of fingerprint ID (node) 3208 (see also Figure 22, page 72), rooms 2.150-2.180
all have this fingerprint. When a user would like to navigate from the Entrance/Exit to Room 2.180,
the following happens:








The user selects Entrance/Exit as departure
The user selects Room 2.180 as destination
The user presses Start Navigation
The application looks up the fingerprint node for Entrance/Exit: 2001
The application looks up the fingerprint node for Room 2.180: 3208
The application calculates the route between nodes 2001 and 3208.
The application provides information which nodes are required to traverse upon going from node
2001 to 3208
The application recodes the node numbers to location information, such as “reception” and
“stairs”.
In the prototype, the fingerprint ID’s are displayed along with the to be selected room; these can be
left out in future developments.
16.3. Location determination and geocoding in Java
Box AP. 6 in the annex (A.5) contains some of the code of the location determination algorithm. First,
the Wi-Fi needs to set up, which is done in lines 36-39 and lines 152-189. This allows the application
to search for Wi-Fi signal strengths, along with their MACs, or Base SSIDs. Second, the to be searched
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 92
MACs have to be declared, which has been done in lines 42-43 (in Box AP. 6, only two have been
displayed; one can add as many MACs as desirable). Third, the lines 47-50 creates the search space
for which the received signal strength should match the recorded database, which is done in lines
209-214. The fourth step is to retrieve information from that looked up location. This can also be
done via a database search, which is being carried out in lines 219-231.
In fact, the query itself exists of three parts: a first part, which includes the selection of the mac
addresses and locations (line 196, also displayed below in Box 7), the middle part, in which the
recorded signal strength is being compared with the lower and upper boundaries of the search
space, and the closing part, which counts the matching nodes (lines 203-216), and groups it by node
number (or fingerprint ID, the FID, line 197). The query is being put together, and from this new
query, the FID with the maximum matches has to be the designated location (lines 229-231 to be
precise).
Finally, this has to be displayed on screen, but this has to be sent in a message, rather than
independent parameters, which are carried out in lines 236-256. First the parameters are being
looked up and packed as a message, then the message is being decoded into visible texts again,
which is carried out in lines 102-147.
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
public void OnRuleAchieved(RuleEvent event) {
WifiManager wifimanager =
(WifiManager)activity.getSystemService(Context.WIFI_SERVICE);
wifimanager.startScan();
List<ScanResult> ls = wifimanager.getScanResults();
//A loop to customize the query according to the measured values
String queryHead = "Select fid, count(*) from table_mac where ";
String queryTail = " group by fid";
//Vector<String> vQuery = new Vector<String>();
int counter = 0;
String queryMid = "";
if(ls.size() > 0){
for(ScanResult sr: ls){
//Prepare
String mac = sr.BSSID;
int SS = sr.level;
if(counter == 0){
queryMid = "(mac = '"+mac+"' and rssi_avg < "+(SS+4)+" and
rssi_avg > "+(SS-4)+")"; // First MAC with upper and lower
}
else{
queryMid += " or (mac = '"+mac+"' and rssi_avg <
"+(SS+4)+" and rssi_avg > "+(SS-4)+")"; // Subsequent MAC with upper and lower
}
counter++;
} //End of loop scan results
String finalQuery = queryHead + queryMid + queryTail; // Query buildup
int max = 0;
int selectedFID = 0;
Cursor cursor = db.rawQuery(finalQuery, null);
if(cursor.getCount() > 0){
while(cursor.moveToNext()){
int fid = cursor.getInt(0);
int count = cursor.getInt(1);
if(count > max){
max = count;
selectedFID = fid;
}//If some records are equal, top one is selected
Box 7. Snippet overview of queries for searching for nodes within the search space
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 93
16.4. Navigation algorithm
The user is able to determine a route before traveling. A simple insertion of both departure and
destination can be fulfilled without the sensor technique (similar to the classic route planners before
GPS was being utilized). A Dijkstra route is being implemented in the application, to preserve the
simplicity of general navigation.
The Dijkstra algorithm is being used to compute the shortest path between to nodes in an edge. The
algorithm searches for the shortest path from one node to all connected nodes and eliminates the
paths that are longer than the currently calculated value. The algorithm will stop until the destination
has been reached, in which the shortest path can be displayed.
Vogel (2009) has developed a simplistic Java code, in the sense that the Dijkstra algorithm will be
used inversely for the purpose of preserving overview in the codes. The codes from Vogel (2009) are
displayed in annex A.6.
In order to let the algorithm work, another Java code has to be written, in which all nodes and edges
are being loaded, and from which the origin and destination nodes are being selected. This is visible
in Box 8.
First, the nodes have to populate the Vertex.java. This is being done with the database recall
functions, visible in lines 12-35: the FID (the fingerprint ID’s: the navigable nodes) is thus being
retrieved. The second step is to populate the edges for Edge.java, which is being done in lines 37-66;
in this case the near entire database table is being looked up. Further processing to build the graph
itself is being done in a separate class, in lines 96-105. The third step is using the information from
the nodes and edges to the graph (Graph.java), which can be seen in lines 70-93: a path is being
calculated using the Graph, which in turns makes use of the DijkstraAlgorithm.java (line 72-75).
Regarding intelligence in route calculation, it is possible to add extra edges between distant nodes, if
they nodes follow the same route. For instance, if on a straight line between node A and B, the nodes
C, D, E and F are located, this would be normally a line between A-C, C-D, D-E, E-F and F-B. It is
possible to add an extra node between A and B directly, with a slightly lower impedance value
assigned to it, so that upon a continuous path, directly the direction A-B can be provided. However,
as the aim is to check whether the user is on route or not, this might not be the most efficient way to
do so, as there can be only false checks for the intermediate points, even though in reality the user
traverses those points.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 94
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
public class RouteCalc {
protected SQLiteDatabase db;
protected Cursor cursor;
private List<Vertex> nodes;
private List<Edge> edges;
public RouteCalc(Context c) {
db = (new DatabaseHelper(c).getWritableDatabase());
}
public LinkedList<Vertex> findRoute(int fromNode, int toNode) {
nodes = new ArrayList<Vertex>();
edges = new ArrayList<Edge>();
ArrayList<Integer> nodesFids = new ArrayList<Integer>();
Cursor cursor = db.rawQuery("SELECT fid, locprop FROM table_location",
null);
// Log.i("Roomlist", cursor.getCount()+"Amount found");
if (cursor.getCount() >= 1) {
while (cursor.moveToNext()) {
Vertex location = new Vertex(cursor.getInt(0),
cursor.getString(1), null);
nodes.add(location);
nodesFids.add(cursor.getInt(0));
Log.d("routecalc", cursor.getString(0) + " added (="
+ location.getRoomName() + ")");
}
}
cursor.close();
Cursor cursor2 = db
.rawQuery(
"SELECT cid, fid_link_01, fid_link_02,
length, direction, restricted, stairs, elevator, ramp, door, objects_left, objects_right
FROM table_connectivity",
null);
// Log.i("Roomlist", cursor.getCount()+"Amount found");
if (cursor2.getCount() >= 1) {
while (cursor2.moveToNext()) {
addLane(cursor2.getInt(0),
nodesFids.indexOf(cursor2.getInt(1)),
nodesFids.indexOf(cursor2.getInt(2)),
cursor2.getInt(3), cursor2.getString(4),
cursor2.getString(5), false, false, false,
false,
cursor2.getString(10),
cursor2.getString(11));
Log.d("routecalc", "Connection " + cursor2.getString(0) + "
added ... length =" + cursor2.getInt(3));
Log.i("routecalc", "... From "+ cursor2.getInt(1) + "=" +
nodesFids.indexOf(cursor2.getInt(1)) + " to " + cursor2.getInt(2) + "=" +
nodesFids.indexOf(cursor2.getInt(2)));
}
} else {
Log.e("routecalc", "No connections??");
}
cursor2.close();
db.close();
Graph graph = new Graph(nodes, edges);
DijkstraAlgorithm dijkstra = new DijkstraAlgorithm(graph);
dijkstra.execute(nodes.get(nodesFids.indexOf(fromNode)));
LinkedList<Vertex> path =
dijkstra.getPath(nodes.get(nodesFids.indexOf(toNode)));
Log.e("routecalc", "Planning route from node
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 95
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
"+fromNode+"="+nodesFids.indexOf(fromNode));
Log.e("routecalc", "Planning route to node
"+toNode+"="+nodesFids.indexOf(toNode));
if (path.size() == 0) {
Log.e("routecalc", "No route found =(");
}
for (Vertex vertex : path) {
final String tag = "test";
Log.v(tag, "Route: " + vertex.getRoomName());
System.out.println("Route: " + vertex.getRoomName());
}
return path;
}
private void addLane(int cid, int fid_link_01, int fid_link_02, int length,
String direction, String restriction, boolean stairs,
boolean elevator, boolean ramp, boolean door, String object_left,
String object_right) {
Edge lane = new Edge(cid, nodes.get(fid_link_01),
nodes.get(fid_link_02), length, direction, restriction,
stairs,
elevator, ramp, door, object_left, object_right);
edges.add(lane);
}
}
Box 8. Snippet overview of RouteCalc.java
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 96
17. Layout and results
Now the database and location determination algorithms have been established, the application
itself can be constructed regarding the interface, and upon completion, tests can be performed.
Chapter 17.1 deals with the design issues. Finally, in chapter 17.2, investigations of the test runs are
discussed, and chapter 17.3 provides provisory remarks, which are further to be discussed in part IV.
17.1. Layout configuration
The interface consists of three screens: a screen where the user has to insert departure and
destination via an AutoCompleteTextView, a screen in which the user can set route preferences (such
as walker or wheelchair user, or unrestricted access), and a screen with the display of the route in
descriptions. The interface is designed in XML, as shown in Box 9 for example for the Main Screen.
The colour schemes are determined based on the literature recommendations as stated in chapter
12 (in particularly, sub sections 12.4 and 12.5)
As the application is a prototype, the user can only insert rooms or points of interests (POI) from the
OTB building in the first insertion screen (Figure 31). If multiple building were available, another
screen would be necessary to define this. The user can insert rooms by typing in a text field, which
autocompletes the to be typed location. This shows the user directly what is available, and is
applicable for all insert parameters (departure, destination, include via-point, exclude via-point). In
the upper part of this same view, the user gets automatically to see to which room or POI he or she is
located at, so he or she can use this to insert this for a departure parameter.
Figure 31. Main Screen: input and AutoCompleteTextView
In the codes below in Box 9, by default, string texts as @string/noinfo have been placed. As soon as a
location from the gateway service has been provided, these strings will be replaced by this location
information, caused by the Java code:
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 97
TextView textFid = (TextView) findViewById(R.id.currentType);
textFid.setText("Near: " + locprop + " (Node: " + fid + ")");
In this example, a node number (fid) and a location property (locprop) will be provided to replace the
string with no info.
<?xml version="1.0" encoding="utf-8"?>
<LinearLayout
xmlns:android="http://schemas.android.com/apk/res/android"
android:orientation="vertical"
android:layout_width="fill_parent"
android:layout_height="fill_parent"
android:background="#FFC000">
<TextView
android:layout_height="wrap_content"
android:layout_width="match_parent"
android:text="@string/currentlocation"
android:textStyle="bold"
android:textColor="#0070C0"
android:id="@+id/CurrentLocation"></TextView>
<TableLayout
android:padding="3dip"
android:layout_height="wrap_content"
android:layout_width="match_parent"
android:id="@+id/tableLayout1"
android:stretchColumns="1">
<TableRow
android:id="@+id/tableRow1"
android:layout_width="wrap_content"
android:layout_height="wrap_content">
<TextView
android:text="="@string/noinfo "
android:layout_width="wrap_content"
android:textColor="#0070C0"
android:layout_height="wrap_content"
android:id="@+id/currentBuilding"></TextView>
<TextView
android:layout_width="wrap_content"
android:layout_height="wrap_content"
android:textColor="#0070C0"
android:text="="@string/noinfo "
android:id="@+id/currentAddress"></TextView>
</TableRow>
<TableRow
android:id="@+id/tableRow2"
android:layout_width="match_parent"
android:layout_height="wrap_content">
</TableRow>
<TableRow
android:id="@+id/tableRow3"
android:layout_width="match_parent"
android:layout_height="wrap_content">
<TextView
android:text="@string/noinfo"
android:layout_width="wrap_content"
android:textColor="#0070C0"
android:layout_height="wrap_content"
android:id="@+id/currentWing"></TextView>
<TextView
android:text="@string/noinfo"
android:layout_width="wrap_content"
android:textColor="#0070C0"
android:layout_height="wrap_content"
android:id="@+id/currentFloor"></TextView>
</TableRow>
<TableRow
android:id="@+id/tableRow3"
android:layout_width="match_parent"
android:layout_height="wrap_content">
<TextView
android:text="@string/noinfo"
android:layout_width="wrap_content"
android:textColor="#0070C0"
android:layout_height="wrap_content"
android:id="@+id/currentType"></TextView>
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 98
<TextView
android:text="@string/noinfo"
android:layout_width="wrap_content"
android:textColor="#0070C0"
android:layout_height="wrap_content"
android:id="@+id/currentRoom"></TextView>
</TableRow>
</TableLayout>
<TextView
android:id="@+id/textDeparture"
android:layout_height="wrap_content"
android:layout_width="match_parent"
android:textColor="#0070C0"
android:text="@string/departure"
android:textStyle="bold"></TextView>
<AutoCompleteTextView
style="@style/FakeTextField"
android:id="@+id/inputDeparture"
android:background="@drawable/textlines"
android:clickable="true"
android:textColor="#0070C0"></AutoCompleteTextView>
<TextView
android:id="@+id/textDestination"
android:layout_height="wrap_content"
android:layout_width="match_parent"
android:textColor="#0070C0"
android:text="@string/destination"
android:textStyle="bold"></TextView>
<AutoCompleteTextView
style="@style/FakeTextField"
android:id="@+id/inputDestination"
android:background="@drawable/textlines"
android:clickable="true"
android:textColor="#0070C0"></AutoCompleteTextView>
<TextView
android:id="@+id/textExample"
android:layout_width="wrap_content"
android:layout_height="wrap_content"
android:textColor="#0070C0"
android:text="@string/specificationExample"
android:textSize="10sp"
android:paddingBottom="10sp"></TextView>
<Button
android:layout_height="wrap_content"
android:layout_width="match_parent"
android:id="@+id/buttonNavigationStart"
android:text="@string/startnavigation"></Button>
<Button
android:layout_height="wrap_content"
android:layout_width="match_parent"
android:id="@+id/buttonSettings" android:text="@string/settingsLabel"></Button>
</LinearLayout>
Box 9. XML file of the Main Screen
In the second screen, the user is able to define the route parameters. This is being done with radio
buttons for the type of user, and an enabler button for the restriction level. Enhanced features can
be placed here as well, such as a text-to-speech engine, or the display of the graph structures.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 99
Figure 32. Settings screen
In the third and final screen, the user should see the route, along with the current location. As the
application involves a non-geometry feature, no maps are to be displayed. However, the user gets to
see if he or she is at the correct place according to the provided route. In the prototype, the node
number is being displayed as reference. Because the Android Emulator does not support Wi-Fi
measuring, Figure 33 below shows no information about the current position.
Figure 33. Navigation screen: output of a route
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 100
17.2. Testing the application
Upon completion of the application, test runs have been performed, of which the methodology and
results are discussed in this sub chapter. The testing includes application validation (17.2.1), the
location check (17.2.2) and the navigation itself (17.2.3).
17.2.1. Application validation
The functioning of the application itself is the first part of testing, where Wi-Fi does not necessarily
have to be used. This includes the input, processing and output mechanisms without Wi-Fi, as well as
the layout and the interface. For instance, a user might plan ahead and use the application to plan a
journey from the entrance to room 2.250. The application validation is the major part of the
development, and has to work in advance, before a location check comes in.
The prototype has been configured in such way that the node numbers are still visible. That means
that the user can search for the FID number of the nodes/locations. A built-in feature is that it will
also check for empty or incorrect fields: it is not possible to calculate a route between points that do
not exist. Further extensions have been taken into account, such as the possibility to calculate routes
based on user preferences, but are not operational as it is not the aim of the research to look at
various outcomes of user profiles, but rather on localization of the user.
17.2.2. Criteria of location check and test performances
In the prototype, the user first sees what the current location is. This is displayed in the top of the
screen. At various locations, this has been tracked at twenty different timestamps. The location is
considered correct, if the application states that the user is exactly at the right place, or at the place
adjacent to it (with a maximum of two nodes), since the signal strengths may vary over time and
space. This does not apply for a wrong placement on the vertical dimension, since the signal
strengths are significantly different from each other at different heights.
The filtering of search space by ± 4 dBm has been maintained in the next scan results. The first scan
compromised the registration of 12 location nodes, in 20 consecutive time stamps. The results have
been distinguished by:


Close matches:
o Location has an exact match with the database
o Location was within one node away, at the same floor
 E.g. if the actual location is node 3109, then 3108 and 3110 are valid.
o Location was within two nodes away, at the same floor
 E.g. if the actual location is node 3109, then 3107, 3108, 3110 and 3111 are valid.
Bad matches:
o Location was within two nodes away, but at a different floor
 E.g. if actual location is 3109, then 3008-3011, 3208-3211 are valid
o Location was more than two nodes away, but at the same floor
o Location was more than two nodes away, and at a different floor
o Location was physically within two nodes away, but at different part of the building
 This occurs in places with many complex building parts, in which the user is
prompted to be near a location close to another wing, for example.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 101
Two major scans have been performed. In the first series, location information was being retried
based on fingerprints composed by the ten strongest signal strengths at that point, as originally
proposed. This also includes routers that are on a different floor than where the users are. The
second series of scan results were based on eight MAC addresses, whose access points located on
the same floor. In this extent, the physical space of the router is being combined with the physical
location of the fingerprint. This has been made possible, since the routers at the OTB contain four
MAC addresses. To illustrate the routers’ locations and their reaching space on that same floor, see
Figure 34:
Figure 34. New database method: using coverage areas instead of storing best signals
Next, at 12 distinctive locations, scans were performed. The following locations were chosen (Table
16):
Node
2002
2005
2008
3002
3007
3011
3101
3105
3110
3203
3209
3211
Location
Wing 2, Ground Floor, near main exit
Wing 2, Ground Floor, near back exit
Wing 2, Ground Floor, corner of two rooms
Wing 3, Ground Floor, near emergency exit
Wing 3, Ground Floor, centre point
Wing 3, Ground Floor, end of hallway
Wing 3, Floor 1, end of hallway
Wing 3, Floor 1, near kitchen
Wing 3, Floor 1, second staircase
Wing 3, Floor 2, primary staircase
Wing 3, Floor 2, next to glass doors of corr.
Wing 3, Floor 2, near end of hallway
Table 16. Overview of test nodes for location checks
Remark
Open area near nodes 2006-2007-2008
Small corridor near end
Open area near nodes 2002-2006-2007
End of hallway, near router A
Near router B, centre of corridor
Near router C
Near router D
Far away from router D and E
Between routers E and F
Between routers G and H
Near router H
Near router I
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 102
Results on overall scan
Out of the 240 obtained values, only 14 of them had exact matches, which is a mere 5.8% of the scan
results. Moreover, 66 were within two nodes of the actual location, which compromises a 27.5% of
the scan results. This compromises that the user was in one third (33.33%) at or near the appropriate
location. In 47 cases (19.6%), the user was located within two nodes of the actual location, although
it was on a different floor. This was mostly being the case when the user was at the ground or first
floor. This suggests that signal strengths were matched from the floor or below the user, as this
rarely happened on the second floor, where only one floor below could be matched. Still, 26.3% was
matched on the same floor, but beyond two nodes of where the user actually was.
70
60
3211
3209
50
3203
3110
40
3105
3101
30
3011
3007
20
3002
2008
10
2005
2002
0
Exact match
Within 1
places
Within 2
places
Beyond 2
Beyond 2
Within 2
Within 2
places, same places, other places, other places,
floor
floor
floor
behind area
Figure 35. First scan results - grouped by match, specified by location (fingerprint) node, 12 nodes, wing 2+3.
If a closer look is being taken on the results of solely Wing 3 in the test case, it seems that the wrong
allocation on the wrong floor has been affected by the scan results of the locations in Wing 2.
Simultaneously, the total cases of a near correct match dropped to an exact 25% (45 out of 180; with
Wing 2 included, which compromised 80 out of 240). This is caused by the near correct placements of
node 2002 and 2005: 15 out of 20 time intervals, the user was located within 2 places away of node
2002, whereas it happened 17 out of 20 time intervals the user was located within 1 place away of
node 2005.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 103
50
45
40
3211
35
3209
30
3203
25
3110
20
3105
3101
15
3011
10
3007
5
3002
0
Exact match
Within 1
places
Within 2
places
Beyond 2
Beyond 2
Within 2
Within 2
places, same places, other places, other places,
floor
floor
floor
behind area
Figure 36. First scan results - grouped by match, specified by location - wing 3 nodes
The outcomes are also displayed in Table 17 below, and further specified for Wing 3 only in Table 18.
Node
Close matches:
% From total observations
Exact match
3002
3007
3011
3101
3105
3110
3203
3209
3211
2002
2005
2008
2
1
7
1
0
0
14
5
15
16
17
2
TOTAL
80
0,8%
0,4%
2,9%
0,4%
0,0%
0,0%
5,8%
2,1%
6,3%
6,7%
7,1%
0,8%
33,3%
0
1
1
0
0
0
0
4
8
0
0
0
14
0,0%
0,4%
0,4%
0,0%
0,0%
0,0%
0,0%
1,7%
3,3%
0,0%
0,0%
0,0%
5,8%
Within 1 places
2
0
5
0
0
0
0
0
3
1
17
1
29
Within 2 places
0
0
1
1
0
0
14
1
4
15
0
1
37
0,8%
0,0%
2,5%
0,4%
0,0%
0,0%
5,8%
0,4%
2,9%
6,7%
7,1%
0,8%
27,5%
% From total observations
% From total observations
Out of match:
% From total observations
Beyond 2 places, same floor
18
19
13
19
20
20
6
15
5
4
3
18
160
7,5%
7,9%
5,4%
7,9%
8,3%
8,3%
2,5%
6,3%
2,1%
1,7%
1,3%
7,5%
66,7%
2
12
11
3
0
0
3
7
0
4
3
18
63
0,8%
5,0%
4,6%
1,3%
0,0%
0,0%
1,3%
2,9%
0,0%
1,7%
1,3%
7,5%
26,3%
Beyond 2 places, other floor
0
0
0
14
16
4
1
0
0
0
0
0
35
Within 2 places, other floor
1
7
2
2
4
16
2
8
5
0
0
0
47
% From total observations
Within 2 places, behind area
% From total observations
15
0
0
0
0
0
0
0
0
0
0
0
15
6,7%
2,9%
0,8%
6,7%
8,3%
8,3%
1,3%
3,3%
2,1%
0,0%
0,0%
0,0%
40,4%
Table 17. Scan results attempt 1
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 104
Node
Close matches:
% From total observations
Exact match
3002
3007
3011
3101
3105
3110
3203
3209
3211
2
1
7
1
0
0
14
5
15
TOTAL
45
1,1%
0,6%
3,9%
0,6%
0,0%
0,0%
7,8%
2,8%
8,3%
25,0%
0
1
1
0
0
0
0
4
8
14
0,0%
0,6%
0,6%
0,0%
0,0%
0,0%
0,0%
2,2%
4,4%
7,8%
Within 1 places
2
0
5
0
0
0
0
0
3
10
Within 2 places
0
0
1
1
0
0
14
1
4
21
1,1%
0,0%
3,3%
0,6%
0,0%
0,0%
7,8%
0,6%
3,9%
17,2%
% From total observations
% From total observations
Out of match:
% From total observations
Beyond 2 places, same floor
18
19
13
19
20
20
6
15
5
135
10,0%
10,6%
7,2%
10,6%
11,1%
11,1%
3,3%
8,3%
2,8%
75,0%
2
12
11
3
0
0
3
7
0
38
1,1%
6,7%
6,1%
1,7%
0,0%
0,0%
1,7%
3,9%
0,0%
21,1%
Beyond 2 places, other floor
0
0
0
14
16
4
1
0
0
35
Within 2 places, other floor
1
7
2
2
4
16
2
8
5
47
Within 2 places, behind area
15
0
0
0
0
0
0
0
0
15
8,9%
3,9%
1,1%
8,9%
11,1%
11,1%
1,7%
4,4%
2,8%
53,9%
% From total observations
% From total observations
Table 18. Scan results attempt 1 - Wing 3 only
Second test run
In the second test run, the same locations in were being compared, but this time the focus was only
on wing 3. In addition, the database was reconfigured and remeasured (new database) with the
same device as that it is being read, to marginalize the errors in the database caused by using a highend receiver. Instead of an average scan for the best 10 MAC addresses, only the routers on the
same floor as the location node has been used (see Figure 34), to limit the search space. The
rationale is that it will reduce the probability of placing the user on the wrong floor, as the signals
that come through ceilings and floors have been much weakened as such. From the 9 x 20 = 180 scan
results, 81 times the user was located near the appropriate location, from which 46 times was at the
exact location as predicted (=25.6% of the total 180, or 56.8% from all 81 good matches, see also
Table 19). This is an increase of 20% for all (near-)correct nodes. The exact matches have been
increased from 7.8% to 25.6%.
The results at locations which are situated near the end of the hallway or at open spaces are prone
to false location conversions, when taking a closer look at the direct results (see Annex C.1). This is
most likely being the cause of the availability of only the strong signals from a nearby router,
whereas the weaker signals of faraway routers are hard to obtain. In open spaces, it is possible that
many other signals from floors below are easier to receive, as the signals elsewhere need to
penetrate ceilings and floors first, which results in much weaker signals to be interfering the scanning
results. For example, it has been found that there were no fingerprints recorded at the intermediate
floor levels in the staircases, making comparisons not possible. Being halfway, the device picks up
signal strengths and compares it with the database, and places the user on a floor (while the user
finds itself between floors), or in a wrong wing if the location is not too far away from a different
wing.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 105
50
45
40
3211
35
3209
30
3203
25
3110
20
3105
3101
15
3011
10
3007
5
3002
0
Exact match
Within 1
places
Within 2
places
Beyond 2
places, same
floor
Beyond 2
Within 2
places, other places, other
floor
floor
Figure 37. Second scan results – grouped by match, specified by location - wing 3 nodes
Node
Close matches:
% From total observations
Exact match
3002
3007
3011
3101
3105
3110
3203
3209
3211
8
12
8
3
10
10
10
13
7
TOTAL
81
4,4%
6,7%
4,4%
1,7%
5,6%
5,6%
5,6%
7,2%
3,9%
45,0%
3
11
8
1
10
0
0
8
5
46
1,7%
6,1%
4,4%
0,6%
5,6%
0,0%
0,0%
4,4%
2,8%
25,6%
Within 1 places
5
0
0
2
0
8
9
5
2
31
Within 2 places
0
1
0
0
0
2
1
0
0
4
2,8%
0,6%
0,0%
1,1%
0,0%
5,6%
5,6%
2,8%
1,1%
19,4%
% From total observations
% From total observations
Out of match:
12
8
12
17
10
10
10
7
13
99
6,7%
4,4%
6,7%
9,4%
5,6%
5,6%
5,6%
3,9%
7,2%
55,0%
Beyond 2 places, same floor
0
2
0
0
10
5
10
4
13
44
Beyond 2 places, other floor
5
6
10
17
0
4
0
3
0
45
Within 2 places, other floor
7
0
2
0
0
1
0
0
0
10
6,7%
4,4%
6,7%
9,4%
5,6%
5,6%
5,6%
3,9%
7,2%
55,0%
% From total observations
% From total observations
Table 19. Scan results attempt 2 - Wing 3 only
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 106
17.2.3. Indoor Wi-Fi Navigation testing
The navigation, with the Wi-Fi fingerprints as location check, is the core of the application. The user
has planned a journey, and the position is being monitored with the Wi-Fi fingerprints. The user has
to follow the rules and directives provided and should receive a message whether the user is on or
off route.
Routes are being calculated with the graph and Dijkstra algorithm. Neutral directions are provided,
along with the node numbers. If the device picks up the signals and translates that into a node ID,
this is being compared with the list of nodes in the calculated route. If this ID is within the list, the
message “on route” is being displayed, if not, “off route” is being displayed.
Within wing 3 of the OTB building, various routes have been traversed using the application. Five
different type of routes have been travelled upon in different travel directions, as displayed in Table
20. For each route, the number of “off route” messages have been registered, as well as all node
numbers that the device picks up during the journey, which is being compared with the provided list.
Upon registration of the correctness of the information provided, the following scenarios are being
used:



The application detects the user at a location that is provided in the list, and the user finds
himself at the location that should match the list;
o Example: user is at location FID=3001; the device provides message user is at FID=3001.
o This will be registered as fully correct (in Table 20 denoted as “on” – on route).
The application detects the user at a location that is provided in the list, but the user is not at the
location that he should be at;
o Example: user is at location FID=3001; the device provides message user is at FID=3002.
o This will be registered as partly correct (in Table 20 denoted as “on/off”).
The application detects the user at a location that is not provided in the list, and the user does
not find himself at that location;
o Example: user is at location FID=3001; the device provides message user is at FID=2004.
o This will be registered as incorrect (in Table 20 denoted as “off” – off route).
As expected, the results in correct route confirmation are similar to those as the general location
confirmation, since the checks are based on location, although in this case, the test run involved a
swift FID acquisition (maximum of 1 second waiting at a location), instead of waiting for a long
acquisition time (20 seconds as in the previous section). In 41.3% of all route confirmation provisions,
a correct message has been given. Upon examining the variations between the different routes, it
seems that the results were poor on the route that was fully from one end to another on the first
floor (route 2), as well as the route that dealt with travelling within the north end (route 5). From the
former, this can be explained that despite using routers within the same physical space, the device
still picks up signal strengths from floors above or below the user. Since the probability a signal
strength from a floor below the bottom most or above the top most floor is nil, that comes in no
surprise.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 107
Series 1
Route
1
2
3
4
5
Series 2
Route
1
2
3
4
5
Series 3
Route
1
2
3
4
5
Total
Route
1
2
3
4
5
Total
Characteristic
South End 0F - North End 2F
South End 1F - North End 1F
South End 2F - North End 1F
South End 0F - South End 2F
North End 2F - North End 0F
From
3005
3101
3201
3001
3212
Via
3009
3107
3110
3103
3110
To
3212
3112
3112
3201
3011
Nodes
9
11
12
7
7
On
5
4
6
2
2
On/off
2
3
4
1
3
Off
2
4
2
4
2
On
55.6%
36.4%
50.0%
28.6%
28.6%
On/off
22.2%
27.3%
33.3%
14.3%
42.9%
Off
22.2%
36.4%
16.7%
57.1%
28.6%
Characteristic
South End 0F - North End 2F
South End 1F - North End 1F
South End 2F - North End 1F
South End 0F - South End 2F
North End 2F - North End 0F
From
3005
3101
3201
3001
3212
Via
3009
3107
3110
3103
3110
To
3212
3112
3112
3201
3011
Nodes
9
11
12
7
7
On
4
4
5
3
4
On/off
3
2
5
0
0
Off
2
5
2
4
3
On
44.4%
36.4%
41.7%
42.9%
57.1%
On/off
33.3%
18.2%
41.7%
0.0%
0.0%
Off
22.2%
45.5%
16.7%
57.1%
42.9%
Characteristic
South End 0F - North End 2F
South End 1F - North End 1F
South End 2F - North End 1F
South End 0F - South End 2F
North End 2F - North End 0F
From
3005
3101
3201
3001
3212
Via
3009
3107
3110
3103
3110
To
3212
3112
3112
3201
3011
Nodes
9
11
12
7
7
On
3
4
5
4
2
On/off
3
3
1
0
0
Off
3
4
6
3
5
On
33.3%
36.4%
41.7%
57.1%
28.6%
On/off
33.3%
27.3%
8.3%
0.0%
0.0%
Off
33.3%
36.4%
50.0%
42.9%
71.4%
Characteristic
South End 0F - North End 2F
South End 1F - North End 1F
South End 2F - North End 1F
South End 0F - South End 2F
North End 2F - North End 0F
From
3005
3101
3201
3001
3212
Via
3009
3107
3110
3103
3110
To
3212
3112
3112
3201
3011
Nodes
27
33
36
21
21
138
On
12
12
16
9
8
57
On/off
8
8
10
1
3
30
Off
7
13
10
11
10
51
On
44.4%
36.4%
44.4%
42.9%
38.1%
41.3%
On/off
29.6%
24.2%
27.8%
4.8%
14.3%
21.7%
Off
25.9%
39.4%
27.8%
52.4%
47.6%
37.0%
Table 20. Results of route confirmation results within wing 3
A different approach upon testing the route confirmation is not by checking the correct location
confirmation, but on the way the route guidance is provided along the way. In this project, a simple
topological route network guidance has been provided. Without the presence of geometry, it is hard
to tell whether the user has to turn left or right, and therefore it is hard to tell whether the user’s
orientation is correct. This can be relieved if information about correctness is involved. For instance,
if the user traverses a floor from one end to another, he can get confirmation such as “You are doing
correct if you see the add office room numbers are adding up”. There is still no geometry involved,
but still some sort of route confirmation is present. Further research is desirable to do so.
After a short demonstration, test persons have noted many expansions are worthy, which includes
the addition of photos, graphs, images, maps and pictograms. In addition, it was noted that the text
output was rather long or confusing. This confirms the weakness of providing neutral text based
directions. As there is no geometry involved, it is rather difficult to tell whether the user has to turn
left or right. The next step of navigation would indeed be implementing geometric features.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 108
17.3. Remarks and conclusions on the application
How can Wi-Fi fingerprinting determine a location which does not utilize geometric features, and how
can it be implemented in an application as such?
Not utilizing geometric features requires the developer to realize that a navigable graph network
must emphasize on the topology and structure that exist between navigable places.
First of all, the results are influenced by the type of receiver that was being used to record the data
itself. A high-end receiver such as a laptop provides better scan results than the middle-end
smartphone receiver, which was visible from the fingerprint database. As a result, the device that is
being used for the application might over- or underestimate the recorded scan results. It means that
the fingerprint database needs frequent updates, and that there are multiple databases necessary,
for each class of end receivers.
Second, the location methodology is a simplified version of map-matching algorithms, in the sense it
does not uses geometric features. A least sum of square is hard to use, since Android automatically
assigns either non-existent values (-100 dBm) or blank values, which are in both cases sensitive to
providing false location. The count method location determination itself might not be fair as well,
since it only checks the range from the signals that have been recorded and evaluated. As such,
stronger signals have an equal weight as weak signals. With that regard, smart localization should be
implemented.
Third, results on the localization is influenced by the database records: better results have been
found when only MACs from the same physical space have been used in the recording phase,
although that is not a guarantee: still less than 50% provides a correct, or near-correct estimation.
A few suggestions about the application and its development are mentioned: First, automatizing the
process of surveying can be done with the smartphone itself. In the first place, the signal strengths
can be recorded, but by building an application in which the surveyor can add location information to
this signal strength, a graph (and thus navigable network) can be created.
Second, a combination of other techniques could improve the localization method, as well as the
navigation module. For instance, an accelerometer, which is also present at present-day
smartphones, can help detect orientation aspect, and from there, clearer route descriptions can be
given, instead of neutral descriptions. As Wi-Fi is rather unstable due to the variations in signal
strength (see annex C.1 and C.2), further research should be carried out to solve this.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 109
PART IV - Conclusions
Contents





18. Final Conclusion
19. Limitations
20. Recommendations
References
Annexes
In this fourth and final part, the output is being discussed. In chapter 18, the final conclusions are
being drawn and the main research question is it possible to develop an indoor navigation service for
a mobile platform, with the use of Wi-Fi fingerprinting technology and Open Location Services
standards, but without using geometry? will be answered. The results are further being evaluated
with its limitations and recommendations in the chapters 19 and 20 respectively.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 110
18. Final Conclusion
In this thesis project, attempts have been made to investigate the possibilities of indoor navigation
on a smartphone, with the use of Wi-Fi technology, but without using building geometry. The
research question was:
Is it possible to develop an indoor navigation service for a mobile platform, with the use of only Wi-Fi
fingerprinting technology and the framework of the Open Location Services standards, but without
using building geometry?
It is possible to launch an indoor navigation service for a mobile platform, with Wi-Fi fingerprinting
technology. The framework of the Open Location Services were provided as a guideline to set up the
application, but they remained unused, as they cannot be combined with a project which does not
use geometric features such as maps, since the parameters are focussed on that aspect. However,
the location predictions are too unreliable to be used.
Caution has to be taken upon fingerprinting, since fingerprinting assumes that the characteristic
signal strengths are (1) dependent on physical objects present in the navigable space due to
multipath and (2) are changing over time. In addition, their transmitters, the routers or Media Access
Control units (MAC), can be replaced as well. Upon site surveying, it is highly recommended to store
at least a variety of MACs, along with their respective signal strength, at specific locations. In
addition, great care has to be taken upon registering the MACs themselves, since they have to be
combined with location information, which indirectly leads to identifiable persons. This, in turn, can
be prohibited by law in countries such as the Netherlands. Indoor navigation should thus be limited
to public buildings only.
Two types of location determination have been proposed: a count method, and the least sum of
squares method. In Android, both methods are constrained by how access points out of reach are
being treated: non-detectable access points are either assigned with a value of -100 dBm, or as a
blank value. Regarding the least sum of squares (SS) method, in the former case this will lead to the
promotion of recorded weak signal strengths at certain locations. In the latter case, this will lead to
unequal SS value to compare, as one location might have a SS value based on five access points, and
another location might have a SS value based on seven access points, in which it is most likely the
location with the fewest access points prevails. The count method will solely check for matches in a
defined search space. This includes a search margin of roughly 3-5 dBm. In this project the value of 4
dBm as a search space margin has been used.
It has been seen that upon using recorded values from the ten best access points within reach, often
false locations have been provided. In a short live scanning survey, only 25.0% of all 180
measurements were within an acceptable predicted location, which equals the location being
perceived within a maximum of two nodes away, at the same floor. In only 7.8% of all 180 cases, the
location was exactly the same. In 53.9% of all cases, the location estimation was on the wrong floor
and far away. When the database is being replaced with only MAC addresses from access points in
the same physical space (= same floor) as the fingerprint/node, the results seems to improve: 45.0%
of all 180 measurements was now within the acceptable predicted location, and 25.6% of all
measurements were perceived at the exact same spot as where the device was. This has been at the
expense of the estimations far away on wrong floors: this percentage has been decreased to 30.5%
of all cases.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 111
Further conclusions can be drawn regarding the localization algorithm and the database. First of all, it
is desirable that the results obtained from the site surveying should be done with the same device as
the live scanning results, which has been done in the second test runs. This is to prevent the
database to be filled with recordings that are too high or too low, in which the amount of over- and
underestimation can be diminished. Second, certain locations are highly likely to provide false
location information, as they are dependent on the building structure, and thus highly influent by
multipath effects. These locations where fingerprints are stored, includes locations at staircases, or
near the end of the hallway. However, it is rather difficult to realize stable location detection.
The navigation algorithm being used, the Dijkstra algorithm, can be further expanded, with options
regarding user profiles. The navigation output itself has to be improved, since it now uses neutral
descriptions, as there is no orientation aspect, except for recognizable objects that are being used to
provide directions. This is clearly one of the weaker aspects of the project, since it does not use
building geometry at all. It is recommended that previous or future locations, should also be used to
provide route guidance to the user. Furthermore, paths that are in the same direction of extension
are now all treated as single paths. Smart routing in the sense of providing directions such as ‘go to
the end of the hallway near point E’ instead of ‘go towards point A, B, C, D, E’ is highly
recommended. With that respect, the integration of additional technologies, such as a accelerometer
could provide a useful extension, as it will help to incorporate geometry after all.
The representation is also difficult to improve when one considers solely text-based navigation.
Enhances in the application interface are desirable, such as enabling text-to-speech, routing by user
profiles, displaying graphs and active maps, and estimated times of arrival, includes better route
handling and narrowing down location determination algorithms and: this prototype checks whether
the determined location matches with a calculated route, and states whether the user is on or off
route. Ideally, the route has to be recalculated if the user does not find itself at its path predicted. In
addition, by narrowing the location determination, such as ‘only looking in the search space at the
same floor’ or ‘look at only three nodes away from me’ further upgrades the prototype. The
incorporation of other localization techniques, along with its enhancements (such as an
accelerometer), would be helpful.
To answer the main research question, a simple no would suffice, as summed up:







Wi-Fi signals are instable and fluctuate significantly;
Quality of fingerprinting depends on quality of receiver for recording versus live reading;
Fingerprinting databases requires frequent updating;
Emphasis on topology is possible without geometric features, but orientation aspects are difficult
to solve;
Parameters of the OpenLS cannot be used, since they are dependent on geometric features,
although their framework can be used;
Localization methodologies dependent on how Android treats out of reach routers, and on the
how the amount of mac addresses being read are treated equally;
It is proven that despite changing the received signal strengths in favour of the device, below
50% of all location estimations were correct, making the reliability of the whole application poor.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 112
19. Reflection: discussion, limitations and recommendations
In this chapter, limitations and recommendations are being discussed. First, limitations are addressed
in section 19.1, followed by recommendations in section 19.2 regarding the application itself, and in
section 19.3 regarding localization methods.
19.1. Limitations in general
Logging (as in recording the users complete path, not to be confused with tracking, which is following
the users path) and the maintenance are not included in this research, since it is not the goal to allow
users to track their progress, but rather navigating them from one end to another. However, tracking
could be useful in combination with pedometer-like application for instance. Finally, since it is the
purpose to create a prototype, no maintenance is required either.
Unlike the outdoor navigation, data on indoor locations is not widely available. There are no vendors
for networks such as FalkPlan and TeleAtlas, and not all companies are willing to provide floor plans
for public use, or even if there is a provision, it might be possible that not everything is covered to
keep it secret. This might cause less accuracy, or missing data. On the other hand, since the nongeometry is the keyword to this thesis, it might be possible not to work with maps at all, therefore
limiting the data usage in an advantageous way.
Fingerprinting requires a lot of pre-processed work, in the form of site surveying. Also, the fact that it
is capable of handling multi-path makes it a burden at the same time. As soon as large objects,
including people, are moved, the fingerprints will surely be affected. The same applies for the
fixation of MAC addresses. Since the received signal strengths are dependent on the MAC address,
patterns might not be found, as soon as an access point has been replaced by another. An application
thus has to be constantly updated, and from time to time a new survey has to take place. Warnings
about imprecise location estimation as such should be present.
Text descriptions itself are useful, as they are capable of steering the user on what he or she has to
do. However, as Hsu (2011) already pointed out, is that the attention of users is in the visualization.
Photos, pictograms and maps are in that sense better options. The application can thus further
enhanced with visual media as such. Yet, when maps are being used, the concept of localization with
fingerprinting must be changed to the concept of positioning with map matching, which requires
different implementation algorithms. Ideally, the reverse geocoding from fingerprints to location
information can be re-geocoded again, to known, absolute locations. In that sense, indoor navigation
can be further enhanced, and it is highly recommended to do so in future research.
The aim of this thesis was to investigate possibilities regarding indoor Wi-Fi without building
geometry. It was not the intention to develop a full and complete product, but rather on how to use
the various aspects of navigation. As such, the result is that the location determination, and the
navigation are standalone versions as well. Ideally, those two have to be combined for real live
navigation as one is known for car navigation, with the addition of the confirmation of where the
user is – either on route, or off route. If this incentive can be reached, the use of maps or graphs can
be further enhanced, as the device might show the user where the location on the graph is.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 113
19.2. Regarding the application
More recommendations can be discussed for the prototype itself.
A first recommendation is that the database should be stored on a remote server, so that users will
not need to update the application constantly regarding changes of macs and signals. The downside
however, is that it requires at least a 3G wireless connection, in case the user cannot establish a data
connection via the Wi-Fi access points. 3G cannot be guaranteed in rooms which blocks the signals of
mobile telecommunication. In this occasion, it is likely that Wi-Fi signals cannot reached either.
Further recommendations are suggested for the user interface. Enhancements not related maps, but
to the input and output, can be done in certain ways. A few are named below:
1. Only a departure and destination can be selected, but it is possible to include via waypoints, or to
exclude via waypoints. For instance, a user would like to go from one point in the building, to
another point, but the user would like to have a copy machine on his way, and the user would
like to avoid a certain section, since there are ongoing refurbishments in that particular part of
the building.
2. The inclusion of routes for impaired users, such as wheelchair users. In the database, it has been
taken into account which routes are navigable for wheelchairs, as for each edge, information is
stored whether a ramp is available, or where the elevators are located. In the prototype, there
has been room left for this, as in the Settings screen this has already been created.
3. The inclusion of text-to-speech. Android has a text-to-speech engine, in which the user can hear
the directions the application provides. This might help the user, when the user prefers to have
his smartphone socketed in his pocket for instance.
There are also limitations and recommendations regarding privacy and legal issues, which are
explained in the next chapter.
19.3. Regarding the localization methodologies
In this project, two localization methodologies have been proposed, in which one has been used: the
count method was the preferred method over the least sum of squares due to issues in the way how
Android treats weak and bad signals. Aside from that, the number of entries per location node
influences the outcome of the least sum of squares, as well as the maximum number of valid records.
Further research on both methodologies in thus necessary.
These methods behave different than the map-matching algorithms as discussed in the literature
research, since there is no geometry to match with. This assumes that there is a given location, in
which both location information and signal strength characteristics can be stored. By combining this
information, a node can be created, and eventually, a graph, in which various network algorithms can
be applied. This can be further extended by explicitly stating which adjacent rooms belong to which
node. This way, geometry is not being used, but rather the connectivity and topology is preserved.
The importance of topology must be emphasized in further research: since in route guidance, no
left/right can be distinguished, alternative solutions must be found. User awareness plays a role
here: route confirmation can also made possible by using logics from the real environment, such as
“you are still on route, if you can see the signs of the office room numbers are adding up”.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 114
In addition, MAC addresses themselves are dynamic and can be changed at will. This leads to
unstable databases in time: it is recommended that upon up-scaling to larger coverage areas,
updates should occur frequently. Aside from that, the signal strength characteristics still remain
prone to multipath effects, which results in false location information. This can be countered by
effectively using the MACs: for instance: only use MACs that are at the same floor as the fingerprint.
This prevents looking for outlying signal strengths, which can be unstable anyway.
The search space in which the signal strengths have to be looked up from, can be changed at will as
well. The larger the search space, the more problems are foreseen with matching locations: as the
larger it is, the more locations can be made available. Moreover, it is recommended that the
locations to be registered should not be too far away from each other, as otherwise, intermediate
locations are being detected as somewhere else, while that has not be the case per say. On the other
hand, the to be fingerprinted locations should not be too close to each other, since it is then difficult
to perceive which location belongs to what. It is noteworthy, that the intention of this project was to
map locations and not map each single possible position – there is simply too much differentiation in
signal strength to do so.
Further automatizing is possible by developing an application, which is capable of directly registering
average signal strengths from MACs, in which location information can be assigned. By repeating this
for several spots, and by adding information about which spots are connected in physical space, it is
easy to build up a graph.
Finally, the databases are dependent from the device being used. A high-end receiver, such as a
laptop provides different outcomes than using the smartphone directly for both recording as
receiving. A clear disadvantage however, is that multiple databases are needed: each database that
should be adapted to the quality of the Wi-Fi receiver.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 115
20. Additional remarks on privacy and legal issues
The application faces multiple issues regarding privacy and legal aspects, which are addressed below.
In this chapter a MAC (Media Access Control) refers to the address from a transmittable access point
(routers), as non-transmittable devices, such as computers also contain MACs.
20.1. Personal information
Originally, the intention was to include personal information in the application. This is related to the
available cataloguing options in the application. In the case a user is looking for the office of a certain
person, the user can search for the name of the person, in which the proper room is being given.
Thus, the name of a staff member was available in combination with the room. However, this implies
that a significant amount of personal information has to come with the application, resulting in an
application which can be data intensive, and is vulnerable to hacking as well. Yet, if not too much
effort can be done to retrieve personal information via one information channel (such as the office
rooms of personnel, which can be browsed by a corporate website), it is probably not too sensitive,
hence, it can be used in the application.
20.2. MAC as personal information
Late April 2011, Google was fined €250,000 by a Dutch lawsuit, as Google deliberately recorded all
MAC addresses they detected during their coverage of Google Street View (NRC, 2011). Recording
the MAC address (the Base SSID, or BSSID) itself is not punishable by law, as long as personal
information is not combined with it, according to the College Bescherming Persoonsgegevens (CBP),
a quasi-non-governmental organization in the Netherlands, which concerns about the protection of
personal data and privacy. However, Google combined the MAC address with a WGS84 Coordinate,
along with street names, postal codes, places and more information, in which then the MAC address
itself turns into personal data (CBP, 2010; CBP, 2011). Thus a MAC address can be considered as
proprietary data. This resulted in a forced deletion of the acquired information, as demanded by the
CBP. The law (WBP Article 8.f) is roughly translated (CBP, 2011):
Data about persons may only be processed when data processing is crucial for the (representation of)
the justifiable interest for the responsible person, institute or third member to whom the data will be
issued, unless the interest or fundamental rights and freedom of the involved person prevails, with
special attention regarding the right to the private life.
This issue is strongly related with the application, since it utilizes MAC information, which is being
correlated with the received signal strengths. To avoid lawsuits, the application should not be
coupled with other personal information.
20.3. Coverage
Correlated with the MAC as personal information depicted and personal information in general in the
previous sections, the coverage area itself is prone to privacy and legal issues as well. Since MACs are
proprietary (as they can be bought for a specific individual or group), it is reasonable to limit the
coverage of a service deployment to public buildings, such as hospitals, conference centres, shopping
malls and governmental buildings. This will reduce the amount of personal and sensitive information.
Finally, it is most unlikely a user is trying to navigate through very small and refined rooms such as in
private objects as houses and apartments. As required by the CBP, upon site surveying, the
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 116
persons/institutions who own or work with their MAC, should be informed their MAC is being used
for location information.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 117
References
Albrecht-Buehler, C., B. Watson & D.A. Shamma (2005), Visualizing Live Text Streams Using Motion
and Temporal Pooling. IEEE Computer Society, pp. 52- 59.
Android Developers (2011), Developers Guide - Using Databases. Retrieved [August 18 2011]
http://developer.android.com/guide/topics/data/data-storage.html#db.
BIPM, Bureau International des Poids et Mesures (2008), International vocabulary of metrology –
Basic and general concepts and associated terms (VIM). Retrieved [February 24 2011]
http://www.bipm.org/utils/common/documents/jcgm/JCGM_200_2008.pdf.
Boguslawski, P, C.M. Gold & H. Ledoux (2011), Modelling and analysing 3D buildings with a
primal/dual data structure. ISPRS Journal of Photogrammetry and Remote Sensing (66), pp. 188-197.
Bowditch, N. (2002), The American Practical Navigator – An Epitome Of Navigation. 2002
Bicentennial Edition. Bethesda, MD: National Imagery and Mapping Agency.
Chiou, Y., C. Wang, S. Yeh & M. Su (2009), Design of an adaptive positioning system based on WiFi
radio signals. Computer Communications 32, pp. 1245-1254.
CBP, College Bescherming Persoonsgegevens (2010), Definitieve bevindingen. Onderzoek CBP naar
de verzameling van Wifi-gegevens met Street View auto’s door Google. Z2010-00582, openbare
versie. Retrieved [April 19 2011]:
http://www.cbpweb.nl/downloads_rapporten/rap_2011_google.pdf
CBP, College Bescherming Persoonsgegevens (2011), Last onder dwangsom / opdracht tot
vernietiging payload data. Retrieved [April 19 2011]:
http://www.cbpweb.nl/downloads_pb/pb_20110419_brief_google_lod.pdf
Dawes, B. & K. Chin (2011), A comparison of deterministic and probabilistic methods for indoor
localization. The Journal of Systems and Software 84, pp. 442-451.
De Koning, M. (2010), Quality assessment of GSM positioning. GSM fingerprinting versus Cell-ID
positioning. MSc Thesis made available upon request.
De La Roche, G., P. Flipo, Z. Lai, G. Villemaud, J. Zhang & J. Gorce (2010), Implementation and
validation of a new combined model for outdoor to indoor radio coverage predictions. Journal on
Wireless Communications and Networking, pp. 1-9.
Döllner, J., B. Hagedorn & S. Schmidt (2007), An approach towards Semantics-Based Navigation in 3D
City Models on Mobile Devices. Lecture Notes in Geoinformation & Cartography 2 (4), pp. 357-368.
Dovis, F., R. Lesca, G. Boiero & G. Ghinamo (2010), A test-bed implementation of an acquisition
system for indoor positioning. GPS Solutions 14, pp. 241-253.
ETSI (European Telecommunication Standard Institute) (2010), On-demand Base Maps on the Web
generalized according to User Profiles. Internal document submitted.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 118
Foerster, T. (2010), Web-based architecture for on-demand maps – integrating meaningful
generalization processing. Retrieved [August 18 2011]
http://www.itc.nl/library/papers_2010/phd/foerster.pdf
Haenselmann, T. (2005), Lecture on Sensor Networks. Retrieved [February 24 2011]
http://www.slideshare.net/netfet/localization-presentation.
Hsu, C.C. (2011), Factors affecting webpage’s visual interface design and style. Procedia Computer
Science 3, pp. 1315-1320.
Ibrahim, A. & D. Ibrahim (2010), Real-time GPS based outdoor WiFi localization system with map
display. Advances in Engineering Software 41, pp. 1080-1086.
IEEE (2011), IEEE Standard for Information technology — Telecommunications and information
exchange between systems — Local and metropolitan area networks — Specific requirements — Part
11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications. Retrieved
[March 1 2011] http://standards.ieee.org/about/get/802/802.11.html.
Intini, A.L. (2000), Orthogonal Frequency Division Multiplexing for Wireless Networks - Standard IEEE
802.11a. Retrieved [March 16 2011] http://www.create.ucsb.edu/ATON/01.01/OFDM.pdf.
Jan, S., L. Hsu & W. Tsai (2010), Development of an Indoor Location Based Service Test Bed and
Geographic Information System with a Wireless Sensor Network. Sensors 10, pp. 2957-2974.
Kessler, B. (2006), Direct Sequence Spread Spectrum and You. Presentation Retrieved [March 14
2011] http://www.eecs.tufts.edu/~bkesslr/EE194SDR/Final/DSSS.ppt.
Klukas, R., O. Julien, L. Dong, E. Cannon & G. Lachapelle (2004), Effects on building materials on UHF
ranging signals. GPS Solutions 8 (1), pp. 1-8.
Kolodziej, K. & J. Hjelm (2006), Local Positioning Systems. LBS Applications and Services. Boca Raton,
FL, USA: CRC Press – Taylor & Francis Group.
Kraak, M.J. & F.J. Ormeling (2003), Cartography – Visualization of Geospatial Data. Second edition.
Harlow, UK: Pearson Education Limited.
Le, M.H.V., D. Saragas & N. Webb (2009), Indoor Navigation System for Handheld Devices. A Major
Qualifying Project Report. Retrieved [May 20 2011] http://www.wpi.edu/Pubs/E-project/Available/Eproject-102209-164024/unrestricted/Indoor_Navigation_System_for_Handheld_Devices.pdf.
Lee, K.-C. (2007), Localization Systems Using Signal Strength Fingerprinting. Retrieved [July 11 2011]
https://circle.ubc.ca/bitstream/handle/2429/28750/ubc_2010_fall_lee_kungchung.pdf.
Lim, H. L. Kung, J.C. Hou & H. Luo (2010), Zero-configuration indoor localization over IEEE 802.11
wireless infrastructure. Wireless Networks 16, pp. 405-420.
Lu, F. & P. Lai (2006), A Shortest Path Algorithm Based On Limited Search Heuristics. Lecture Notes
on Computer Science (3967), pp. 487-497.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 119
Manodham, T., L. Loyola & T. Miki (2008), A Novel Wireless Positioning System for Seamless Internet
Connectivity based on the WLAN Infrastructure. Wireless Personal Communications 44, pp. 295-309.
Mehmood, H., N.K. Tripathi & T. Tipdecho (2010), Indoor Positioning System Using Artificial Neural
Network. Journal of Computer Science 6 (10), pp. 1219-1225.
Milioris, D., L. Kriara, A. Papakonstantinou, G. Tzagkarakis, P. Tsakalides & M. Papadopouli (2010),
Empiral Evaluation of Signal-Strength Fingerprint Positioning in Wireless LANs. Publication from
Institute of Computer Science. Retrieved [July 11 2011]
http://www.ics.forth.gr/mobile/publications/mswim2010.pdf.
Nagi, R.S. (2004), Cartographic visualization for mobile applications. Thesis. Retrieved [March 11
2011] http://www.itc.nl/library/papers_2004/msc/gfm/nagi.pdf.
NRC (2011), Dwangsom voor Google om verzamelen wifi-gegevens. News Article Retrieved [April 19
2011]: http://www.nrc.nl/nieuws/2011/04/19/dwangsom-voor-google-om-verzamelen-wifigegevens/
OGC (Open Geospatial Consortium) (2008), Open GIS Location Services (OpenLS): Core Services.
Retrieved [January 29 2011] http://www.opengeospatial.org.
OGC (Open Geospatial Consortium) (2008), Open GIS Location Services (OpenLS): Part 6-Navigation
Service. Retrieved [March 10 2011] http://www.opengeospatial.org.
OGC (Open Geospatial Consortium) (2010), Requirements and Space-Event Modeling for Indoor
Navigation. How to simultaneously address route planning, multiple localization methods, navigation
contexts and different locomotion types. Retrieved [January 11 2011]
http://www.opengeospatial.org.
Reichenbacher, T. (2005), Chapter 10: Adaptive egocentric maps for mobile users. In: L. Meng, A. Zipf
& T. Reichenbacher (2005), Map-Based Mobile Services. Theories, Methods and Implementations,
pp. 141-158. Springer Berlin: Heidelberg.
Roos, T., P. Myllymäki, H. Tirri, P. Misikangas & J. Sievänen (2002), A probabilistic Appraoch to WLAN
User Location Estimation. International Journal of Wireless Information Networks 9 (3), pp. 155-163.
Roto, V. & A. Kaikkonen (2003), Perception of Narrow Web Pages on a Mobile Phone. Proceedings of
the 19th International Symposium on Human Factors in Telecommunication (HFT 2003), pp. 205-212.
Sarjakoski, L.T., T. Koviula & T. Sarjakoski (2007), A Knowledge-Based Map Adaptation Approach for
Mobile Map Services. Lecture Notes in Geoinformation & Cartography 2 (3), pp. 247-264.
Shum, K.C.Y, Q.J. Cheng, J.K.Y Ng & D. Ng (2011), A Signal Strength based Location Estimation
Algorithm with a Wireless Network. 2011 International Conference on Advanced Information
Networking and Applications, pp. 509-516.
Sturgeon, T. et al. (2011), Wi-Fi Virtual Laboratory. Retrieved [March 11 2011] http://wifi.cs.standrews.ac.uk/wififrame.html.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 120
Swangmuang, N. & P. Krishnamurthy (2008), An effective location fingerprint model for wireless
indoor localization. Pervasive and Mobile Computing 4, pp. 836-850.
Usher, J.M. & L. Strawderman (2010), Simulation operational behaviors of pedestrian navigation.
Computer & Industrial Engineering (59), pp. 736-747.
Vogel, L. (2009), Dijkstra’s Shortest Path Algorithm in Java. Retrieved [May 1 2011]
http://www.vogella.de/articles/JavaAlgorithmsDijkstra/index.html
Wang, A.H. & C.H. Chen (2003), Effects of screen type, Chinese typography, text/background color
combination, speed, and jump length for VDT leading display on users’ reading performance.
International Journal of Industrial Ergonomics 31, pp. 249-261.
Wi-Fi Alliance (2009), Wi-Fi Certified n: Longer-Range, Faster throughput, Multimedia-Grade Wi-Fi
Networks. Retrieved [March 1 2011] http://www.wi-fi.org/register.php?file=wp_WiFi_CERTIFIED_n_Industry.pdf
Winter (2002), Modeling Costs of Turns in Route Planning. GeoInformatica 6 (4), pp. 345-361.
Woo et al. (2011), Application of Wi-Fi-based indoor positioning system for labor tracking at
construction sites: A case study in Guangzhou MTR. Automation in construction 20, pp. 3-13.
Wu, Z., Q. Zeng & X. Hu (2009), Mining Personalized User Profile Based on Interesting Points and
Interesting Vectors. Information Technology Journal 8 (6), pp. 830-838.
Yim, J. (2008), Introducing a decision tree-based indoor positioning technique. Expert Systems with
Applications 34, pp. 1296-1302.
Yim, J., S. Jeong, K. Gwon & J. Joo (2010), Improvement of Kalman filters for WLAN based indoor
tracking. Expert Systems with Applications 37, pp. 426-433.
Zhang, D. et al. (2010), Localization Technologies for Indoor Human Tracking. Paper submitted to the
5th International Conference on Future Information Technology (FutureTech), May 2010, Busan,
Korea. Retrieved [January 26 2011] http://arxiv.org/ftp/arxiv/papers/1003/1003.1833.pdf.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 121
Annexes
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 122
A. Tables and scripts
A.1. Table of overview possible sensor technologies for location based services
Technique
(Assisted) Global Positioning
System
Range
Anywhere
outdoor
Accuracy
6.0 m - 12.0 m
Remarks
 Satellite based positioning
+ Low barrier entry
- Energy consumptive (Assisted GPS in a lesser extent)
- Slow computation and processing time
- Very susceptible to reflectance and multi-paths
Allocation methods
Trilateration
Global System for Mobile
Communication (GSM) / Universal
Mobile Telecommunication
System (UMTS)
Infrared (IR)
≈ 35.0 km
Cell-based
 Standard data- and telephone communication radio waves.
+ Globally available
- Cell-based accuracy
Cell-ID
Signal Strength
0.7 m – 2.5 m
Dependent on
range of
application
 Device tagging with ID’s
- Short range of detection limits infrastructure
- No penetration of materials / multipath
- Line of sight
- Signal can be disturbed easily
Cell-ID
Infrared Data Association (IrDA)
≈ 2.5 m
1.0 m – 2.0 m
 Same method as IR, yet with higher speed
Cell-ID
 Radio waves exchange between reader and tag, using a
transmitter, transceiver with decoder and unique information.
+ Real-time location systems
+ High-speed response time
+ read/write capabilities
- No communication network
- No positioning information
- Manual programming
Cell-ID
Radio Frequency Identification
(RFID)
Active RFID
< ≈ 100 m
 Tags contain internal power source, signalling upon entering
range (toll boots)
Passive RFID
1.5 m – 2.0 m
 No internal power source, only reading/writing upon entering
range/connection (smartcards)
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 123
Technique
Range
Accuracy
Remarks
Allocation methods
Ultrasound
Dependent on
environment
3.0 cm – 1.0 m
 Pulse (radar) emittance
- No penetration of materials / multipath
- Line of sight
- Excessive manual intervention
- Bad reception with intervening noises – extremely sensitive to
environment
Trilateration
Bluetooth
< ≈ 100 m
10 m – 20m
 Frequency hopping for connecting devices and creating Personal
Area Networks (PAN), with high security
+ High speed data transfer
- Explicit links between devices required
- Its mobility also limits positioning and topology
Trilateration
Signal Strength
Ultra Wide Band (UWB)
≈ 50 m (with
reduced
accuracy,
normal: chips
about 9 m, due
to cap of power
amount)
15 cm – 4 m
 Extremely short emittance of pulses
+ Multipath immunity
+ Inherent precision for TOA/TDOA
+ Low power
+ High speed data (nearly 10xWi-Fi)
- Large uncertainty about health effects
- Not everywhere legal
- Economically expensive
Trilateration
Signal Strength
IEEE 802.11 (Wi-Fi)
≈ 32 m (indoor)
≈ 95 m
(outdoor)
1m–5m
 Radio waves transfer at the 2.4 GHz band
+ Large scale available over the world
+ Economical viable
- High power consumption
- Security can be weak
- Slightly multipath susceptible
+ 802.11n at 5 GHz band, range indoor 91m, outdoor 182m
Trilateration
Signal Strength
Table AP. 1. Extended overview of sensor techniques used for allocation of receiver.
Table based on Manodham et al. (2008) and Kolodziej & Hjelm (2006) and Lim et al. (2010).
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 124
A.2. Navigation flows
The OGC (2008b) has listed a set of 30 possible cases (flows) which determines or influences the
navigation. They are listed below:
Flow
Main flow
1
2
3
Flow description
User specifies routing preferences.
User requests a route from the client a route from
origin to destination.
Client requests current position.
Client returns network link(s) (and possibly locations
along) as origin.
5
User specifies destination.
6
Client notes the returned network link(s) (and possibly
locations along) as the destination.
7
Client requests navigation service to provide a route
from origin to destination.
8
Client informs user and guidance function route has
been generated.
9
Client displays route.
Extension: navigation request detour
10
User requests detour re-route after starting route
guidance
11
User requests traffic information and client makes the
request of navigation or traffic server
Remark
In this project detection of
fingerprint
4
User supplies detour information (avoidances)
Client request navigation or route service to provide a
new route based on detour
14
Navigation client displays new route on the map
15
User accepts new route
16
Client informs route guidance function a detour route
has been generated
Extension: rerouting with new waypoints
17
While route guidance is underway, user sets additional
waypoint(s)
18
Client returns location(s) as additional waypoint(s)
19
Client requests navigation (or route) service to provide
a new route, from current location to original
destination, including the new waypoints not yet
visited
20
Client informs user and route guidance function a waypointed route has been generated
21
Client displays the new route
Extension: recovery reroute
22
User travels off current guided route
Geocoding service
In general not applicable for
indoor navigation, unless in
crowded buildings, such as
supermarkets
12
13
Can be done to include certain
points to the route
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 125
Flow
23
Flow description
Client detects user is off-route
24
Client request navigation service to provide a recovery
route from the current user position back onto the
current route
25
Client informs user and route guidance recovery route
has been generated
26
Client displays the recovery route
Extension: change of routing preference
27
User requests change of routing preferences
28
Client provides list of preferences supported
29
User selects items and submits
30
Client verifies selection and establish them as new
current settings, and to re-plan route if necessary
Table AP. 2. OGC Navigation flows
Source: OGC (2008b)
Remark
Assumes that fingerprints
does no longer correlate
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 126
A.3. Database contents
The relations of the tables can be found in Figure 26. The fields are explained in the tables below.
Table: table_location (nodes)
Field
Fid
Type_exit
Type_room
Type_junction
Type_stairs
Type_elevator
Type_aisle
Building
Address
Place
Floor
Wing
Useraccess
Locprop
Type
Number
Boolean
Boolean
Boolean
Boolean
Boolean
Boolean
String
String
String
String
String
Boolean
String
Description
Fingerprint location ID
Exit available at id?
Room(s) available at id?
Junction(s) available at id?
Staircase(s) available at id?
Elevator(s) available at id?
Aisle(s) available at id?
Name of building fingerprint measured
Address of building fingerprint measured
Place of building fingerprint measured
Floor of building fingerprint measured
Wing of building fingerprint measured
Restricted or unrestricted at point measured
Objects to be seen at fingerprint
Table AP. 3. Database: fingerprint locations
Table: table_connectivity (edges)
Field
cid
Fid_link_01
Fid_link_02
Length
Direction
Restricted
Stairs
Elevator
Ramp
Door
Objects_left
Objects_right
Type
Number
Number
Number
Number
String
Boolean
Boolean
Boolean
Boolean
Boolean
String
String
Description
Number of connectivity link ID
Link between FID (1)
Link between FID (2)
Length of link in metres
Movement of link in reality: straight, up, down
Restricted or unrestricted movement
Movement along stairs possible?
Movement along elevator possible?
Movement along ramp possible?
Movement through door possible?
Description of objects left of the node
Description of objects right of the node
Type
Number
Number
String
String
String
String
String
String
String
Boolean
String
Description
Number of room ID
Proximity of Fingerprint ID
Name of room (browsable)
Room in wing of building
Room on floor of building
Building name
Department present in room
Office present in room
Type of room (office, kitchen, toilet etc.)
Restricted or unrestricted to staff
Orientation regarding topology
Table AP. 4. Database: connectivity
Table: table_rooms (directory)
Field
Rid
Fid
Room_name
Wing
Floor
Building
Department
Office
Room_type
Restricted
Orientation
Table AP. 5. Database: room information
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 127
Table: table_POI (directory, unused)
Field
Type
Poid
Number
Poigroup
Number
Poisubgroup
Number
Fid
Number
Poispecification
String
Table AP. 6. Database: POI information
Description
Number of POI
Category of POI
Sub category of POI
POI located at Fingerprint ID
Comments on POI
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 128
A.4. Database samples
The tables below provide examples of the SQLite database in the form of statements in XML files.
The full files can be found in the attached xml files, to be found under Indoor Nav/res/raw.
<sql>
<statement>
CREATE TABLE IF NOT EXISTS table_location (
fid INTEGER PRIMARY KEY,
type_exit BOOLEAN,
type_room BOOLEAN,
type_junction BOOLEAN,
type_stairs BOOLEAN,
type_elevator BOOLEAN,
type_aisle BOOLEAN,
building VARCHAR(50),
address VARCHAR(100),
place VARCHAR(100),
floor INTEGER,
wing VARCHAR(50),
useraccess VARCHAR(50),
locprop_01 VARCHAR(200),
locprop_02 VARCHAR(200),
locprop_03 VARCHAR(200))
</statement>
<statement>INSERT INTO table_location VALUES(2001,1,0,0,0,0,0,'OTB','Jaffalaan
9','Delft',0,2,'Unrestricted','Outside Entrance','','');</statement>
<statement>INSERT INTO table_location VALUES(2002,1,0,1,0,0,1,'OTB','Jaffalaan
9','Delft',0,2,'Unrestricted','Entrance','','');</statement>
<statement>INSERT INTO table_location VALUES(2003,0,0,1,0,0,1,'OTB','Jaffalaan
9','Delft',0,2,'Unrestricted','Path to Students Affairs and Back Exit','','');</statement>
<statement>INSERT INTO table_location VALUES(2004,0,0,0,0,0,1,'OTB','Jaffalaan
9','Delft',0,2,'Unrestricted','Hall to back exit','','');</statement>
Box AP. 1. Sample database for table_location
<sql>
<statement>
CREATE TABLE IF NOT EXISTS table_mac (
fid INTEGER,
mac VARCHAR(20),
rssi_avg INTEGER,
CONSTRAINT table_mac
PRIMARY KEY (fid, mac)
)
</statement>
<statement>INSERT
<statement>INSERT
<statement>INSERT
<statement>INSERT
INTO
INTO
INTO
INTO
table_mac
table_mac
table_mac
table_mac
Box AP. 2. Sample database for table_mac
VALUES(2001,'00:1A:A2:C0:0D:70',-76);</statement>
VALUES(2001,'00:1A:A2:C0:0D:72',-78);</statement>
VALUES(2001,'00:1A:A2:C0:0D:74',-76);</statement>
VALUES(2001,'00:1A:A2:C0:0D:75',-77);</statement>
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 129
<sql>
<statement>
CREATE TABLE IF NOT EXISTS table_connectivity (
cid INTEGER
PRIMARY KEY,
length INTEGER,
fid_link_01 INTEGER,
fid_link_02 INTEGER,
direction VARCHAR(10),
restricted BOOLEAN,
stairs BOOLEAN,
elevator BOOLEAN,
ramp BOOLEAN,
door BOOLEAN,
objects_left VARCHAR(200),
objects_right
VARCHAR(200))
</statement>
<statement>INSERT INTO table_connectivity
VALUES(20001,5,2001,2002,'straight',1,0,0,0,1,'','');</statement>
<statement>INSERT INTO table_connectivity
VALUES(20002,5,2002,2001,'straight',1,0,0,0,1,'','');</statement>
<statement>INSERT INTO table_connectivity
VALUES(20003,5,2002,2003,'straight',1,0,0,0,0,'','');</statement>
<statement>INSERT INTO table_connectivity
VALUES(20004,5,2002,2006,'straight',1,0,0,0,0,'','Info Pillar, Sofas');</statement>
Box AP. 3. Sample database for table_connectivity
<sql>
<statement>
CREATE TABLE IF NOT EXISTS table_rooms (
rid INTEGER PRIMARY
KEY,
fid INTEGER,
room_name VARCHAR(50),
wing INTEGER,
floor INTEGER,
building VARCHAR(50),
department VARCHAR(50),
office VARCHAR(50),
room_type VARCHAR(50),
restricted BOOLEAN))
</statement>
<statement>INSERT INTO table_rooms VALUES(2002,2008,'Survey
Room',2,0,'OTB','','Room','Computer Room',1);</statement>
<statement>INSERT INTO table_rooms VALUES(2004,2008,'Grote
Vergaderzaal',2,0,'OTB','','Room','Commission',1);</statement>
<statement>INSERT INTO table_rooms VALUES(2006,2009,'Anneke van
Kootenzaal',2,0,'OTB','','Room','Commission',1);</statement>
<statement>INSERT INTO table_rooms VALUES(2007,2010,'0.070
(Secretary)',2,0,'OTB','','Work Office','Office',1);</statement>
Box AP. 4. Sample database for table_rooms
<sql>
<statement>
CREATE TABLE IF NOT EXISTS table_poi (
poid INTEGER PRIMARY KEY,
poigroup VARCHAR(50),
poisubgroup VARCHAR(50),
fid INTEGER,
poispecification VARCHAR(50))
</statement>
<statement>INSERT INTO table_poi
Cootenzaal');</statement>
<statement>INSERT INTO table_poi
Board');</statement>
<statement>INSERT INTO table_poi
Board');</statement>
<statement>INSERT INTO table_poi
Board');</statement>
VALUES(99901,'Commission Rooms','',2009,'Anneke van
VALUES(99902,'Information','',3006,'Bulletin
VALUES(99903,'Information','',3106,'Bulletin
VALUES(99904,'Information','',3112,'Bulletin
Box AP. 5. Sample database for table_poi (unused)
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 130
A.5. Codes for location determination
The code below contains the location determination as described in section 16.3.4. The numbers to
the left represents the number of lines, which are being described in the main thesis.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
Public class MainScreen extends Activity implements OnClickListener {
String departure;
String destination;
private SQLiteDatabase db;
@Override
public void onCreate(Bundle savedInstanceState) {
super.onCreate(savedInstanceState);
setContentView(R.layout.main);
AutoCompleteTextView dep1 = (AutoCompleteTextView)
findViewById(R.id.inputDeparture);
RoomList roomList = new RoomList(this);
ArrayList<String> items1 = roomList.getAllRooms();
Toast.makeText(this, "Amount of rooms loaded: " +
items1.size(),Toast.LENGTH_SHORT).show();
ArrayAdapter<String> adapter1 = new
ArrayAdapter<String>(this,
R.layout.list_item, items1);
dep1.setAdapter(adapter1);
AutoCompleteTextView des1 = (AutoCompleteTextView)
findViewById(R.id.inputDestination);
ArrayAdapter<String> adapter2 = new ArrayAdapter<String>(this,
R.layout.list_item, items1);
des1.setAdapter(adapter2);
Button StartNavigationButton = (Button)
findViewById(R.id.buttonNavigationStart);
StartNavigationButton.setOnClickListener(this);
Button StartSettingsButton = (Button) findViewById(R.id.buttonSettings);
StartSettingsButton.setOnClickListener(this);
GenericService gs = new GenericService();
// This line initiates an object (gs) for Wi-Fi service
gs.SetActivitiy(this);
// This line lets the object know about your activity
gs.AddAccessPoint("00:1a:a2:c0:0d:70","00:1a:a2:c0:0d:70");
gs.AddAccessPoint("00:1a:a2:c0:0d:72","00:1a:a2:c0:0d:72");
//Add AP to listen, you can add multiple APs by using this method many
times
gs.AddRule("00:1a:a2:c0:0d:70","00:1a:a2:c0:0d:70", Rule.RULE_TYPE_WIFI, 90, -35, 1, null);
gs.AddRule("00:1a:a2:c0:0d:72","00:1a:a2:c0:0d:72", Rule.RULE_TYPE_WIFI, 90, -35, 1, null);
db = (new DatabaseHelper(this).getWritableDatabase());
// Set rules to be informed when given criteria is satisfied by using:
ListenWifi listener= new ListenWifi(db, handler,this);
// Create listeners to catch the events thrown by the service...
gs.AddWifiListener(listener);
// ...and let the object know where your listener is
gs.start();
//and start to wait events
}
public void onClick(View v) {
if (v.getId() == R.id.buttonNavigationStart) {
EditText inputDeparture = (EditText)
findViewById(R.id.inputDeparture);
EditText inputDestination = (EditText)
findViewById(R.id.inputDestination);
if (inputDeparture.getText().length() == 0) {
Toast.makeText(this, "Please input departure",
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 131
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
Toast.LENGTH_SHORT).show();
return;
} else if (inputDestination.getText().length() == 0) {
Toast.makeText(this, "Please input destination",
Toast.LENGTH_SHORT).show();
return;
} else if (inputDeparture == inputDestination) {
Toast.makeText(this, "Departure and Destination equals",
Toast.LENGTH_SHORT).show();
return;
}
Intent n = new Intent(this, Navigation.class);
n.putExtra("departure", inputDeparture.getText().toString());
n.putExtra("destination", inputDestination.getText().toString());
startActivity(n);
} else if (v.getId() == R.id.buttonSettings) {
Intent s = new Intent(this, Settings.class);
startActivity(s);
}
}
@Override
protected void onDestroy() {
db.close();
super.onDestroy();
}
//This is the place where we handle messages coming to the GUI thread
private final Handler handler = new Handler(){
@Override
public void handleMessage(Message msg) {
//gets parameter is the type of the message
String comingMsg = (String)msg.getData().get("2");
if(comingMsg == null){
comingMsg = (String)msg.getData().get("1");
//if there is such a message
if(comingMsg != null){
//try to separate the message again
String fid, floor, wing, building, address, place, locprop;
//separation process
String delims = "[;]";
//split the message into array of strings
String[] data = comingMsg.split(delims);
fid = data[0];
floor = data[1];
wing = data[2];
building = data[3];
address = data[4];
place = data[5];
locprop = data[6];
//Use this variables to set the text now
TextView textBuilding = (TextView)
findViewById(R.id.currentBuilding);
textBuilding.setText("At: " + building + ", " + address + ", " +
place);
TextView textFid = (TextView) findViewById(R.id.currentType);
textFid.setText("Near: " + locprop + " (Node: " + fid + ")");
TextView textWing = (TextView) findViewById(R.id.currentWing);
textWing.setText("Wing: " + wing + ", Floor: " + floor);
}
else{
//No fid is matched!
}
}
super.handleMessage(msg);
}
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 132
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
};
class ListenWifi implements WifiListener{
public static final String MSG_WIFIFOUND = "WifiFound";
public static final String MSG_WIFILOST = "WifiLost";
public static final String MSG_WIFIRULE = "WifiRule";
public static final String MSG_WIFILEVELCHANGED = "WifiChange";
SQLiteDatabase db;
Handler handler;
Activity activity;
ListenWifi(SQLiteDatabase db, Handler handler, Activity activity){
this.db = db;
this.handler = handler;
this.activity = activity;
}
private void SendMessage(String type, String str){
Bundle b = new Bundle();
b.putString(type, str);
Message msg = new Message();
msg.setData(b);
handler.sendMessage(msg);
}
@Override
public void OnStateChange(WifiEvent event){
}
@Override
public void OnAPFound(WifiEvent event) {
event.GetAccessPoint().getName();
}
@Override
public void OnAPLost(WifiEvent event) {
}
@Override
public void OnRuleAchieved(RuleEvent event) {
WifiManager wifimanager =
(WifiManager)activity.getSystemService(Context.WIFI_SERVICE);
wifimanager.startScan();
List<ScanResult> ls = wifimanager.getScanResults();
//A loop to customize the query according to the measured values
String queryHead = "Select fid, count(*) from table_mac where ";
String queryTail = " group by fid";
//Vector<String> vQuery = new Vector<String>();
int counter = 0;
String queryMid = "";
if(ls.size() > 0){
for(ScanResult sr: ls){
//Prepare
String mac = sr.BSSID;
int SS = sr.level;
if(counter == 0){
queryMid = "(mac = '"+mac+"' and rssi_avg < "+(SS+4)+" and
rssi_avg > "+(SS-4)+")"; // First MAC with upper and lower
}
else{
queryMid += " or (mac = '"+mac+"' and rssi_avg <
"+(SS+4)+" and rssi_avg > "+(SS-4)+")"; // Subsequent MAC with upper and lower
}
counter++;
} //End of loop scan results
String finalQuery = queryHead + queryMid + queryTail; // Query buildup
int max = 0;
int selectedFID = 0;
Cursor cursor = db.rawQuery(finalQuery, null);
if(cursor.getCount() > 0){
while(cursor.moveToNext()){
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 133
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
int fid = cursor.getInt(0);
int count = cursor.getInt(1);
if(count > max){
max = count;
selectedFID = fid;
}//If some records are equal, top one is selected
}
}
if(selectedFID > 0){
String query = "select fid, floor, wing, building, address, place,
locprop from table_location where fid = " + selectedFID ;
//New cursor is defined to handle new query results
Cursor cursorFID = db.rawQuery(query, null);
if(cursorFID.getCount() > 0){
cursorFID.moveToNext();
String msg = cursorFID.getString(0)+ ";" +cursorFID.getString(1) +
";" + cursorFID.getString(2) + ";" + cursorFID.getString(3) + ";" +
cursorFID.getString(4) + ";" + cursorFID.getString(5) + ";" + cursorFID.getString(6);
SendMessage("1", msg);
}
else{
String msg = "!";
SendMessage("2", msg);
}
}
}
}
@Override
public void OnAPLevelChanged(WifiEvent event) {
}
}// end of listener class
Box AP. 6. Overview of code for location determination (MainScreen.java and Navigation.java)
Source: M. Yildirim (2011)
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 134
A.6. Codes for navigation algorithm (Dijkstra’s algorithm)
The tables/codes below provide an overview on how the scripts work. Nodes and edges are being
loaded via Vertex.java and Edge.java, in which a graph will be constructed in Graph.java. Finally,
those three inputs are being used in DijkstraAlgorithm.java, in which the calculation of the shortest
path is being made.
Public class DijkstraAlgorithm {
private final List<Vertex> nodes;
private final List<Edge> edges;
private Set<Vertex> settledNodes;
private Set<Vertex> unSettledNodes;
private Map<Vertex, Vertex> predecessors;
private Map<Vertex, Integer> distance;
public DijkstraAlgorithm(Graph graph) {
// create copy of array
this.nodes = new ArrayList<Vertex>(graph.getVertices());
this.edges = new ArrayList<Edge>(graph.getEdges());
}
public void execute(Vertex fid_link_01) {
settledNodes = new HashSet<Vertex>();
unSettledNodes = new HashSet<Vertex>();
distance = new HashMap<Vertex, Integer>();
predecessors = new HashMap<Vertex, Vertex>();
distance.put(fid_link_01, 0);
unSettledNodes.add(fid_link_01);
while (unSettledNodes.size() > 0) {
Vertex node = getMinimum(unSettledNodes);
settledNodes.add(node);
unSettledNodes.remove(node);
findMinimalDistances(node);
// fid_link_01 = source, fid_link_02 = destination
}
}
private void findMinimalDistances(Vertex node) {
List<Vertex> adjacentNodes = getNeighbors(node);
for (Vertex target : adjacentNodes) {
if (getShortestDistance(target) > getShortestDistance(node) +
getDistance(node, target)) {
distance.put(target, getShortestDistance(node) +
getDistance(node, target));
predecessors.put(target, node);
unSettledNodes.add(target);
}
}
}
private int getDistance(Vertex node, Vertex target) {
for (Edge edge : edges) {
if (edge.getFid_link_01().equals(node) &&
edge.getFid_link_02().equals(target)) {
return edge.getLength();
// Length = weight in this case
}
}
throw new RuntimeException("should not happen");
}
private List<Vertex> getNeighbors(Vertex node) {
List<Vertex> neighbors = new ArrayList<Vertex>();
for (Edge edge : edges) {
if (edge.getFid_link_01().equals(node) &&
!isSettled(edge.getFid_link_02())) {
neighbors.add(edge.getFid_link_02());
}
}
return neighbors;
}
private Vertex getMinimum(Set<Vertex> vertices) {
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 135
Vertex minimum = null;
for (Vertex vertex : vertices) {
if (minimum == null) {
minimum = vertex;
} else {
if (getShortestDistance(vertex) < getShortestDistance(minimum))
{
minimum = vertex;
}
}
}
return minimum;
}
private boolean isSettled(Vertex vertex) {
return settledNodes.contains(vertex);
}
private int getShortestDistance(Vertex fid_link_02) {
Integer d = distance.get(fid_link_02);
if (d == null) {
return Integer.MAX_VALUE;
} else {
return d;
}
}
/*
* Returns path from source (fid_link_01) to selected target and NULL if no path
*/
public LinkedList<Vertex> getPath(Vertex target) {
LinkedList<Vertex> path = new LinkedList<Vertex>();
Vertex step = target;
// Check if a path exists
if (predecessors.get(step) == null) {
return null;
}
path.add(step);
while (predecessors.get(step) != null) {
step = predecessors.get(step);
path.add(step);
}
// Put it into the correct order
Collections.reverse(path);
return path;
}
Box AP. 7. Code for DijkstraAlgorithm.java
Edited after: L. Vogel (2009)
public class Graph {
public final List<Vertex> vertices;
public final List<Edge> edges;
public Graph(List<Vertex> vertices, List<Edge> edges) {
this.vertices = vertices;
this.edges = edges;
}
public List<Vertex> getVertices() {
return vertices;
}
public List<Edge> getEdges() {
return edges;
}
}
Box AP. 8. Code for Graph.java
Edited after: L. Vogel (2009)
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 136
public class
final
final
final
Vertex {
private Integer fid;
private String room_name;
private ArrayList<Room> rooms;
public Vertex(Integer fid, String room_name, ArrayList<Room> rooms) {
this.fid = fid;
this.room_name = room_name;
this.rooms = rooms;
}
public int getFid() {
return fid;
}
public String getRoomName() {
return room_name;
}
public ArrayList<Room> rooms() {
return this.rooms;
}
@Override
public int hashCode() {
final int prima = 31;
int result = 1;
result = prima * result + ((fid == null) ? 0 : fid.hashCode());
return result;
}
@Override
public boolean equals(Object obj) {
if (this == obj)
return true;
if (obj == null)
return false;
if (getClass() != obj.getClass())
return false;
Vertex other = (Vertex) obj;
if (fid == null) {
if (other.fid != null)
return false;
} else if (!fid.equals(other.fid))
return false;
return true;
}
@Override
public String toString() {
return room_name;
}
}
Box AP. 9. Code for Vertex.java
Edited after: L. Vogel (2009)
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 137
public class Edge {
private final
private final
private final
private final
private final
private final
private final
private final
private final
private final
private final
private final
int cid;
Vertex fid_link_01;
Vertex fid_link_02;
int length;
String direction;
String restriction;
boolean stairs;
boolean elevator;
boolean ramp;
boolean door;
String object_left;
String object_right;
public Edge(int cid, Vertex fid_link_01, Vertex fid_link_02, int length, String
direction, String restriction, boolean stairs, boolean elevator, boolean ramp, boolean door,
String object_left, String object_right) {
this.cid = cid;
this.fid_link_01 = fid_link_01;
this.fid_link_02 = fid_link_02;
this.length = length;
this.direction = direction;
this.restriction = restriction;
this.stairs = stairs;
this.elevator = elevator;
this.ramp = ramp;
this.door = door;
this.object_left = object_left;
this.object_right = object_right;
}
public int getCid() {
return cid;
}
public Vertex getFid_link_01() {
return fid_link_01;
}
public Vertex getFid_link_02() {
return fid_link_02;
}
public int getLength() {
return length;
}
public String getDirection() {
return direction;
}
public String getRestriction() {
return restriction;
}
public boolean getStairs() {
return stairs;
}
public boolean getElevator() {
return elevator;
}
public boolean getRamp() {
return ramp;
}
public boolean getDoor() {
return door;
}
public String getObject_left() {
return object_left;
}
public String getObject_right() {
return object_right;
}
@Override
public String toString() {
return fid_link_01 + " " + fid_link_02;
}
}
Box AP. 10. Code for Edge.java
Edited after: L. Vogel (2009)
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 138
B. Background literature
B.1. Background information on pedestrian movements and navigation algorithms
Another difference with outdoor navigation, is that (indoor) pedestrian movement is often not
variable in speed, making the shortest path the fastest path. Usher and Strawderman (2010)
investigated the pedestrian movement, in where they noticed some remarkable properties: (1)
people move with an average speed of 1.57 m/s (5.7 km/h) for fair open places and 1.47 m/s (5.3
km/h) for narrow places, depending on the preferred walking speed, and depending on how crowded
a place can be; (2) people try to change the angular movement as limited as possible, keeping their
route as smooth as possible. This implies that not always the shortest (fastest) path in indoor
environments will be followed. In addition, Winter (2002) researched the modelling of turn costs in
route planning for the outdoor navigation. This includes the disallowance of turning in some
situations, where a modality need to look for alternatives, such as driving to the nearest roundabout,
make a U-turn and still drive in the desired direction. Pedestrian movements are in general not
restricted to such features, as they freely move from one point to another. Beside physical
constraints, a pedestrian is not allowed to turn when the designated space is constrained by
authority, such as opening hours. The information above can be used to determine estimated times
of arrival.
Figure AP. 1 summarizes the above contents. In building A, it takes 20 units from a pedestrian to walk
from one end to another end. If the building was structured as a U-shape (B), then it would take 30
units to walk. If there would be suspended corridor on another floor in this same building (C), one
could consider taking the staircases. However, the travel costs for using the staircase, might depend
on how the pedestrian perceives such vertical movement. For instance, if the pedestrian does not
bother using the stairs, a value of 1 for one vertical movement can be assigned, ending up in a total
of 3+1+12+1+3=20 units. If the pedestrian finds it annoying, a value of 5 could be assigned, ending up
in a total of 3+5+12+5+3=28 units. In addition, the risk of getting lost (the second comment of Usher
& Strawderman (2010), as stated above), might be incorporated as well, with the result that this
shortest path might be inconvenient.
A possible inclusion of the exterior to be used for indoor positioning and the shortest paths is
recently mentioned by Boguslawski, Gold & Ledoux (2011). Sometimes, it might pay off to leave the
building, and enter the same building through another door at another side. This could be the case in
Figure AP. 1B if this would be located on the ground floor, and if exits/entrances would exist at the
northern edges. One could easily walk from on and to another. If the Wi-Fi access points could reach
that area, then it is surely possible to use such outdoor fingerprints in the model.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 139
Figure AP. 1. Shortest paths in buildings
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 140
B.2. Mobile cartography and the user
The user profile has influence on the visualization and use of maps Sarjakoski et al. (2007), underlines
the importance of user adaptive maps. According to them, the final map on mobile phones is
dependent on the following parameters:








Use case, also to be considered for which purpose the map should serve;
Level of detail, expressed in the resolution of the map;
Time, the map’s time stamp (the up-to-datedness);
Age, as in the age group of the user;
Device, related to the capabilities and constraints of the mobile device;
Centre, as in the centre point of the map;
Scale, which is the display scale; and
Position, which is the position of the user on the map.
This assumes there is a strong correlation of the user’s knowledge of maps and map usage in general.
It also determines which data should be displayed on the map first. Sarjakoski et al. (2007), proposes
to use a map specification knowledge base (MSKB) to store all kinds of combinations and
possibilities, so the actual map representation can be adjusted accordingly. These will affect the
placement and layout of:





Topographic features to be displayed;
Points/Areas/Lines of Interest;
Level of detail;
Other visualization operators, such as icon placement; and
Other map visualization, such as colours and line widths.
In fact, these ideas are similar as described by Reichenbacher (2005), who made the following
graphic interpretation (Figure AP. 2), on how user, information, situation, activities and the device
are interrelated.
Figure AP. 2. Reichenbacher's Context Model for Mobile Cartography
As can be seen in the number of links, the user again largely defines the cartography. Adaption of the
application (in particular activities, information and situation) to the user is considered to be the key
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 141
factor in mobile, geographic applications. These adaptations must fit within the original intended
designed application and maps. Focus on specific elements on maps can be further extended by
highlighting it in colour, centring the object, emphasize it with outlines, increasing the sharpness
(while making the rest blurry), enhancing the level of detail of the object and by animating the object
(such as blinking, size increase and rotation) (Reichenbacher, 2005). In this project, the same
interpretation and visualization can be used for texts instead.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 142
B.3. Background information on cartographic rules
Kraak & Ormeling (2003) provided renewing interesting in the cartographic rules set by Bertin in
1967, which assembled existing conventionalities and practices of centuries of cartographic habits in
a logical structure. The phrase “How do I say what to whom and is it effective?” is the underlying
principle upon successful map designing, using cartographic rules, aside from visualization of
backgrounds and the map structure itself. The how involves the use of maps, symbology, texts or
charts; the what resembles the data; the whom is the target audience; and the effective implies to in
what extent the three have to be combined for understanding. The what and the effective is being
emphasized in this section.
Not only symbology and cartography changes with its scale (e.g. a city can be represented as a point,
but also as a polygon), but the data variable itself is largely influencing how symbols should be used.
Data can be nominal (classification, categorical), ordinal (when a ranking can be made), interval
(where variables are distinguishable from each other by numeric values) or ratio (same as interval,
except with a fixed zero point). From here, six graphical variables can be used on top of the data
classification, to distinguish areas, namely by differences size, lightness/colour (value), grain/texture,
hue/colour (differentiation), orientation and shape (Kraak & Ormeling, 2003, seeFigure AP. 3). Bertin
related the graphical variables with the data variables to see how effective the representation is, as
shown in Table AP. 7 and Table AP. 8.
Figure AP. 3. Bertin's geographic variables
Source: Kraak & Ormeling (2003)
Dashes
Patches
X
X
X
Dots
X
X
X
Ratio
Interval
Size
Colour value
Grain/texture
Colour dif.
Orientation
Shape
Ordinal
Nominal
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 143
X
4
3
2
7
4
-
4
4
4
7
2
-
5
5
5
8
-
X
X
X
Table AP. 7. Relation of graphic variables and visual isolation
Source: Kraak & Ormeling (2003)
Ratio
Interval
Ordinal
Nominal
Overall impression
Selection
Differentiation
Proportional
Distance
Order
Differentiation
High
Middle
Low
Size
Colour value
Grain/texture
Colour dif.
Orientation
Shape
Definition: characteristic
Definition: implicit
Definition: possible
Perception: strong
Perception: weak
Table AP. 8. Information perception of cartographic information
Source: Kraak & Ormeling (2003)
From the above figures, Kraak & Ormeling (2003) emphasizes on the constraints of human
perception in cartographic visualization. For example, there is a maximum of colours humankind can
perceive to easily distinguish data values. If this maximum is exceeded, it becomes much harder to
do so. In the case of colour, extra psychological and cultural associations play a fundamental role in
judging a map. A red colour is usually associated with something negative, except for temperatures,
where this implies higher (positive) temperatures. It can also be associated with political parties or
other cultural references.
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 144
C. Results
C.1. Scan results for location estimation
The tables below provide information which location nodes have been obtained. The table headers
indicate at which location the user/device was actually located.
Node
Remark
T01
T02
T03
T04
T05
T06
T07
T08
T09
T10
T11
T12
T13
T14
T15
T16
T17
T18
T19
T20
3002
Scan
ALL
2013
3003
2013
2013
2013
2014
3201
2013
2014
2013
2014
2013
3003
2014
2013
2011
2011
2013
2013
2013
3007
Scan
ALL
3107
3007
3005
3107
3107
3005
3004
3107
3005
3010
3004
3107
3004
3107
3010
3005
3004
3005
3107
3005
3011
Scan
ALL
2001
3006
2001
3112
3009
2001
2001
3009
2001
2001
3112
2001
2014
3009
3009
3011
3009
2001
2001
3010
3101
Scan
ALL
3105
2014
2012
2011
2012
3105
2014
2012
2012
2011
2014
2012
2011
2014
3105
2014
2012
3103
3202
3001
3105
Scan
ALL
2011
2011
2011
2011
2011
2011
2011
2011
2011
2011
2011
3206
2011
3206
2011
2011
3206
2011
2011
3206
3110
Scan
ALL
2001
3009
3009
3009
3011
3010
3211
3009
3011
2001
3011
3010
2001
3011
3211
3211
3009
3211
2001
3209
3203
Scan
ALL
3205
3205
3105
3209
3209
3104
2011
3205
3205
3206
3205
3205
3205
3205
3205
3205
3205
3205
3205
3205
3209
Scan
ALL
3209
3111
3108
3206
3206
3111
3108
3206
3206
3108
3111
3111
3209
3206
3111
3209
3207
3206
3209
3206
3211
Scan
ALL
3209
3211
3111
3211
3211
3111
3209
3211
3111
3212
3211
3212
3211
3212
3211
3211
3111
3209
3209
3111
2002
Scan
ALL
2004
2004
2004
2003
2004
2004
2004
2004
2004
2004
2011
2004
2004
2004
2011
2004
2004
2011
2011
2004
2005
Scan
ALL
2011
2004
2011
2004
2004
2004
2004
2004
2004
2004
2004
2004
2011
2004
2004
2004
2004
2004
2004
2004
2008
Scan
ALL
2006
2003
2004
2003
2002
2003
2002
2003
2003
2003
2003
2004
2003
2003
2003
2003
2003
2003
2007
2002
Table AP. 9. Scan results for all locations - best 10 macs in vicinity
Node
Remark
T01
T02
T03
T04
T05
T06
T07
T08
T09
T10
T11
T12
T13
T14
T15
T16
T17
T18
T19
T20
3002
SCAN
FLOOR
3104
3103
3101
3003
3104
3104
3101
3101
3003
3003
3101
3104
3001
3103
3002
3104
3002
3003
3101
3002
3007
SCAN
FLOOR
3007
3003
3101
3013
3007
3007
3007
3105
3007
3007
3104
3007
3009
3104
3105
3007
3103
3007
3007
3007
3011
SCAN
FLOOR
3011
3011
3011
3011
3005
3011
3003
3003
3003
3112
3108
3108
3112
3002
3002
3108
3011
3011
3011
3002
3101
SCAN
FLOOR
3102
3204
3101
3003
3004
3102
3004
3207
3204
3205
3204
3004
3205
3204
3204
3204
3206
3204
3004
3204
3105
SCAN
FLOOR
3105
3105
3105
3105
3102
3101
3105
3102
3105
3105
3102
3105
3102
3101
3101
3105
3102
3101
3105
3101
3110
SCAN
FLOOR
3010
3108
3106
3109
3003
3109
3109
3003
3104
3109
3105
3109
3109
3204
3204
3104
3112
3109
3107
3109
3203
SCAN
FLOOR
3207
3207
3204
3207
3207
3204
3206
3205
3207
3207
3204
3206
3204
3204
3204
3206
3206
3204
3204
3204
Table AP. 10. Scan results for wing 3 locations - only routers from same floor registered
3209
SCAN
FLOOR
3210
3209
3208
3209
3209
3209
3209
3205
3205
3209
3102
3205
3209
3002
3205
3209
3001
3208
3208
3210
3211
SCAN
FLOOR
3212
3211
3211
3201
3211
3204
3212
3211
3211
3206
3206
3204
3204
3204
3206
3201
3204
3206
3206
3203
M S c T h e s i s I n d o o r N a v i g a t i o n w i t h W i - F i F i n g e r p r i n t i n g | 145
C.2. Scan results for route prediction
The table below shows the outcomes for the locations provided by the device upon walking a specific
route, which has been repeated three times.
Route
1
Nodes
1
3005
3006
3007
3008
3009
3110
3210
3211
3212
Located
3005
3004
3007
3009
3007
3109
3210
3211
3212
2
3101
3102
3103
3104
3106
3107
3108
3109
3110
3111
3112
Located
3101
3105
3103
3101
3005
3203
3004
3109
3109
3112
3112
3
3201
3202
3203
3204
3206
3207
3208
3209
3210
3110
3111
3112
Located
3105
3202
3203
3105
3206
3207
3208
3206
3107
3108
3112
3112
4
3001
3002
3003
3103
3203
3202
3201
Located
3002
3002
3101
3103
3205
3101
-
5
3212
3211
3210
3110
3009
3010
3011
Located
3212
3212
3211
3211
3009
3108
3010
Route
9
11
7
7
2
Nodes
1
3005
3006
3007
3008
3009
3110
3210
3211
3212
Located
3007
3007
3007
3008
3006
3109
3107
3211
3212
2
3101
3102
3103
3104
3106
3107
3108
3109
3110
3111
3112
Located
3101
3102
3104
3105
3106
3004
3109
3109
3204
3208
3007
3
3201
3202
3203
3204
3206
3207
3208
3209
3210
3110
3111
3112
Located
3202
3202
3203
3203
3206
3207
3207
3206
3108
3108
3112
3112
4
3001
3002
3003
3103
3203
3202
3201
Located
3001
3002
3003
3102
3204
3204
3101
5
3212
3211
3210
3110
3009
3010
3011
Located
3212
3211
3108
3108
3009
3009
3011
Route
12
9
11
12
7
7
3
Nodes
1
3005
3006
3007
3008
3009
3110
3210
3211
3212
Located
3006
3006
3006
3008
3007
3109
3210
3107
3107
9
2
3101
3102
3103
3104
3106
3107
3108
3109
3110
3111
3112
Located
3002
3006
3102
3104
3106
3106
3006
3109
3105
3110
3112
3
3201
3202
3203
3204
3206
3207
3208
3209
3210
3110
3111
3112
Located
3205
3202
3203
3203
3206
3208
3108
3108
3108
3108
3111
3112
4
3001
3002
3003
3103
3203
3202
3201
Located
3001
3002
3003
3104
3204
3202
3202
5
3212
3211
3210
3110
3009
3010
3011
Located
3212
3211
3107
3109
3101
3003
3002
Table AP. 11. Scan results for 3 series of 5 different routes in wing 3
11
12
7
7
On
On/off
Off
5
2
2
55,6%
22,2%
22,2%
4
3
4
36,4%
27,3%
36,4%
6
4
2
50,0%
33,3%
16,7%
2
1
4
28,6%
14,3%
57,1%
2
3
2
28,6%
42,9%
28,6%
On
On/off
Off
4
3
2
44,4%
33,3%
22,2%
4
2
5
36,4%
18,2%
45,5%
5
5
2
41,7%
41,7%
16,7%
3
0
4
42,9%
0,0%
57,1%
4
0
3
57,1%
0,0%
42,9%
On
On/off
Off
3
3
3
33,3%
33,3%
33,3%
4
3
4
36,4%
27,3%
36,4%
5
1
6
41,7%
8,3%
50,0%
4
0
3
57,1%
0,0%
42,9%
2
0
5
28,6%
0,0%
71,4%
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertising