wireless positioning system

Seminar II

On

Wi-Fi based Indoor positioning

Submitted By

Vaibhav Gaikwad Exam seat no.

Seminar Guide

Mrs. Trupti Wagh

Department of Electronics and Telecommunication

PG Communication Network

D. Y. Patil College of Engineering

Akurdi, Pune - 411044

2014-2015

D. Y. PATIL COLLEGE OF ENGINEERING

AKURDI, PUNE - 411044

DEPARTMENT OF ELECTRONICS & TELECOMMUNICATION

ENGINEERING

CERTIFICATE

This is to certify that Vaibhav Gaikwad - Roll No.11

of M.E. Communication Network has successfully completed the

Seminar II titled


towards the partial fulfillment for the requirements of the Masters

Degree of Engineering course under the University of Pune during the

academic year 2014-2015.

Mrs. Trupti Wagh Mrs.Padma Lohiya Prof.(Dr.)Mrs.P.Malathi

Seminar Guide PG Coordinator HOD (E. & T.C.)

Acknowledgement

I express my sincere gratitude towards the faculty members who makes this

seminar a successful.

I would like to express my thanks to my guide Mrs. Trupti Wagh and Head

of Department Prof.(Dr).Mrs.P. Malathi for her whole hearted co-operation and

valuable suggestions, technical guidance throughout the seminar work. And special

thanks for her kind official support given and encouragement.

I am also thankful to my PG coordinator Mrs. Padma Lohiya for her

valuable guidance.

Finally, I would like to thank to all our staff members of Electronic and

Telecommunication Department who helped me directly or indirectly to complete this

work successfully.

ii

Contents

Certificate i

Acknowledgement ii

1 Introduction 1

2 Literature Survey 4

3 Theory 8

3.1 Fuzzy nature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

3.1.1 Fuzzy Membership Function . . . . . . . . . . . . . . . . . . . . 8

3.1.2 Fuzzy Signal Distribution . . . . . . . . . . . . . . . . . . . . . 9

3.2 POSITIONING METHODOLOGIES . . . . . . . . . . . . . . . . . . . 10

3.2.1 Testing the Stability of the Signal . . . . . . . . . . . . . . . . . 12

3.2.2 Estimation Testing Within the Trust-Region . . . . . . . . . . . 12

3.2.3 K-Mean Location Fingerprinting Algorithm . . . . . . . . . . . 13

3.3 System Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3.3.1 Analyzing accelerometer and compass data . . . . . . . . . . . . 15

3.3.2 Particle filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.3.3 Map constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3.3.4 Room-level localization . . . . . . . . . . . . . . . . . . . . . . . 17

3.3.5 Updating the database through crowd-sourcing . . . . . . . . . 18

3.4 Implemented project by Google . . . . . . . . . . . . . . . . . . . . . . 19

4 Application 21

5 Conclusion 23

6 References 24

iii

List of Figures

1 LF WPS Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Fuzzy Membership function . . . . . . . . . . . . . . . . . . . . . . . . 9

3 Flowchart of positioning approach . . . . . . . . . . . . . . . . . . . . . 11

4 System design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

5 Collision is detected . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

iv

List of Abbreviations

WPS Wi-Fi Positioning System

GPS Global Positioning System

LF Location Fingerprinting (LF)

RSS received signal strength

AOA angle-of arrival

TOA time-of arrival

MS Mobile station

TR Newton Trust Region

AP Access points

SINR Signal-to-Interference-plus-Noise-Ratio

PDF probability density function

MAC Medium access controller

AP Access Point

v


1 Introduction

In an era where smartphones are becoming an essential part of our life, location

based services like GPS and GLONASS are being extensively used. These services

detect the user equipment (Mobile/GPS device) with the help three or more satellites

As most of the current UE (Mobile,GPS devices) depend on GPS but the quality of

GPS indoor positioning is still poor because GPS and GLONASS cant work without

a direct visibility of sky. The two major problems of current indoor positioning tech-

nologies are accuracy and cost. The new method which is becoming popular in indoor

positioning that is Wi-Fi Positioning System (WPS), it is used where GPS is inade-

quate due to various causes including multipath and signal blockage indoors. Wi-Fi

positioning takes advantage of the rapid growth in the early 21st century of wireless ac-

cess points in urban areas. It is now widely acknowledged, although Google, Apple, and

various phone makers and carriers have compiled their own very extensive databases

of Wi-Fi access point locations by correlating Wi-Fi access points with GPS locations

of cell phone, smartphone, and in some cases, tablet computer users. Anonymously

determining users location. WPS may be combined with cell phone tower triangu-

lation and GPS to provide reliable and accurate position data under a wide range of

conditions. The data needed to be collected for location based service and is collected

by cell towers and WiFi access points in variety of ways, including UE information.

These data are aggregated and store in public domain. This requires installation of

access points which supply their location information, which is also costly. On the

other hand, cellular networks have a wide coverage on the scale of kilometers and can

serve a whole building. However, measurement accuracy based on cellular signals is

not high enough to locate which shop or restaurant a person is in, and does not meet

the requirements of most indoor applications[10].

There are two typical indoor positioning approaches, Propagation-based and Loca-

tion Fingerprinting (LF) based. Propagation-based approaches estimate the position

by measuring the received signal strength (RSS) with path loss. The drawbacks of

these approaches lie in the requirement to have a strong Wi-Fi coverage in order to

DYPCOE, Dept.of E&TC, PG (Communication Network) Seminar II 1


compute every condition that the signal can blend. The LF-based approaches such as

locate a device by comparing its coordinates with the received signal strengths (RSSs)

in a pre-recorded database. The drawbacks of these approaches are highly affected by

internal building infrastructure changes, presence of humans, and interference among

other devices. All these lead to unstable Wi-Fi coverage and inaccurate localization[9].

Figure 1: LF WPS Architecture

As shown in figure:1, the location is computed by WiFi RSS and calibrated by WPS

database for positioning in LF method. The fingerprinting technique is relatively simple

to deploy compared to the other techniques such as angle of arrival (AOA) and time of

arrival (TOA). There is no specialized hardware required at the mobile station (MS).

Any existing wireless LAN infrastructure can be reused for this positioning system.

The deployment of fingerprinting based positioning systems can be divided into two

phases. First, in the oine phase, the location fingerprints are collected by performing

a site-survey of the received signal strength (RSS) from multiple access points (APs).

The entire area is covered by a rectangular grid of points. The RSS is measured with

enough statistics to create a database or a table of predetermined RSS values on the

points of the grid. The vector of RSS values at a point on the grid is called the

location fingerprint of that point. Second, in the on-line phase, a MS will report a

sample measured vector of RSSs from different APs to a central server (or a group of

APs will collect the RSS measurements from a MS and send it to the server). The server

uses an algorithm to estimate the location of the MS and reports the estimate back to



the MS (or the application requesting the position information). The most common

algorithm is to estimate the location computes the Euclidean distance between the

measured RSS vector and each fingerprint in the database. The coordinates associated

with the fingerprint that provides the smallest Euclidean distance is returned as the

estimate of the position. The orientation filter and the Newton Trust Region (TR)

algorithm to enhance the traditional LF by filtering the noisy signal. Although the

average distance error is 1.82 m, this was not completely effective because they still

suffer from the poor Wi-Fi coverage region [9].



2 Literature Survey

There are several indoor positioning technologies used in smartphones.Research is

going on in Infra based,network terminal based, GPS cast based as well as Wi-Fi based

positioning system. And the most popular is WiFi positioning technology.

[1] Binghao Li1, James Salter, Andrew G. Dempster1 and Chris Rizos1,

Indoor Positioning Techniques Based on Wireless LAN, IEEE Confer-

ence, .

In this paper author has proposed approach for location fingerprinting method one

is deterministic and probabilistic. In deterministic the structure of the fingerprint

database is relatively simple and the feature of the RP is only determined by the av-

erage RSSs of each APs. Each fingerprint summarizes the data as the average signal

strength to visible access points, based on a sequence of signal strength values recorded

at that location.

[2] J. Kwon, B. Dundar, and P. Varaiya, Hybrid algorithm for indoor

positioning using wireless LAN, IEEE 60th Vehicular Technology Confer-

ence, VTC, vol. 7, 2004

In this paper the development of a hybrid method that combines the strength of two

methods which are progation and LF. First authors formulated the RF propagation

loss in a nonlinear, censored regression model and adjusts the regression function to the

observed signal strength in the fingerprint dataset. It balances flexibility and accuracy

of the two traditional methods, makes intelligent use of missing values, produces error

bounds, and can be made dynamic.

[3] K. Kaemarungsi and P. Krishnamurthy, Modeling of indoor position-

ing systems based on location fingerprinting, INFOCOM. Twenty-third

Annual Joint Conference of the IEEE Computer and Communications So-

cieties, vol. 2, 2004.

In this paper analytical models for analyzing positioning systems was developed. The



framework for analyzing a simple positioning system that employed the Euclidean dis-

tance between a sample signal vector and the location fingerprints of an area stored

in a database. And then analyzed the effect of the number of access points that are

visible and radio propagation parameters.

[4] N. Swangmuang and P. Krishnamurthy, Location Fingerprint Analy-

ses Toward Efficient Indoor Positioning, Sixth Annual IEEE International

Conference on Pervasive Computing and Communications, pp. 101109,

2008.

The authors proposed a new analytical model that employed proximity graphs for

predicting performance of indoor positioning systems based on location fingerprinting.

This model allows computation of an approximate probability distribution of error dis-

tance given a location fingerprint database based on received signal strength and its

associated statistics.

[5] S. Fang, T. Lin, and P. Lin, Location Fingerprinting In A Decor-

related Space, Knowledge and Data Engineering, IEEE Transactions on,

vol. 20, no. 5, pp. 685691, 2008.

The authors proposed by projecting the measured signals into a decorrelated signal

space, the positioning accuracy can be improved. This approach achieves a more

efficient information compaction and provides a better scheme to reduce the online

computational complexity. The whole APs information can be utilized. And an addi-

tional advantage of technique is that fewer training samples are required to build the

localization system.

[6] Tatsuya Iwase and Ryosuke Shibasaki, Infra-free Indoor Positioning

Using only Smartphone Sensors, International Conference on Indoor Po-

sitioning and Indoor Navigation, 28th-31th October 2013.

A solution to reduce the accumulative error of pedestrian dead reckoning carried out

with only the low cost sensors and WiFi in smartphones by realizing cooperative posi-

tioning among multiple pedestrians. Authors proposed a method to introduces linkage



structures to simplify trajectories of pedestrians. The structures work as a constraint

to reduce the number of variables to be estimated. Since this constraint makes posi-

tioning problems solvable with a small number of observations, this proposed method

screened WiFi data by the signal strength and selected only strong signals for use in

positioning an indoor positioning technique which uses sensors on smartphones without

using any infrastructure.

[7] Yang Liu1, Marzieh Dashti, Mohd Amiruddin Abd Rahman and Jie

Zhang, Indoor Localization using Smartphone Inertial Sensors, IEEE2014.

In this paper, an economic and easy-to-deploy indoor localization model suitable for

ubiquitous smartphone platforms have been established. The method processes em-

bedded inertial sensors readings through a inertial localization system. A particle filter

is developed to integrate the building map constraints and inertial localization results

to estimate users location. To increase the algorithm convergence rate, the users

initial/on-line room-level localization is achieved using WiFi signals. To achieve room-

level accuracy, only very few training WiFi data, i.e. one per room or per segment of

a corridor, are required.

[8] Igor Bisio, Fabio Lavagetto, Mario Marches and Andrea Sciarrone,

Energy Efficient WiFi-based Fingerprinting for Indoor Positioning with

Smartphones, Globecom 2013-Wireless Networking Symposium

The paper presents an energy efficient WiFi-based indoor positioning algorithm, based

on the probabilistic fingerprinting method, suited to be used over smartphone plat-

forms. The work proposed here a simple algebraic approach aimed at reducing the

computational and energy loads of the probabilistic fingerprinting, which is employed

to carry out the position of a smartphone on the basis of the captured WiFi Access

Points signal strengths in an indoor area. The presented solution does not apply any

kind of approximation with respect to the traditional approach, so avoiding accuracy

detriment. The idea is to factoring out the parts of the probabilistic fingerprint for-

mulae that can be computed a-priori, so reducing the computational burden of the

positioning process.



[9]Eddie C. L. Chan and George Baciu, S.C. Mak, Using Fuzzy Color

Maps to Increase the Positioning Accuracy in Poor Wi-Fi Coverage Re-

gions, IEEE 7th International Conference on Wireless and Mobile Com-

puting, Networking and Communications

Many Researchers used the orientation filter and the Newton Trust Region (TR) algo-

rithm to enhance the traditional LF by filtering the noisy signal. This paper, proposed

approach is divided into four phases. In the first phase that detect the IEEE 802.11b

Wi-Fi signal strength and collect the LFs into a training database. In the second phase

to create a fuzzy color map to visualize the distribution of Wi-Fi signal. To imposed

the Wi-Fi signal distribution color-coded as follows: red represents strong signals and

blue represents weak signals. The semantic expressivity of Fuzzy Logic makes easier

the best fingerprints extraction. Then, this fuzzy color map is used to improve the K-

mean positioning algorithm by selecting location fingerprints from those access points

in the red region. Finally, to apply the orientation filter and the Trust-Region method

to estimate the position.



3 Theory

Wi-Fi based positioning system can be divide into three first one is Fuzzy Signal Color

Mapping, Secondly the positioning methodologies and and third system design

3.1 Fuzzy nature

Due to inherent uncertainty or fuzziness involved in defining Received signal strength,

the RSS can be treated as fuzzy variable. Thus Fuzzy Logic can be used to model the

wireless received signal strength. The fuzzy set theory provides a mechanism for repre-

senting linguistic constructs such as many, low, medium ,often, few. The

fuzzy logic provides an inference structure that enable appropriate human reasoning

capabilities. The theory of fuzzy logic is based upon the notion of relative graded

membership. The utility of fuzzy set lies in their ability to model uncertain or am-

biguous data. The WLAN signal distribution color-coded as follows: red represents

strong signals and blue represents weak signals. The produced fuzzy signal graph could

give a very concrete visualization and useful pre-positioning information that helps to

estimate the location.

3.1.1 Fuzzy Membership Function

By using Fuzzy Logic mapping of RSS from a 0 to 1 fuzzy membership function can

be achieved. This approach does not use a numeric value. It uses fuzzy logic to broadly

categorize the RSS as strong, normal, or weak.

P (x) = 1

2e222 99K (1)

where P(x) is the probability function, x is the normalized RSS, is the standard

deviation of normalized signal normalized strength in a region, is the mean of signal

strength in a region. The membership function of term set, (RSS Density) is equal

to the set of Red,Green,Blue. Red means the signal strength density is strong; green

means the signal strength is normal; and blue means the signal strength density is

weak. The fuzzy set interval of blue is [0, 0.5], [0, 1] is green and [0.5, 1] is red.



For the blue region, = 0.5, = 0.

Blue(0 < x < 0.5) =22pie2x

2 99K (2)

For the green region substitute, = 0.5, = 0.5.

Green(0 < x < 1) =22pie2(x

12 )2 99K (3)

For the Red region substitute, = 0.5, = 1.

Red(0.5 < x < 1) =22pie2(x1)

2 99K (4)

Figure 2: Fuzzy Membership function

Figure 2 shows the fuzzy membership function. X-axis represents the normalized

signal strength from 0 to 1 (from -93dBm to -15dBm). The width of membership func-

tion depends on the standard deviation of the RSS. The overlap area can be indicated

by mixed colors.

3.1.2 Fuzzy Signal Distribution

Using different colored regions to represent the WLAN RSS distribution. Concep-

tually a spatio-temporal region is defined as follows: Assume that B is a finite set

of RSS vectors belonging to a particular color region, where B = b1, ....bn |bi Rn i.ebi S,S R, and S [l, u] where l is the lower bound of fuzzy interval and u isa upper bound of fuzzy interval. To analyze the distribution surfaces S, there always

exists a spatio-temporal mapping, q : B Sq(x) =

sh(x)b(S)dS 99K (5)



h(x) =

1, if x S0, x / S99K (6)

where h(x) is the characteristic function of S, i.e., b(S) is a weight function that

specifies a prior on the distribution of surfaces Explicitly define b(S) by the signal

propagation loss algorithm which calculates the received signal strength (RSS) with

path loss as follows: R = r0 10 log10(d) wallLoss 99K (7)where r0 is initial RSS, d is a distance from access points (APs) to a location, is

the path loss exponent (clutter density factor) and wallLoss is the sum of the losses

introduced by each wall on the line segment drawn at Euclidean distance d. Initially,

r0 is the initial RSS at the reference distance of d0 is 1 meter.

3.2 POSITIONING METHODOLOGIES

Orientation filter, Location Fingerprinting and the Newton Trust- Region methods

are used to estimate the position when the signal is stable. But when a person enters

in a poor Wi-Fi coverage region, the positioning accuracy drops dramatically. To solve

the inaccurate estimation due to the unstable signal, By selecting good candidates of

AP from the red region of the map that to estimate the position. Figure 3 shows the

flowchart which illustrated the process of each step of the the proposed positioning

approach. As shown in the flowchart,first testing of stability of signal by the Signal-to-

Interference-plus-Noise-Ratio (SINR) is done. The estimated position is too far away

from the Trust-Region. If either the signal is unstable or the estimation point is too far

away from the Trust-Region, instead of picking K-nearest fingerprints, the fingerprints

from the AP in the red region is chosen. Testing the Stability of the Signal In following

subsection, by using the Signal-to-Interference plus- Noise-Ratio (SINR) to determine

whether the signal is stable. SINR is commonly used in wireless communication as a

way to measure the quality of wireless connections.The energy of a signal fades with

distance, which is referred to as a path loss in wireless networks. The definition of



Figure 3: Flowchart of positioning approach

SINR is usually defined for a particular receiver (or user). In particular, for a receiver

located at some point x in space (usually, on the plane), then its corresponding SINR

given by as

SINR(x) = PI+N

where P is the power of the incoming signal of interest, I is the interference power



of the other (interfering) signals in the network, and N is some noise term, which may

be a constant or random. Like other ratios in electronic engineering and related fields,

the SINR is often expressed in decibels or dB.

3.2.1 Testing the Stability of the Signal

In following subsection, by using the Signal-to-Interference plus- Noise-Ratio (SINR)

to determine whether the signal is stable. SINR is a very common indicator to measure

interference and defined as follows:

SINR = Rb(c)

R+n

99K (8)

where Rb is the highest RSS after path loss calculation. R is the remaining set of

RSS after path loss calculation. n is the noise signal strength. Rb, R, n are in dBm

unit. Usually, n should have the value of -100dbm. The interference-level function is

defined as follows.

(C) = max(0, 1 kC) 99K (9)where C is the absolute channel difference and k is the non-overlapping ratio of

two channels. and C are in Db unit. For example, if SINR 0.3, the signal suffersfrom a large interference which make it unstable.

3.2.2 Estimation Testing Within the Trust-Region

To test whether the estimated point is too far away from the Trust-Region (TR).

k+1 =

2k, if Pk 2k, if Pk [2, 1]

1k, if Pk 199K (10)

where Pk represents the TR fidelity, k represents the TR radius.2 and 1] represent

the lower and upper bound of TR fidelity. 1 and 2 represent the changing ratio of

the TR radius. If Pk > 32 holds, it implies the estimated point is three times away

from the region.



3.2.3 K-Mean Location Fingerprinting Algorithm

The K-Mean Location Fingerprinting algorithm requires the online RSS to search for

K closest matches of known locations in signal space from the previously-built database

according to root mean square errors principle. By averaging these K location candi-

dates with or without adopting the distances in signal space as weights, an estimated

location is obtained via weighted K-NN or unweighted K-NN. K-NN classifies a new

data point based on the majority of its K-nearest neighbors. In the simplest case

(K=1), the algorithm finds the single closest match and use that fingerprints loca-

tion as prediction. There are two sets of data for the Location-Fingerprint approach.

The first set of data is the oine samples of RSS from N APs in the area. Each ele-

ment in a vector is an independent RSS (in dBm) collected from APs in the location.

W = 1...n|i Re is a set of online sampling LF vectors in database. Instead ofusing entire set of oine LF vectors, one can use the LFs from APs in the red region

where W B S,S [0.5, 1], B is now a finite set of RSS in the red color regionand defined. The second set of data contains online RSSs,R = R1...Rn|Ri Re fromn APs at a particular location. The K-NN algorithm requires the collection of data

(i, dq), i, q N , for n locations in the site, where dq is the known location of the qthmeasurement and the vector w i is the fingerprint of the location di. When areceiver in unknown location A becomes aware of a new fingerprint r, it searches for

the fingerprint wi that is closest to r and then estimates the location. The unknown

location for r is decided by a majority vote from the K shortest distance fingerprints.

Estimation the location dq by clustering the distance between online received LF vec-

tor r and oine sampling LF vector i as. vector r and oine sampling LF vector i

as dq(r, w) = (n

i=1 |r i|q)1/q 99K (11) dq is called Manhattan distance if q=1 andEuclidean distance if q=2 the accuracy does not necessarily higher as q increases. Let

wij be jth sample RSS in the ith access points. m is the number of access points. n is

the number of sample data. The distance between w i and wij is defined asdj =

mi=1 ij i j=1,2....n

Electing K samples since the smallest value and calculate average coordinates as

outputs in following equation: (x, y) = 1k

ki=1(xi, yi) where (xi, yi is coordinates cor-



responding to ith sample.

3.3 System Design

The overall vision of proposed positioning system is shown in Fig.4. The working

Figure 4: system design

process consists of four main phases:

1) localization with room-level accuracy using Wi-Fi signals.

2) step detection using accelerometer sensor.

3) turning angle calculation using digital compass sensor.

4) considering building floor plan constraints to avoid collision to the walls.

Finally the particle filter fuses this information to provide a robust tracking system.

Particle filter employs particles to estimate users position and motion trends. The

key idea is that the particles should not occupy impossible positions given the map

constraints and users motion model. since the user cannot pass through walls, the

particles that cross a wall are weighted down. When a new step is detected, users

motion model is updated by compass readings. Finally, the users position is estimated

to be the average of particles positions.



3.3.1 Analyzing accelerometer and compass data

Due to repetitive nature of walks, when a user walks continually, similar pattern

for every step is found on accelerometer reading, a step detection algorithm based

on computing the cross-correlation of the previous sequence of accelerometer readings

samples with current sequence of accelerometer readings samples on the direction of

walks is presented. In this system, similar approach is utilized. Every time a new

step is detected, particle filter estimates the new position status by combining building

floor map constraints information and compass measurements. The compass reading

samples are stored in buffer and processed at the right time, i.e., the moments that

ever step is detected.

3.3.2 Particle filter

Particle filter is a practical implementation of recursive Bayesian filter with the se-

quential Monte Carlo methods. Bayesian filter estimates a system state vector by con-

structing the posterior probability density function (pdf) of that state. The posterior

pdf should reflect all the up-to-date knowledge about the system state vector. When

new information is received by recursive Bayesian filter, propagating and updating of

the previous posterior pdf will be implemented. Particle filter approaches probability

density function by distributing different weights to large amount of randomly gener-

ated samples. All the randomly generated samples are propagated and updated with

the motion model and measurement model, respectively. Unlike Kalman filter, particle

filter does not requires a prior knowledge of the exact position of user. Also, the model

of the inherent noise embedded in newly entered information and measurement used by

particle filter is not limited to Gaussian distribution. The particle filter implemented in

this system and the assumption used are similar to . The difference is the mechanism

of filtering particles. Convergence rate of particle filter in this system is faster and

reliable as well. The contribution is from implementing on-line room-level localization

technique.



3.3.3 Map constraints

Building floor map constraints are considered to calculate possible intersections of

particles trajectory with walls. A collision between moving particle and wall represents

that a possible user moving path intersects walls. As a result, the particle should be

filtered. An easy-to-implement collision detection algorithm is introduced. Given the

floor plan of the building, the coordinates of every walls and partitions are recorded in

the cloud server. To determine if the users trajectory collides a specific wall or not,

a novel approach is applied as following: Given both the coordinates of users current

location, (x1, y1), and possible next location,(x2, y2) as motion endpoints coordinates,

a linear equation representing the users trajectory can be written out , in the form

of y = kx + b. Similarly, linear equations for wall vectors could be built like that as

well. Then, by substituting two motion endpoints 2-D coordinates for x and y in wall

linear function, two values of M1 andM2 are obtained. Similarly, by substituting two

wall end points coordinates for x and y in linear motion function, two values of W1

and W2 are obtained. Three different scenarios may happen:

Figure 5: Collision is detected

M1andM2 have the same polarities (they are both negative or both positive val-ues) AND the polarities of W1 andW2 are the same as well: in this scenario users

trajectory does not collide the wall. The black dashed line shown in Fig. 4.

M1andM2 have different polarities (one of them is a negative value and the other



is positive) AND W1andW2 have different polarities as well: in this scenario users

trajectory collide the wall. See blue dashed line shown in Fig. 5.

either M1andM2 have different polarities AND W1andW2 have same polaritiesOR M1 and M2 have same polarities AND W1 and W2 have different polarities

: in this scenario users trajectory does not collide the wall. See red line shown

in Fig. 5. . . .

3.3.4 Room-level localization

1) Initial room: In the oine phase, few Wi-Fi training data (RSS fingerprints)

are collected only at the center of the rooms, and center of each segments of long

corridors. These few training data limits the initial range of particles. Consequently

the filter convergence to the exact users location more quickly. The basic idea is that

the users online Wi-Fi scans are compared to all few training data to find the nearest

training data in signal space. In this system, using the algorithm presented in to get

the comparison results. Figure 5, A collision is detected in the second scenario. That

gives an initial room level accuracy of the users position. Then, particles are uniformly

distributed in and only in the users estimated initial room.

2) On-line room-level localization: As it is described, particle filter fuses inertial

data and map constraints and accordingly updates the particles continuously. At every

step, the users location is calculated as the average of survived particles positions.

Often when users go out from a room, experience a turn at corridor and then go along

the corridor, survived particles will be distributed in several different room or corridor

segments, so called multi-clustered particles. Clustered particles far from ground true

position of user, as shown in Fig. 5, have serious negative impact on location estimate.

To address this problem, on-line room-level localization algorithm is applied. This

algorithm only runs when a dramatic change in users status is observed, e.g., when

compass readings at two consecutive steps vary significantly. The algorithm aims at

limiting the particles in one room or one segment of a corridor. This task is also

finished by comparing on-line Wi-Fi scans with training fingerprints. The room or

corridor segment holding the smallest difference in signal space will be selected to be



on-line room localization result. The particles beyond determined room or corridor

segment range are filtered out. These useless particles are always filtered when their

future movements intersects with walls even if on-line room-level localization technique

is not used i.e., the application of on-line room-level localization technique accelerates

the process of dismissing these particles, which saves the computational load involved

in propagation of and collision detection of useless particles as well. Moreover, on-line

user location estimation error is further reduced by filtering these particles by on-line

room-level localization technique.

3.3.5 Updating the database through crowd-sourcing

Wi-Fi infrastructures may be added/removed/ or moved to new places according to

wireless coverage requirements. The un-updated RSS database causes inaccurate room-

level positioning. Also, Collecting signal fingerprints deliberately to build a training

database covering all points on the interested indoor area is a time-consuming effort.

For tackling this problem, a simple crowd-sourcing approach, so-called back tracking

is developed to keep the database updated. To implement the back-tracking approach,

all particles are labelled uniquely at the start point of the trajectory. The number

of survived particles reduces by time due to filtering mechanism from on-line room-

level localization and collision detection. Hence at the end point of the trajectory, the

final survived particles are more exactly converge to the exact users position. The

algorithm track the survived (and labelled) particles backward from the end of the

trajectory to the start point. Trough back propagation of final survived particles,

the position information of these tagged particles throughout the whole trajectory are

obtained. Average of particles position at every step is calculated as users location.

The back-tracking algorithm is applied in the off-line analysis phase after users arrive

at the desired locations. In this way, the Wi-Fi fingerprints database covering whole

interested indoor space could be built after several hours trial and updated as well.

Then, on-line localization accuracy could be benefited from comparing on-line Wi-Fi

scans with the fingerprints data in the database.



3.4 Implemented project by Google

Google has successfully implemented Wi-Fi based positioning system by using omni-

directional radio antenna. This antennae receives publicly broadcast wifi radio signals

within range of the vehicle. The vehicle travels at normal road speeds, and so spends

only a very short amount of time within the range of any given wifi access point.

The signals are initially processed onboard in the car, using software including the

standard Kismet open source application. The data is then further processed when

transferred to servers within a Google Data Centre, and used to compile the Google

location based services database. The equipment within the vehicle operates passively,

receiving signals broadcast to it but not The information visible to the equipment is

that which is publicly broadcast over the radio network, using the 802.11 standard.

This includes the 802.11b/g/n protocols. The equipment is able to receive data from

all broadcast frames. This includes, from the header data, SSID and MAC addresses.

All data payload from data frames are discarded, so Google never collects the content

of any communications. In addition, the operator of the access point can choose to

restrict the SSID from broadcast, and in many cases this will mean that the SSID is

not received (although this may vary depending on the way the access point broadcasts

data). The equipment also separately records the signal strength and channel of the

broadcast at the point at which it was received by equipment, and is able to establish

the protocol used (i.e. 802.11b/g/n). It is possible to identify from the data received

if an access point is encrypted - this may be included in the data sent in the frame

header but in any event will be self-evident from the presence of encryption within the

frames generally.

The data which is collected used to provide location based services within Google

products and to users of the Geolocation API. For example, users of Google Maps for

Mobile can turn on My Location to identify their approximate location based on

cell towers and wifi access points which are visible to their device. Similarly, users of

sites like Twitter can use location based services to add a geolocation to give greater

context to their messages. Google currently uses 2 pieces of the data collected during



the driving operation to build its database and provide location based services - the

MAC address of the access point and the GPS co-ordinates of the vehicle at the point

at which the access point was visible. This data is stored in aggregate form, and is

used to provide the location based service. Google location based services using wifi

access point data work as follows:

The users device sends a request to the Google location server with a list ofMAC addresses which are currently visible to the device;

The location server compares the MAC addresses seen by the users device withits list of known MAC addresses, and identifies associated geocoded locations

(i.e. latitude /longitude);

The location server then uses the geocoded locations associated with visible MACaddress to triangulate the approximate location of the user;

This approximate location is geocoded and sent back to the users device. . . .

The only data which Google discloses is a triangulated geocode which is an approximate

location of the users device. At no point does Google publicly disclose MAC addresses

from its database (in contrast with some other providers in Germany and elsewhere).

There has been speculation that Google will make available a map or list of wifi access

points, including identifying the SSID of each access point and/or identifying those

which are open.



4 Application

WPS is a fast-growing, technologically sophisticated field, with potential applications

in many different industries. The most common civilian applications to date have been

land, air and marine navigation, and surveying. More recent applications include agri-

culture, aircraft precision approach, robotics, intelligent traffic systems, construction,

resource extraction, and geographic information systems (GIS). Developments in hy-

brid GPS and WPS mean increased reliability and accuracy, and even more widespread

possibilities for tomorrow.

Automobile Navigation : An automotive navigation system incorporates GPS +WPS and route guidance technology to give accurate real-time route direction

and guidance. Major automobile manufactures now offer optional navigation sys-

tems, including General Motors, which first offered, GuideStar from Oldsmobile

and later OnStar from Cadillac.

BMW is offering a navigation system with a liquid-crystal monitor includes

controls for the navigation system, cellular phone, audio system, on-board com-

puter, security system and automatic ventilation system. BMW owners will have

an emergency function that automatically calculates the vehicles exact location

and displays it on the monitor.

Fleet Tracking: Automatic Vehicle Location (AVL) is the combination of GPS+WPS and computer technology for use with plans, automobiles, trucks and ships.

Vehicles equipped with fleet tracking can be tracked in real time or trips can be

replayed later. You can determine the exact track of each journey and identify

the exact position and time of specific events such as deliveries, extended stops,

etc. Citycab, one of Singapores primary taxi companies, has deployed innovative

satellite technology to provide a solution to the citys transportation dilemma.

The Hilton Hotel chain has applied this idea to its 27 airport hotels in the US.

In addition to tracking each vans exact location, improving driver efficiency and



providing advance guest registration, the system is also being used to rapidly

locate a guests automobile upon checkout for immediate front-door delivery.

Mapping & Surveying : WPS + GPS Surveying is the ideal way to surveypositions in a short period of time. Each point surveyed is independent of the

other, which eliminates time consuming traverse. With GPS, an unlimited area

can be covered using satellite transmissions. One person is capable of conducting

a field survey on foot or in a vehicle. Integrated with GIS systems, maps can be

computer generated directly from the field survey data and WPS added more for

indoor.

Mining: Using the WPS gear linked to a computer-designed mine plan, this helpsthe company achieve crucial new levels of efficiency. . . .



5 Conclusion

An economic and reliable indoor positioning apporach suitable for smartphone plat-

forms which provides a more accurate and robust mechanism to solve the traditional

poor indoor GPS problem by Wi-Fi based positioning system.In Propagation-based

approache estimated the position by measuring the received signal strength (RSS) so

for poor RSS the system performed poor.The LF method is highly affected by internal

building infrastructure changes, presence of humans, and interference among other de-

vices. So in Wi-Fi based positioning system in fuzzy signal color map helps to improve

the K-mean positioning algorithm by sampling location fingerprints from access points

in the red regions.



6 References

[1] Binghao Li1, James Salter2, Andrew G. Dempster1 and Chris Rizos1, Indoor

Positioning Techniques Based on Wireless LAN, IEEE Conference, .

[2] J. Kwon, B. Dundar, and P. Varaiya, Hybrid algorithm for indoor positioning

using wireless LAN, IEEE 60th Vehicular Technology Conference, VTC, vol. 7, 2004

[3] K. Kaemarungsi and P. Krishnamurthy, Modeling of indoor positioning systems

based on location fingerprinting, INFOCOM. Twenty-third Annual Joint Conference

of the IEEE Computer and Communications Societies, vol. 2, 2004.

[4] N. Swangmuang and P. Krishnamurthy, Location Fingerprint Analyses To-

ward Efficient Indoor Positioning, Sixth Annual IEEE International Conference on

Pervasive Computing and Communications, pp. 101109, 2008.

[5] S. Fang, T. Lin, and P. Lin, Location Fingerprinting In A Decorrelated Space,

Knowledge and Data Engineering, IEEE Transactions on, vol. 20, no. 5, pp. 685691,

2008.

[6] Tatsuya Iwase and Ryosuke Shibasaki, Infra-free Indoor Positioning Using only

Smartphone Sensors, International Conference on Indoor Positioning and Indoor Nav-

igation, 28th-31th October 2013.

[7] Yang Liu1, Marzieh Dashti, Mohd Amiruddin Abd Rahman and Jie Zhang,

Indoor Localization using Smartphone Inertial Sensors, IEEE2014.

[8] Igor Bisio, Fabio Lavagetto, Mario Marches and Andrea Sciarrone, Energy Effi-

cient WiFi-based Fingerprinting for Indoor Positioning with Smartphones, Globecom

2013-Wireless Networking Symposium

[9]Eddie C. L. Chan and George Baciu, S.C. Mak, Using Fuzzy Color Maps to

Increase the Positioning Accuracy in Poor Wi-Fi Coverage Regions, IEEE 7th Inter-

national Conference on Wireless and Mobile Computing, Networking and Communi-

cations.

[10] Google Help - Location-based services-How do I opt out? Obtained 2012-05-30.

[11] website: http : //gps.about.com/od/glossary/g/wifiposition.htm


CertificateAcknowledgementIntroductionLiterature SurveyTheoryFuzzy natureFuzzy Membership FunctionFuzzy Signal Distribution

POSITIONING METHODOLOGIESTesting the Stability of the SignalEstimation Testing Within the Trust-RegionK-Mean Location Fingerprinting Algorithm

System DesignAnalyzing accelerometer and compass dataParticle filterMap constraintsRoom-level localizationUpdating the database through crowd-sourcing

Implemented project by Google

ApplicationConclusionReferences

wireless positioning system

Documents