TELKOM
NIKA
, Vol.14, No
.2, June 20
16
, pp. 748~7
5
6
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v14i1.2901
748
Re
cei
v
ed O
c
t
ober 2
1
, 201
5; Revi
se
d Decem
b
e
r
19, 2015; Accept
ed Ja
nua
ry 4,
2016
Improved Indoor Location Systems in a Controlled
Environments
Selcuk Helhel*
1
, Atala
y
Koca
kus
a
k
2
Akden
iz Univ
er
sit
y
, Eng
i
ne
eri
ng F
a
cult
y, De
partment of E.E.E., 07058, Antal
y
a, T
u
rke
y
T
e
lephon
e: +
90 242 3
10 6
3
9
3
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: selcukh
e
lh
el
@akd
eniz.e
du.
tr
1
, atala
y
ak
oc
akusak
@gma
il
.com
2
A
b
st
r
a
ct
T
he prec
ise l
o
cali
z
a
ti
on
by u
s
ing W
i
-F
i Acc
e
ss Poi
n
t (AP) has b
e
co
me
a
very i
m
port
ant
issue for
ind
oor l
o
catio
n
base
d
servic
e
s
such as
mar
k
eting, p
a
tie
n
t follow
up
and
so on. Pres
ent
AP local
i
z
a
ti
o
n
systems
are w
o
rkin
g o
n
sp
ec
ially
d
e
sig
n
e
d
Wi-Fi units, a
n
d
the
i
r a
l
gor
ith
m
s
usin
g ra
di
o
sign
al str
engt
h
(RSS) exhi
bit (relativ
e
ly) hi
gh
errors, so ind
u
stry
looks
mo
re precis
e an
d
fast adaptab
le
meth
ods. A n
e
w
mo
de
l cons
ide
r
ing/e
l
i
m
i
natin
g
strong RSS l
e
vels in
ad
ditio
n
to clos
e dist
ance
error e
l
i
m
i
nati
on a
l
gor
i
t
hm
(CDEEA) co
mbin
ed w
i
th me
dia
n
filters ha
s been pro
p
o
s
ed in ord
e
r to increas
e the perfor
m
a
n
ce
of
conve
n
tio
nal
R
SS bas
ed l
o
ca
tion syste
m
s.
Coll
ectin
g
l
o
ca
l
sign
al stre
ngt
hs by
means
o
f
an or
din
a
ry
WiFi
units pr
esent o
n
any l
apto
p
a
s
a receiv
er is
follo
w
ed
by the
appl
icati
on of
CDEEA to el
i
m
inate stro
ng R
SS
levels. Me
dia
n
filter is then
appl
ied to th
ose
el
i
m
in
ated
values, an
d
AP based
pat
h loss mod
e
l
is
gen
erate
d
, ada
ptivelly. F
i
na
lly
, the
propose
d
algorit
hm pr
ed
icts locatio
n
s w
i
thin a max
i
mum
me
an error
of
2.96
m for 90
%
precisi
on l
e
ve
l. T
h
is achi
eve
m
e
n
t w
i
th an
ordi
nary w
i
fi u
n
its prese
n
t o
n
any co
mmer
c
ial
lapto
p
is co
mp
arab
ly at very goo
d lev
e
l in l
i
terature.
Ke
y
w
ords
: Ind
oor rad
i
o pro
p
a
gatio
n, locati
ng
algor
ith
m
s, lap
t
op as a receiv
er, signa
l an
aly
s
is, W
i
F
i
Copy
right
©
2016 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Locating i
s
a
n
e
sse
ntial t
e
ch
nolo
g
y which fin
d
s lot
s
of in
du
stria
l
appli
c
ation
s
su
ch
as
marketing, rehabilitation
campus
es, mil
i
tary and
security ap
plications, and so on
[1-5]
.
Ins
t
e
ad
locatin
g
tech
nologi
es a
nd
its appli
c
ation
s
are ei
ther ti
me based o
r
radio
sign
al strength (RSSI)
based
syste
m
s, mo
st
of the te
ch
n
o
logi
es prefer RS
SI
based systems.
GPS a
nd cellul
a
r ba
se
d
system
s have very
satisfi
ed lo
cation preci
s
ion
capability at out
door, but they
have
almost no
cap
ability (G
PS) or limite
d
cap
ability (cellular) to
cov
e
r for i
ndo
or
positio
ning. I
t
shoul
d be
n
o
ted
that cellul
a
r t
e
ch
nolo
g
ies
are
also op
erator d
epe
nde
nt, not flexible, and
not a
p
p
lica
b
le to i
n
door
locatin
g
. In
orde
r to
in
crease the
pe
rforman
c
e
of
indoo
r p
o
siti
oning
services, m
o
st i
n
d
o
o
r
locatin
g
sy
stems
use mult
ilateration
on
fingerprintin
g
po
sitionin
g
method
s [2].
Fixing a lo
cat
i
on
requi
re
s
so
me referen
c
e fram
es to
de
scrib
e
p
o
sition
s
rel
a
tive to tho
s
e
pre-d
e
termi
n
e
d
referen
c
e
s
, a
nd fram
es are co
mmonly
calle
d a
s
co
ordin
a
te
syst
ems i
n
which
any lo
cation
is
specified with respect to it
s orig
i
n
. Indo
or locating te
ch
nologi
es
are rapidly g
r
o
w
in
g a
s
a
re
sult
of
an increa
se i
n
popula
r
ity of mobile eq
uipment. Thi
s
popula
r
ity require
s that d
e
velopme
n
t of a
prop
er
pro
p
a
gation mo
del
and lo
catin
g
algo
rithm
s
are e
s
senti
a
ls for
unint
errupted
and
/or
pre
c
ise locating syste
m
s i
n
an indo
or
environ
m
ent.
Requi
rem
e
n
t
to use of a prop
er mo
d
e
l
force
s
scient
ists to
inve
stigate propa
gation m
e
ch
anism
s [1
-7].
The
r
e i
s
al
so treme
ndo
us
increase in
Wi-Fi l
o
calization
system
a
pplication
s
in
an auton
om
ously n
a
vigating ro
bot proj
ect
[3, 4]. Suc
h
models
,
bas
i
c
a
lly, us
e Wi-Fi
s
i
gnature map
with geomet
ric
c
o
ns
traints
and
introdu
ce a
continuo
us p
e
rceptu
a
l mode
l of t
he environment gen
erated from the
discrete g
r
ap
h-
based
Wi-F
i sign
al stren
g
th sam
p
ling.
Contin
uou
s l
o
cali
zatio
n
te
chni
que
s refe
rrin
g
to kno
w
n
referen
c
e poi
nts are
slightl
y
different than the ce
rtain locatio
n
identi
f
ication at an
y time.
RSSI based i
n
building po
sitionin
g
syst
ems are gro
w
ing very ra
p
i
dly in two ways. One
way of ind
o
o
r
positio
ning i
s
pa
ssive-lo
calizatio
n an
d
the othe
r on
e
is
called
acti
ve-localizatio
n.
Passive
-lo
c
ali
z
ation
syste
m
s are ba
si
cally
ba
sed
on Ra
dio Freque
ncy Ide
n
tification RFID
techn
o
logie
s
whi
c
h d
o
n
o
t allow two
way comm
uni
cations, b
u
t p
r
ovide
com
p
arably
accu
rate
locatio
n
[5, 6]
. Ozd
eni
zci,
e
t
al., [6] pre
s
ent a
ca
se
st
udy for th
e
system
req
u
ire
m
ents giving
the
desi
gn d
e
tail
s. They
co
mpared thei
r prop
osed
a
ppro
a
ch with
existing in
d
oor n
a
vigati
on
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Im
proved Ind
oor Lo
catio
n
System
s in a
Controlled En
viro
nm
ents (Selcu
k Hel
hel
)
749
system
s, an
d
notice
d
that t
heir
pr
o
p
o
s
ed
model i
s
co
st effective.
Active-locali
z
ati
on sy
stem
s a
r
e
based
on
Wi
-Fi AP, GSM, CDMA a
nd
sma
r
t ph
one
appli
c
ation
s
allowin
g
bro
a
d
ba
nd t
w
o
-
way
data com
m
u
n
icatio
ns. Th
e indu
strial
Scientif
ic Me
dical (ISM
) 2.4 GHz na
rrowban
d ind
oor
cha
nnel i
s
al
so used in m
a
ny contem
po
rary medi
cal
a
pplication
s
,
such as wirele
ss physi
ologi
cal
sen
s
o
r
netwo
rks [7]. Another study
presents a visual
localization
approa
ch tha
t
is suitable for
dome
s
tic
an
d indu
strial
environ
ment
s as it e
nab
les a
c
cu
rate,
reliabl
e an
d ro
bu
st po
se
estimation.
T
hey u
s
ed
inn
o
vative artificial lan
d
ma
rks (clu
sters
of
high i
n
ten
s
ity RGB
LE
Ds)
on
the ceiling which creat
e
cancell
a
tion [7].
There a
r
e
stu
d
ies [8
-1
0] using Wi
-Fi
ba
sed
sy
stem
s f
o
r in
doo
r lo
cating. Ca
mpo
s
, et al.,
[8] pre
s
e
n
ted
a
system
for multi-floo
r i
n
door p
o
sitioni
ng
whi
c
h
co
n
s
ide
r
s a
r
chite
c
tural
a
s
p
e
ct
s.
They propo
sed a
Data
Co
rrel
a
tion Met
hod
combi
n
e
d
with n
eural
netwo
rk
appli
c
ation
s
o
n
th
em.
They compa
r
ed m
e
a
s
u
r
e
d
Ra
dio Sig
nal Stren
g
th
(RSS
) level
s
by a
pplyin
g
natu
r
al d
a
ta
clu
s
terin
g
a
nd d
a
ta
co
rrelation
met
hod
s. RSS
based te
ch
n
i
que
s d
o
n
o
t re
quire
any
synchro
n
ization li
ke
Time
of Arrival
(T
O
A
) an
d
Tim
e
Differen
c
e
of Arriv
a
l (T
DO
A)
meth
od
s
[
11,
12]. Mani, et
al., [13] pro
posed a
stu
d
y that par
a
m
eteri
z
e
s
a polari
m
etri
c diffuse scatte
ring
model in an i
ndoo
r environ
ment, sin
c
e d
i
ffuse or
d
ense multipath compon
ents pl
ay an importa
nt
role
in determining
the p
o
lari
zation be
havior
of
wireless tra
n
smi
ssi
on
chan
ne
ls. Thei
r anal
ysis
reveal
s
that diffuse scattering si
gnifi
ca
ntly depola
r
izes the i
m
ping
ing wave in i
ndoo
r sce
n
a
r
i
o
s.
Their meth
o
d
b
r
iefly tells that MIMO
appli
c
ation
s
will b
e
m
o
re
ben
eficial
for
better ind
o
o
r
locali
zation, since 3D ele
c
troma
gneti
c
field
in
the
air
will bri
ng extra gain
and
extra adva
n
tage
t
o
desi
gne
rs [14
-
15].
In this
study;
a medi
an filte
r
was ap
plied
to
RSSI valu
es
coll
ected
by wifi unit
s
pre
s
ent
on a
n
y com
m
ercial l
apto
p
, -5
5dBm le
vel wa
s
pre
s
et as the th
resh
old val
u
e
(n
amed
a
s
clo
s
e
distan
ce
bo
rder
or eq
uall
y
clo
s
e
dista
n
ce
erro
r eli
m
ination
pr
o
c
e
s
s) in
the
algorith
m
, an
d a
corre
c
tion fu
nction
ext
has
been
adde
d
to well kno
w
n p
a
th loss mod
e
l in
Section 3.2
.
Propo
se
d m
o
del
con
s
id
eri
ng
scann
ed
RSSI level in
additio
n
to
well kno
w
n
n-i
ndex b
a
sed
pat
h
loss formul
a is a ne
w com
m
ent to literature.
2. Materials
and Method
s
Wi-Fi t
r
an
smi
tters, hol
ding
omnidi
re
ction
a
l anten
na
s, operating b
e
twee
n 22
00 M
H
z
an
d
2600 M
H
z
were u
s
ed a
s
a part of mea
s
ureme
n
t se
t
up. The ante
nna had a
n
o
u
tput power o
f
2
1
dBm. A wifi u
n
it pre
s
ent
o
n
any o
r
dina
ry comme
rcial
laptop
wa
s
use
d
a
s
a
ra
dio re
ceive
r
with
our
own
software. Safe (clea
r) IP d
e
p
ende
nt
frequ
enci
e
s
we
re
sele
cted
not to be inte
rrept
ed b
y
addition
al co
mmuni
cation
traffic in the
system.
Ad
ditional traffic may affect the propa
gati
on
mech
ani
sm
s. Tran
smitting
antenn
as were
pla
c
ed j
u
st belo
w
th
e
roof that they
are
abo
ut 2.
4m
above the ground a
nd 30
cm belo
w
the
roof. The
accuracy of RF
receiver
wa
s assume
d as
1
(dBm), it is the averag
e re
ceived
sign
al powe
r
t
hat, in total, ~100
sampl
e
s
we
re taken
withi
n
a
cycle. Kim, e
t
al., [16] proposed a
stud
y to m
easu
r
e
small
-
scal
e fading an
d note
d
that 0.125
cm
measurement
interval at 2
.
4 GHz is e
n
ough. Sin
c
e
the used receiver spee
d in this stu
d
y is
much
faste
r
than Kim’
s st
udy; sam
p
lin
g interval i
s
greate
r
tha
n
0.125
cm, an
d 100
sa
mpl
e
s
were take
n in
one cal
c
ul
ation cycl
e in order to re
move small scal
e affects.
a) Co
rri
do
r for verificatio
n
measurement
s
b) Floo
r plan
Figure 1. Dist
ribution of
4
Wi-Fi trans
m
itters
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 2, June 20
16 : 748 – 75
6
750
Thre
e set
s
of measu
r
eme
n
ts had be
e
n
con
d
u
c
ted in total; By using first set’
s data,
stand
ard
pat
h loss me
asurem
ents
(when the
r
e
i
s
no ob
stru
cti
on bet
ween
transmitter a
n
d
receiver) had
been cond
u
c
ted in two
d
i
fferent co
rri
d
o
rs
of an en
ginee
ring fa
culty as se
en
in
Figure 1 for slope
cal
c
ul
ation, and th
e re
st one
wa
s u
s
ed fo
r mod
e
l verif
i
cation a
nd f
o
r
determi
ning
ext
value
a
s
mentione
d in
Section
2.1.
Shado
win
g
factor
ext
is normally a
rand
om va
ria
b
le d
epen
din
g
on
an
environment, b
u
t cal
c
ulate
d
on
e
al
so gua
ra
ntees/
c
overs that
rand
om varia
b
le.
3. Indoor Propaga
tion a
nd Loca
t
ing Algorithms
3.1. Slope of En
v
i
roment
Instead, th
ere
are
varying
i
ndoo
r p
r
op
ag
ation mo
del
s
valid in the
literatu
r
e [1
7, 1
8
], very
comm
on m
o
d
e
l used fo
r in
door path
loss
cal
c
ulatio
n
usin
g Radio
Signal Stre
ng
th (RSS
) is gi
ven
in Equation (1) [4].
ext
d
d
n
d
P
d
P
0
0
log
10
(1)
Whe
r
e
)
(
d
P
is the
sign
al st
ren
g
th obtain
ed
by t
he re
ceiv
er at a
dista
n
ce of
d an
d
)
(
0
d
P
is the
received
sign
al stre
ngth at
1m dista
n
ce (mean
s re
fe
re
nce
sign
al) b
o
th in dBm, n is the path lo
ss
index for ind
o
o
rs,
and
ext
is
n
o
rmally th
e ra
ndom va
ria
b
l
e
de
scribi
ng
sha
d
o
w
ing
fa
ctor [9, 10].
Four diffe
ren
t
Wi-Fi tran
smitters were locate
d aro
u
nd the co
rn
e
r
of measure
m
ent cam
pai
gn
pre
s
ente
d
in
Section 3.2. Corrid
ors
u
s
e
d
a
s
a
mea
s
urem
ent
cam
paign
is 7.2
m
wid
e
a
nd
50m
length. Since
n value is a value whi
c
h
strict
ly depend
s
on intere
sted
camp
aign’
s di
mensi
o
n
s
an
d
stru
ctures, m
easure
m
ent
s were
con
d
u
c
ted in or
der to calculate
IP depende
nt n values. In
logarith
m
ic f
o
rm of path
loss, n (sl
o
p
e
of pat
h loss)
can b
e
a
pproxim
ated
by Equation
(2).
Experimental
studie
s
sho
w
us th
at this equ
at
ion it
self ha
s not
enoug
h cap
ability to obtain
precise locati
on informations. As
it will
be detail
ed i
n
further sect
i
ons; thi
s
equation need to be
improve
d
by con
s
id
erin
g a
dditional pa
ra
meters or ad
ditional condit
i
ons.
0
10
/
log
10
d
d
n
(
2
)
Figure 2. Tra
n
smitter atten
uation an
d sl
ope (n
) calcul
ation
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Im
proved Ind
oor Lo
catio
n
System
s in a
Controlled En
viro
nm
ents (Selcu
k Hel
hel
)
751
Figure
2 sho
w
s path
lo
ss attenuation
v
e
rsus
dist
ance in 10
*log fo
r AP-A a
s
an
example
.
Similar p
a
th l
o
ss di
strib
u
tio
n
s
we
re
obtai
ned fo
r oth
e
r
three AP
s. Path lo
ss
mea
s
ureme
n
t poi
nts
are
align
e
d
b
y
60cm a
p
a
r
t
from e
a
ch
other that 1
2
0
different l
o
cations were d
e
termin
ed i
n
o
r
der
to calcul
ate the slop
e n prese
n
ted at Table 1.
A dataset sto
r
e
s
those path lo
ss index “n” values
for future u
s
a
ge. Table 1 g
i
ves IP depe
ndent calcul
a
t
ed n values t
hat they are they are
slight
ly
different then
from ea
ch oth
e
r.
Table 1. n(slo
pe) Ta
ble
WLAN (AP Code
)
Slope (n)
A 2.192
B 2.115
C 2.404
D 2.363
After obtaini
n
g
“n
”, o
ne
ne
eds to d
e
term
ine/de
scrib
e
shad
owi
ng
factor pa
ram
e
ter
ext
.
Instead
we
might have determi
ned t
h
is value fro
m
the literature referring
to the similar
measurement
camp
aign,
we p
r
eferre
d
to gene
ra
te
it. The value of this p
a
ram
e
ter
wa
s
gene
rated fro
m
control data (more
than
2,000 data di
stribute
d
at 576 di
fferent lo
cation
s) a
s
8.
62
by usin
g k-mean
s cl
uste
ring e
r
ror
cal
c
ulatio
n [
19]. It has to b
e
noted th
at there
we
re
fou
r
different
ext
para
m
eters valu
e
s
related
to e
a
ch
AP, and
t
hey are ve
ry
clo
s
e to
ea
ch
other.
Fin
a
l
value was ob
tained by m
e
dian filter
ap
plicatio
n. Ob
tained valu
e
can
co
mpe
n
sate sh
ado
win
g
affects which norm
a
lly has
to be determi
ned ra
ndo
mly.
3.2. Deriv
a
tion of Ne
w
E
m
piric Model for Indoor En
v
i
ronment
Although
the
aforementio
ned
path
lo
ss m
odel
is
a
goo
d startin
g
point for path loss
estimation,
such
a sim
p
le
model fo
r the
path loss o
n
l
y
exists in sp
ecial
ca
se
s. It was ob
se
rv
ed
that the
cal
c
ulation
of pat
h lo
ss
u
s
ing
Equation
(2
)
results in
so
me deviatio
n
, and
this sm
all
deviation in calcul
ated di
stance (to be
use
d
in mult
ilateral location cal
c
ulation) brings very bi
g
unexpe
cted
e
rro
r in
rel
a
tive co
ordinate
s
. By the
way, one
nee
ds t
o
take
into a
c
count/con
sid
e
r
addition
al co
rre
ction p
a
ra
meters to improve
Eq
u
a
tion (2
). Experim
ent
ally observed th
a
t
a
distan
ce
calculated by
u
s
i
ng stro
ng sig
nal
level
s
(RSSI >-5
5dBm
)
h
a
ve bi
g
rel
a
tive co
ordin
a
te
errors. Thi
s
i
s
mo
stly be
cause of tran
smitting ant
en
na ne
ar field
affects. Thi
s
observation
i
s
the
starting point that
thos
e
strong si
gnal
s n
eed to be eli
m
inated fr
o
m
referen
c
e di
stance cal
c
ul
atio
n
lis
t whic
h is
named as
c
l
ose dis
t
anc
e
error el
imination approac
h
(CDEEA). In a s
i
milar
way
but
not the sam
e
level; wea
k
RSS si
gna
ls scan
ned a
t
long distan
ce
s fail to predict a relati
ve
distan
ce a
s
expecte
d. Co
ordin
a
tes/p
o
sitions
were
calcul
ated ba
sed on Equ
a
tion (2
), and
we
observed d
e
v
iated coo
r
di
nates. Fig
u
re
3 demon
stra
tes initial co
ordin
a
te devi
a
tion dist
ribut
ion
function
with
respe
c
t to
RSSI, and we
descri
be
a fu
nction
nam
ed
as initial
coo
r
dinate
deviat
i
on
distrib
u
tion fu
nction
a
s
(m)
in term
s
of p
o
we
r
(P) i
n
d
B
m. The te
r
m
initial in
th
e tex
t
refe
r t
o
starting valu
e
s
whi
c
h i
s
goi
ng to be co
rrected by som
e
appli
c
ation
furhe
r
in this text.
923
.
818
636
.
39
63617
.
0
0033741
.
0
2
3
P
P
P
(
3
)
RSS level
s
referri
ng th
ose critical two
points
(one
for
clo
s
e
di
sta
n
ce
an
d
one
for
wea
k
sign
al)
can
b
e
estim
a
ted b
y
taking th
e d
e
rivati
ve of in
itial deviation
distri
bution f
unctio
n
and
equatin
g it to zero a
s
in Eq
uation (4). Ze
ros
of
this e
q
uation give
s
a RSS level
of about -55d
Bm
as cl
ose dista
n
ce e
r
ror elim
ination limit (bord
e
r) and
-70dBm a
s
we
ak RSSI limit (bo
r
de
r).
0
P
(
4
)
Observing ini
t
ial coordinat
e deviation
s in Fi
gure 3, and by usin
g result
s of Equ
a
tion (4
)
force
s
u
s
to
obtain a
co
rrection fu
nctio
n
ext
(o
r eq
uivalently regi
on
based)
co
nsi
derin
g RS
S
level in dBm as de
scrib
ed
in Equation (5
). Finally, ge
nerate
d
co
rre
c
tion fun
c
tion
ext
is comi
ned
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 2, June 20
16 : 748 – 75
6
752
with Equatio
n
(2), a
nd a fin
a
l path lo
ss
d
i
stan
ce in
te
rms of po
we
r i
n
dBm an
d di
stan
ce in m
e
ter
can b
e
de
scri
bed a
s
in Equ
a
tion (6
).
Figure 3. Distribution fun
c
tion
col
o
re
d b
y
blue
dBm
P
P
P
P
dBm
P
dBm
P
ext
70
,
7842
9
.
319
36
.
4
0198
.
0
55
70
,
065
.
2
057
.
0
2
3
(5)
dBm
P
dBm
P
used
be
to
allowed
Not
d
ext
N
d
P
d
P
final
ext
55
10
55
)
(
)
(
0
(6)
Figure 4 sho
w
s lo
cation di
stributio
ns in
comp
ar
is
on
w
i
th
r
e
a
l
g
eome
t
r
i
c
loc
a
tion
s
.
R
e
d
circled val
u
e
s
are b
e
lon
g
in
g to app
roa
c
h co
nsi
deri
n
g
both clo
s
e
di
stan
ce e
r
ror
elimination li
mit
(bo
r
de
r) an
d
wea
k
RSSI limit (borde
r). Blue circle
d values a
r
e
belongin
g
cl
assic p
a
th loss
cal
c
ulatio
ns.
Figure 4. Dev
i
ation from Real Geom
etri
c Lo
cation
-8
5
-8
0
-75
-7
0
-6
5
-60
-5
5
-50
-20
-15
-10
-5
0
5
10
15
RSS L
e
v
e
ls
in dB
m
I
n
it
i
a
l D
e
v
i
a
t
io
n
(
m
)
Test
1
Test
2
Test
3
Test
4
E
s
ti
m
a
ti
o
n
-1
0
-5
0
5
10
-1
0
-5
0
5
10
D
e
vi
at
i
o
n
i
n
x
D
i
r
e
c
t
i
o
n
(
m
)
D
e
v
i
at
io
n
in
y
D
i
re
ct
i
o
n
(
m
)
Al
l
-5
5
d
B
m
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Im
proved Ind
oor Lo
catio
n
System
s in a
Controlled En
viro
nm
ents (Selcu
k Hel
hel
)
753
Figure 5
de
monst
r
ate
s
t
he flo
w
chart
for
di
sta
n
ce
cal
c
ul
ation
betwe
en ta
rg
et and
a
certai
n tran
smitter Tx. System
s scan
RSS leve
ls based on
WLAN
ID followe
d b
y
an application
of median
filter on th
em. Median
filtered
dat
a a
r
e
feeding
clo
s
e dista
n
ce e
rro
r elimi
nati
o
n
approach (CDEEA) unit for maki
ng deci
sion. Th
i
s
first deci
s
ion determines either scanned
WLAN I
D
ba
sed RSS sig
n
a
l will be in
cl
uded into lo
cation cal
c
ul
ation or n
o
t. If n
o
t, that scan
n
ed
RSS value is ignore
d
from
the location
cal
c
ulatio
n
for that location
. Otherwi
se it is recorded f
o
r
further
cal
c
ul
ations. Depe
nding o
n
th
e re
sult
s o
b
t
ained from
Equation (4),
two differe
nt
corre
c
tion
fu
nction
s
are
g
enerated.
Wi
th gen
erat
e
d
corre
c
tion
fu
nction
s
and
p
r
eviou
s
ly sto
r
ed
building b
a
se
d path index
n, system
ma
ke
s dista
n
ce cal
c
ulatio
n.
ext
ext
Figure 5. Flowchart for di
stance
cal
c
ulat
ion for a ce
rta
i
n Tx
4. Results
To evaluate t
he model
pe
rforma
nce, it is
com
pared
with other p
r
opo
se
d mod
e
ls. As
prop
osed in [17-2
3
], performance is qu
a
n
tified by
usi
ng root me
an
squa
re (RM
S
) error, defi
ned
by Equation (7).
n
E
E
n
i
i
rms
1
2
(
7
)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 2, June 20
16 : 748 – 75
6
754
Whe
r
e
i
E
is th
e differen
c
e
betwee
n
e
s
timated a
n
d
mea
s
urem
ent value at
i
th
point of
measurement
in dBm a
nd
n is the
num
b
e
r of m
e
a
s
urement p
o
ints.
By using
IP depe
ndent
pa
th
loss
equations
with c
l
ose
dis
t
ance error eliminat
ion algorithm (CDEEA),
mobile equipment
locatio
n
wa
s cal
c
ulated.
Cal
c
ulate
d
lo
cation
s in x
and y directi
ons
are
com
pare
d
with
real
geomet
ric l
o
cation i
n
dat
a ba
se. It is observed t
hat, while m
ean e
r
ror
wi
thout pro
p
o
s
ed
algorith
m
is 3
.
34 (m
) diffe
rence in total
combi
ned
di
stance, it
sligh
t
ly decrea
s
ed
to 2.96
(m
),
as
sho
w
n i
n
Ta
ble 2(a). Ta
bl
e 2(b
)
indi
cat
e
s lo
ngi
tudin
a
l mean
erro
r in (m
) a
nd t
r
an
sverse m
ean
error in
(m) t
h
rou
gho
ut the co
rri
dor. Th
ese
are
com
para
b
ly at very good l
e
vel
s
compa
r
ing
wit
h
the literature.
Table 2
(
a). M
ean Error Ta
ble
Approach
Mean
Erro
r
Median Filter itself
3.34(m)
Median Filter
w
i
t
h
CDEEA
3.24 (m)
Proposed Model
w
i
th
ext
2.96 (m)
Table 2
(
b). M
ean erro
r in b
o
th axes
Expression
Longitudinal mea
n
erro
r in (m)
T
ransverse mea
n
erro
r in (m)
Convetional Path Loss
2.17 1.93
Proposed Model
2.20
1.83
Table
3 i
s
co
mpari
s
o
n
tabl
e that
sho
w
s
porp
o
sed
mo
del in
compa
r
ison
with
the
result
s
in the lite
r
at
ure. F
o
r 90
% pre
c
i
s
ion
level, p
r
op
ose
d
mo
del
accu
ra
cy i
s
the
se
con
d
be
st
achi
evement.
Table 3. Wi
-F
i Based Syste
m
s’ Co
mpa
r
ison Table
S
y
stem Name
Accur
a
cy
(
m
)
Pr
ecision (
%
)
Horus
2,10
90
Proposed Model
2,96
90
DIT
3,00
90
TIX 5,40
90
Microsoft RADAR
5,90
90
5. Conclusio
n
Rapi
d develo
p
ment of location syste
m
s ha
s
mad
e
indoo
r lo
catin
g
system
s a
r
e quite
popul
ar a
nd
wide
sp
rea
d
such th
at they find lots
of
comm
ercial a
pplication
s
. Propo
se
d stu
d
y
use
s
wifi unit
s
of any o
r
di
nary comme
rcial la
ptop a
s
a
re
ceviver
combi
ned
wit
h
clo
s
e
dista
n
ce
error eli
m
inati
on ap
pro
a
ch
that the se
co
nd be
st lo
cating pe
rform
a
n
c
e (l
ocation e
rro
r is l
e
ss th
en
3m in total) h
a
s be
en a
c
hi
eved for 90%
preci
s
io
n level in the litera
t
ure.
A model co
nsiderin
g RSS l
e
vel and cl
ose dist
an
ce e
r
ror elimin
ation
algorithm
co
mbine
d
with medi
an filters in
crea
ses
the p
e
rfo
r
mance of co
nventional
RSS based l
o
cation syste
m
s by
amount
of 1
2
%. Minimu
m Tra
n
sve
r
se mean
-e
rror
has bee
n
o
b
tained as 1.83(m
)
.
Altho
ugh
prop
osed mo
del
give
s a good
pe
rformance within
the
scope
of this
study,
it nee
ds to
be
improve
d
. Future
work
wi
ll focus
on wa
lking live
s
affecting in
doo
r
prop
agatio
n model a
s
well
as
indoo
r lo
catin
g
algo
rithms
and
close an
tenna
wa
ll ef
fects o
n
prop
agation m
o
d
e
l, and we a
r
e
trying
to co
mbine both WiFi ba
sed
system
s with
ultra
s
oni
c lo
cating
sy
ste
m
s
name
d
Hybrid
Indoor L
o
cating Systems
(HiLOS
) [22].
It is quite cl
e
a
r that u
s
ag
e
of profe
s
sio
nal
well de
si
gned
WiFi
re
ceivers
will al
low u
s
to
determi
ne mo
re preci
s
e results. Since ou
r startin
g
poin
t
was ba
se
d on the usage
of ordina
ry wi
fi
units
on
any l
aptop,
we
did not try them.
Those
well designed an
d accessi
ble WiFi
unit
s
will
be
u
s
ed
w
i
th
H
i
LO
S s
t
u
d
i
es
.
.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Im
proved Ind
oor Lo
catio
n
System
s in a
Controlled En
viro
nm
ents (Selcu
k Hel
hel
)
755
Ackn
o
w
l
e
dg
ements
This
study i
s
sup
porte
d by
Akde
niz
Uni
v
ersi
ty, Sc
ie
ntific
R
e
s
e
arc
h
Pr
o
j
ec
ts
Supp
o
r
ting
Unit (BAPYB) and granted
by TÜB
İ
TAK 2209
-A (PN:
1919B0
113
0
3278
). Mea
s
u
r
eme
n
t facilities
are belo
ng
to
EMUMAM Near Field
A
n
tenna Lab
orat
ory g
r
anted
b
y
State Plann
ing O
r
ga
nization
(200
7K12
053
0-DPT
)
.
Referen
ces
[1]
Md S
y
edu
l Am
in, Mamun B
i
n
Ibne Re
az, Sa
l
w
a Sh
ei
kh Na
sir.
Integrated Vehi
c
l
e Acci
de
nt Detectio
n
and
Loc
atio
n
S
y
stem.
T
E
LK
OMNIKA T
e
lec
o
mmunic
a
tio
n
Co
mp
uting
Ele
c
tronics a
n
d
C
ontrol.
201
4;
12(1): 73-
78.
[2]
Mi
w
a
S T
agashira, H Matsud
a,
T
Sutsui, Y
Araka
w
a, A F
u
kuda.
A Multil
e
r
ation Bas
ed L
o
cali
z
a
tio
n
Sche
me for
Adhoc W
i
re
le
ss Position
ing
Netw
orks Used in Infor
m
a
t
ion ori
ente
d
Constructi
on.
Procee
din
g
s o
f
IEEE27th Internetio
nal
Co
nfer
en
ece o
n
Advanc
ed In
formation N
e
t
w
o
r
kin
g
a
n
d
Appl
icatio
ns (AINA). 2013: 69
0-69
5.
[3]
M Ocaña,
LM
Bergas
a, MA
Sotelo, J
Nu
ev
o, R
Flor
es. IEEE ISIE 200
5. Dubr
ovn
i
k, Cr
oatia.
200
5:
154
5-15
50.
[4]
I Co
x. Bla
n
ch
e-an
e
x
p
e
rime
nt in
gu
ida
n
ce
an
d n
a
vig
a
tio
n
of a
n
auto
n
o
mous r
o
b
o
t v
ehicl
e.
IEEE
Tra
n
s
a
c
ti
on
s on
R
o
bo
ti
cs a
nd Au
to
ma
ti
o
n
.
1
991; 7(2): 1
93-
204.
[5]
F
a
rhan M
anz
o
o
r, Yi H
uan
g,
Karsten M
enz
el.
Passiv
e
RF
ID-Based I
ndo
or Positi
on
ing
System, An
Algorit
h
m
ic A
p
proac
h
. Pro
g
r
a
m for th
e IE
EE Internati
o
n
a
l C
onfer
enc
e
on
RF
ID-T
echno
log
y
a
nd
Appl
icatio
ns. Guan
gzh
ou, Chi
na. 201
0.
[6]
Busra Ozdenizci, Vedat Co
sk
un, Ker
e
m Ok. NF
C Intern
al:
An Ind
oor
Na
vigati
on S
y
ste
m
.
Sensors
.
201
5; 15: 257
1
-
259
5.
[7]
S
ł
a
w
om
ir H
a
u
s
man,
Ł
uk
asz
Januszk
ie
w
i
c
z
. Impact of Indo
or Envir
o
n
m
ent on P
a
th
Loss i
n
Bo
d
y
Area Net
w
o
r
ks
.
Sensors.
201
4; 14: 195
51-1
956
0.
[8]
Andr
y Ma
yk
ol
G Pinto, A Paul
o More
ira,
P
aul
o G Cost
a. Indoor
Loc
alizati
on S
y
ste
m
base
d
o
n
Artificial
La
nd
marks a
n
d
Mo
nocu
l
ar V
i
si
on.
T
E
LKOMNIKA T
e
lec
o
mmun
ic
ation
Co
mputi
n
g El
ectronics
and C
ontrol
. 2
012; 10(
4): 609
-620.
[9]
RS Camp
os, L Lov
isol
o, M Luiz R
de C
a
mpos. W
i
-F
i
multi-floor
in
d
oor p
o
sitio
n
in
g
consid
eri
n
g
architectur
a
l a
s
pects and co
ntroll
ed comp
u
t
ationa
l compl
e
xit
y
.
Exp
e
rt Systems w
i
th Appl
icatio
ns.
201
4; 41: 621
1
–62
23.
[10]
Helh
el S, Ko
caku
ş
ak A, Y
o
l Kay
b
ı
.
İ
ndeks Hesab
ı
Ve W
i
fi T
aban
l
ı
Kontrolsü
z
Ge
z
g
in
lerd
en
Ar
ı
nd
ı
r
ı
lm
ı
ş
Bi
na
İ
çi Konum
Tespit Sistem
i
.
VII URSI-
T
ü
rkiy
e B
ilims
e
l Ko
ngres
i. KAYSER
İ
, T
ÜRKIYE.
201
4; 1(1): 1-4
.
[11]
Helh
el S, K
o
ca
ku
ş
ak A.
Al
gor
ithms F
o
r
Indo
or Loc
ali
z
at
io
n
On W
l
an N
e
t
w
orks Applic
ati
ons
. PIERS
Guangz
au, Ch
i
na. 201
4; 1(1): 1-2.
[12]
AA W
ahab, A Khattab, YA F
ahm
y
.
T
w
o-W
a
y T
O
A
w
i
th Li
mite
d Dea
d
Recko
nin
g
for GPS-F
r
ee
Vehic
l
e
L
o
ca
li
z
a
ti
on Usin
g Sing
le
RS
U.
Procee
din
g
s o
f
the Internati
ona
l Co
nfere
n
c
e on IT
S
T
e
lecommunic
a
tions. 20
13: 2
44-2
49.
[13]
S Ueb
a
y
as
hi,
M Shimiz
u, T
F
u
ji
w
a
r
a
.
A St
udy of T
D
OA
Positio
n
in
g Us
i
ng UW
B
Refle
c
ted W
a
ves.
Procee
din
g
s of
the 78th IEEE Vehic
u
lar
T
e
chnol
og
y
Co
nfere
n
ce. 201
3: 1-5.
[14]
Parameter
i
zati
on of a Pol
a
rimetric Diffu
se Scattering
Model i
n
Indoor Env
i
ron
m
ents.
IEEE
T
r
ansactio
n
s On Antenn
as An
d Propa
gati
o
n
. 201
4; 62(8): 43
61-4
364.
[15]
Selçuk
Hel
h
e
l
, Huse
yi
n Goks
u,
Sukru Oze
n
. Investig
ation
of 3D
el
ectrom
agn
etic fie
l
d
di
stributio
n fo
r
d
y
nam
ic pol
ari
z
ation d
i
versit
y.
IEEE Antennas and Wire
les
s
Propag
atio
n Letters.
200
8; 7: 171-1
74.
[16]
Selcuk
He
lh
el,
Sukru
Ozen,
I Bah
a
d
i
r Ba
s
y
i
g
it, Osman
Kurnaz,
Y
u
nus
E Yor
u
k, Murat Bitirgan,
Ze
y
n
ep
Co
lak.
Ra
diate
d
s
u
sc
eptib
ilit
y
of me
dical
e
qui
pme
n
t
in
hea
lth c
a
re
un
its: 2G a
nd
3G mob
ile
pho
nes as a
n
i
n
terferer.
Micro
w
ave and Opti
cal T
e
chn
o
lo
gy
Letters.
2011;
11(5
3
): 265
7-2
661.
[17]
Kim D, In
gram
MA, Smith W
W
. Measurem
e
n
ts
of Sma
ll-Sc
ale
F
adi
ng
a
n
d
Path
Loss
for
Lon
g R
a
n
g
e
RF
T
ags.
IEEE T
r
ansactions
On Antenn
as And Prop
ag
ation
. 2003; 51(
8): 1740-
174
9.
[18]
Yong
xian
g Z
h
ao, Meifa
ng
Li,
F
eng S
h
i. Ind
oor R
adi
o Pro
pag
atio
n Mod
e
l
Base
d o
n
D
o
mina
nt Path.
Int. J. Commun
i
catio
n
s, Netw
ork and Syste
m
Scienc
es.
201
0; 3: 330-3
37.
[19]
Çelik
S, Kur
n
a
z
O, Alba
yr
ak
Y, Bitirga
n
M,
Bas
y
ig
it IB, He
lhel
S. An
Ind
o
o
r T
o
Outdoor
Propa
gat
ı
on
Mode
l At Gsm900 Gsm1
8
00 A
nd C
d
m
a21
00.
Inter
n
a
t
iona
l Jo
urna
l
Of Engin
eer
i
ng & A
p
p
lie
d
Scienc
es (IJEAS)
. 2013; 5(2): 9-17.
[20]
Lin M
a
, Yubi
n
Xu. R
e
ce
ived
Sign
al Stren
g
t
h
Rec
o
ver
y
i
n
Green W
L
AN I
ndo
or Positi
on
i
ng S
y
stem
Using S
i
ng
ul
ar Valu
e T
h
reshol
din
g
.
Sensors.
201
5; 15: 129
2
-
131
1.
[21]
YS Lin. T
a
ich
ung, R
C
Ch
e
n
, YC Li
n.
An
ind
oor l
o
catio
n
id
entificati
o
n
system b
a
se
d on
neur
al
netw
o
rk
and g
enetic alg
o
rith
m.
3rd Inter
nat
ion
a
l C
onf.on
A
w
a
r
e
ness Sci
ence
and T
e
chn., (iCAST
).
201
1: 27-3
0
.
[22]
Kurnaz O, H
e
l
hel S. N
ear G
r
oun
d Pro
pag
ation M
o
d
e
l F
o
r Pin
e
T
r
ee F
o
rest Enviro
n
m
ent.
AEU-
Internatio
na
l Journ
a
l of Electr
onics a
nd C
o
mmu
n
ic
ations
. 2
014; 1(1): 2-5.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 2, June 20
16 : 748 – 75
6
756
[23]
Selcuk H
e
lh
el,
Atala
y
k
o
cak
u
sak, Yalc
ı
n
Alba
yr
ak, Sukru Ozen.
W
L
AN Based Ind
o
o
r Locati
n
g
Systems En
ha
nced
by Ultra
sonic Se
nsors.
H
y
brid In
do
or Locater S
y
st
ems (HILoS),
Progress i
n
Electroma
gneti
c
Researc
h
S
y
mposi
u
m PIERS. Prag. 2015.
[24]
Harin
d
ra W
i
sn
u Prad
han
a, Sur
y
on
o, Ach
m
ad
W
i
do
do.
Estimating Ob
j
e
ct Locati
on u
s
ing P
a
rticl
e
Clusteri
ng o
n
Rotatin
g
Sonar Detection.
TELKOMNIKA.
2014; 12(
2): 291
-296.
Evaluation Warning : The document was created with Spire.PDF for Python.