Localiza ion me hod o low-powe wi eless
senso ne wo ks
Diego Fco. La ios, Julio Ba bancho, Fco. Ja ie Molina and Ca los Le´on Senio Membe , IEEE.
Depa men o Elec onic Technology, “Escuela Poli ´ecnica Supe io ”
Uni e si y o Se ille, Se ille, Spain,
Phone: +34 954 55 28 38
Fax: +34 954 55 28 33
Email: d la io[email p o ec ed], jba [email protected], jmolin[email p o ec ed], [email p o ec ed]
Abs ac —Con ex awa eness is an impo an issue in am-
bien in elligence o an icipa e he desi e o he use and,
in consequence, o adap he sys em. In con ex awa eness,
localiza ion is e y impo an o enable a esponsi e en i-
onmen o he use s.
Focusing on his issue, his pape p esen s a localiza ion
sys em based on he use o Wi eless Senso Ne wo ks
de ices. In con as o a adi ional RFID, hese de ices o e
he possibili y o a collabo a i e sensing and p ocessing o
en i onmen al in o ma ion.
The p oposed sys em is a ange- ee localiza ion algo i hm
ha uses uzzy in e ence o p ocess he RSSI measu emen
and o es ima e he posi ion o mobile de ices. The main
goal o he algo i hm is o educe he powe consump ion
and he cos o he de ices, especially o he mobiles ones,
main aining he accu acy o he in e ed posi ion.
Index Te ms—Fuzzy sys em, WSN, localiza ion, RSSI, cen-
oid, AmI
I. INTRODUCTION
Ambien In elligence (AmI) e e s o elec onic en i-
onmen s ha imp o e he use expe ience o he en i on-
men by esponding o hem in elligen ly. I implies he
use o ad anced ne wo king compu ing echnology ha
is awa e o human p esence, pe sonali ies and needs. In
ambien in elligence, de ices may wo k o help people in
hei e e yday ac i i ies. These de ices would bo h sense
and eac based on he en i onmen al con ex .
In elligen homes a e a ypical AmI applica ion a ea,
because hey inc ease secu i y and com o o he use s
[1]. In he no ion o ambien in elligence, he main keys
[2] a e he ollowings ( igu e 1):
•Embedded. Many ne wo ked de ices a e in eg a ed
in o he en i onmen o sensing and con olling i .
Ideally, hese de ices a e deployed h oughou he
en i onmen , bu hei p esence mus be unde ec ed
o he use s. The de ices sha e in o ma ion o
This pape is based on “Loca ing senso wi h uzzy logic algo i hms,”
by D. F. La ios, J. Ba bancho, F. J. Molina and C. Le´on, which appea ed
in he P oceedings o he 24 h IEEE Symposium on P oceedings o he
8 h In e na ional Join Con e ence on e-Business and Telecomunica ions
(ICETE), Se ille, Spain, July 2011.
This esea ch has been suppo ed by he “Conseje ´ıa de Inno aci´on,
Ciencia y Emp esa”, “Jun a de Andaluc´ıa”, Spain, h ough he excel-
lence p ojec ARTICA ( e e ence numbe : P07-TIC-02476) and by he
“C´a ed a de Tele ´onica, In eligencia en la Red”, Se ille, Spain, h ough
he p ojec ICARO.
Figu e 1. Ambien In elligence keys.
inc ease use ’s com o . Communica ion is essen ial
o ob ain an adequa e esponse in compu ing, whe e
con ol in elligence is sp ead all o e he ne wo k.
WSN is an example o echnology ha complies wi h
his key.
•Con ex awa eness. The sys em would ecognize he
si ua ional con ex o he use s. This implies ha he
de ices can ecognize he use [3], his posi ion and
he posi ions o hose de ices, he use s can in e ac
wi h.
•Pe sonalized. The sys em mus adap o he needs o
he use s. The sys em mus be in elligen enough o
adap o he desi es o he use in unc ion o his
na u al eac ions.
•Adap i e. The sys em mus lea n he use ’s needs.
Fo example, lea ning he empe a u e o he illu-
mina ion ha use s eel like ha ing in each oom,
adjus ing hem au oma ically.
•An icipa o y. The sys em mus an icipa e use ’s
needs wi hou conscious media ion. An icipa o y is
ela ed o he use in e aces in he ambien in elli-
gence and i sea ches o an easy in e ac ion be ween
machines and use s [4]. Ideally, hese in e aces
would no be pe cei ed by he use s [5].
JOURNAL OF NETWORKS, VOL. 8, NO. 1, JANUARY 2013
45
© 2013 ACADEMY PUBLISHER
doi: 10.4304/jnw.8.1.45-58
The ambien in elligence pa adigm is o en buil on
pe asi e compu ing, ubiqui ous compu ing and com-
pu a ional in elligence [6]. AmI pa adigm s a ed o be
de eloped wi h he challenge launched by he Eu opean
Commission in 2001, aiming ha he esea che s in es i-
ga e in his a ea [7].
In con ex awa eness, localiza ion sys em is essen ial o
enable a mo e eac i e use awa e esponsi e en i onmen s
[8]. T acking use s o e s basic in o ma ion use ul o lea n
abou hei habi s, o o p edic ollowing displacemen s
[9]. Fo example, be o e a use en e s a oom, he ai con-
di ioning would be adjus ed o his com o au oma ically.
This pape aces he p oblem o loca ing use s based
in he AmI pa adigm, educing he powe consump ion
and he cos o he de ices. Ene gy e iciency is especially
ele an o mobile de ices. I imp o es ba e y au onomy
and inc eases he use com o . The p oposed me hod
is based on a Wi eless Senso Ne wo k (WSN) ha
can moni o and con ol he en i onmen in elligen ly in
unc ion o he use desi es.
The es o his pape is o ganized as ollows: sec ion
II sums up he s a e o he a abou WSN applied o
ambien in elligence. Sec ion III desc ibes he localiza ion
algo i hm. Sec ion IV e alua es he ad an ages o he p o-
posed localiza ion algo i hm o e he powe consump ion.
The ou come o he localiza ion algo i hm pe o mance
is de eloped by simula ions, and he esul s a e shown in
sec ion V. Finally, in Sec ion VI we p esen concluding
ema ks.
II. LOCALIZATION TECHNIQUES
A wi eless senso ne wo k (WSN) consis s o lo s o
small de ices deployed in a physical en i onmen o i s
s udy. Each node has special capabili ies, such as wi eless
communica ions wi h i s neighbou s, sensing, da a s o age
and p ocessing.
WSN has been widely used in many a eas [10], such as
en i onmen al moni o ing [11] and con ol [12], heal h-
ca e and medical esea ch [13], na ional de ense and
mili a y a ai s [14] [15], e c.
The use o WSN in ambien in elligence o e s many
ad an ages [16]. I uses has been p oposed by many
au ho s [17] and ejec ed by o he s because o he cos
and ba e y consump ion ha hese de ices p esen [18].
Bu nowadays, he cos has been educed signi ican ly,
and he ene gy consump ion can be educed wi h sui able
so wa e and ha dwa e design. In his way, i is possible
o ge yea s o ba e y au onomy o he mobile nodes.
Using he mic ocon olle and ex e nal senso s and
ac ua o s, he cu en nodes used on WSNs can con ol
he com o and sa e y o he use s. Fo example, a de ice
can make eme gency calls in case o an acciden [19].
T adi ionally, in ambien in elligence, he localiza ion
solu ion is based on passi e o ac i e RFID echnology,
bu RFID has a e y limi ed capaci y o moni o he
en i onmen compa ed o WSNs. Mo eo e , he p ices o
RFID ags and some WSNs nodes a e cu en ly simila
(20-30 $ [20]). Mo eo e , he in WSN ags and eade s
sha e he same economic ha dwa e, bu in RFID a eade
is much mo e expensi e han a ag (a ound 1500 $) and
mo e expensi e han a WSN de ice.
In his jou nal, we p opose a sys em o con ol he
localiza ion using WSN de ices only. This educes he
cos and he complexi y o he sys em and inc eases i s
capabili ies and unc ionali ies.
1) Localiza ion Techniques o WSN: Fo his appli-
ca ions, we a e going o conside he ollowing ypes o
nodes:
•Ancho nodes: loca ed on a ixed posi ion in he
house. Theses de ices a e used o ou e he in o -
ma ion o he Base S a ion and o he localiza ion
algo i hm. These de ices ac as he eade s o he
RFID echnology.
•Tags: a e he small de ices deployed wi h he ob-
jec o use s o loca e. Thei posi ions a e ini-
ially unknown and he algo i hm would ind hem.
These de ices can ob ain en i onmen al in o ma ion
h ough senso s, such as empe a u e o accele ome-
e s. These de ices a e also called non-ancho nodes.
•Base S a ion: This is a special ancho node ha ac s
ou ing he WSN in o ma ion om he ne wo k o a
PC. This PC p o ides he in o ma ion acqui ed by
he ne wo k o he es o he AmI de ices.
Localiza ion algo i hms p esen ed in he li e a u e can
be classi ied in o wo ca ego ies:
•Range-based: These echniques es ima e, poin - o-
poin , he dis ance be ween all he nodes using sen-
so s such as ul asound [21]. Wi h his in o ma ion,
using echniques such as iangula ion, he absolu e
posi ion o he non-ancho nodes can be es ima ed.
Gene ally, hese echniques equi e addi ional ha d-
wa e. The mos common ones a e Recei ed Signal
S eng h Indica ion (RSSI) [22], Time O A i al
(TOA) [23] and angle o a i al (AOA) [24].
•Range- ee: In hese echniques, he posi ion o
non-ancho nodes is ob ained acco ding o implici
in o ma ion p o ided by ancho nodes, usually based
on messages exchanged, commonly called beacons.
This in o ma ion is usually made up o di e en as-
pec s, such as adio co e age membe ship o numbe
o hops be ween de ices. The mos common ones a e
Cen oid (CL) [25] and DV-Hop [26].
In gene al, he ange-based ones o e good accu acy,
bu addi ional ha dwa e is o en needed. The e o e, he
weigh , he cos and he powe consump ion o node
de ices inc ease, and make hese so o echniques un-
sui able. RSSI ange-based echniques a e an excep ion
o his because mos o he cu en anscei e s p o ide
his measu e by de aul . Howe e , RSSI echniques a e
e y sensi i e o noise and in e e ences. Figu e 2 shows
expe imen s ealized by he au ho s o e alua e he e-
la ionship be ween RSSI and he dis ance in di e en
si ua ions: ee-space wi hou obs acles and long u ban
a ea wi h obs acles.
The esul s do no ma ch wi h any known models, such
as he F iis equa ion, bu in e y ideal condi ions. In ac ,
46
JOURNAL OF NETWORKS, VOL. 8, NO. 1, JANUARY 2013
© 2013 ACADEMY PUBLISHER
-45
-40-40
-50
-55
-60
-65
5
-70
RSSI (dBm)
Dis ance (me e s)
10 15 20 25 30 35 40 45 50
RSSI s. Dis ance
F ee space
Wi h obs acles
Figu e 2. RSSI s. dis ance.
in common scena ios, wi h obs acles and/o inside he
buildings, he e is no di ec ela ionship be ween dis ance
and RSSI. In ac di e en dis ances p oduce he same
RSSI alue.
Ins ead o using a ma hema ical model, he p oposed
solu ion uses a uzzy-logic-based sys em o de i e he
dis ance om RSSI le el. This is mo e obus in noisy and
complex scena ios. This idea is no new. Compu a ional
In elligence has been p oposed o localiza ion in se e al
pape s, such as [27] ha uses p obabilis ic neu onal ne -
wo ks, [28] ha applies a uzzy sys em and [29] ha uses
uzzy neu ons. In gene al, all hese algo i hms ack down
cu en posi ions based on es ima e posi ion changes. Bu
none o hem conside he p oblem o powe consump ion.
III. DESCRIPTION OF THE PROPOSAL SOLUTION
To loca e he impo an elemen s (de ices, use s, e c)
o e a WSN, we p oposed LIS (Localiza ion Based on
In elligen Sys ems). LIS is a echnique ha de e mines
he localiza ion using a uzzy sys em. LIS use RSSI, bu
no associa ing i o a dis ance es ima ion. Due o i , i is
a ange- ee echnique, such as weigh ed cen oid.
The inpu s o he uzzy sys em a e he RSSI measu e-
men s ela ed o he non-ancho node. These RSSI alues
a e measu ed when ancho nodes ecei e a message sen
om he ag.
LIS was designed o be de eloped in a building. In ac ,
a p o o ype is cu en ly deploying in he “Red building” o
he Uni e si y o Se ille. This building is in he “Campus
de Reina Me cedes” and has 12 class ooms and 38 o ices
sp ead o e h ee loo s and a basemen . This p o o ype
will consis o a mesh o ixed ancho nodes sp ead
h oughou he building ( igu e 3), and any small non-
ancho nodes si ua ed oge he o he elemen s o loca e.
The aim o his ne wo k is o con ol he posi ion o any
esou ces, such as lap ops, moni o s, p in e s, p ojec o s,
e c, o p e en he .
Wi h LIS, all he in o ma ion is s o ed and imes amped
in a PC a ached o he Base S a ion. This PC s o es
wo di e en classes o in o ma ion ga he ed om he
ne wo k:
Ancho nodes
Non-ancho nodes
Communica ion link
Message ansmission
Use 1
Use 2 Use 4
Use 3
Figu e 3. In as uc u e o LIS
•Ta ge loca ion: This in o ma ion desc ibes he po-
si ion.
•Senso s in o ma ion: This in o ma ion is he en-
i onmen al measu emen , such as he empe a u e,
humidi y, e c.
Despi e ha ange- ee and ange-based echniques
ha e been ex ensi ely s udied, nowadays he e a e some
aspec s ha con inue o be a challenge:
•The use o addi ional ha dwa e o lo s o beacons
inc eases powe consump ion.
•Cen alized p ocessing (i.e. on Base S a ions) e-
qui es a la ge amoun o messages. Con e sely, p o-
cessing in ags nodes educes he ba e y o hese
de ices signi ican ly.
•Scalabili y. Many o ange-based algo i hms a e ha d
o ex end o big senso ne wo ks.
LIS has been especially designed o ace all o he
abo e p oblems. As a esul , he p oposed algo i hm is
scalable and he powe consump ion and ne wo k au on-
omy a e op imized. LIS combines: (I) a uzzy sys em
o es ima e (ac ually o quali y) he dis ance be ween
ansmi e and ecei e om RSSI measu emen s, (II)
a ubiqui ous algo i hm execu ed in ancho nodes ha
ecei e he beacon o he non-ancho nodes, o de e mine
ela i e posi ions o hem, and (III) a coope a i e algo-
i hm o de i e he mos likely loca ion unning a he
Base S a ion.
LIS algo i hm consis s o ou s ages:
S1: Ancho nodes wai o non-ancho nodes ( ags)
beacons.
S2: The ag node b oadcas s a beacon.
S3: Recei e ancho nodes measu e RSSI, and execu e
he Ubiqui ous P ocessing (UP) o ela i e and
JOURNAL OF NETWORKS, VOL. 8, NO. 1, JANUARY 2013
47
© 2013 ACADEMY PUBLISHER
1 (B.S.) 23
45
6
7
8
(X?,Y?)
Ancho nodes
Non-ancho node
Communica ion link
Limi o adio co e age
Message ansmission
(a)
1 (B.S.) 23
4
5
6
7
8
(b)
1 (B.S.) 23
45
6
7
8
(X 7,Y 7)
(X 4,Y 4) (X 5,Y 5)
(c)
1 (B.S.) 23
45
678
(X 4,Y 4) (X 5,Y 5)
(X 7,Y 7)
(Xes ,Yes )
(Xes ,Yes )=cen oid(X i,Y i)
(d)
Figu e 4. S eps o LIS algo i hm: (a) S1; (b) S2; (c) S3; (d) S4
pa ial posi ioning.
S4: Ancho nodes send pa ial solu ions o he Base
S a ion, whe e he loca ion is inally de e mined
wi h he Coope a i e P ocessing (CP).
Figu e 4 illus a es hese s ages. When a non-ancho
node b oadcas s a beacon o any o he so o message,
he localiza ion p ocess s a s. Jus ancho ecei e nodes
pa icipa e in he p ocess. The es o he nodes can swi ch
o he adio anscei e o hold in a low powe s a e.
Fo localiza ion, he ags pe iodically send b oadcas
messages o he ancho nodes. The equency o he mes-
sages would be adjus ed in unc ion o he cha ac e is ics
o he sys em.
A. Ubiqui ous P ocessing (UP)
LIS uses he measu ed RSSI o a node and i s neigh-
bou s o de e mine he a ea whe e he non-ancho node
could be loca ed. This algo i hm is based on a uzzy
sys em dis ibu ed on e e y ancho node o he ne wo k.
Acco ding o he algo i hm s ages, once an ancho
node ecei es a beacon, i es ima es he posi ion o he
non-ancho nodes. The localiza ion algo i hm has been
designed o dis ibu e he compu a ion consump ion o e
he ne wo k. The a ea whe e he non-ancho node could
be loca ed wi h a ce ain p obabili y is called Rep esen-
a i e A ea (RA). A Sec o is he minimum a ea o med
by h ee ancho -node neighbou s. A Rep esen a i e A ea
(RA) ep esen he es ima ion, in e ms o sec o s, whe e a
node es ima e a ag would be. as desc ibed below, RA can
be made up o one o mo e sec o s wi h an UP p ocessing
highe han a ce ain h eshold.
Ancho nodes mus execu e he dis ibu ed Fuzzy ica-
ion Algo i hm (FA) measu emen o e e y su ounding
sec o . Figu e 5 shows an example wi h i e sec o s in
which, he uzzy p ocessing a e execu ed i e imes.
Ancho nodes
Comunica ion link
Neighbou 1
Sec o
1
Sec o
2
Sec o
3
Sec o
4
Neighbou 2
Neighbou 3
Neighbou 4
Sec o
5
Neighbou 1
Figu e 5. Example o node wi h 5 neighbou s.
E e y node ha ecei es a beacon, i measu es and
b oadcas s he RSSI le el o i s neighbou s ( igu e 4.b). In
his way, he closes ancho nodes elabo a e a able wi h
he RSSI measu ed by hem and hei neighbou s.
The RSSI able is p ocessed by FA ( igu e 6) o e alua e
he ep esen a i e a ea no ma e he numbe o sec o s.
This a ea can be o med by he union o one o mo e
sec o s ( igu e 4.c). A sec o is conside ed pa o he
ep esen a i e a ea i i s membe ship deg ee is highe
han he h eshold. This alue is adjus ed expe imen ally.
48
JOURNAL OF NETWORKS, VOL. 8, NO. 1, JANUARY 2013
© 2013 ACADEMY PUBLISHER
Ou pu
Fuzzi ica ion
De uzzi ica ion
In e ence engine
Base o
ules
Fuzzy
se s
da a class
uzzy
c ips
Impu s
}
Node RSSI Possi ion
Cen al
Neib. 1
Neib. 2
Value (X,Y)
Value (X,Y)
Value (X,Y)
Neib. N Value (X,Y)
.
.
.
.
.
.
Membe ship o
ac ual sec o (0 - 1)
Figu e 6. In e ence uzzy sys em.
Ou simula ions show ha a h eshold o 0.1 manages a
good ade-o be ween noise immuni y and localiza ion
pe o mance. The esul s o Rep esen a i e A eas (RA)
a e sen om he ancho nodes o he Base S a ion o
compu e he inal solu ion ( igu e 4.d).
An RA is emp y i i does no con ain signi ican
sec o s, his is i he membe ship deg ee o all o hem is
lowe han he h eshold. In his case, o sa e ene gy, he
esul is disca ded and he algo i hm will inish un il he
nex beacon a i es ( igu e 8). This is especially impo an
in huge ne wo ks, whe e he ene gy needed o mul i-hop
ansmissions is high. This issue is discussed in de ail in
sec ion IV.
med
RSSIsens RSSITX powe
Inpu alue
Membe ship deg ee
LOW MEDIUM HIGH
RSSI
0
1
Figu e 7. Se s o he uzzy impu s.
1) The Fuzzy Sys em Inpu s: RSSI ables ep esen he
signal le el ecei ed in ei he local and neighbou nodes.
Th ee uzzy se s quali y he RSSI as High, Medium and
Low o each inpu , as appea in igu e 7. This igu e is a
simpli ica ion. The a ea o he uzzy impu se s a ies in
unc ion o he es ima ed RSSI. Due o i , i may be no
symme ical.
The LOW RSSI uzzy se is ep esen ed by a apezoid.
Maximum membe ship deg ee (1) is assigned i powe
alls down he sensibili y h eshold o he emi e node
(RSSIsens). As he powe inc eases, membe ship deg ee
dec eases linea ly un il ze o. The ollowing equa ion
de ines his uzzy se :
µ(x) = max min 1,RSSImed −x
RSSImed −RSSIsens ,0
(1)
The MEDIUM RSSI uzzy se is ep esen ed by a
iangle whe e maximum membe ship deg ee co esponds
o he medium RSSI alue (RSSImed). Ze o membe ship
is eached o powe RSSI alues lowe han he sensi-
bili y h eshold o close o he maximum ansmission
(RSSIT Xpowe ). In his pape , medium RSSI alue mus
be compu ed o e e y sec o using he F iis model
equa ion (equa ion 2) and assuming he emi e ag is
loca ed a he cen e. This compu a ion only needs o be
execu ed once because ancho nodes a e loca ed a ixed
posi ions.
Mo e complex models, such as Two-Ray g ound model
o ITU indoo model a e no conside ed in his pape o
minimizing he complexi y o he algo i hm, Bu in can
be used o inc ease he accu acy o he sys em. The e o
wi h led o he use a simpli ied model is compensa ed
wi h he edundancy o in o ma ion and he use o FA
and UP Algo i hms.
PRX
PT X
=GT X ·GRX ·λ
4πR2
(2)
Whe e GT X and GRX a e he gain o TX and RX an-
ennas, Ris he dis ance be ween ansmi e and ecei e
and λ he wa eleng h.
Nex exp ession de ines he uzzy se o MEDIUM
RSSI:
µ(x) = max min x−a
b−a,c−x
c−b,0(3)
Whe e a=RSSIsens,b=RSSImed and c=
RSSIT Xpowe .
Fuzzy se o HIGH RSSI alues is a apezoid wi h a
lineal inc easing om 0 o 1 o RSSI powe alues ang-
ing be ween RSSImed and RSSIT Xpowe . This uzzy se
is de ined by he nex exp ession:
JOURNAL OF NETWORKS, VOL. 8, NO. 1, JANUARY 2013
49
© 2013 ACADEMY PUBLISHER

Nodes wi h UP ou pu
below a h eshold do
no send in o ma ion o
he base s a ion, sa ing
ene gy. This
in o ma ion is disca ed.
Ancho nodes Non-ancho node
Communica ion link Limi o adio co e age
7
Message ansmission
(B.S.) n hops
Cen aliced Algo i hm
7
(B.S.) n hops
Dis ibu ed Algo i hm
In ex ended ne wo ks, edi ec ed messages would
was e a lo o ene gy
Any node ha
ecei es a beacon
sends he associa ed
in o ma ion o he
base s a ion, whe e
i is p ocessed
A node decides wi h his neighbou i i has
use ul in o ma ion
Figu e 8. Dis ibu ed algo i hm would sa e Powe Ene gy on ex ended ne wo ks.
µ(x) = max min x−RSSImed
RSSIT Xpowe −RSSImed
,1,0
(4)
2) The Fuzzy Sys em Ou pu s: The Fuzzy Sys em
o e s an ou pu o each and e e y sec o . The ou pu
associa ed o a sec o is a [0, 1] anged alue ha
ep esen s he con idence deg ee ha he ag is ac ually
loca ed in ha sec o .
LOW MEDIUM HIGH
0 0.5 1
Con idence deg ee
Ou pu alue
0
1
1.5-0.5
Figu e 9. Se s o he uzzy ou pu .
As igu e 9 shows, he LOW ou pu uzzy se is a
iangle wi h he cen al poin a ze o and he co ne s a
-0.5 and 0.5. The nex exp ession de ines his uzzy se :
µ(x) = max min x+ 0.5
0.5,0.5−x
0.5,0(5)
MEDIUM ou pu is ep esen ed by a iangle wi h he
cen al poin a 0.5 and co ne s a 0 and 1. Ma hema ically
i can be exp essed by he ollowing equa ion:
µ(x) = max min x
0.5,1−x
0.5,0(6)
HIGH ou pu quali ie is also de ined by a iangle wi h
he cen al poin a 1 and he co ne s a 0.5 and 1.5. This
uzzy se is de ined by he nex exp ession:
µ(x) = max min x−0.5
0.5,1.5−x
0.5,0(7)
3) In e ence Engine: The in e ence engine is a Man-
dani’s ules based one wi h a cen oid de uzzi ica ion
me hod and a single on inpu uzzi ica o . The uzzy
engine e alua es he an eceden o e e y ule by he in e -
sec ion o he uzzy inpu s, using he minimum unc ion
o he AND ope a o (Eq. 8), and he maximum unc ion
o he OR ope a o (Eq. 9). The implica ion be ween
inpu s and ou pu s applies he minimum unc ion.
AND(a, b) = min(µ(a), µ(b)) (8)
OR(a, b) = max(µ(a), µ(b)) (9)
As men ioned, he ules mus be e alua ed o e e y
single sec o o es ima e he con idence deg ee, aking
in o accoun he uzzy quali ica ions o RSSI alues o
ei he he cu en sec o nodes and he su ounding ones.
The ules summed up in able I ha e been de i ed om
mul iple simula ions in o de o ob ain he bes ade-o
be ween p ecision and noise immuni y.
B. Coope a i e P ocessing (CP)
The Base S a ion collec s he pa ial solu ions om he
ancho nodes, and i p ocesses hem cyclically as ollows:
CP-S1: The Base S a ion wai s o ecei e he i s pa ial
solu ion.
CP-S2: On a i al, he pa ial solu ion is sa ed and a ime
s a s unning.
CP-S3: While he ime is unning, he nex pa ial solu ions
a e sa ed in a able as hey we e ecei ed.
CP-S4: When he ime expi es, he sys em will compu e
he inal posi ion as he cen oid o all hese pa ial
solu ions ( iangle sec o s). The cen oid compu a-
ion o a ini e se o poin s ~
P1,~
P2,··· ~
PNcan be
simpli ied as:
50
JOURNAL OF NETWORKS, VOL. 8, NO. 1, JANUARY 2013
© 2013 ACADEMY PUBLISHER
TABLE I.
RULES OF THE INFERENCE ENGINE.
RSSI node RSSI Neighbou s Ou pu
High. All medium. High
Low. All low. Low
Medium. All medium. High
Medium. All low. Low
High. All high. Medium
Medium. Medium in cu en sec o . High
Low in he es .
Medium. High in any sec o excep he cu en one. Low
Low in he es .
High. High in a neighbou o he cu en sec o . Medium
Low in he es .
High. High in a neighbou , excep on he cu en
sec o . Low
Low in he es .
Medium. Medium in a neighbou o he cu en
sec o . Medium
Low in he es .
Medium. Medium in a neighbou , excep on he
cu en sec o . Low
Low in he es .
Posi ion =PN
i=1 ~
Pi
N(10)
P e ious algo i hms can be easily ex ended o loca ing
mul iple ags, by simply associa ing a ag iden i ie o he
ansmi ed beacons.
The inal es ima ed posi ion is imes amped and sa ed
in he Base S a ion o make i accessible h ough In e ne .
IV. POWER CONSUMPTION OF LIS
Gene ally, powe consump ion is a s ong cons ain
in a WSN applica ion, especially o non-ancho nodes
( ags), whe e mobili y and addi ional cons ain s like size
and weigh do no allow use o high capaci y ba e ies.
Mos node powe consump ion is caused by adio ans-
missions. As an example, Telosb pla o m consumes 41
mW in ac i e mode (POn). The mic ocon olle consumes
only 3mW and he emainde powe consump ion is
caused by he adio anscei e ha equi es 38 mW in
ecep ion mode and 35 mW in ansmission mode [30].
Figu e 10 ep esen s a localiza ion algo i hm compu ed
in he non-ancho node. This is he ypical execu ion
phase o ange-based o ange- ee algo i hm, such as
cen oid.
As i can be obse ed, a e he ag node b oadcas s
a beacon ( igu e 10.a), i wai s o he esponse o all
he ancho nodes placed in he adio ange ( igu e 10.b).
This phase akes a long ime because o he numbe o
su ounding nodes and because o he collisions. A e
ha , he ag node execu es he localiza ion algo i hm and
deli e s he esul o he Base S a ion ( igu e 10.c). Du ing
all o his ime, he adio anscei e mus be in he
ac i e s a e. I was es a lo o ene gy and i s au onomy is
conside ably educed o he non-ancho node. I is in ac ,
he de ice wi h he highes ene gy cons ain s. Figu e 11
depic d his in o ma ion. I shows he ene gy consump ion
in a gene ic ag ha compu es a localiza ion algo i hm,
Ancho nodes
Non-ancho node
Communica ion link
Message ansmission
(B.S.) (B.S.)
23
4
5
6
7
23
4
5
6
7
(B.S.)
23
4
5
6
7
(a) (b) (c)
Figu e 10. Localiza ion algo i hm using he non-ancho node o
es ima ing i posi ion.
conside ing he powe consump ion o TelosB node [30].
In his simula ion he adio anscei e is main ained in
he [10 s - 60 s] ange, sending messages in he [1 - 60]
pe hou ange. In hese g aphics, Tx, Rx and idle powe
consump ion o he node a e conside ed.
0,00E+00
5,00E+02
1,00E+03
1,50E+03
2,00E+03
2,50E+03
3,00E+03
3,50E+03
4,00E+03
24
72
120
168
216
264
312
360
408
456
504
552
600
648
696
744
792
840
888
936
984
1032
1080
1128
1176
1224
1272
1320
1368
1416
10
20
40
60
Always ON
Ene gy (J)
Messages pe day
Time ON (s)
Figu e 11. Ene gy consump ion in non-ancho nodes execu ing local-
iza ion algo i hms.
As his igu e shows, he ene gy consump ion o he
ag’s p ocesso in idle mode is negligible in compa ison
wi h he powe consump ion o he adio anscei e .
Mo eo e , he powe consump ion o he node wi h
he adio anscei e in ansmission mode (35 mW ) is
simila o he powe consump ion o he node wi h he
anscei e in ecep ion mode (38 mW). The e o e, he
powe consump ion in a WSN node would be es ima ed
as equa ion 11 shows.
Enode,day ∼
=TOn,day ·POn (11)
I is impo an o poin ou ha he powe consump ion
is e y high ei he in ansmission and also in ecep ion
mode. Because o ha , o educe he powe consump ion
in ags i is necessa y o educe he numbe o exchange
messages, bu i is also necessa y o s op all he node
ac i i y enabling low powe modes and swi ching o he
adio anscei e . The e o e, a sui able ac i i y manage
is needed.
LIS akes his issue in o accoun , also ha ancho nodes
ha e mo e powe supply esou ces han he ags. The
algo i hm has been designed o be execu ed mainly in he
ancho nodes. Fu he mo e, he adio anscei e o he
ag is ac i a ed o a sho ime, jus enough o b oadcas
he beacon. In he emaining pe iod o ime, he ag will
JOURNAL OF NETWORKS, VOL. 8, NO. 1, JANUARY 2013
51
© 2013 ACADEMY PUBLISHER
be in a idle s a e and i s adio anscei e o . The ime
wi h he adio anscei e on (Ton) can be es ima ed, in
a wo s case, a ending equa ion 12.
Ton =Tac +Tmsg +Tdeac ≤17 ms (12)
Whe e Tmsg is he ime needed o send a beacon
(Tmsg ≤5ms, conside ing he p eamble and a 128
by es message leng h). Tac and Tdeac a e he imes o
ac i a ing and deac i a ing he adio anscei e (Tac =
Tdeac = 6 ms in elosB nodes). Conside ing his iming,
he ene gy consump ion o a ag sending a message
(Emsg) can be ob ained a ending equa ion 13.
Emsg =Tac ·PRx +Tmsg ·PT x +Tdeac ·PRx (13)
Whe e PRx is he powe consump ion o he mic o-
con olle wi h he adio anscei e in ecep ion mode
(PRx = 41 mW) and PT x is he powe consump ion o
he mic ocon olle wi h he adio anscei e in ansmis-
sion mode (PT x = 38 mW ) [30].
As i was explained in sec ion III, wi h LIS non-ancho
node only sends a beacon and ecei e no hing. This is why
he ene gy consump ion o ecep ion is no conside ed.
Figu e 12 shows hese esul s, and depic s he ene gy
consump ion o LIS in a non-ancho node, sending a
message in a [1 - 60] pe hou ange. As i can be seen,
i s powe consump ion is e y low, p ac ically equal o
he powe consump ion o main aining always he ags
in sleep mode. This ep esen s he bes case o sa ing
ene gy, whe e To >> Ton.
0,00E+00
5,00E-01
1,00E+00
1,50E+00
2,00E+00
2,50E+00
24
72
120
168
216
264
312
360
408
456
504
552
600
648
696
744
792
840
888
936
984
1032
1080
1128
1176
1224
1272
1320
1368
1416
Ene gy (J)
Messages pe day
Figu e 12. Ene gy consump ion in non-ancho nodes execu ing LIS.
Howe e LIS also educes he powe consump ion in
ancho nodes. I implemen s a ubiqui ous and dis ibu ed
algo i hm ha sp eads he localiza ion p ocessing amongs
he nodes su ounding he ag. In a cen alized algo i hm,
all he in o ma ion ecei ed by he ancho nodes mus
be deli e ed o he Base S a ion ( igu e 13). By con as ,
ou p oposed algo i hm sa es powe ene gy because only
signi ican in o ma ion is deli e ed ( igu e 14).
In he wo s case, LIS deli e s p ac ically he same
numbe o messages han a cen alized algo i hm. Bu
o low dense deploymen s, o example when medium
numbe o nodes ha a beacon ecei es is lowe han
he medium numbe o hops necessa y o each he Base
S a ion, he sa ed ene gy is signi ican .
Ancho nodes
Non-ancho node
Communica ion link
Message ansmission
(B.S.) (B.S.)
2
3
4
5
6
7
2 3
4
5
6
7
(a) (b)
8
8
Figu e 13. Example o cen alized algo i hm.
Ancho nodes
Non-ancho node
Communica ion link
Message ansmission
(B.S.) (B.S.)
2
3
4
5
6
72 3
4
5
6
7
(B.S.)
2
3
4
5
6
7
(a) (b) (c)
8
8
8
Figu e 14. Example o dis ibu ed algo i hm.
As a consequence, he ene gy sa ed wi h he dis ibu ed
algo i hm a ies wi h he densi y and complexi y o he
ne wo ks. Figu e 15 shows a s udy abou he numbe o
sa ed messages in unc ion o he numbe nodes wi h
use ul in o ma ion. F om his, i can be de i ed ha in
case all he in o ma ion ob ained by he ancho nodes
we e use ul, bo h me hods send p ac ically he same
numbe o messages. Bu he mo e numbe o nodes wi h
useless in o ma ion, he ene gy sa ing pe o mance o
he dis ibu ed p ocessing inc eases d as ically, due o he
educ ion o ansmissions.
Addi ional sa ings can be managed clus e ing he ne -
wo ks, and using he clus e heads as Base S a ions. This
is, ecei ing an p ocessing pa ial es ima ions om i s
clus e nodes. Figu e 16 shows his idea.
Fo his case, he algo i hms mus be modi ied as
ollows:
S1: Ancho nodes wai o non-ancho node beacons.
S2: The ag node b oadcas s a beacon.
S3: Recei e ancho nodes measu e RSSI, and execu e
bo h, uzzi ica ion algo i hm and ubiqui ous p o-
cessing o ela i e and pa ial posi ioning.
S4: Ancho nodes send pa ial solu ions o he clus e -
head, whe e he loca ion is inally de e mined.
S5: Clus e head node execu es he coope a i e posi ion-
ing algo i hms and deli e s he inal posi ion o he
Base S a ion.
52
JOURNAL OF NETWORKS, VOL. 8, NO. 1, JANUARY 2013
© 2013 ACADEMY PUBLISHER
-15
-10
-5
0
5
10
15
20
25
1 2 3 4 5 6 7 8 9 10
Numbe o nodes
Numbe o message sa ed
10 8 6
4 2 0
Numbe o nodes wi h
use ul in o ma ion:
Message sa ed wi h dis ibu ed
p ocessing: 3 hops
(a)
-20
-10
0
10
20
30
40
50
60
1 2 3 4 5 6 7 8 9 10
Numbe o nodes
Numbe o message sa ed
10 8 6
4 2 0
Numbe o nodes wi h
use ul in o ma ion:
Message sa ed wi h dis ibu ed
p ocessing: 6 hops
(b)
-20
0
20
40
60
80
100
1 2 3 4 5 6 7 8 9 10
10 8 6
4 2
0
Numbe o nodes
Numbe o message sa ed
Numbe o nodes wi h
use ul in o ma ion:
Message sa ed wi h dis ibu ed
p ocessing: 10 hops
(c)
Figu e 15. Messages sa ed e sus he numbe o nodes: a) Th ee hops; b) Six hops; c) Ten hops.
Ancho nodes
Non-ancho node
Message ansmission
Base S a ion
CH i
CH j
Clus e Head nodes
CH 1
Figu e 16. Example o he use o clus e s.
S6: Base S a ion execu es he same coope a i e posi-
ioning algo i hm han he clus e head nodes, bu
using he in o ma ion deli e ed om hese clus e -
heads. In his way, i he ag posi ioning comes om
jus one clus e head, his posi ion will be conside ed
as he inal solu ion. Bu , i i is ecei ed om
mo e han one clus e head, he cen oid es ima ion
is applied o all o hem.
De e mining when he use o clus e ing sa es mo e
ene gy is no i ial. I depends on he size and complexi y
o he ne wo k. Bu in gene al, i is easonable o hink
ha clus e ing echniques a e be e o wide and complex
ne wo ks.
Addi ionally, a dis ibu ed p ocessing such as he one
p oposed in his pape , inc eases h oughpu and educes
he esponse delay, because a ic bo leneck and colli-
sions close o he Base S a ion a e a oided. Dis ibu ed
p ocessing also sp eads compu a ional load o e he ne -
wo k. This is especially impo an o wide ne wo ks o
wi h mul iple non-ancho nodes.
V. ACCURACY OF LIS
We ha e compa ed he accu acy o LIS e sus he
classic CL algo i hm [25] using di e en simula ions. The
es ed ne wo k was made up o 25 ancho de ices wi h
a adio ange o 200 me e s and sepa a ed also by abou
200 me e s. Aniso opic adia ion pa e n is assumed. The
simula o has been de eloped in C++. I allows he selec-
ion o adio ange, adia ion pa e n, noise, sensibili y,
ne wo k deploymen and ancho loca ion. F iis equa ion
is used o calcula e he ecei ed RSSI le el in e e y node.
All pa ame e s, in he es s ha e been selec ed o model
Telosb de ices.
The esul s o he simula ions a e p esen ed in he nex
subsec ion.
A. E o s. Posi ion
The ollowing expe imen s include a mo ing ag. The
noise has been neglec ed and he e o is exp essed in
me e s. Figu es 17 and 18 show he posi ion e o in on
o he posi ion o he non-ancho node. Maximum and
medium e o s o LIS algo i hm a e conside ably smalle
han he ones es ima ed wi h he (CL) Cen oid classic
algo i hm.
As we can see in igu e 18, wi h LIS he e o s always
emain less han a 25 %o he dis ance be ween ancho
nodes.
B. E o s. Co e age
In his es , adio ange is inc eased while he dis ance
be ween nodes emains cons an .
Figu e 19 shows he in luence o adio ange. As
he adio ange inc eases, he numbe o non-ancho
nodes ha ecei e a beacon also inc eases, and he e o
dec eases. In all he cases, LIS ge s smalle e o s han
cen oid algo i hm.
Simila esul s a e ob ained ixing he co e age a ea o
he non-ancho nodes and educing he dis ance be ween
ancho de ices.
C. E o s. Noise
In his es we ha e analysed he e o ob ained in
a ixed posi ion o he non ancho nodes when adding
Gaussian noise o he sys em. All analyses use a co e age
a ea o 200 m o he ag. We ha e made 1000 analysis
a e e y poin o he simula ions.
The numbe o e o s ( igu e 21) ep esen s he numbe
o absolu e e o s bigge han 100 m (1/2 o he co e age
a ea) ob ained in he localiza ion. As i can be seen, LIS
always has less numbe o e o s bigge han 100 m. han
he classic CL algo i hm.
This esul shows ha despi e he in e e ences, LIS
con inues o o e good accu acy o localiza ion.
JOURNAL OF NETWORKS, VOL. 8, NO. 1, JANUARY 2013
53
© 2013 ACADEMY PUBLISHER