IMPLEMENTATION OF AN
EVENT-TRIGGERED SMART SENSOR
NETWORK ARCHITECTURE BASED ON THE
IEEE 802.15.4 STANDARD
E ico Meneses Le˜ao ∗Luiz A onso Guedes ∗
F ancisco Vasques ∗∗
∗Depa men o Compu e Enginee ing and Au oma ion -
UFRN - Na al - B azil
∗∗ Depa men o Mechanical Enginee ing - Uni e si y o
Po o - Po o - Po ugal
Abs ac : A sma ansduce is he in eg a ion o a senso /ac ua o elemen , a
p ocessing uni , and a ne wo k in e ace. Sma senso ne wo ks a e composed o
sma ansduce nodes in e connec ed h ough a communica ion ne wo k. This
pape p oposed an e en d i en a sma senso ne wo k a chi ec u e (asynch onous
da a) and i s espec i e implemen a ion based in he IEEE 802.15.4 s anda d. The
e en s a e de i ed om a da a comp ession algo i hm embedded in o he sma
senso , which comp esses da a om he senso . The a chi ec u e also suppo s
con igu a ion and moni o ing ac i i ies o all dis ibu ed sys em.
Keywo ds: sma senso , ne wo k, a chi ec u e, e en s, IEEE 802.15.4,
comp ession algo i hm.
1. INTRODUCTION
Au oma ion ac i i ies a e essen ial o he com-
pe i i eness inc ease in all indus ial sec o s. F om
a sys emic app oach, indus ial au oma ion can be
cha ac e ized as a se o echniques ha enable
he cons uc ion o ac i e subsys ems wi h he
capabili y o in e ac wi h he indus ial p ocesses
o con ol, moni o ing, and supe ision p oposes.
In he las ew decades, he au oma ion echnolo-
gies ha e been e ol ing om a s ongly cen al-
ized echnology o an essen ially dis ibu ed ech-
nology. Wi hin his new app oach, componen s
a e in e connec ed by digi al communica ion ne -
wo ks. Thus, he adi ional senso s based in he
4-20mA s anda d a e being eplaced by digi al
de ices. These de ices may ha e senso s, ac ua-
o s, and con ol unc ionali ies and a e endowed
o digi al p ocesso s and communica ion sys ems.
Wi hin his con ex , a sma senso is de ined as
he in eg a ion o an analog o digi al senso o an
ac ua o elemen , a p ocessing uni , and a ne wo k
in e ace (Elmen eich, 2006).
In his way, he a chi ec u es o he lowes le el o
indus ial au oma ion a e cha ac e ized o using
a se o he sma ansduce s, usually connec ed
h ough a communica ion ne wo k wi h eal ime
p ope ies. The dis ibu ed app oach p o ides a
signi ican imp o emen in he lexibili y and scal-
abili y aspec s o he indus ial p ocesses; how-
e e , i also b ough new scien i ic and echnolog-
ical challenges, such as he need o new models
and algo i hms o eal ime and sa e communi-
ca ions, conside ing inancial and en i onmen al
es ic ions.
The sma senso design mus deal wi h in e -
changes among de ices, and also wi h in e op-
e abili y and a ailabili y o in o ma ion in eal
ime. The ne wo ked ope a ion o a sma senso
ha uses s anda dized in e aces allows sha ing
in o ma ion and esou ces. Thus, i allows he in-
eg a ion o all con ol and supe ision p ocesses,
o example.
The objec i e o his pape is o p opose an e en -
igge ed sma senso ne wo k a chi ec u e and
i s implemen a ion based IEEE 802.15.4 s an-
da d (IEEE 802.15.4 S anda d, 2006). This A -
chi ec u e is simila o he OMG (Objec Manage-
men G oup) s anda d (Kope z and Wien, 2003).
Howe e , di e en ly o he o iginal ime- igge ed
OMG p oposal, he p oposed app oach uses an
asynch onous e en - igge ed mechanism o ans-
mi he da a om he sma senso s. This asyn-
ch onous beha io is a consequence o he imple-
men a ion o a da a comp ession algo i hm in o
he sma senso s, which sends only he ele an
da a om he aw da a mass. Thus, he p o-
posed e en - igge ed app oach sa es an impo -
an amoun o communica ion bandwid h.
The es o he pape is o ganized as ollows:
Sec ion 2 desc ibes he p oposed e en - igge ed
sma senso ne wo k a chi ec u e and i s cha -
ac e is ics. In o de o alida e he p oposed a -
chi ec u e, he Sec ion 3 p esen s a p o o ype im-
plemen a ion using he IEEE 802.15.4 s anda d
and some expe imen al esul s ha we e ob ained
om he p o o ype. The pape is concluded in
Sec ion 4.
2. EVENT-TRIGGERED SMART SENSOR
NETWORK ARCHITECTURE
The p oposed e en - igge ed sma senso a -
chi ec u e was b ie ly in oduced in (Le˜ao e
al., 2007). In his pape , a mo e de ailed p esen-
a ion is made. I is based on he OMG’s (Objec
Managemen G oup) s anda d. The choice o his
s anda d was mo i a ed by i s simple and well
de ined da a access in e ace. Howe e , di e en ly
o he ime- igge ed OMG s anda d (Kope z and
Wien, 2003), he p oposed a chi ec u e ollows an
e en - igge ed app oach. Tha is, he sma sen-
so s send h ough he ne wo k only he ele an
in o ma ion abou he p ocess. The ans e o his
in o ma ion is igge ed by asynch onous e en s.
Figu e 1 shows he basic componen s o he p o-
posed a chi ec u e. This a chi ec u e is composed
by clus e s. In u n, each clus e can ha e one
mas e node and up o 255 sla e nodes, which
a e in e connec ed by a ield ne wo k. The mas e
node is a mos powe ul p ocessing de ice and
i is designed o manage he clus e . I can also
communica e wi h o he mas e nodes h ough
a supe ision ne wo k. The e may be edundan
shadow mas e s o suppo aul ole ance. Thus,
i is he esponsible o managemen , con ol, and
con igu a ion ac i i ies h oughou he sys em.
...
Mas e DM
Clien
CP
RTDB
and
DB
...
Mas e DM
RTDB
and
DB
Supe isionNe wo k
Clus e A Clus e B
Clien
RS
FieldNe wo k
Sla e Sla e Sla e Sla eSla eSla eSla eSla e
Clien
RS
Clien
CP
Fig. 1. E en -T igge ed Sma Senso A chi ec-
u e.
2.1 P oposed Mas e Node S uc u e
The mas e node has wo da abase ypes: a eal
ime da abase (RTDB) and a adi ional da abase
(DB). The RTDB has o gua an ee he em-
po al deadlines o i s ansac ions. Thus, some
da a will be alid jus o a speci ic ime in e -
al (Ramam i ham, 1993). The mas e node logs
in i s da abases he mos ele an in o ma ion
om he sma senso s. These da abases a e ac-
cessed h ough he diagnos ic and main enance in-
e ace (DM) using he supe ision ne wo k. The
mas e node accesses in o ma ion om he sla e
nodes using he RT clien ( eal ime clien ) and
he CP clien (con igu a ion and planning clien ).
2.2 P oposed Sla e Node S uc u e
A sla e node equi es wo ypes o in e aces o
accessing i s da a, a comp ession algo i hm and a
bu e o s o e empo a y da a. Such empo a y
da a will be applied o he da a comp ession al-
go i hm, as shown in Figu e 2. The comp ession
algo i hm is esponsible o selec ing he ele an
da a in o ma ion, and hus i gene a es an asyn-
ch onous low o da a. This beha io leads o un-
p edic able iming in e als be ween consecu i e
da a ans e s om he senso . Howe e , he ime
o sending a da um is no o ally unp edic able,
due o he minimum and maximum ime p o ided
he comp ession algo i hm o he ansmission,
as will be shown ahead. So, an e en - igge ed
app oach would be e icien o he communica ion
o hese con ol- ela ed da a.
The sma senso will ha e a la ge au onomy, as i
will send only he ele an da a. The e o e, he e
is a signi ican educ ion o he ans e ed da a,
comp ession
Algo i hm
Bu e
RS In e ace CP In e ace
Fig. 2. Sma T ansduce wi h a Embedded Com-
p ession Algo i hm.
esul ing in a smalle bandwid h u iliza ion. This
way, i is possible o connec a la ge amoun o
sma senso s o he ne wo k. Howe e , he sma
senso s become mo e complex and he e o e de-
mand a la ge p ocessing capabili y. Ne e heless,
wi h he g owing echnological p og ess i is pos-
sible o design low-cos sma senso s wi h high
p ocessing capabili ies. Being so, he embedded
comp ession algo i hm is one o he esea ch a -
ge s ha mus be add essed o he a chi ec u al
ne wo k o e en - igge ed sma senso s.
Senso s equi e wo ypes o in e aces o access
he da a (RS and CP), whe e he communica ion
is suppo ed by wo di e en communica ion mod-
els (publishe -subsc ibe and clien -se e ):
•RS in e ace - eal- ime se ice in e ace.
I is used o ans e eal- ime da a o he
clus e . The da a gene a ed by he sma
senso s will be published in he ne and
consumed by he unc ions o he sys em.
•CP in e ace - con igu a ion and planning
in e ace. Th ough his in e ace i is possible
o iden i y new nodes connec ed o he ne -
wo k, o ans e new con igu a ion pa ame-
e s o he senso , as alues o comp ession
de ia ion, maximum and minimum ime o
he comp ession algo i hm, besides in o ma-
ion as he iden i ica ion o he senso in he
dis ibu ed sys em.
2.2.1. Publishe -Subsc ibe Model The commu-
nica ion model used by he RS in e ace is he
eal- ime publishe -subsc ibe model (RTPS). This
model o exchanging da a a o s he message ex-
change wi h ime pa ame e s amid de ices be-
ween wo en i ies: he publishe , esponsible by
sending he messages, and he subsc ibe s, espon-
sible o consuming hese messages, in case hey
in e es hem (Dolejs e al., 2004; Oce a, 2002).
The messages sen by a sla e node h ough he
RS in e ace will be consumed by he a ious unc-
ions o he sys em conce ned by such da a. These
messages can be send as commands o he ac u-
a o s o s o ed in da abases in acco dance wi h
hei equisi es; hey can be eal- ime da abases
(RTDB) o adi ional da abases (DB), as shown
in Figu e 3.
Senso RS
subsc ibe s
publishe
diagnos ic()
con ol_ unc ion()
senso _his o ic()
ope a ion_ ime()
Con olle
DB
RTDB
RTDB
Fig. 3. Publishe -Subsc ibe Model o P oposed
A chi ec u e.
The da a packe s sen by sma senso s h ough
he RS in e ace con ain a ious ields wi h impo -
an in o ma ion om he sma senso s, which
will be consumed by he sys em unc ions o di -
e en inali ies. Fo example, he ope a ion ime()
unc ion will access he TIME ield and will s o e
his da um in he eal- ime da abase. Such unc-
ion in o ms he supe ision ne wo k abou he
unc ioning ime o a de e mined sma senso .
The senso his o ic() unc ion will access he in-
o ma ion in he DATA ield, in o de o make
a ailable he his o ic ile o lags o a sma sen-
so . The packe ’s TAG ield is esponsible o iden-
i ying he senso ; he TYPE ield by he iden i i-
ca ion o he ype o he packe o be ans e ed,
while he QUALITY ield will con ain in o ma-
ion abou he quali y lag o he da um (good,
egula , bad, no de e mined). The CRC ield is
esponsible o con olling he packe ’s e o .
2.2.2. Clien -Se e Model The CP in e ace is
accessed di ec ly by he mas e node, which can
pe o m con igu a ion ac i i ies, can exchange pa-
ame e s and can econ igu e o new nodes (Fig-
u e 4). As an example o he con igu a ion o a
new node, a mas e node h ough i s unc ion
new de ice() s ays moni o ing he sys em in o de
o sea ch new nodes connec ed o he ne wo k.
Thus, a new sma senso will ask i s inclusion in
he sys em and will be gi en by he mas e node,
h ough he unc ion con igu e(), he necessa y
pa ame e s o s a unc ioning in he ne wo k.
F om his poin onwa ds, a e con i ming he e-
cei ed pa ame e (con i m()) and ha ing ecei ed
he au ho iza ion o ans e i s da a (s a ()), he
senso will s a sending i s da a packe s h ough
he RS in e ace.
Figu e 5 shows he econ igu a ion p ocedu e
o a sma senso . The mas e node using he
eques econ igu e(ID) unc ion econ igu es he
pa ame e o a pa icula sla e node. Then, he
sla e node is au ho ized o send i s packages.
Some impo an sys em unc ions a e p esen ed as
ollows:
• eques _inclusion(): sla e node equisi es
o mas e node i s inclusion in he ne wo k.
CP
con igu e()
...
(Pa ame e s)
(Id)
Mas e
Sla e
eques _inclusion()
econ igu e()
Fig. 4. Clien -Se e Model o P oposed A chi-
ec u e.
MASTER SLAVE
eques _ econ igu e()
econ igu e(pa ame e s)
con i m()
s a ()
ansmi (packe )
ansmi (packe )
Fig. 5. Recon igu a ion P oceeding o a Sla e
Senso .
•con igu e(): mas e node sends con igu a-
ion pa ame e s o a new sla e node.
• econ igu e(): mas e node sends econ ig-
u a ion o a sla e node.
• ansmi (): sla e node sends a de e mined
packe o he ne wo k.
•ope a ion_ ime(): esponsible o he ime
unc ioning o a sma senso .
3. CASE STUDY
This sec ion p esen s a case s udy implemen a-
ion o he p oposed e en - igge ed sma sen-
so ne wo k a chi ec u e. The implemen a ion
uses he IEEE 802.15.4 (IEEE 802.15.4 S an-
da d, 2006) s anda d as communica ion in as-
uc u e. The p o o ype has been implemen ed
upon he F eescale Semiconduc ion de elopmen
ki .
3.1 Implemen a ion based on IEEE 802.15.4
The IEEE 802.15.4 s anda d de ines he physical
and MAC laye s o low cos , low powe , and low
a e de ices (Ba on i e al., 2007). The embedded
so wa e was de eloped in acco dance o SMAC
(Simple MAC ) p emisses.
3.2 De ini ion o he Tes En i onmen
Figu e 6 shows he en i onmen de ined o ex-
pe imen ally assess his a chi ec u e, whe e i is
de ined wo sla e sma nodes and one mas e
sma node connec ed h ough a ield ne wo k.
The supe ision sys em can manage he sma
senso ne wo k h ough he mas e node.
RS-232
RS-232
RS-232
MASTER
SLAVERS
DB
Senso
Values
Senso
Values
Fig. 6. S uc u e o a Sma Senso .
Fo ou applica ion, a eal senso has been em-
ula ed using a PC equipped wi h a se ial po
and he ne wo k in e ace has been a wi eless
communica ion boa d o F eescale.
To implemen he emula ion p ocess o he eal
senso s, we used wo da abases con en eco ds ex-
ac ed om a eal- ime gas dis ibu ion moni o -
ing sys em; speci ically, we choose an ou le a i-
able, wi h 10000 eco ds. Fo es ing e ec s, a p o-
g am w i en in ANSI C was de eloped, i is capa-
ble o ead each eco d om he da abase and o
send i h ough RS-232 po (se ial inpu /ou pu )
o he wi eless communica ion boa d o F eescale.
The mas e node, on he o he hand, is composed
o a F eescale wi eless communica ion boa d con-
nec ed in a PC equipped wi h a so wa e able o
s o e he ecei ed alues by mas e boa d in he
da abase and gene a es he cu e o he ele an
alues o each sla e senso connec ed upon he
ield ne wo k.
3.3 De ini ion o he P o o ype Func ions
The sla e nodes suppo he plug-and-play pa adigm.
This pa adigm is easily implemen ed because each
sla e node has an only MAC add ess as i s iden i i-
ca ion on he ne wo k. Thus, he mas e node can
de ec he sla e nodes in au oma ic way. When a
sla e node is connec ed o he ne wo k i equi es
o he mas e node i s con igu a ion. Then he
mas e node eply wi h he ollowing pa ame e s:
senso ID, minimum and maximum ime and com-
p ession de ia e o he comp ession algo i hm.
All he sla e nodes ha e an implemen a ion o he
comp ession algo i hm Swinging Doo (B is ol,
1990). Thus, when a new da a is gene a ed, he
comp ession algo i hm decides i i is ele an o
no , hen his da a can be send o no o he
mas e boa d. To send a da a o mas e node, he
sla e node mus build a packe wi h he ollowing
ields: senso iden i ica ion (ID), cu en ime and
alue. Then, when he mas e node ecei es any
da a om he sla e node, i mus s o e ha da a
in i s da abase o he ac i i ies o supe ision
ne wo k.
Ano he impo an unc ion implemen ed in he
p o o ype is he econ igu a ion o a sla e senso .
Fo example, h ough he mas e node, i is pos-
sible o econ igu e he comp ession algo i hm o
any sma senso connec ed o ield ne wo k.
3.4 Tes Scena ios
To alida e he p o o ype, we de ined wo es
scena ios. Theses scena ios a e desc ibed as ollow.
3.4.1. Scena io 1 The scena io 1 is de ined as
ha ing only one sma senso connec ed o he
ne wo k wi h pa ame e s econ igu a ion. In ha
way, he sla e senso eques s o he mas e node
i s ini ial con igu a ion. The ollowing pa ame e s
a e passed o he sla e node:
•ID: 1;
•Minimum ime: 3 s;
•Maximum ime: 10 s;
•Comp ession de ia e: 5 m3/day.
The sla e senso is econ igu ed a e 3150 sec-
onds. The new con igu a ion o he sla e senso is
isualized as ollow:
•Minimum ime: 5 s;
•Maximum ime: 20 s;
•Comp ession de ia e: 9 m3/day.
3.4.2. Scena io 2 The scena io 2 de ines he
execu ion o wo sla e senso , in which he second
sla e senso is connec ed 2000 seconds la e han
i s sla e senso . Sla e senso 1 is con igu ed wi h
he pa ame e s as ollow:
•ID: 1;
•Minimum ime: 3 s;
•Maximum ime: 10 s;
•Comp ession de ia e: 5 m3/day.
Sla e senso 2 is con igu ed wi h he pa ame e s
as ollow:
•ID: 2;
•Minimum ime: 5 s;
•Maximum ime: 20 s;
•Comp ession de ia e: 5 m3/day.
The mas e node is esponsible by s o ing he
g aph gene a ion senso da a wi h he ele an
da a om each sma senso . The sma senso
1 uses he da a s o ed in da abase 1 while sma
senso 2 uses he da a s o ed in da abase 2.
3.5 Resul s
This sec ion p esen s he ob ained esul s om
all es scena ios desc ibed abo e. Figu es 7 and 8
show he aw da a om he senso 1 and senso
2, espec i ely.
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
460
480
500
520
540
560
580
600
620
640
660
680
O iginalDa a
ins an (seconds)
alues(cubicme e s/day)
O iginalValues
Fig. 7. O iginal Da a om he Sla e Senso 1.
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
540
560
580
600
620
640
660
680
O iginalDa a
ins an (seconds)
alues(cubicme e s/day)
O iginalValues
Fig. 8. O iginal Da a om he Sla e Senso 2.
3.5.1. Scena io 1 In his scena io was de ined
he ope a ion o only one sla e wi h econ igu a-
ion du ing i s ac i i y. The supe ision sys em
eques s he econ igu a ion o he sla e node com-
p ession pa ame e s 3150 seconds a e i s ini ia-
ion. Fo his scena io, i is calcula ed he com-
p ession a es and he mean squa e e o s o all
senso ac i i y and be o e senso econ igu a ion
and a e senso econ igu a ion ac i i ies.
Figu e 9 shows he sla e senso ele an da a. Fo
his scena io he comp ession o all senso ac i i y
was 90.76% wi h a mean squa e e o o 44.25. The
comp ession a e be o e he econ igu a ion was o
85.52% wi h a mean squa e e o o 86.76. A e
he sla e senso econ igu a ion, he comp ession
a e was o 93.16% wi h a mean squa e e o o
24.71.
3.5.2. Scena io 2 In his scena io was de ined
he ope a ion o wo sla e sma senso . The sla e

0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
460
480
500
520
540
560
580
600
620
640
660
680
Comp essedValues
Comp essionRa eo 90.76%
ins an (seconds)
alues(cubicme e es/day)
Fig. 9. Rele an Da a o Sla e Senso 1 wi h
econ igu a ion.
senso 2 was connec ed 2000 seconds a e he
sla e senso 1. Fo his scena io, i is calcula ed he
comp ession a e and he mean squa e e o o
each sla e senso . The comp ession a e o sla e
senso 1 was 85.80% wi h he mean squa e e o
o 41.13. The comp ession a e o sla e senso 2
was 90.16% wi h he mean squa e e o o 32.58.
Figu e 10 and 11 shows he ele an da a o he 1
and 2 sla e senso , espec i ely.
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
460
480
500
520
540
560
580
600
620
640
660
680
Comp essedValues
Comp essionRa eo 85.80%
ins an (seconds)
alues(cubicme e es/day)
Fig. 10. Rele an Da a o Sla e Senso 1.
4. CONCLUSION
The pape p esen ed an e en - igge ed sma sen-
so ne wo k a chi ec u e and i s implemen a ion
based IEEE 802.15.4 s anda d. The main ea u e
o his a chi ec u e is he in eg a ion o a da a
comp ession algo i hm wi h simple implemen a-
ion in o he sma senso s. This app oach sa es
ne wo k bandwid h, because he senso s only send
ele an da a.
Wi h he achie ed esul s h ough he ealized
expe imen s, i is possible o conclude ha he
local comp ession p ocess in each sma senso can
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
540
560
580
600
620
640
660
680
Comp essedValues
Comp essionRa eo 90.16%
ins an (seconds)
alues(cubicme e es/day)
Fig. 11. Rele an Da a o Sla e Senso 2.
dec ease conside ably he da a exchanges in he
communica ion ne wo k.
REFERENCES
Ba on i, Paolo, P ashan Pillai, Vince Chook, S e-
ano Chessa, Albe o Go a and Y. Fun Hu
(2007). Wi eless Senso Ne wo ks: A Su -
ey on he s a e o he a and he 802.15.4
and ZigBee S anda ds. Compu e Communi-
ca ions 30, 1655–1695.
B is ol, E. H. (1990). Swinging Doo T ending:
adap i e end eco ding?. In: ISA Na ional
Con . P oc.. pp. 749–753.
Dolejs, Ond ej, Pe Smolik and Zdenek Hanza-
lek (2004). On he E he ne Use o Real-
Time Publish-Subsc ibe Based Applica ions.
In: P oceedings o IEEE In e na ional Wo k-
shop on Fac o y Communica ion Sys ems.
pp. 39–44.
Elmen eich, Wil ied (2006). Time-T igge ed
Sma T ansduce Ne wo ks. IEEE T ansac-
ions on Indus ial In o ma ics 2, 192–199.
IEEE 802.15.4 S anda d (2006). Pa 15.4: Wi e-
less Medium Access Con ol (MAC) and
Physical Laye (PHY) Speci ica ions o
Low-Ra e Wi eless Pe sonal A ea Ne wo ks
(WPANs). Speci ica ion.
Kope z, H. and TU Wien (2003). OMG Sma
T ansduce Speci ica ion II. Speci ica ion.
Le˜ao, E ico Meneses, Luiz A onso Guedes
and F ancisco Vasques (2007). An E en -
T igge ed Sma Senso Ne wo k A chi ec-
u e. In: P oceedings o 5 h IEEE In e na-
ional Con e ence on Indus ial In o ma ics,
INDIN 2007, Vienna, Aus ia.
Oce a (2002). Wp2 - a chi ec u e speci ica ion.
deli e able d2.1 - a chi ec u e and compo-
nen s in eg a ion.
Ramam i ham, K i hi (1993).
Real-Time Da abases. In e na ional Jou nal
o Dis ibu ed and Pa allel Da abases 1, 199–
226.