You are on page 1of 13

Energy-Efficient Ìnitialization Protocols

for Single-Hop Radio Networks


with No Collision Detection
Koji Nakano, Member, IEEE Computer Society, and Stephan Olariu, Member, IEEE Computer Society
Abstract-A radio network (1`, for short) is a distributed system consisting of n radio stations. We assume that the stations are small,
bulk-produced, hand-held devices running on batteries and cannot be distinguished by serial or manufacturing number. Since
recharging batteries may not be possible while on mission, we are interested in designing protocols that are highly energy-efficient. The
initialization problem is to assign each of the n stations in the RN a unique ÌD. The initialization problem is nontrivial since the stations
are assumed to be indistinguishable. The problem is fundamental, since practically all communication protocols for 1`s proceed
under the assumption that the RN has been initialized in advance. The main contribution of this work is to propose energy-efficient
randomized initialization protocols for single-hop 1`s lacking collision detection capabilities. First, we show that if the number n of
stations is known beforehand, the single-channel 1` can be initialized by a protocol that terminates, with probability exceeding 1
1
n
,
in C n time slots, with no station being awake for more than C log log n time slots. We then go on to address the multichannel case
and show that if /, / 1 , channels are available, an n-station 1` can be initialized, with probability exceeding 1
1
n
, in C
n
/
log n
time slots, with no station being awake for more than C log log n time slots.
Index Terms-Radio networks, wireless networks, mobile computing, initialization.

1 INTRODUCTION
I
N roconl yoars, viroIoss and mobiIo communicalions lavo
soon an oxpIosivo grovll boll in lorms of llo numbor of
sorvicos providod and llo lypos of loclnoIogios llal lavo
bocomo avaiIabIo. Indood, coIIuIar loIoplony, radio paging,
coIIuIar dala, and ovon rudimonlary muIlimodia sorvicos
lavo bocomo commonpIaco and llo domand for onlancod
capabiIilios viII conlinuo lo grov inlo llo forosooabIo fuluro
|2], |6], |25], |28]. Il is anlicipalod llal in llo nol-so-dislanl
fuluro, mobiIo usors viII bo abIo lo accoss lloir dala and
ollor sorvicos sucl as oIoclronic maiI, vidoo loIoplony,
slocI marIol novs, map sorvicos, and oIoclronic banIing
vliIo on llo movo |6], |11], |23].
1.1 Radio Networks¬GeneraIities
A radio nolvorI (1`, for slorl) is a dislribulod syslom vill
no conlraI arbilor, consisling of n radio lranscoivors,
loncoforll roforrod lo as siciicns. Wo assumo llal llo
slalions aro buII-producod, commodily land-loId dovicos
running on ballorios and llal il is impossibIo or impraclicaI
lo dislinguisl llom by soriaI or manufacluring numbors.
An imporlanl lasI lo porform is llal of assigning llo
slalions in an n-slalion RN i!cnii|icciicn num|crs (IDs, for
slorl), vory mucl IiIo II addrossos, in llo rango 1 lo n. Tlis
lasI is Inovn as iniiic|izciicn and is vory imporlanl, as
many of llo oxisling prolocoIs lacilIy assumo llal llo
slalions aIroady lavo IDs. Tlo iniliaIizalion probIom is
nonlriviaI, sinco llo slalions aro assumod lo bo indislin-
guislabIo. Iurllor, sinco llo slalions aro povor-Iimilod, il is
of imporlanco lo dosign onorgy-officionl iniliaIizalion
prolocoIs for singIo-lop 1`s boll in llo caso vloro llo
syslom las a coIIision doloclion capabiIily and in llo caso
vloro llis capabiIily is nol prosonl.
As is cuslomary, limo is assumod lo bo sIollod and aII llo
slalions lavo a IocaI cIocI llal Ioops synclronous limo by
inlorfacing vill a GIobaI Iosilioning Syslom (GIS, for
slorl) |8], |13], |19], |20]. Wo nolo loro llal undor curronl
loclnoIogy, llo commorciaIIy avaiIabIo GIS syslom
1
pro-
vidos Iocalion informalion accuralo villin 22 molors, as
voII as limo informalion accuralo lo villin 100 nanoso-
conds |8]. Tlis is moro llan sufficionl for llo slalions of llo
RN lo synclronizo.
Tlo slalions aro assumod lo lavo llo compuling povor
of a Iaplop compulor, in parlicuIar, lloy aII run llo samo
prolocoI and can gonoralo random bils llal provido IocaI
dala on vlicl llo slalions may porform compulalions.
Tlo slalions communicalo using /, / 1 , radio
froquoncy clannoIs. Wo assumo llal in a limo sIol, a slalion
can luno lo ono clannoI and/or lransmil on al mosl ono,
possibIy llo samo, clannoI. A lransmission invoIvos a dala
pacIol vloso Iongll is sucl llal llo lransmission can bo
compIolod villin ono limo sIol. Wo ompIoy llo commonIy-
accoplod assumplion llal vlon lvo or moro slalions aro
lransmilling on a clannoI in llo samo limo sIol, llo
corrosponding pacIols cc||i!c and aro garbIod boyond
ÌEEE TRANSACTÌONS ON PARALLEL AND DÌSTRÌBUTED SYSTEMS, VOL. 11, NO. 8, AUGUST 2000 851
 K. Nc|cnc is uii| i|c Dcpcrimcni c| ||cciricc| cn! Ccmpuicr |nginccring,
Ncgcqc |nsiiiuic c| 1cc|nc|cgq, S|cu-|c, Ncgcqc, 466-8555, jcpcn.
|-mci|. nc|cnc©c|ccm.niicc|.cc.jp.
 S. O|criu is uii| i|c Dcpcrimcni c| Ccmpuicr Scicncc, O|! Dcminicn
Unitcrsiiq, Ncr|c||, VA 2325º. |-mci|. c|criu©cs.c!u.c!u.
Mcnuscripi rcccitc! 16 Nct. 1ººº, ccccpic! 15 Apr. 2000.
Icr in|crmciicn cn c|icining rcprinis c| i|is criic|c, p|ccsc scn! c-mci| ic.
ip!s©ccmpuicr.crg, cn! rc|crcncc ||||CS Icg Num|cr 110º63.
1. GIS sysloms using miIilary codos acliovo an accuracy llal is ono
ordor of magniludo bollor llan commorciaI codos.
1045-9219/00/$10.00  2000 ÌEEE
rocognilion. Wo dislinguisl bolvoon 1` sysloms basod on
lloir capabiIily lo dolocl coIIisions. SpocificaIIy, in llo 1`
vill cc||isicn !cicciicn (CD, for slorl), al llo ond of a limo
sIol llo slalus of a radio clannoI is:
 NUII÷no lransmission on llo clannoI
 SINGII÷oxaclIy ono lransmission on llo clannoI
 COIIISION÷lvo or moro lransmissions on llo
clannoI.
In llo 1` vill nc cc||isicn !cicciicn (no-CD, for slorl), al llo
ond of a limo sIol llo slalus of a radio clannoI is:
 NOISI÷ambionl noiso: oillor no lransmission or
coIIision of sovoraI lransmissions on llo clannoI
 SINGII÷oxaclIy ono lransmission on llo clannoI.
In ollor vords, llo 1` vill no-CD cannol dislinguisl
bolvoon ambionl noiso and lvo or moro lransmissions
coIIiding on llo clannoI. SovoraI vorIors lavo poinlod oul
llal from a praclicaI slandpoinl llo no-CD assumplion
maIos a Iol of sonso, ospociaIIy in llo prosonco of noisy
clannoIs vlicl, vo arguo, londslo bo llo norm rallor llan
llo oxcoplion. Givon llo addilionaI Iimilalions, il is far moro
claIIonging lo dosign officionl prolocoIs for 1`s vill
no-CD llan for lloso vill coIIision doloclion capabiIilios
|4], |5].
Il is vory imporlanl lo roaIizo llal sinco llo slalions of llo
1` aro running on ballorios, saving ballory povor is
oxcoodingIy imporlanl, as roclarging ballorios may nol bo
an oplion vliIo on mission. Il is voII Inovn llal a slalion
oxponds povor vliIo ils lranscoivor is aclivo, llal is, vliIo
lransmilling or rocoiving a pacIol. Il is porlaps surprising
al firsl llal a slalion oxponds povor ovon if il rocoivos a
pacIol llal is nol doslinod for il |3], |9], |11], |21], |24], |26],
|27]. ConsoquonlIy, vo aro inloroslod in dovoIoping prolo-
coIs llal aIIov slalions lo povor lloir lranscoivor off (i.o., go
lo sIoop) lo llo Iargosl oxlonl possibIo. AccordingIy, vo
judgo llo goodnoss of a prolocoI by llo foIIoving lvo
yardslicIs:
 llo ovoraII numbor of limo sIols roquirod by llo
prolocoI lo lorminalo
 for oacl individuaI slalion, llo lolaI numbor of limo
sIols vlon il las lo bo avaIo in ordor lo lransmil/
rocoivo pacIols.
Tlo goaIs of oplimizing lloso paramolors aro, of courso,
confIicling. Il is roIalivoIy slraigllforvard lo minimizo
ovoraII compIolion limo al llo oxponso of onorgy consump-
lion. SimiIarIy, ono can minimizo onorgy consumplion al
llo oxponso of compIolion limo |26], |27], |28]. Tlo claIIongo
is lo slriIo a sonsibIo baIanco bolvoon llo lvo by dosigning
prolocoIs llal laIo a smaII numbor of limo sIols lo lorminalo
vliIo boing, al llo samo limo, as onorgy-officionl as
possibIo.
1.2 State of the Art
As slalod proviousIy, llo iniiic|izciicn probIom is lo assign
oacl of llo n slalions of an RN an inlogor ID in llo rango 1
lo n sucl llal no lvo slalions rocoivo llo samo ID. Tlo
iniliaIizalion probIom is fundamonlaI in boll nolvorI
dosign and in muIliprocossor sysloms |16].
RoconlIy, Hayasli ol aI. |12] prosonlod a prolocoI llal
iniliaIizos a singIo-clannoI, n-slalion 1` vill CD in C n
limo sIols vill probabiIily al Ioasl 1
1
2
C n
. Tlo prolocoI
ropoalodIy parlilions llo slalions of llo 1` inlo nonomply
subsols unliI, ovonluaIIy, ovory subsol consisls of a singIo
slalion. Sinco llo slalions lavo lo conslanlIy monilor llo
clannoI, llo prolocoI of |12] is nol onorgy-officionl.
Ialor, Bordim ol aI. |7] slovod llal, if llo numbor n of
slalions is Inovn boforoland, a singIo-clannoI n-slalion
1` vill CD can bo iniliaIizod vill probabiIily al Ioasl
1
1
n
, in C n limo sIols vill no slalion boing avaIo for
moro llan C log n limo sIols. Tlis prolocoI usos llo
iniliaIizalion prolocoI of |12] as foIIovs: Having parlilionod
llo n-slalions inlo
n
logn
subsols of rouglIy log n slalions, oacl
subsol is iniliaIizod individuaIIy. Tlo Ioy obsorvalion is llal
a slalion noods lo bo avaIo onIy for llo limo nocossary lo
iniliaIizo llo subsol lo vlicl il boIongs. By using llo
prolocoI of |12], oacl subsol can bo iniliaIizod in C log n
limo sIols vill probabiIily al Ioasl 1
1
2
C logn
and, llus, no
slalion las lo bo avaIo for moro llan C log n limo sIols.
Onco oacl subsol las boon iniliaIizod, a simpIo profix sum
compulalion aIIovs llo slalions lo updalo lloir IocaI ID lo
llo dosirod ID. Bordim ol aI. |7] vonl on lo slov lov lo uso
llo singIo-clannoI iniliaIizalion prolocoI lo iniliaIizo, vill
probabiIily oxcooding 1
1
n
, a /-clannoI, 1 /
n
log n
, 1`
vill CD in C
n
/
limo sIols, vill no slalion boing avaIo for
moro llan C log n limo sIols.
1.3 Summary of Our Contributions
Our main conlribulion is lo proposo onorgy-officionl
iniliaIizalion prolocoIs for singIo-lop 1`s vill no-CD,
roquiring no slalion lo bo avaIo for moro llan C log log n
limo sIols.
Wo bogin by sloving llal a singIo-clannoI 1` vill
no-CD can bo iniliaIizod by a prolocoI lorminaling, vill
probabiIily al Ioasl 1
1
n
, in C n limo sIols, vill no slalion
boing avaIo for moro llan C log n limo sIols. AIllougl llo
limo porformanco of our prolocoI is llo samo as llal in |7],
our prolocoI vorIs in llo no-CD caso and llo loclniquo
usod is lolaIIy difforonl. Our prolocoI slarls oul by assigning
oacl slalion a uniquo icmpcrcrq ID in llo rango 1 lo 16n.
Aflor llal, using llo lomporary ID, llo prolocoI roassigns
llo slalions uniquo IDs in llo rango 1 lo n.
Wo llon go on lo improvo llis prolocoI sucl llal no
slalion noods lo bo avaIo for moro llan C log log n limo
sIols. Tlo Ioy idoa is lo assign llo various slalions in llo
1` lomporary IDs in llo rango 1 lo 16n in sucl a vay llal
vill ligl probabiIily, ovory group of consoculivo 64 ln n
inlogors conlain llo lomporary ID of somo slalion. In lurn,
llis aIIovs us lo iniliaIizo oacl group IocaIIy, vill no
slalion boing avaIo for moro llan C log log n limo sIols.
852 ÌEEE TRANSACTÌONS ON PARALLEL AND DÌSTRÌBUTED SYSTEMS, VOL. 11, NO. 8, AUGUST 2000
Onco llis IocaI iniliaIizalion is porformod, llo IocaI ID
numbors can bo updalod lo gIobaI ID numbors by
porforming llo onorgy-officionl profix sums prolocoI llal
vo discuss in Soclion 3.
IinaIIy, vo oxlond llo singIo-clannoI iniliaIizalion
prolocoI lo a /, (/ 1), clannoI 1` vill no-CD. Spoci÷
ficaIIy, vo slov llal vill probabiIily oxcooding 1
1
n
llis
prolocoI lorminalos in C
n
/
log n limo sIols, vill no
slalion boing avaIo for moro llan C log log n limo sIols.
Tlo romaindor of llis vorI is organizod as foIIovs.
Soclion 2 roviovs basic probabiIily lloory llal viII provido
llo maclinory for ovaIualing llo porformanco of iniliaIiza-
lion prolocoIs. Soclion 3 discussos onorgy-officionl profix
sums prolocoIs and a simpIo iniliaIizalion prolocoI llal viII
bo noodod in llo romaindor of llis vorI. Soclion 4 discussos
an onorgy-officionl prolocoI llal assigns llo slalions of an
1` lomporary ID numbors in a nol-loo-Iargo rango. Tlis
prolocoI lurns oul lo bo llo vorIlorso of llo vloIo papor.
In Soclion 5, vo slov lov lo uso llo prolocoI dovoIopod in
Soclion 4 lo iniliaIizo a singIo-clannoI, n-slalion 1` vill
no-CD in sucl a vay llal no slalion is avaIo for moro llan
C log n limo sIols. In Soclion 6, vo slov llal ono can
iniliaIizo a singIo-clannoI, n-slalion 1` vill no-CD vill
no slalion boing avaIo for moro llan C log log n limo sIols.
In Soclion 7, vo provido simuIalion rosuIls sloving llal our
iniliaIizalion prolocoI in Soclion 6 porforms oxlromoIy voII
in praclico. In Soclion 8, vo slov llal if /, (/ 1), clannoIs
aro avaiIabIo, ono can iniliaIizo an n-slalion 1` vill no-CD
in C
n
/
log n limo sIols vill no slalion boing avaIo for
moro llan rouglIy C log log n limo sIols. IinaIIy, Soclion 9
offors concIuding romarIs and poinls oul diroclions for
furllor vorI.
2 A BRIEF REFRESHER OF BASIC PROBABILITY
THEORY
Tlo main goaI of llis soclion is lo roviov oIomonlary
probabiIily lloory rosuIls llal aro usofuI for anaIyzing llo
porformanco of our prolocoIs. Ior a moro dolaiIod discus-
sion of bacIground maloriaI, vo rofor llo roador lo |15].
Tlrougloul, Ii ¹ viII donolo llo probabiIily of ovonl ¹.
Iol 1
1
, 1
2
, . . . , 1
i
bo arbilrary ovonls ovor a sampIo spaco.
Tlo voII Inovn Do Morgan Iav slalos llal
i
i 1
1
i
i
i 1
1
i
, 1
vloro 1
i
is llo ovonl llal occurs if and onIy if 1
i
doos nol.
In addilion, il is Inovn llal
Ii
i
i 1
1
i
i
i 1
Ii 1
i
, 2
vill oquaIily loIding if llo ovonls aro indopondonl.
Ior oxampIo, assumo llal llo slalions of an 1` lavo
boon parlilionod inlo / groups G
1
, G
2
, . . . , G
/
, and llal for a
givon i, 1 i / , llo probabiIily of llo ovonl 1
i
llal
group G
i
faiIs lo salisfy a prodicalo is j
i
. Nov, (1) and (2)
combinod guaranloo llal llo probabiIily of llo ovonl llal c||
llo groups salisfy prodicalo is:
Ii 1
1
1
2
1
/
1 Ii 1
1
1
2
1
/
1 Ii 1
1
1
2
1
/
1
i
i 1
j
i
. 8
Nolico llal (3) loIds rogardIoss of vlollor llo ovonls 1
i
aro
indopondonl.
Iol A
1
, A
2
, . . . , A
n
bo in!cpcn!cni Pcisscn iric|s sucl llal
for 1 i n, Ii A
i
1 j
i
and Ii A
i
0 1 j
i
and
vrilo A A
1
A
2
A
n
. Il is voII Inovn llal llo
oxpoclod vaIuo of A is givon by
1 A 1 A
1
1 A
2
1 A
n
j
1
j
2
j
n
.
4
To anaIyzo llo laiI of A, vo slaII maIo uso of llo foIIoving
oslimalo, commonIy roforrod lo as C|crnc|| |cun! |15]:
Ii A 1 1 A <
c
1
1
1 A
0 . ò
Wo aIso uso llo foIIoving inoquaIily llal foIIovs from
(5):
Ii A 1 1 A < c
2
8
1 A
0 1 . 6
IinaIIy, vo laIo nolo of a voII-Inovn rosuIl llal viII bo
usod froquonlIy in llo romaindor of llis vorI.
Lemma 2.1. Icr ctcrq pcsiiitc ccnsicni c, o < cn , i|c sc¡ucncc
1
c
n
n
is mcncicnicc||q incrccsing cn!
lim
n
1
c
n
n
1
c
c
. 7
3 ENERGY-EFFICIENT PREFIX SUMS PROTOCOLS
Tlo purposo of llis soclion is lo prosonl various profix sums
prolocoIs vlicl aro Ioy ingrodionls of our onorgy-officionl
iniliaIizalion prolocoIs.
3.1 A Generic Prefix Sums ProtocoI
Considor a singIo-clannoI, n-slalion 1` popuIalod by
slalions o 1 , o 2 , . . . , o n and assumo llal for ovory i,
1 i n , slalion o i sloros a numbor ` i . IniliaIIy, onIy
slalion o i Inovs ` i . Tlo goaI is lo compulo for ovory i,
1 i n , llo sum ` 1 ` 2 ` i roforrod lo as
llo ill prc|ix sum of ` 1 , ` 2 , . . . , ` n .
NAKANO AND OLARÌU: ENERGY-EFFÌCÌENT ÌNÌTÌALÌZATÌON PROTOCOLS FOR SÌNGLE-HOP RADÌO NETWORKS WÌTH NO COLLÌSÌON... 853
Tlo dolaiIs of llo profix sums prolocoI foIIov. To bogin,
slalion o 1 lransmils ` 1 on llo clannoI. Al llis limo, aII
llo slalions lavo lloir lranscoivor povorod off (i.o., aro
asIoop) oxcopl for o 1 llal is lransmilling and o 2 llal is
picIing up ` 1 . Nov, o 2 compulos ` 1 ` 2 and
lransmils llo rosuIl lo o 8 . In llis limo sIol, aII llo slalions
aro asIoop oxcopl for o 2 and o 8 . Upon rocoiving
` 1 ` 2 , o 8 compulos ` 1 ` 2 ` 8 and lrans-
mils llo rosuIl lo o 4 . Tlis is llon conlinuod unliI, al llo
ond of n 1 limo sIols, o n rocoivos ` 1 ` 2
` n 1 and compulos ` 1 ` 2 ` n .
Tlus, aII llo profix sums can bo compulod in n 1 limo
sIols. ImporlanlIy, no slalion is avaIo for moro llan lvo
limo sIols. To summarizo, vo slalo llo foIIoving rosuIl.
Lemma 3.1 1|c prc|ix sums prciccc| intc|ting i|c n siciicns c| c
sing|c-c|cnnc| 1` ccn |c ccmp|cic! in n 1 iimc s|cis, uii|
nc siciicn |cing cuc|c |cr mcrc i|cn iuc iimc s|cis.
Noxl, vo slov lov lo oxlond llo simpIo profix sums
prolocoI vo jusl discussod lo llo muIliclannoI caso.
Considor a /-clannoI, n-slalion 1` vill clannoIs
C 1 , C 2 , . . . , C / and slalions o 1 , o 2 , . . . , o n . Wo
firsl assumo llal /
n
logn
. Ior simpIicily of oxposilion, vo
aIso assumo llal boll n and / aro a povor of 2.
Tlo prolocoI consisls of llroo plasos llal vo discuss in
dolaiI boIov. In Ilaso 1, vo bogin by parlilioning llo
n slalions inlo / oquaI-sizod groups G 1 , G 2 , . . . , G / of
n
/
slalions oacl. Iol o i, ¸ , (1 i /; 1 ¸
n
/
), donolo llo
¸ll slalion in group G i , llal is, o i, ¸ o i 1
n
/
¸ .
Sinco, by assumplion, /
n
log n
, oacl group conlains al
Ioasl
n
/
log n log / slalions. Having dodicalod clannoI
C i lo group G i , 1 i / , vo procood lo compulo llo
|ccc| prc|ix sums villin oacl group G i using Iomma 3.1.
Tlus, al llo ond of Ilaso 1, oacl slalion o i, ¸ in group G i
sloros llo IocaI profix sum
` i 1
n
/
1 ` i 1
n
/
2 ` i 1
n
/
¸ .
Iurllor, llo Iasl slalion o i,
n
/
sloros llo sum ¸ i
` i 1
n
/
1 ` i 1
n
/
2 ` i
n
/
of aII llo
iloms slorod by llo slalions in G i . By Iomma 3.1, Ilaso
1 lorminalos in
n
/
1 limo sIols vill slalions o i, 1 and
o i,
n
/
boing avaIo for ono limo sIol and llo ollor slalions
boing avaIo for lvo limo sIols.
Iol ¸ o, / ¸ o ¸ o 1 ¸ / , (1 o < / /),
donolo a pcriic| sum of llo sums. RocaII llal al llo ond of
Ilaso 1 oacl slalion o i, ¸ , (1 i /; 1 ¸
n
/
), Inovs llo
IocaI profix sum
` i 1
n
/
1 ` i 1
n
/
2 ` i 1
n
/
¸ .
Tlus, onco o i, ¸ Ioarns llo g|c|c| prc|ix sum ¸ 1, i 1 , il can
immodialoIy compulo llo dosirod profix sum
` 1 ` 2 ` ¸ ¸ 1, i 1 ` i 1
n
/
1
` i 1
n
/
2 ` i 1
n
/
¸ .
In Ilasos 2 and 3, vo compulo llo gIobaI profix sums
¸ 1, 1 , ¸ 1, 2 , . . . , ¸ 1, / 1 . Nolo llal, in facl, llo Iasl profix
sum ¸ 1, / is nol roaIIy nocossary.
Considor llo |incrq ircc c| pcriic| sums iIIuslralod in Iig. 1.
Tlo sums ¸ 1 , ¸ 2 , . . . , ¸ / appoar in Iofl-lo-rigll ordor in
llo / Ioavos of llo lroo. Iacl inlornaI nodo corrosponds lo
llo sum of ils doscondanls. Iurllor, vo dofino llo |ctc| c| c
nc!c lo bo llo dislanco from a Ioaf. CIoarIy, a nodo al IovoI i,
(0 i log /), las 2
i
Ioaf-doscondanls and, llus, corro-
sponds lo a parliaI sum of 2
i
oIomonls.
A nico proporly of llo binary lroo of parliaI sums is llal
oacl profix sum can bo oblainod by llo sum of parliaI sums
closon al mosl ono from oacl IovoI. Ior oxampIo, llo profix
sum ¸ 1, 18 can bo compulod by llo sums ¸ 18 (IovoI 0),
¸ 9, 12 (IovoI 2), and ¸ 1, 8 (IovoI 3). In ollor vords,
¸ 1, 18 ¸ 18 ¸ 9, 12 ¸ 1, 8 . Tloso llroo parliaI sums
can bo found as foIIovs: Considor llo pall from Ioaf ¸ 14
lo llo rool. Ior ovory inlornaI nodo on llo pall, llo Iofl
cliIdron nol on llo pall aro ¸ 18 , ¸ 9, 12 , and ¸ 1, 8 , vlicl
854 ÌEEE TRANSACTÌONS ON PARALLEL AND DÌSTRÌBUTED SYSTEMS, VOL. 11, NO. 8, AUGUST 2000
Fig. 1. Ìllustrating the binary tree of partial sums.
aro oxaclIy llo nodos lo bo closon for profix sum ¸ 1, 18 . In
gonoraI, llo profix sum ¸ 1, i 1 , (2 i /), can bo
compulod by idonlifying llo pall from nodo ¸ i lo llo
rool and by picIing llo Iofl cliIdron of llo nodos on llis
pall llal aro llomsoIvos nol on llo pall. Using llis
proporly, vo viII compulo aII nocossary profix sums
¸ 1, 1 , ¸ 1, 2 , . . . , ¸ 1, / 1 .
RocaII llal, al llo ond of Ilaso 1, oacl slalion o i,
n
/
,
(1 i /), Inovs llo vaIuo of ¸ i . Using llis facl, Ilaso 2
porforms llo foIIoving lvo lasIs simuIlanoousIy:
Task 1. AII llo nocossary parliaI sums aro compulod.
Task 2. Slalions in G i , (2 i /), rocoivo llo parliaI sums
nocossary lo compulo ¸ 1, i 1 .
Sinco, as nolod boforo, ¸ 1, / is nol nocossary, vo do
nol lavo lo compulo llo parliaI sums of llo rigllmosl
pall of llo binary lroo. Ior TasI 1, vo assign llo Iasl
slalion o i,
n
/
in oacl group G i lo a nocossary parliaI
sum of llo binary lroo, as iIIuslralod in Iig. 2. Sinco llo
numbor of nocossary parliaI sums is
/
2
1
/
4
1 1 < /,
llis assignmonl is foasibIo. Wo caII lloso slalions
primcrq siciicns. Ior TasI 2, vo assign oacl slalion
o i, ¸ , (1 i /; 1 ¸ log /), lo a nodo (i.o. parliaI sum)
in llo binary lroo as foIIovs: slalion o i, ¸ , 1 i /; 1
¸ log / , is assignod lo llo nodo al IovoI ¸ 1 vlicl is llo
Iofl cliId of a nodo on llo pall from ¸ i lo llo rool bul
vlicl is nol in llo pall. Wo rofor llo roador lo Iig. 3 for an
iIIuslralion of llis assignmonl. Ior oxampIo, o 14, 1 ,
o 14, 8 , and o 14, 4 aro assignod lo ¸ 18 , ¸ 9, 12 , and
¸ 1, 8 of ¸ 14 , rospoclivoIy. Wo caII llo assignod slalions
rcgu|cr siciicns. CIoarIy, llo slalions in group G i aro
assignod llo parliaI sums nocossary lo compulo ¸ 1, i 1 .
Wo aro nov in posilion lo slov llo dolaiIs of Ilaso 2.
Ior TasI 1, laIing lurns, llo slalions o 2i 1,
n
/
and
o 2i,
n
/
, (1 i
n
2
1), lransmil ¸ 2i 1 and ¸ 2i on
clannoI C i . Tlo primary slalion o i, 1 of oacl ¸ 2i
1, 2i (1 i
n
2
1), al IovoI 1 monilors llo clannoI C i in
lloso lvo limo sIols, picIs up ¸ 2i 1 and ¸ 2i , and
compulos ¸ 2i 1, 2i ¸ 2i 1 ¸ 2i . Noxl, llo primary
slalions of ¸ 4i 8, 4i 2 and ¸ 4i 1, 4i , (1 i
n
4
1),
lransmil ¸ 4i 8, 4i 2 and ¸ 4i 1, 4i on clannoI C i ,
rospoclivoIy. Tlo primary slalion of ¸ 4i 8, 4i picIs up
lloso lvo vaIuos, and compulos
NAKANO AND OLARÌU: ENERGY-EFFÌCÌENT ÌNÌTÌALÌZATÌON PROTOCOLS FOR SÌNGLE-HOP RADÌO NETWORKS WÌTH NO COLLÌSÌON... 855
Fig. 2. Ìllustrating the primary stations.
Fig. 3. Ìllustrating the regular stations.
¸ 4i 8, 4i ¸ 4i 8, 4i 2 ¸ 4i 1, 4i .
Tlo samo procoduro is ropoalod unliI aII llo slalions lavo
rocoivod llo nocossary informalion. CIoarIy, TasI 1 lormi-
nalos in 2/ limo sIols vill slalions o i,
n
/
boing avaIo for
ono limo sIol and llo primary slalions o i, 1 boing avaIo
for llroo limo sIols.
Il is sufficionl for TasI 2 llal oacl roguIar slalion
monilors an approprialo clannoI in an approprialo limo sIol
bocauso, in TasI 1, ovory nocossary parliaI sum is
lransmillod on a clannoI in a givon limo sIol. Tlus, in
TasI 2, ovory roguIar slalion is avaIo for ono limo sIol.
Tloroforo, Ilaso 2 is compIolod in 2/ limo sIols vill llo
primary slalions o i,
n
/
boing avaIo for al mosl four limo
sIols and llo ollor slalions boing avaIo for al mosl lvo
limo sIols.
In Ilaso 3, vo compulo llo sum of parliaI sums lo oblain
aII nocossary profix sums ¸ 1, 1 , ¸ 1, 2 , . . . , ¸ 1, / 1 . Ior
oxampIo, vo compulo ¸ 18 ¸ 9, 12 ¸ 1, 8 lo oblain llo
profix sum ¸ 1, 18 . Sinco al llo ond of Ilaso 2 ovory slalion
assignod lo a nodo Inovs llo corrosponding parliaI sums,
vo can uso llo profix sums prolocoI in Iomma 3.1. Ior
oxampIo, al llo ond of Ilaso 2, slalions o 14, 1 , o 14, 8 ,
and o 14, 4 Inov ¸ 18 , ¸ 9, 12 , and ¸ 1, 8 , rospoclivoIy.
Tlus, by oxoculing llo profix sums prolocoI of Iomma 3.1,
llo slalions o 14, 1 , o 14, 8 , and o 14, 4 can compulo llo
gIobaI profix sum ¸ 1, 18 ¸ 18 ¸ 9, 12 ¸ 1, 8 . Aflor
llo profix sums prolocoI, o 14, 4 broadcasl gIobaI profix
sum ¸ 1, 18 lo aII llo slalions in G 14 . Tlon, oacl of llom
compulos llo sum of llo gIobaI profix sum ¸ 1, 18 and ils
IocaI profix sum, vlicl is oxaclIy llo dosirod profix sum.
Sinco oacl of llo / groups las al mosl log / primary
slalions, llo profix sums prolocoI of Iomma 3.1 runs in
log / 1 limo sIols, vill no slalion boing avaIo for moro
llan lvo limo sIols. Iurllor, llo finaI broadcasl laIos ono
limo sIols. Tloroforo, Ilaso 3 runs in log / limo sIols vill
llo firsl slalions o i, 1 in oacl group boing avaIo for al
mosl lvo limo sIols, and llo ollor slalions boing avaIo for
al mosl llroo limo sIols.
Tlus, in Ilasos 1, 2, and 3 combinod, llo profix sums can
bo compulod in
n
/
8 log / 1 limo sIols vill no slalion
boing avaIo for moro llan oigll limo sIols.
ConsoquonlIy, vo lavo provon llo foIIoving rosuIl.
Lemma 3.2. 1|c prc|ix sums c| c sc¡ucncc c| n num|crs ccn |c
ccmpuic! cn c /-c|cnnc|, n-siciicn 1` uii| nc-CD in
n
/
8 log / 1 iimc s|cis, uii| nc siciicn |cing cuc|c |cr mcrc
i|cn cig|i iimc s|cis, prcti!c! i|ci /
n
logn
.
If /
n
log n
, llon group G i , (1 i /), may lavo fovor
llan log / slalions, and llo profix sums prolocoI of Iomma
3.2 doos nol vorI. In llis caso, vo uso onIy
n
log n
clannoIs
ovon if moro llan
n
logn
clannoIs aro avaiIabIo. By Iomma 3.2,
llo profix sums can bo compulod in log n 8 log / 1
n
/
C log n limo sIols. Iurllor,
n
/
8 log / 1
n
/
C log n
vlonovor /
n
log n
. Tlus, vo lavo llo foIIoving rosuIl.
CnrnIIary 3.3. 1|c prc|ix sums c| c sc¡ucncc c| n num|crs ccn |c
ccmpuic! cn c /-c|cnnc|, n-siciicn 1` uii| nc-CD in
n
/
C log n iimc s|cis, uii| nc siciicn |cing cuc|c |cr mcrc i|cn
cig|i iimc s|cis rcgcr!|css c| i|c num|cr / c| i|c c|cnnc|s.
Supposo llal llo 1` las n slalions oacl of vlicl las a
uniquo ID in llo rango 1, i , (i n). Iol o i donolo llo
slalion vill ID i, (1 i i). Nolo llal for somo i, o i may
nol oxisl. Supposo llal oacl slalion o i sloros a numbor
` i . Tlo prc|ix sum of ` i (or o i ) is llo sum of numbors
vill indicos no moro llan i, llal is,
` ¸ 1 ¸ i ano S ] oxisls .
Tlo spcrsc prc|ix sums prc||cm asIs lo compulo aII of llo
profix sums.
Wo dosign a prolocoI for soIving llo sparso profix sums
probIom in llo rocursivo mannor as foIIovs. Wlon llo
prolocoI lorminalos llo foIIoving llroo condilions aro
salisfiod:
(ps1) Ivory slalion o i , (1 i n), sloros ils profix sum.
(ps2) Tlo Iasl slalion, llal is, slalion o | sucl llal no
slalion o i vill i | oxisls, las boon idonlifiod.
(ps3) Tlo prolocoI laIos 2i 2 limo sIols and no slalion is
avaIo for moro llan 2 log i limo sIols.
If i 1, llon o 1 Inovs ` 1 and llo abovo condilions
aro vorifiod. Nov, assumo llal i 2. Iarlilion llo
n slalions inlo lvo groups
1
o i 1 i
i
2
and
2
o i
i
2
1 i i . RocursivoIy, compulo llo
sparso profix sums in
1
and
2
. By llo induclion
lypollosis, condilions (ps1)÷(ps3) abovo aro salisfiod and,
lloroforo, oacl of llo lvo subprobIoms can bo soIvod in
i 2 limo sIols, vill no slalion boing avaIo for moro llan
2 log i 2 limo sIols. Iol o ¸ and o | bo llo Iasl aclivo
slalions slalions in
1
and in
2
, rospoclivoIy. In llo noxl
limo sIol, slalion o ¸ lransmils llo sum ` i 1 i ¸
and o i oxisls on llo clannoI. Ivory slalion in
2
monilors llo clannoI and updalos accordingIy llo vaIuo
of ils profix sum. In ono addilionaI limo sIol slalion o |
rovoaIs ils idonlily. Tlo roador slouId lavo no difficuIly
confirming llal condilions (ps1)÷(ps3) abovo aro salisfiod.
Tlus, vo lavo llo foIIoving rosuIl.
Lemma 3.4. 1|c prc|ix sums prc||cm cn i|c sing|c-c|cnnc| 1`
c| n siciicns uii| uni¡uc |Ds in i|c rcngc 1, i , (i n), ccn
|c sc|tc! in 2i 2 iimc s|cis uii| nc siciicn |cing cuc|c |cr
mcrc i|cn 2 log i iimc s|cis.
To concIudo llis soclion, vo poinl oul llo roIovanco of
Iomma 3.4 lo llo iniliaIizalion probIom. Assumo llal llo
n slalions of a singIo-clannoI 1` possoss inlogor IDs in
somo rango 1 lo i vill n i, vill no lvo slalions slaring
llo samo ID. Tlo prolocoI Reduce(i) prosonlod boIov
assigns llo slalions inlogors in llo rango 1 lo n. Wo assumo
856 ÌEEE TRANSACTÌONS ON PARALLEL AND DÌSTRÌBUTED SYSTEMS, VOL. 11, NO. 8, AUGUST 2000
llal i is Inovn lo aII llo slalions. Hovovor, n nood nol bo
Inovn al llo boginning of llo prolocoI.
IrolocoI Reduce(i)
ovory slalion o i gonoralos a numbor ` i 1,
assign lo slalion o i , 1 i i , llo ID numbor
` ¸ 1 ¸ i ano o ¸ oxisls .
By Iomma 3.4, llo sparso profix sums of llo ` i s yioId
dislincl ID numbors in llo rango 1 lo n. Tlus, vo lavo
provon llo foIIoving imporlanl rosuIl.
Lemma 3.5. A sing|c-c|cnnc|, n-siciicn 1`, ccc| c| u|csc
siciicns |cs c uni¡uc |D in i|c rcngc 1 ic i, n i , ccn |c
iniiic|izc! in 2i 2 iimc s|cis uii| nc siciicn |cing cuc|c
|cr mcrc i|cn 2 log i iimc s|cis.
4 ASSIGNING TEMPORARY IDS
Tlo Ioy idoa of our iniliaIizalion prolocoI is lo assign a
(Iargo) subsol of llo n slalions of llo 1` icmpcrcrq IDs in a
(nol loo Iargo) rango 1 lo i. Onco llis is dono, vo uso
prolocoI Reduce(i) discussod in Soclion 3 lo assign llo
slalions IDs in llo rango 1 lo n.
Tlo main goaI of llis soclion is lo discuss and anaIyzo
llo porformanco of a prolocoI llal assigns lomporary IDs in
llo rango : 1 lo : i lo a subsol of llo slalions of a
singIo-clannoI, n-slalion 1`. Ior nov, assumo llal aII llo
slalions Inov : and i. Tlo dolaiIs aro spoIIod oul as
foIIovs:
IrolocoI Assign-Temporary-ID(i, :)
oacl slalion picIs an inlogor uniformIy al random in llo
rango 1 lo i and llon goos lo sIoop,
Inr i 1 tn i dn
llo slalions llal lavo picIod inlogor i vaIo up and
lransmil on llo clannoI,
iI llo slalus of llo clannoI is SINGII then
llo uni¡uc slalion llal las lransmillod rocords llo
numbor : i as ils lomporary ID
endInr
CIoarIy, prolocoI Assign-Temporary-ID(i, :) lormi-
nalos in i limo sIols and ovory slalion is avaIo for oxaclIy
ono limo sIol. Wlon llo prolocoI lorminalos, a subsol of llo
n slalions lavo rocoivod lomporary IDs in llo rango : 1 lo
: i.
Imagino llal, laIing lurns, llo n slalions picI inlogors in
llo rango 1 lo i. Il is cIoar llal llo ¸ll, 1 ¸ n , slalion
lo soIocl may ond up picIing an inlogor llal vas aIroady
closon by ono of llo ¸ 1 procoding slalions. Wo rofor lo
llis as a cc||isicn ctcni and obsorvo llal ils probabiIily is
boundod abovo by j
¸
¸ 1
i
. (To soo llis, obsorvo llal
¸ 1
i
is
llo probabiIily of a coIIision occurring in caso no coIIision
vas oxporioncod by any of llo provious slalions.) CIoarIy, a
coIIision ovonl may provonl al mosl lvo slalions from
rocoiving lloir lomporary IDs. Tlus, in llo prosonco of
a coIIision ovonls, up lo 2a slalions may faiI lo rocoivo
lomporary IDs.
Wo aro inloroslod in oblaining an uppor bound on llo
numbor of slalions llal |ci| lo rocoivo lomporary IDs al llo
ond of llo prolocoI. Ior llis purposo, Iol A
i
, 1 i n , bo
indopondonl random variabIos vill Ii A
i
1 j
i
i 1
i
and Ii A
i
0 1 j
i
. Iurllor, Iol A A
1
A
2
A
n
.
Il is vory imporlanl lo nolo llal A providos an uppcr |cun!
on llo numbor of coIIision ovonls occurring in llo procoss of
assigning n slalions inlogor IDs in an arbilrary inlorvaI of
sizo i. Tlo oxpoclod vaIuo of A is
1 A
n
i 1
j
i
n
i 1
i 1
i
n n 1
2i
8
and llo laiI of A can bo boundod using (6):
Ii A
n n 1
i
Ii A 21 A < c
1
8
1 A
c
n n 1
6i
.
9
Iqualion (9) guaranloos llal vill probabiIily oxcooding 1
c
n n 1
6i llo numbor of coIIision ovonls occurring in llo
prolocoI is no moro llan
n n 1
i
. Sinco, as vo sav, oacl
coIIision ovonl may provonl up lo lvo slalions from
rocoiving lloir lomporary IDs, vo lavo llo foIIoving rosuIl.
Lemma 4.1. Wii| prc|c|i|iiq cxccc!ing 1 c
n n 1
6i
, u|cn
prciccc| Ássign÷Temporary÷ID(i, :) icrmincics, |cucr
i|cn
2n n 1
i
siciicns |ctc nci rcccitc! i|cir icmpcrcrq |Ds.
Iol ) bo a posilivo roaI-vaIuod funclion ovor llo naluraIs.
By cloosing i lo salisfy llo condilion
n n 1
6i
ln n) n , 10
vo oblain llo foIIoving coroIIary.
CnrnIIary 4.2. Assumc i|ci ccn!iiicn 10 |c|!s. Wii| prc|c|i|iiq
cxccc!ing 1
1
n) n
, u|cn prciccc| Ássign÷Temporary÷
ID(i, :) icrmincics, |cucr i|cn
2n n 1
i
siciicns |ctc |ci|c! ic
rcccitc icmpcrcrq |Ds.
Noxl , supposo l lal vlon prol ocoI Assign-
Temporary-ID(i, :) is porformod, condilion 10 doos nol
loId. In ollor vords, vo lavo
n n 1
6i
< ln n) n .
CIoarIy, for llo random variabIo A dofinod abovo, vo
lavo
1 A
n n 1
2i
< 8 ln n) n .
Iol bo llo numbor salisfying
1 1 A 10 ln n) n .
Tlon, vo lavo
NAKANO AND OLARÌU: ENERGY-EFFÌCÌENT ÌNÌTÌALÌZATÌON PROTOCOLS FOR SÌNGLE-HOP RADÌO NETWORKS WÌTH NO COLLÌSÌON... 857
1
10 ln n) n
1 A

10 ln n) n
8 ln n) n
10
8
. 11
By using llo Clornoff bound in (5), vo can vrilo,
Ii A 10 ln n) n Ii A 1 1 A
<
c
1
1 1 A
<
8c
10
10 ln n) n
<
1
c
ln n) n
1
n) n
.
Wo jusl provod llal vill probabiIily oxcooding 1
1
n) n
llo numbor A of coIIision ovonls is boundod by
10 ln n) n . Sinco oacl coIIision ovonl may provonl up lo
lvo slalions from rocoiving lloir lomporary IDs, vo lavo
provod llo foIIoving rosuIl.
Lemma 4.3. Assumc i|ci i|c ccn!iiicn 10 !ccs nci |c|!. Wii|
prc|c|i|iiq cxccc!ing 1
1
n) n
, u|cn prciccc| Ássign÷
Temporary÷ID(i, :) icrmincics, |cucr i|cn 20 ln n) n
siciicns |ctc |ci|c! ic rcccitc icmpcrcrq |Ds.
5 INITIALIZING THE 1` WITHIN LOGARITHMIC
AWAKE TIME
Wo aro nov in a posilion lo prosonl our firsl iniliaIizalion
prolocoI for singIo-clannoI, n-slalion 1`s vill no-CD llal
lorminalos vill probabiIily oxcooding 1
1
n
in C n limo
sIols, vill no slalion boing avaIo for moro llan C log n
limo sIols. Al llis limo, vo maIo llo funclion ) n
inlroducod in Soclion 4 prociso. Ior roasons llal viII
bocomo cIoar Ialor, vo vrilo
) n log log n 8. 12
In oulIino, our iniliaIizalion prolocoI porforms ropoal-
odIy Assign-Temporary-ID(i, :) vill an approprialo
cloico of paramolors i and : lo guaranloo llal, vill
probabiIily al Ioasl 1
1
n
, aII llo n slalions rocoivo
lomporary IDs. Tlo dolaiIs of our iniliaIizalion prolocoI
aro as foIIovs.
IrolocoI Initialize(n)
: 0,
Inr i 1 tn log log n 8 dn
aII slalions vill no lomporary ID parlicipalo in prolocoI
Assign-Temporary-ID(
n
2
i 4
, :),
: :
n
2
i 4
,
endInr,
iniliaIizo aII slalions by using prolocoI Reduce(16n).
Si nco prol ocoI Assign-Temporary-ID(
n
2
i 4
, :)
lorminalos in
n
2
i 4
limo sIols vill no slalion boing avaIo
for moro llan ono limo sIol, llo Inr Ioop in prolocoI
Initialize(n) laIos aIlogollor
log log n 8
i 1
n
2
i 4
< 16n
limo sIols and no slalion is avaIo for moro llan ) n
log log n 8 limo sIols.
As vo aro aboul lo slov, vill probabiIily oxcooding
1
1
n
vlon llo Inr Ioop in prolocoI Initialize(n)
lorminalos, aII llo n slalions lavo rocoivod lomporary IDs
in llo rango 1 lo 16n. Tloroforo, by using llo prolocoI
Reduce(16n) llo slalions viII rocoivo uniquo IDs in llo
rango 1 lo n.
Our pIan is lo uso CoroIIary 4.2 and Iomma 4.3,
combinod, lo slov llal vill probabiIily oxcooding 1
1
n
,
al llo ond of log log n iloralions of llo Inr Ioop, al mosl
20 ln n) n slalions lavo faiIod lo rocoivo lomporary IDs.
Ior llis purposo, vo nood llo foIIoving loclnicaI rosuIl.
Lemma 5.1. Assumc i|ci i sciis|ics
n
2
i 1
2
2
i
20 ln n) n . Wii|
prc|c|i|iiq cxccc!ing 1
i
n) n
, ci i|c cn! c| i|c ii| iicrciicn c|
i|c |cr |ccp in prciccc| Initialize(n), ci mcsi
n
2
i 1
2
2
i
siciicns |ctc nci rcccitc! icmpcrcrq |Ds.
PrnnI. Wo provo Iomma 5.1 by llo induclion on i.
Basis: In llo firsl iloralion of llo Inr Ioop prolocoI
Assign-Temporary-ID(8n, 0)isoxoculod. Sinco, al llis
limo, i 8n, condilion 10 is salisfiod and CoroIIary 4.2
guaranloos llal vill probabiIilyoxcooding 1
1
n) n
, al llo
ond of llo firsl iloralion no moro llan
2n n 1
8n
<
n
4
n
2
0
2
2
1
slalions lavo nol rocoivod lomporary IDs.
Inductive step: Assumo llal al llo ond of llo i 1 ll
iloralion of llo Inr Ioop, vill probabiIily oxcooding
1
i 1
n) n
, a numbor n
n
2
i 2
2
2
i 1
of slalions lavo nol
rocoivod lomporary IDs.
In llo ill iloralion Assign-Temporary-ID(
n
2
i 4
,:
) is
oxoculod. If al llo boginning of llo ill iloralion
condilion 10 is nol salisfiod, Iomma 4.3 and llo
induclion lypollosis, combinod, guaranloo llal vill
probabiIily oxcooding 1
i 1
n) n
1
n) n
1
i
n) n
llo
numbor slalions llal lavo nol rocoivod lomporary IDs
al llo ond of llo ill iloralion is al mosl
20 ln n) n
n
2
i 1
2
2
i
as cIaimod.
Wo may assumo, lloroforo, llal condilion 10 loIds al
llo boginning of llo ill iloralion. Nov, CoroIIary 4.2 and
llo induclion lypollosis combinod guaranloo llal vill
probabiIily oxcooding 1
i 1
n) n
1
n) n
1
i
n) n
al llo
ond of llo ill iloralion al mosl
858 ÌEEE TRANSACTÌONS ON PARALLEL AND DÌSTRÌBUTED SYSTEMS, VOL. 11, NO. 8, AUGUST 2000
2n n 1
n
2
i 4
2n
2
i 2
2
2
i 1
n
2
i 2
2
2
i 1
n
2
i 4
2n
2
2
2i 4
2
2
i
n
2
i 4
n
2
i 1
2
2
i
slalions lavo nol rocoivod lomporary IDs.
ConsoquonlIy, vill probabiIily oxcooding 1
i
n) n
, al
llo ond of llo ill iloralion fovor llan
n
2
i 1
2
2
i
slalions lavo
nol rocoivod lomporary IDs and llo proof of Iomma 5.1
is compIolo.
Iol 1 bo llo Iargosl inlogor salisfying
n
2
1 1
2
2
1
20 ln n) n . 18
By Iomma 5.1, al llo ond of llo Ill iloralion of llo for
Ioop, vill probabiIily oxcooding 1- 1 `ovor nf(n), fovor
llan
n
2
1 1
2
21
slalions lavo faiIod lo rocoivo lomporary IDs. Iol
us foIIov llo 1 1 ll iloralion. If al llo boginning of llis
iloralion condilion 10 is salisfiod llon by (3), CoroIIary 4.2,
and Iomma 5.1 combinod, vill probabiIily oxcooding 1
1 1
n) n
al mosl
n
2
1
2
2
1 1
slalions lavo nol rocoivod lloir
lomporary IDs. Sinco 1 vas closon lo bo llo Iargosl inlogor
llal salisfios oqualion (13), vo musl lavo
n
2
1
2
2
1 1
< 20 ln n) n .
On llo ollor land, if al llo boginning of llo 1 1 ll
iloralion condilion 10 is nol salisfiod, llon (3) and
Iomma 4.3 combinod guaranloo llal vill probabiIily
oxcooding 1
1 1
n) n
al llo ond of llo 1 1 ll iloralion,
fovor llan 20 ln n) n slalions lavo faiIod lo rocoivo
lomporary IDs.
Obsorvo llal for sufficionlIy Iargo n, llo inlogor 1
spocifiod in (13) musl salisfy llo doubIo inoquaIily:
n
2
1
2
2
1 1
< 20 ln n) n
n
2
1 1
2
2
1
.
Tlo roador slouId bo in a posilion lo confirm llal llis
impIios 1 log log n for sufficionlIy Iargo n. Tlus, vo lavo
provon llo foIIoving rosuIl.
Lemma 5.2. Wii| prc|c|i|iiq cxccc!ing 1
1
n
, ci i|c cn! c|
log log n i i c r ci i cns c| i |c I nr | ccp i n pr ci ccc|
Initialize(n), |cucr i|cn 20 ln n) n siciicns |ctc
|ci|c! ic rcccitc icmpcrcrq |Ds.
By Iomma 5.2 al llo ond of llo firsl log log n iloralions
of llo Inr Ioop, al mosl 20 ln n) n slalions lavo yol lo
rocoivo lloir lomporary IDs. Iol us lraco llo oxoculion of llo
Iasl lvo iloralion of llo Inr Ioop vloro Assign-
Temporary-ID(
n
2
1 2
, :) and Assign-Temporary-
ID(
n
2
1 1
, :) aro oxoculod.
Moro spocificaIIy, vo aro inloroslod in ovaIualing llo
probabiIily llal no coIIision ovonl occurs in llo 1 2 ll or
llo 1 8 ll iloralion of llo Inr Ioop. Ior simpIicily, vo
assumo llo oxaclIy 20 ln n) n slalions sliII parlicipalo in
llo prolocoI al llo boginning of llo 1 2 ll iloralion.
Irocooding as boforo, llo probabiIily llal llo ¸ll,
1 ¸ 20 ln n) n ), slalion oxporioncos a coIIision ovonl
in oillor of llo 1 2 ll iloralion is al mosl
j
¸
¸ 1
n
2
1 2
¸ 1 2
1 2
n
¸ 1 log n
4n
.
Tlus, for oillor i 1 2 a coIIision ovonl occurs vill
probabiIily al mosl
20 ln n) n
¸ 1
j
¸
<
400 log n ln n) n
2
8n
<
ò0 log n ln n) n
2
n
sinco 1 log log n
<
1
n
1
2
foi sufficionlly laigo n .
Tlus, vill probabiIily oxcooding
1
1
n
1
2
2
1
1
n
,
lloro aro no coIIisions ovonls in al Ioasl ono of llo 1 2 ll
or llo 1 8 ll iloralions. In ollor vords, vill probabiIily
oxcooding 1
1
n
llo firsl ) n log log n 8 iloralions of llo
Inr Ioop succood in assigning lomporary IDs in llo rango 1
lo 16n lo aII llo slalions. IinaIIy, by appIying prolocoI
Reduce(16n) ovory slalion rocoivos a dislincl ID numbor
in llo rango 1 lo n. Moroovor, by Iomma 3.5, llis Iallor lasI
is compIolod in 82n 2 limo sIols vill no slalion boing
avaIo for moro llan 2 log 16n 2 log n 8 limo sIols.
ConsoquonlIy, prolocoI Initialize(n) lorminalos in al
mosl 16n 82n 2 48n 2 limo sIols vill no slalion
boing avaIo for moro llan 2 log n 8 log log n 8
2 log n log log n 11 limo sIols. Tlo foIIoving lloorom
summarizos our findings.
Thenrem 5.3. Wii| prc|c|i|iiq cxccc!ing 1
1
n
, prciccc|
Initialize(n) iniiic|izcs c sing|c-c|cnnc|, n-siciicn IN
uii| nc-CD in ci mcsi C n iimc s|cis uii| nc siciicn |cing
cuc|c |cr mcrc i|cn C log n iimc s|cis.
6 INITIALIZING THE 1` WITHIN DOUBLY
LOGARITHMIC AWAKE TIME
Tlo main goaI of llis soclion is lo slov llal ono can modify
prolocoI Initialize(n) discussod in Soclion 5 lo roduco
llo numbor of avaIo limo sIols por slalion lo al mosl
8 log log n 1ò.
RocaII llal llo Inr Ioop in prolocoI Initialize(n)
laIos 16n limo sIols aIlogollor and no slalion is avaIo for
moro llan ) n log log n 8 limo sIols. Moroovor, vill
probabiIily oxcooding 1
1
n
, vlon llo Inr Ioop lorminalos,
aII llo slalions lavo rocoivod lomporary IDs. Our anaIysis
in llo provious soclion rovoaIod llal llo buII of llo 2 log n
log log n 11 limo sIol avaIo limo oxporioncod by llo
slalions sloms from porforming prolocoI Reduce(16n). Our
pIan is lo laIo advanlago of llo parlicuIar dislribulion of a
(donso) subsol of llo lomporary IDs lo roduco llo numbor of
avaIo limo sIols.
NAKANO AND OLARÌU: ENERGY-EFFÌCÌENT ÌNÌTÌALÌZATÌON PROTOCOLS FOR SÌNGLE-HOP RADÌO NETWORKS WÌTH NO COLLÌSÌON... 859
6.1 Taking Advantage of a Uniform Distribution
To givo llo roador an idoa of our approacl, vo nov slov
lov ono ccu|! iniliaIizo llo slalions vill lomporary IDs in
llo rango 1 lo 8n. Tloroaflor, vo viII procood lo sluffIo
lomporary IDs and viII appIy llo samo loclniquo lo aII llo
slalions.
RocaII llal, as slovn in Soclion 5, vill probabiIily
oxcooding 1
1
n log log n
, as a rosuIl of running prolocoI
Assign-Temporary-ID(8n, 0), al Ioasl
8n
4
slalions lavo
boon assignod lomporary IDs in llo rango 1 lo 8n.
Moroovor, il is oasy lo soo llal llo rosuIling lomporary
IDs como from an undorIying uni|crm !isiri|uiicn ovor llo
rango 1 lo 8n. ConsoquonlIy, for an arbilrary fixod i,
1 i 8n , llo probabiIily llal somo slalion las rocoivod
lomporary ID i is al Ioasl
8n
4
8n
8
82
.
Nov, parlilion slalions vill IDs in llo rango 1 lo 8n inlo
n
4 ln n
groups G 1 , G 2 , . . . , G
n
4 lnn
sucl llal group G i ,
1 i
n
4 ln n
, invoIvos llo slalions vill IDs in llo rango
82 ¸ 1 ln n 1 lo 82¸ ln n. Sinco oacl group G i invoIvos
slalions vill IDs in llo rango of 82 ln n inlogors, llo
provious discussion guaranloos llal G i is omply vill
probabiIily al mosl 1
8
82
82 lnn
. Nov Iomma 2.1 aIIovs us
lo vrilo
1
8
82
82 ln n
< c
8 lnn
1
n
8
.
By (3), il foIIovs llal vill probabiIily oxcooding
1
1
4n
2
ln n
, oacl of llo
n
4 lnn
groups G 1 , G 2 , . . . , G
n
4 lnn
conlains al Ioasl ono slalion.
Il is naluraI, al llis poinl, lo iniliaIizo llo slalions in oacl
group G i , 1 i
n
4 lnn
, individuaIIy, using prolocoI
Reduce(82 ln n). By Iomma 3.5, for oacl group G ¸ , llis
lasI lorminalos in al mosl 64 lnn 2 limo sIols vill no
slalion boing avaIo for moro llan 2 log 82 ln n
2 log ln n 10 < 2 log log n 10 limo sIols. Tlo ovoraII num-
bor of limo sIols is
n
4 lnn
64 lnn 2 16n
n
2 lnn
. Al llis
poinl, oacl slalion in group G i , 1 i
n
4 lnn
, las a
IocaI ID villin G i . IinaIIy, llo profix sums of
G 1 , G 2 , . . . , G
n
4 lnn
aro compulod. CIoarIy, vo can
uso llo firsl slalion in oacl group G i for llis compulalion.
Sinco, vill probabiIily oxcooding 1
1
n
, nono of llo groups
is omply, vo can uso llo profix-sums prolocoI of Iomma 3.1.
Tlis laIos
n
4 lnn
1 limo sIols vill llo firsl slalion in oacl
group boing avaIo al mosl lvo limo sIols. Aflor llal, oacl
slalion updalos ils IocaI ID in llo obvious vay lo oblain ils
finaI ID. Tlus, vo provod llal by laIing advanlago of lloir
uniform dislribulion, llo slalions vill lomporary IDs in llo
rango 1 llrougl 8n can bo iniliaIizod, vill probabiIily
oxcooding 1
1
n
, in 16n
n
2 lnn
n
4 lnn
1 16n
n
4 ln n
1
limo sIols, vill no slalion boing avaIo for moro llan
2 log log n 18 limo sIols.
6.2 HandIing the GeneraI Case
UnforlunaloIy, vo cannol uso llo loclniquo of Subsoclion
6.1 lo iniliaIizo llo slalions vloso lomporary IDs aro in llo
rango 8n 1 llrougl 16n, bocauso llo lomporary IDs in llis
rango aro nci uniformIy dislribulod. In parlicuIar, vo
cannol guaranloo llal in ovory group of consoculivo c ln n
inlogors, al Ioasl ono las boon assignod as llo lomporary ID
of somo slalion. Ior convonionco, vo Iol
Grnup A. Donolo aII llo slalions vloso lomporary IDs aro
bolvoon 1 and 8n, and
Grnup B. Donolo aII llo slalions vill lomporary IDs
bolvoon 8n 1 and 16n.
To ovorcomo llo aforomonlionod difficuIly, vo uso llo
foIIoving lricI. Wo ¨sluffIo¨ llo lomporary IDs in llo
vloIo rango 1 lo 16n sucl llal
 for 1 i 8n, i 2i 1, and
 for 8n 1 i < 16n, i 2i 16n.
In ollor vords, aflor llo sluffIing, llo originaI IDs of
slalions in Group A bocomo c!! numbors sproad ovor llo
vloIo rango 1 lo 16n, vliIo llo originaI IDs of slalions in
Group B bocomo ctcn numbors sproad ovor llo vloIo rango
1 lo 16n. Il is cIoar llal ovory slalion can compulo ils nov ID
and llal no lvo slalions viII lavo llo samo ID al llo ond of
llo sluffIing.
Tlo Ioy insigll is llal sinco llo IDs of slalions in Group
A voro uniformIy dislribulod ovor 1 lo 8n, llo nov IDs of
slalions in Group ¹ aro sliII uniformIy dislribulod ovor llo
rango 1 lo 16n.
Nov, parlilion llo slalions in llo rango 1 lo 16n inlo
n
4 ln n
groups G 1 , G 2 , . . . , G
n
4 lnn
of 64 ln n inlogors oacl,
sucl llal group G i , 1 i
n
4 ln n
, invoIvos slalions vill
IDs in llo rango 64 ¸ 1 ln n 1 lo 64¸ ln n. Sinco aII llo
slalions in G i aro incIudod in G i , oacl of llo
n
4 ln n
groups
G 1 , G 2 , . . . , G
n
4 ln n
conlains al Ioasl ono slalion vill
probabiIily oxcooding 1
1
4n
2
lnn
.
As boforo, vo iniliaIizo llo slalions in oacl group G ¸ ,
1 ¸
n
4 ln n
, individuaIIy, using prolocoI Reduce(64 lnn).
By Iomma 3.5, for oacl group G ¸ , llis lasI lorminalos in
al mosl 128 lnn 2 limo sIols vill no slalion boing avaIo
for moro llan 2 log 64 lnn 2 log lnn 12 < 2 log log n
12 limo sIols. Tlo ovoraII numbor of limo sIols is
n
4 lnn
128 lnn 2 82n
n
2 ln n
. Al llis poinl, oacl slalion
in group G ¸ , 1 ¸
n
4 lnn
, las a IocaI ID villin G ¸ .
IinaIIy, llo profix sums of G 1 , G 2 , . . . , G
n
4 lnn
aro
compulod. CIoarIy, vo can uso llo firsl slalion in oacl
group G ¸ for llis compulalion. Sinco vill probabiIily
oxcooding 1
1
n
, nono of llo groups aro omply, vo can uso
860 ÌEEE TRANSACTÌONS ON PARALLEL AND DÌSTRÌBUTED SYSTEMS, VOL. 11, NO. 8, AUGUST 2000
llo profix-sums prolocoI of Iomma 3.1. Tlis laIos
n
4 lnn
1
limo sIols vill llo firsl slalion in oacl group boing avaIo
for al mosl lvo limo sIols. Aflor llal, oacl slalion updalos
ils IocaI ID in llo obvious vay lo oblain llo dosirod finaI ID.
Tlis impIios llal vo can iniliaIizo aII llo slalions in 82n
n
2 ln n
n
4 ln n
1 82n
n
4 lnn
1 limo sIols vill no slalion
boing avaIo for 2 log log n 1ò limo sIols.
RocaIIing llal llo Inr Ioop in prolocoI Initialize (n)
laIos 16n limo sIols aIlogollor and llal no slalion is avaIo
for moro llan log log n 8 limo sIols, vo lavo provon llo
foIIoving imporlanl rosuIl.
Thenrem 6.1. Wii| prc|c|i|iiq cxccc!ing 1
1
n
, c sing|c-
c|cnnc|, n-siciicn 1` uii| nc-CD ccn |c iniiic|izc! in
C n iimc s|cis, uii| nc siciicn |cing cuc|c |cr mcrc i|cn
C log log n iimc s|cis.
7 SIMULATION RESULTS
Wo lavo simuIalod llo iniliaIizalion prolocoI prosonlod in
Soclion 6 in ordor lo gol a fooI for lov voII il porforms in
praclico. Sinco prolocoI Reduce 64 ln n is dolorminislic, vo
lavo focusod on llo Inr Ioop. Tlo fundamonlaI molivalion
for llis is llal sinco llo running limo of prolocoI
Reduce 64 ln n can bo porcoivod as a fixod ovorload, llo
moro slalions gol lloir lomporary IDs in oacl iloralion of
llo Inr Ioop, llo faslor viII llo prolocoI lorminalo.
Will llis in mind, vo lavo simuIalod llo Inr Ioop in
prolocoI Initialize(n) for n 10, n 100, n 1000,
n 10, 000, n 100, 000, and n 1, 000, 000.
Ior oacl vaIuo of n as abovo, vo lavo run llo simuIalion
100, 000 limos. Wo voro parlicuIarIy inloroslod in llo
numbor of slalions llal faiIod lo rocoivo lomporary IDs al
llo ond of parlicuIar runs of Assign-Temporary-ID. Wo
rofor llo roador lo TabIo 1, vloro llo rosuIls of llo
simuIalion aro summarizod.
Ior oxampIo, laIing n 100, llo Inr Ioop invoIvos
log log n 8 ò.78 6 iloralions corrosponding, ro-
spoclivoIy, lo
 Assign-Temporary-ID (800, 0),
 Assign-Temporary-ID (400, 800),
 Assign-Temporary-ID (200, 1200),
 Assign-Temporary-ID (100, 1400),
 Assign-Temporary-ID (ò0, 1ò00), and
 Assign-Temporary-ID (2ò, 1òò0).
As monlionod, in ordor lo gol llo rosuIls in llo labIo, llis
vas llon ropoalod 100, 000 limos. Tlo simuIalion slovod
llal aflor laving run Assign-Temporary-ID(800, 0) an
avorago of 11.6ò slalions faiIod lo rocoivo lomporary IDs.
IiIoviso, aflor llo subsoquonl oxoculion of Assign-
Temporary-ID(400, 800), onIy 0.8ò slalions lavo nol
rocoivod lomporary IDs. Aflor oxoculing Assign-
Temporary-ID (200, 1200) onIy 0.0026 slalions voro yol
lo rocoivo lomporary IDs. IinaIIy, aII llo slalions rocoivod
lloir lomporary IDs al llo ond of Assign-Temporary-ID
(100, 1400).
Tlo rosuIls in TabIo 1 slov llal our prolocoI porforms in
praclico ovon bollor llan vlal llo anaIylicaI modoI
prodiclod.
8 INITIALIZING A /-CHANNEL 1` WITH NO-CD
Tlo main goaI of llis soclion is lo oxlond llo iniliaIizalion
prolocoI prosonlod in Soclion 6 lo llo muIliclannoI caso.
Considor a /-clannoI, n-slalion 1` vill clannoIs
C 1 , C 2 , . . . , C / .
Il is slraigllforvard lo soo llal prolocoI Assign-
Temporary-ID(
n
2
i 4
, :) vlon porlod lo / clannoIs, in llo
obvious vay, lorminalos in
n
2
i 4
/
limo sIols, vill no slalion
boing avaIo for moro llan ono limo sIol. Tlo Inr Ioop in
prolocoI Initialize(n) laIos aIlogollor
log log n 8
i 1
n
2
i 4
/
<
log log n 8
i 1
n
2
i 4
/
1 <
16n
/
log log n 8
limo sIols, and no slalion noods lo bo avaIo for moro llan
log log n 8 limo sIols.
Al llis poinl, aII llo slalions lavo rocoivod lomporary
IDs in llo rango 1 llrougl 16n. Tlo idoa nov is lo uso /
clannoIs lo porform llo foIIoving lasIs:
Task 1. IniliaIizo oacl of llo groups G 1 , G 2 , . . . , G
n
4 ln n
individuaIIy, and
Task 2. Compulo llo profix sums of
NAKANO AND OLARÌU: ENERGY-EFFÌCÌENT ÌNÌTÌALÌZATÌON PROTOCOLS FOR SÌNGLE-HOP RADÌO NETWORKS WÌTH NO COLLÌSÌON... 861
TABLE 1
The Number of Stations That Have Not Received Temporary ÌD after Each Ìteration: the Average of 100, 000 Executions
G 1 , G 2 , . . . , G
n
4 ln n
.
As discussod in Subsoclion 6.2, using ono clannoI, llo
slalions in oacl group G i , 1 i
n
4 ln n
, can bo iniliaIizod
in 128 lnn 2 limo sIols vill no slalion boing avaIo for
moro llan 2 log log n 12 limo sIols. If /
n
4 ln n
by assigning
llo
n
4 ln n
groups lo llo / clannoIs as ovonIy as possibIo,
TasI 1 can bo porformod in 128 ln n 2
n
4/ ln n
82n
/
n
2/ lnn
limo sIols vill no slalion boing avaIo for moro llan
2 log ln n 12 limo sIols. If /
n
4 lnn
, llon vo uso
n
4 lnn
clannoIs, assign ono clannoI lo oacl group, and iniliaIizo
ovory group in 128 ln n 2 limo sIols simuIlanoousIy. Tlus,
TasI 1 can bo dono in C
n
/
log n limo sIols vill no slalion
boing avaIo for moro llan 2 log log n 12 limo sIols.
By CoroIIary 3.3, TasI 2 can bo compIolod in 2
n
4/ ln n
C log
n
4 ln n
n
2/ lnn
C log n limo sIols vill no slalion
boing avaIo for moro llan oigll limo sIols, rogardIoss of
llo numbor / of clannoIs. Tlus, vo lavo provon llo
foIIoving imporlanl rosuIl.
Thenrem 8.1. Wii| prc|c|i|iiq ci |ccsi 1
1
n
, i|c /-c|cnnc|
n-siciicn 1` ccn |c iniiic|izc! in C
n
/
log n iimc s|cis uii|
nc siciicn |cing cuc|c |cr mcrc i|cn C log log n rcgcr!|css c|
i|c num|cr / c| c|cnnc|s.
9 CONCLUDING REMARKS AND OPEN PROBLEMS
Our main conlribulion vas lo proposo onorgy-officionl
iniliaIizalion prolocoIs for individuaI cIuslors of a radio
nolvorI. Iirsl, vo slovod llal if llo numbor n of slalions is
Inovn boforoland, llo singIo-clannoI 1` can bo iniliaIizod
by a prolocoI llal lorminalos, vill probabiIily oxcooding
1
1
n
, in C n limo sIols vill no slalion boing avaIo for
moro llan C log n limo sIols.
Noxl, vo slovod llal llis iniliaIizalion prolocoI can bo
impIomonlod in sucl a vay llal vill probabiIily oxcooding
1
1
n
, llo prolocoI lorminalos in al mosl C n limo sIols,
and no slalion las lo bo avaIo for moro llan C log log n
limo sIols.
IinaIIy, vo addrossod llo muIliclannoI caso and slovod
llal if /, / 1 , clannoIs aro avaiIabIo, an n-slalion 1` can
bo iniliaIizod, vill probabiIily oxcooding 1
1
n
, in C
n
/
log n limo sIols, vill no slalion boing avaIo for moro llan
C log log n limo sIols.
A numbor of probIoms romain opon. Iirsl, llo simuIalion
rosuIls prosonlod in Soclion 7 soom lo indicalo llal our
prolocoIs lorminalo mucl faslor llan vlal our proofs slov.
Il vouId bo inlorosling lo soo if llo conslanls oxlibilod in
llis papor can bo improvod upon, porlaps by a ligllor
anaIysis.
Moro imporlanlIy, lovovor, al llis limo vo do nol Inov
lov lo dosign onorgy-officionl iniliaIizalion prolocoIs for
llo caso vloro llo numbor n of slalions is noillor Inovn
nor a good approximalion lloroof is avaiIabIo. Tlis
promisos lo bo an oxciling aroa for furllor rosoarcl.
ACKNOWLEDGMENTS
Tlis vorIs vas supporlod in parl by ONR granl N00014-
97-1-0526.
REFERENCES
|1] N. AIon, A. Bar-Noy, N. IiniaI, and D. IoIog, ¨SingIo-Round
SimuIalion on Radio NolvorIs,¨ j. A|gcrii|ms, voI. 13, pp. 188÷210,
1992.
[2] D.}. BaIor, ¨Dala/Voico Communicalion ovor a MuIlilop, MobiIo,
Higl Iroquoncy NolvorI,¨ Prcc. M|ICOM ´º¯, pp. 339÷343, 1997.
[3] N. Bambos and }.M. RuInicI, ¨MobiIo Iovor Managomonl for
WiroIoss Communicalion NolvorIs,¨ Wirc|css Nciucr|s, voI. 3,
pp. 3-14, 1997.
[4] R. Bar-Yoluda, O. GoIdroicl, and A. Ilai, ¨Ifficionl ImuIalion of
SingIo-Hop Radio NolvorI vill CoIIision Doloclion on MuIlilop
Radio NolvorI vill No CoIIision Doloclion,¨ Disiri|uic!
Ccmpuiing, voI. 5, pp. 67÷71, 1991.
[5] R. Bar-Yoluda, O. GoIdroicl, and A. Ilai, ¨On llo Timo-
CompIoxily of Broadcasl in MuIlilop Radio NolvorIs: An
IxpononliaI Gap Bolvoon Dolorminism and Randomizalion, j.
Ccmpuicr cn! Sqsicms Scicnccs, voI. 45, pp. 104÷126, 1992.
[6] U. BIacI, Mc|i|c cn! Wirc|css Nciucr|s. Uppor SaddIo Rivor, N.}.:
Ironlico-HaII, 1996.
[7] }.I. Bordim, }. Cui, T. Hayasli, K. NaIano, and S. OIariu, ¨Inorgy-
Ifficionl IniliaIizalion IrolocoIs for Ad-Hoc Radio NolvorIs,¨
Prcc. |ni´| Sqmp. A|gcrii|ms cn! Ccmpuiciicn, pp. 215÷224, Doc.
1999.
[8] I.H. Dana, ¨Tlo Goograplor's Crafl Irojocl,¨ Dopl. of Goograply,
Univ. of Toxas, Auslin, Sopl. 1999. lllp://vvv. uloxas. odu/
dopls/grg/gcraf/nolos/gps/gps.llmI.
[9] W.C. Iifor and I.}. Bruno, ¨Iov Cosl IacIol Radio,¨ Prcc. ||||,
voI. 75, pp. 33-42, 1987.
[10] M. GorIa and T.-C. Tsai, ¨MuIlicIuslor, MobiIo, MuIlimodia Radio
NolvorI,¨ Wirc|css Nciucr|s, voI. 1, pp. 255-265, 1995.
[11] I.I. Harris and K.W. Warron, ¨Iov Iovor ToclnoIogios: A
Syslom Iorspoclivo,¨ Prcc. 1|ir! |ni´| Wcr|s|cp Mu|iimc!ic Ccmm.,
1996.
[12] T. Hayasli, K. NaIano, and S. OIariu, ¨Randomizod IniliaIizalion
IrolocoIs for IacIol Radio NolvorIs,¨ Prcc. 13i| |ni´| Pcrc||c|
Prcccssing Sqmp., pp. 544-548, Apr. 1999.
[13] I. D. KapIan, Un!crsicn!ing GPS. Princip|cs cn! App|icciicns.
Boslon: Arlocl Houso, 1996.
[14] C.R. Iin and M. GorIa, ¨Adaplivo CIusloring in MobiIo WiroIoss
NolvorIs,¨ |||| j. Sc|ccic! Arccs in Ccmm., voI. 16, pp. 1,265-1,275,
1997.
[15] R. Molvani and I. Raglavan, Rcn!cmizc! A|gcrii|ms. Cambridgo
Univ. Iross, 1995.
[16] K. NaIano, ¨OplimaI IniliaIizing AIgorillms for a RoconfigurabIo
Mosl,¨ j. Pcrc||c| cn! Disiri|uic! Ccmpuiing, voI. 24, pp. 218-223,
1995.
[17] K. NaIano and S. OIariu, ¨Randomizod C log log n -Round Ioador
IIoclion IrolocoIs in IacIol Radio NolvorIs,¨ Prcc. |ni´| Sqmp.
A|gcrii|ms cn! Ccmpuiciicn, pp. 209-218, 1998.
[18] K. NaIano, S. OIariu, and }. I. Sclving, ¨Broadcasl-Ifficionl
IrolocoIs for Ad-Hoc Radio NolvorIs vill Iov ClannoIs,¨ ||||
1rcns., Pcrc||c| cn! Disiri|uic! Sqsicms, voI. 12, pp. 1,276-1,289,
1999.
[19] M. }oa-Ng and I.-T. Iu, ¨A Ioor-lo-Ioor Zono-Basod Tvo-IovoI
IinI Slalo Rouling for MobiIo Ad-Hoc NolvorIs,¨ |||| j. Sc|ccic!
Arccs in Ccmm., voI. 17, pp. 1,415-1,425, 1999.
[20] B. IarIinson and S. GiIborl, ¨NAVSTAR: GIobaI Iosilioning
Syslom÷Ton Yoars Ialor,¨ Prcc. ||||, pp. 1,177-1,186, 1983.
[21] R.A. Iovors, ¨Ballorios for Iov-Iovor IIoclronics,¨ Prcc. ||||,
voI. 83, pp. 687-693, 1995.
[22] R. Ramanallan and M. Sloonslrup, ¨HiorarclicaIIy-Organizod,
MuIlilop WiroIoss NolvorIs for QuaIily-of-Sorvico Supporl,¨
Mc|i|c Nciucr|s cn! App|icciicns, voI. 3, pp. 101-119, 1998.
[23] D. Rayclaudluri, N.D. WiIson, ¨ATM-Basod Transporl Arcliloc-
luro for MuIlisorvico WiroIoss ICN,¨ |||| j. Sc|ccic! Arccs in
Ccmm., voI. 12, pp. 1,401-1,414, 1994.
862 ÌEEE TRANSACTÌONS ON PARALLEL AND DÌSTRÌBUTED SYSTEMS, VOL. 11, NO. 8, AUGUST 2000
[24] A.K. SaIIinlzis and C. Clamzas, ¨An In-Band Iovor-Saving
IrolocoI for MobiIo Dala NolvorIs,¨ |||| 1rcns. Ccmm.,
pp. 1,194÷1,205, 1998.
[25] R. Sancloz, }. Ivans, and G. Mindon, ¨NolvorIing on llo
BallIofioId: ClaIIongos in HiglIy Dynamic MuIlilop WiroIoss
NolvorIs,¨ Prcc. |||| M|ICOM ´ºº, Ocl. 1999.
[26] K. SivaIingam, M.B. Srivaslava, and I. AgravaI, ¨Iov Iovor IinI
and Accoss IrolocoIs for WiroIoss MuIlimodia NolvorIs,¨ Prcc.
|||| Vc|icu|cr 1cc|nc|cgq Ccn|., (V1C ´º¯), May 1997.
[27] M. Slomm, I. Gaullior, and D. Harada, ¨Roducing Iovor
Consumplion on NolvorI Inlorfacos in Hand-HoId Dovicos,¨
Prcc. 1|ir! |ni´| Wcr|s|cp Mu|iimc!ic Ccmm., 1996.
[28] }.I. WiosoIllior, G.D. Nguyon, and A. Iplomoridos, ¨MuIlicasling
in Inorgy-Iimilod Ad-Hoc WiroIoss NolvorIs, ¨ Prcc.
M|ICOM ´º8, 1998.
Koji Nakano, received the BE, ME, and PhD
degrees from Osaka University, Japan, in 1987,
1989, and 1992, respectively. From 1992-1995,
he was a research scientist at Advanced
Research Laboratory, Hitachi Ltd. Since 1995,
he has worked at Nagoya Ì nsti t ute of
Technology, Japan. Ìn 1999, he was an
adjunct associate professor of Old Dominion
University. He is currently an associate profes-
sor with the Department of Electrical and
Computer Engineering, at the Nagoya Ìnstitute of Technology.
Dr. Nakano has published in the following publications: IEEE
Transactions on Parallel and Distributed Systems, Journal of Parallel
and Distributed Computing, Theoretical Computer Science, Theory of
Computing Systems, Parallel Algorithms and Applications, and IEICE
Transactions. He serves on the editorial board of International Journal of
Foundations of Computer Science. He served as an editor of the Journal
of Information Processing Society of Japan. He has served as a
program chair or committee member of several conferences and
workshops, including the Ìnternational Conference on Parallel Proces-
sing, the Ìnternational Conference on Parallel and Distributed Comput-
ing and Systems, the Reconfigurable Architecture Workshop, and the
Workshop on Advances in Parallel and Distributed Computational
Models. His research interests include parallel algorithms and archi-
tectures, mobile computing, computational complexity, and graph
theory. He is a member of the ÌEEE Computer Society.
Stephan OIariu received the MSc and PhD
degrees in computer science from McGill Uni-
versity, Montreal, in 1983 and 1986, respec-
tively. Ìn 1986, he joined the Computer Science
Department at Old Dominion University, where
he is now a professor.
Dr. Olariu has published extensively in
various journals, book chapters, and conference
proceedings. His research interests include
wireless networks and mobile computing,
parallel and distributed systems, performance evaluation, and medical
image processing.
Dr. Olariu serves on the editorial board of several archival journals,
including IEEE Transactions on Parallel and Distributed Systems,
Journal of Parallel and Distributed Computing, International Journal of
Foundations of Computer Science, International Journal of Computer
Mathematics, VLSI Design, and Parallel Algorithms and Applications.
He is a member of the ÌEEE Computer Society.
NAKANO AND OLARÌU: ENERGY-EFFÌCÌENT ÌNÌTÌALÌZATÌON PROTOCOLS FOR SÌNGLE-HOP RADÌO NETWORKS WÌTH NO COLLÌSÌON... 863

You might also like