ContentslistsavailableatScienceDirect
Ad Hoc Networks
journalhomepage:www.elsevier.com/locate/adhoc
An adaptive, energy-aware and distributed fault-tolerant topology-control algorithm for heterogeneous wireless sensor networks
Fatih Deniz
a,∗, Hakki Bagci
a, Ibrahim Korpeoglu
b, Adnan Yazıcı
aa Department of Computer Engineering, Middle East Technical University, Ankara, Turkey
b Department of Computer Engineering, Bilkent University, Ankara, Turkey
a rt i c l e i n f o
Article history:
Received 2 December 2015 Revised 20 February 2016 Accepted 22 February 2016 Available online 3 March 2016 Keywords:
Topology control Fault-tolerance Energy efficiency Prolonged network lifetime k -connectivity
Heterogeneous wireless sensor networks
a b s t ra c t
Thispaperintroducesanadaptive,energy-awareand distributedfault-toleranttopology- control algorithm, namelythe AdaptiveDisjoint PathVector (ADPV) algorithm, forhet- erogeneouswirelesssensornetworks.Inthisheterogeneousmodel,wehaveresource-rich supernodes aswell as ordinarysensor nodes thataresupposedto beconnectedto the supernodes.UnlikethestaticalternativeDisjoint PathVector(DPV)algorithm,the focus ofADPVistosecuresupernode connectivityinthepresenceofnodefailures, andADPV achieves thisgoalby dynamicallyadjustingthesensor nodes’transmission powers.The ADPV algorithminvolvestwo phases:a single initializationphase, which occursat the beginning,andrestorationphases,whichareinvokedeachtimethenetwork’ssupernode connectivityisbroken.Restorationphasesutilizealternativeroutesthatarecomputedat theinitializationphasebythehelpofanoveloptimizationbasedonthewell-knownset- packingproblem.Throughextensivesimulations,wedemonstratethatADPVissuperiorin preservingsupernode connectivity.Inparticular,ADPVachieves thisgoaluptoafailure of95%ofthesensor nodes;whiletheperformance ofDPVislimitedto 5%.Inturn,by ouradaptive algorithm,weobtainatwo-foldincreaseinsupernode-connected lifetimes comparedtoDPValgorithm.
© 2016 Elsevier B.V. All rights reserved.
1. Introduction
Wireless sensor networks (WSNs) are typically com- posedof largenumbers oftiny sensor nodesthat areca- pable of sensing, processing and transmitting data over wireless channels. Such networkscan be used innumer- ous fields, such as battlefield surveillance [1–3], environ- mental monitoring [4–6] and traffic control [7–9]. Sen- sor nodes collaborate in a distributed, autonomous and
∗ Corresponding author. Tel.: +90 5531866846.
E-mail addresses: fatih.deniz@tcmb.gov.tr , fatihdeniz@gmail.com (F. Deniz), hakkibagci@gmail.com (H. Bagci), korpe@cs.bilkent.edu.tr (I. Korpeoglu),yazici@ceng.metu.edu.tr (A. Yazıcı).
self-organizedmanner to accomplish a certain task,usu- allyinanenvironmentwithnoinfrastructure.
Sensornodes inWSNsshould be low-cost and should have small form-factor. This restricts sensor nodes in many ways as they have limited energy, short transmis- sionrange, relativelyslow CPU andsmallmemory. These limitations bring out many challenges unique to WSNs, suchasvery low powerconsumption.Since sensornodes are battery powered and these batteries are usually not rechargeable,comingupwithsolutionsthatreduceenergy consumption and prolong network lifetime are very im- portant. Numerous studies address this problem [10–13]
in literature. According to Li andMohapatra [14], 90% of asensornetwork’senergyisstillavailable afterfirstnode
http://dx.doi.org/10.1016/j.adhoc.2016.02.018 1570-8705/© 2016 Elsevier B.V. All rights reserved.
dies.Despitethissubstantialamountofremaining energy, theexistence ofhighly-loadedandbottlenecknodescause earlynetworkdemise.Therearenumerousstudiesthatad- dressbalancing energyconsumption amongnodesto en- sure that all nodes will run out of energy at about the sametime[15].Low-energyAdaptiveClusteringHierarchy (LEACH) [16] is a well-known early study that uses dy- namictransmissionrangesto betterbalance theloadand prolongnetworklifetime.Therearealsosomerecentstud- ieswhichre-establishlostconnectivitybyadjustingtrans- mission ranges. CoRAD [17] and RESP [32] can be listed assome of those studies. With the recent developments inthehardwareofWSNs,dynamictransmissionrangeas- signmenthasbecomeevenmoreeffective[18].
Fault-tolerance is another critical issue in WSNs. Due totheerror-pronenatureofwirelesscommunication,links mayfail,packetscan getcorrupted orcongestionmayoc- cur [19,20]. There are also factors that cause long-term faultsinsensornodes,suchasenergydepletion,hardware failure, link breaks, malicious attacks. Multi-hop commu- nication multiplies the chances of faulty incidents for a packet stream traveling from a source to a destination.
Therefore,fault-tolerancemethods,includingfault-tolerant topologycontrol,areessentialforimprovingWSNreliabil- ityaswellasnetworklifetime.
As stated by Liu et al. [21], most existing works on fault-toleranttopology-controlaimtoobtaink-vertexcon- nectivitybetweenanytwosensornodes,wherethetopol- ogyisguaranteedtoremainconnecteduntilthefailureof thekthsensornode.
In this study, the focus is on two-tiered heteroge- neous WSNs, where the network consists of two differ- ent typesof nodes: resource-rich supernodesand simple sensornodeswithlimitedbattery power.In thisnetwork model,sensor nodes are connectedto the set of supern- odesvia multi-hop paths. Toreflect thisasymmetry,[22]
proposesk-vertexsupernodeconnectivity,whereeachsen- sor is connected to at least one supernode by k vertex- disjointpaths.Insuchtopologies,thesensornodesremain connectedtothesupernodesaslongasatmostk− 1sen- sornodesfail.
Moststudiesonfault-toleranceproposestaticsolutions, that is, they do not adapt the topology to the changing network conditions. Bagci et al. [23] propose a static al- gorithmcalledtheDisjoint PathVector(DPV) tooptimize total transmissionpower fora givenk-vertex supernode- connectednetwork.Thatstudydoesnot considerresidual batteryenergyanddisregardsthe unbalancedloaddistri- bution on sensor nodes. As a result, k-vertex supernode connectivity is achieved but may not be preserved for a sufficientamountoftime.
In this study, we propose a novel adaptive and dis- tributedtopology-controlalgorithm,AdaptiveDisjointPath Vector (ADPV), which efficiently constructs a k-vertex supernode-connected network topology and adapts the topologytonodefailures,whichinturnincreasesnetwork lifetime. The contribution is two-fold. First, the residual batterypowerlevelsofindividualsensornodesareconsid- eredto prolong k-vertex supernode connectivity. Second, anadaptivesolutionisproposedtorestore,ifnecessary,k- vertexsupernodeconnectivityafteranodefailure.
The remainder of the paper is organized as follows:
Section 2 gives some background information and dis- cusses the related studies. In Section 3, we present our proposed adaptive topology-control solution. The results forsimulationexperimentsare presentedinSection4.Fi- nally,Section5concludesthepaper.
2. Relatedwork
Inthissection,wegiveabriefoverviewofsomeofthe prominentrecentworkaddressingfault-tolerance,connec- tivityrestorationandheterogeneityinWSNs.Wealsogive abriefoverviewoftheDPValgorithm[23].
Fault-tolerancetechniquescanbe categorizedintofour [24]:prevention,detection,isolation,andrecovery.Preven- tion attains network connectivity and establishes redun- dant links/nodeswhen necessary.Detection monitorstraf- ficandsendsalertswhenanyindicationoffaulthappens, such as a decrease in packet delivery rate, which would imply a packet loss,interruption, ordelay. Isolation diag- nosesandidentifiesthealert.Asforrecovery,afterdetect- ingandidentifyingthefault,thesystemshouldbeableto recoverineitheracentralizedordistributedmanner.Note that dueto thenature ofWSNsit isessential forthe re- coveryschemetobeadistributedmethod.
The replication and redundancy of components prone tofailureisthemostcommonlyusedmethodforfaultpre- ventionandrecovery[21].Forinstance,ifsomenodeshave problemsandfailtosensetheenvironment,theredundant nodesinthevicinitycanstillprovidedata.Keepingredun- dant links ormultiple paths also provides fault-tolerance when some communicationlinksarebroken dueto node failuresorcommunicationerrors.
2.1. ConnectivityrestorationinWSNs
There are three approachesto connectivity restoration in WSNs: mobile node relocation, relay node placement andtopology-controlviatransmissionrangeadjustment.In thefirstapproach,asthenodesaremobile,themainidea istorepositiontheexistingalivenodestorestoreconnec- tivity.OneexampleofthismethodisPADRA,developedby Akkayaetal.[25].Inthisapproach,eachnodechoosesone ofitsneighborstobe thefailurehandler,whichwillstart recovery ifthenodedies.Therestorationprocessonlyoc- cursiftheentirenetworkgetsdisconnected,inwhichcase theclosestnodethatcantakethedeadnode’splaceisre- locatedtothatposition.
Intherelaynode placementapproach[26–31],theob- jectiveistoplaceaminimumnumberofrelaynodesina regionwheresensornodesarerandomlydeployedsothat theresultingnetworktopologyisfault-tolerant.
These first two approaches, that is, mobile node relo- cation andrelaynode placement,maynot be practical in real-world scenarios because sensor nodes are often de- ployedinremote andinhospitable regionswithharshen- vironments that render manual node placement or relo- cation infeasible.Note that duetothe dynamicnature of WSNs,nodeplacementand/orrelocationmustberepeated periodically. In addition, these approachesrequire overall
Fig. 1. Initial network.
Fig. 2. Optimized network.
network information,something that is also not suitable formostreal-worldapplications.
As a remedy to the problems discussed above, topology-control emerges asathird approachforconnec- tivity restoration. In this approach, the topology is con- trolledbyadjustingthesensornodes’transmissionranges.
One example of this method is RESP [32], which is an energy-aware topology-control algorithm that ensures k- edge connectivity for flat networks. The RESP algorithm assumes sensor nodes are aware of their location in- formation via GPS or other localization techniques, and periodicallyupdatesthenetworktopologytoadapttosen- sor nodes’ residual battery power levels. Because of en- suring k-edge connectivity, but not k-vertex connectiv- ity, RESPcannot keep the networkconnected up tok− 1 node failures.Anotherrecentapproach,Energy-harvesting Heterogeneous WSN (EHWSN) [33], also aims to pre- serve k-vertex supernode-connectivity for heterogeneous WSNs.EHWSNisacentralizedapproachandignoresresid- ual battery power levels, therefore not scalable and not energy-aware.
2.2. DPValgorithm
The aimofthe DPValgorithm [23] istominimize the total transmission power of a WSN while maintaining k- vertexdisjointpaths fromeach sensornode to theset of supernodes.TheDPValgorithmgetsak-vertexsupernode- connected network topology asan input andgenerates a subnetworkconsistingofthesamesetofsensorsbutfewer connections. The output of the DPV algorithm is a to- taltransmissionpoweroptimizedandk-vertexsupernode- connectednetwork topology.Considertheexample topol- ogy giveninFig. 1,whichconsists ofone supernode and threesensornodes.Whentheaimistoprovideone-vertex supernodeconnectivity,DPVremovesthreeedgesandop- timizesthegivennetworktopology,asinFig.2.Themain contributionofDPVisitsefficiencyincomputingsuchnet- worktopologies.TheDPValgorithmrequiresO(n2)mes- sage transmissions, whereas the best alternative [22] in- curs O(5) messages, where n is the number of sensor nodes and refers to the maximum degree of a sensor
node.Note thatwe assume adense network,where is sufficientlylarge.TheDPV algorithmconsistsoffivemain stages:
1.Collecting path information and calculating disjoint paths,
2.Calculatingthesetofrequiredneighbors,
3.Notifying thenodesinthe disjointpaths andupdating therequiredneighbors,
4.Removingthenon-requiredneighborsand
5.Reducing the powerlevel to apoint sufficient only to reachthefarthestrequiredneighbor.
2.3.Powerconsumptionmodel
OurADPV algorithm aims at prolonging network life- time, andthus it should first model the amount oftime untilthebatterypowersofthesensornodesaredepleted.
The ADPValgorithm uses a well-known powerconsump- tion model, proposed by Heinzelman et al. [34,35]. This approachisbasedontheobservationthatthemainfactor inWSNpowerconsumptionisdatacommunication,which consistsoftwo factors:datatransmissionanddatarecep- tion.In thismodel,the power totransmit a bitto a dis- tanceofdis
Pt
(
d)
=α
1+α
2× dn, (1)where
α
1andα
2areparametersthatdependonthetrans- mitter circuitry, and n is the path loss exponent for the environment,whichoftenhasavaluebetween2and4.In ourpowerconsumptionmodel,α
1,α
2,andnareassumed tobe50nJ/bit,100pJ/bit/m2and2,respectively.Inour model,theenergyconsumption fordata recep- tionisaconstantvalueperbit.Werepresentthisconstant with
β
andassumeitequals50nJ/bit.Forour experiments, we assume all sensor nodes are sensingthe environment andgeneratingtraffic at afixed rate.Wealsoassumethat dataaggregationisappliedand thatallnodesonapathcarrythesameload.Therefore,to- talpowerconsumptionforreceivingabitandtransferring ittothenexthopequals:
Pf
(
d)
=β
+α
1+α
2× dn. (2)Iftheresidualbatteryenergylevelofsensornodei isde- notedasei,thenthelifetimeofnodeiequals:
li=ei/
((
rri×β )
+(
rri+ rgi)
×( α
1+α
2× din))
, (3) whererriistheincomingdataratetonodei,rgiisthedata rategeneratedinnodeianddiisthetransmissionrange.3. Adaptivedisjointpathvectoralgorithm
Inthissection, wepresentournoveladaptiveanddis- tributed algorithm, ADPV, which aims to construct and maintainak-vertexsupernode-connectedtopologytopro- longthek-vertexsupernode-connectedlifetimeofthenet- work.TheADPValgorithmcontrolsthetopologybyadjust- ingthetransmissionrangesofsensornodes,andtocomply withreal-lifesituations itconsiders node failures.Theal- gorithm requires only one-hop neighborhood information andconstructs the network topology by a series ofmes- sageexchanges.
Fig. 3. Sample scenario.
TheADPV algorithmconsists oftwo phases:initializa- tionandrestoration.Itcollects necessaryinformationand builds an initial topology during the initialization phase.
Whenevera node failure breaks k-vertex supernode con- nectivity, ADPV restores connectivity within the restora- tionphase.SimilartoDPV,ADPVutilizesdisjointpathsand within each restoration phase each sensor node decides whetherornotto changeitsdisjointpaths.Attheendof eachrestoration phase, sensornodes’transmission ranges areadjustedaccordingtotheintendedtopology.Themain differencesbetweenADPVandDPVareasfollows:
• ADPVisanadaptiveapproach,adaptingtonodefailures and remaining energy levels, whereas DPV is a static one.
• ADPVconsidersresidualbatterypowerlevelsofsensor nodes,thereforeitisanenergy-awaresolution.DPVon theotherhandignoressensornodes’remainingenergy levels.
• ADPV balances energyconsumptionandoptimizes the lifetimeofdisjointpaths,asopposedtoDPV,whichop- timizesthetotaltransmissionpowerofsensornodes.
• ADPV significantly prolongs both one-vertex and k- vertex supernode-connected lifetimes of the network with its solutions for restoration path selection, k- vertexsupernode-connectivityverificationandconnec- tivityrestoration.
3.1. Networkmodel
Consider a mission critical border surveillance system thatisintegratedwithatwo-tieredheterogeneouswireless sensornetwork.Inthisnetwork,thereare supernodeslo- catedoneachtowerandregularsensornodesthatareuni- formlydistributedintothe targetarea asshowninFig.3. In thisnetwork, sensor nodes are responsible for detect- ingpotentialintrusionactivities andinformthetowersby forwardingdatatothesupernodeslocatedatthosetowers.
Sinceitiscommontolosesome sensornodesbecauseof energydepletion, harshenvironmental conditionsorhos- tile activities of intruders, it is desired for every sensor nodetohavemorethanacertainnumberofindependent
paths to the supernodes. Inthe figure, we cansee a sol- dier crossing the border, and a sensor node close-by in- formssometowersviathreedisjointpaths.
Thisnetwork modelisfirstdescribedin[22],andalso used by the DPV algorithm. In this model, the network consistsof Msupernodesthat are deployedatknownlo- cations andNsensornodesthat arerandomly distributed in the 2D plane so that M < <N. We assume the su- pernodes have transmission ranges long enough to com- municate with the base station or any other supernode in the network. Therefore, we do not model andare not concerned with supernode-to-supernode communication.
We areonly interestedinsensor-to-sensorandsensor-to- supernodecommunication.
We represent the initial network topology with an undirected weightedgraphG=(V,E),where Vis the set of nodesand E=
{ v
i,v
j | dist(v
i,v
j)<Rmax}
is theset ofedges; dist(
v
i,v
j) defines the distance between nodesv
iand
v
j.3.2. Problemdefinition
We firstgive the formal definitionofk-vertex supern- odeconnectivity.
Definition 1 (k-vertex supernode connectivity [22]). An heterogeneous WSN is said to be k-vertex supernode- connected ifremoval ofany k− 1 sensornodes doesnot disconnect any sensor node from all the supernode(s), that is, each sensor node is still connected to some supernode(s).
Initially we are given a k-vertex supernode-connected network with M supernodes and N sensor nodes, where thesensornodetransmissionrangecanbeadjustedupto apredefinedconstantRmax.Aswemodelnodefailures,the number ofactive sensor nodes decreases duringthe net- worklifetime.WeuseNttodenotethesetofactivesensor nodesattimet,wheretimeisrepresentedbydiscretetime intervals. Our problem is to determine the transmission rangesofallactivesensornodesatanytime,suchthatthe resultingtopologyisstillk-vertexsupernode-connected,so that networklifetimecan beimproved.Now,we formally statetheproblemofmaximizingfault-tolerantlifetime.
Definition 2 (Fault-tolerantlifetime maximization). Given a k-vertex supernode-connected WSN G=(V,E) with a set M⊂ Vof supernodevertices andaset Nt⊂ Vofactive sensor node vertices, such that M∩Nt=∅, find a set of edges F⊂ E such that G(V,E− F) is k-vertex supernode- connected and|Nt|
i=1li is maximized, where li is the life- time of the minimum lifetime path among the disjoint pathsof
v
∈Nt.3.3. Residualbatterypowerlevel-awaredisjointpath selection
The ADPV algorithm adapts the network topology dy- namicallyduringnetworkoperationby adjustingthe sen- sor nodes’ transmission ranges according to residual en- ergy levels. For instance, if a node has low remaining energy, it should choose closer neighbors; otherwise, it
maychoosefartherones. Inthisway,weattainafairdis- tributionoftotalresidualenergyamongsensornodes.
TheDPValgorithmisnotanenergy-awaresolution,and ignores sensor nodes’ residual energy levels. This design maycauseearly batterydepletion,since anode withlow residual energy may be assigned to a high transmission powerrange.TheADPValgorithm,ontheotherhand,takes residual energy levels into consideration when selecting disjointpaths.Estimatingthelifetimeofeachsensornode onapathliesatthecoreofourapproach.Themotivation behindthismethodisthatachainisonlyasstrongasits weakestlink, andthus apath survivesonlyaslongasall nodessurviveinthepath.Therefore,theshortestnodelife- time onthepathdeterminesthelifetimeofthepath.The ADPV algorithm chooses aset ofdisjoint paths such that theminimumlifetimeofthosepathsismaximized.
Weformallydefinethelifetimeofapathasfollows:Let apathP consistsofnodesn0,n1,..,nl,inwhichn0 isthe startingsensor node andnl is asupernode. Letbidenote the residualenergylevelofsensor node ni anddidenote thedistancebetweenniandni+1foreach0≤ i<l.Then, thelifetimeofPisdefinedas:
Lifetime
(
P)
=min0≤i<l
{
bi/( β
+α
1+α
2× din) }
, (4) whereβ
,α
1,α
2 and n are the constant parameters of powerconsumption,definedinSection2.3.3.4. Initializationphase
Thissectiondescribesourproposedapproachforselect- ing alternativeroutesintheinitializationphaseofthe al- gorithm,wherethoseroutesaretobeusedtorestorecon- nectivity duringrestoration phases. InADPV, each sensor node keepsalternativeroutes, herereferred toasrestora- tionpaths,thatstartwiththatnode.
Theprimary goalistoconsumetheminimumpossible resources while attaining high-quality restoration paths.
Theresourcesincludememory,CPU,andnetwork.Regard- ingmemory,forinstance,ifallpossiblepathsfromsensor nodestosupernodeswere held,thememory requirement wouldbeintractable.In[36],Valiantdiscussestheaverage numberof paths froma node toa givenset ofnodes.In terms of CPU,Bagci et al.[23] show that the complexity ofselectingkdisjointpathsfromapoolofpalternativesis O(pk).Therefore,withahighernumberofrestorationpaths ofsizer,ittakeslongerto computeadisjointpath setof size k duringeach restoration phase. As forthe network, which is last butnot least, we aim to communicate us- ing minimumnumber ofmessages. Eachrestoration path incurs communicationbetween its nodes inorder to up- date its lifetime. As a result, we should maintain a very restrictedsetofrestoration pathsforthe sakeofnetwork performance, but atthe same time, the amount ofthose pathsshouldbehighenoughtorestoreconnectivitywhen- everneeded.
Toovercometheserestrictions andefficientlyconstruct restoration paths, ADPV employs a well-known method, called maximum set packing (MSP) [37]. This method is the optimization version of theset packing (SP) problem and asks for the maximum number of pairwise disjoint
setsamongafamilyofsets.Moreformally,foragivenuni- verseU andafamilyS ofsubsetsofU,MSPisasubfamily C⊆ S ofsets such that all setsinC arepairwise disjoint, andC uses asmany sets aspossible, so that the size of thepacking
C
ismaximum.MaximumsetpackingisNP-
hard[38]andcannotbeapproximatedwithinanyconstant factor[39].
Algorithm3.1 MaximumSetPacking(MSP).
Input: S Output: M
1: M←∅;
2: whileS=∅ do
3: m←MinIntersectingPath(S);
4: M←M
m;
5: for allPathp ∈S do
6: if p
m=∅ then
7: S←S− p;
8: endif
9: end for
10: endwhile
There is a well-known greedy heuristic, shown in Algorithm 3.1, to solve the MSP problem and it runs in polynomial time. We employ this heuristic to construct restorationpaths.Atthebeginning,wehaveapoolofcan- didate paths of a relatively large size. The heuristic per- forms with many iterations, where each iteration selects themostdiversepathfromthepool.Weusethetermdi- verseasbeingdisjointwithothers,thatis,theone thatis disjoint to the largest number of paths among others in the pool. We add the selected path into the restoration pathset andremove all thepaths fromthe poolthat in- tersectwiththeselectedpath.Theiterationscontinueun- tilthepool becomesempty orthe numberof restoration pathsreachesapredefinedthreshold.Sincetheinitialsen- sornodeandthedestinationsupernodedonotviolatedis- joincy,ADPV representseach pathby the setof itsinter- mediatesensornodes.
Algorithm3.2 PathInformationCollectioninADPV.
Input: I,L,k Output: D,R
1: T←∅;
2: R←∅;
3: forallreceivedPathInfomessageI do
4: D←MinDisSet(T,k);
5: c←Cost(D);
6: U←I.T∪T;
7: R←MaxSetPacking(R∪U); (Algorithm3.1)
8: U ←MaxSetPacking(U);
9: Sort(U );
10: T ←
{
pi∈U|
i<=L}
;11: D ←MinDisSet(T ,k);
12: c ←Cost(D );
13: if c <cthen
14: T←T ;
15: TransmitPathInfo(T);
16: end if
17: endfor
Algorithm 3.2 shows path-information-collection and restoration-path-selection procedures. The variables
Table 1 ADPV notations.
I Received pathInfo message
L Maximum number of paths to be stored k Disjoint connectivity degree
R Set of restoration paths T and T Set of local paths D and D Set of disjoint paths
c and c Cost of disjoint paths, which equals the minimum lifetime of the disjoint paths U Union of two path sets
S Set of paths
M Set of paths in MSP m , p , r Variables referencing paths sr Supernode ratio
n Total number of sensors
Maximum degree of a node r Amount of a node’s restoration paths l Average path length in the restoration set n 0 , n 1 , .., n i Number of remaining sensor nodes after each
restoration phase
used in the pseudo codes are defined in Table 1. In Algorithm3.2,eachsensornodemaintainsalocalpathset alongwithdisjointandrestoration pathsets.Asan input, the algorithm takes a ‘PathInfo’ message that contains the localpath set of the sender node and generates two outputs, whichare the disjointpath andrestoration path sets of size k and a relatively large size, respectively.
Local paths are logical paths that are used for informing neighbor nodes about the paths they can use over the sender node. Therefore, local paths have a very critical roleindetermining disjointandrestorationpath setsand need to be selected very carefully. When a sensor node receives a ‘PathInfo’ message containing a local path set, itfirst calculates the unionof the sender’s andreceiver’s localpathsets.ItthenexecutestheMSPprocedureonthis uniontoeliminatepathsthathavetoomanysensornodes in common. The procedure then determines a candidate localpath set T asthe first Lminimum-cost pathof the remainingset.Whiledoingso,theprocedurealsoupdates therestoration paths by executing theMSP procedure on the union of local sets and the current restoration path set.
Using the candidate local path set, the set of dis- joint paths with minimum cost is calculated using Algorithm3.3. In thisalgorithm, all disjoint subsetswith kelementsare traversed andthe onewiththe minimum costisselected.Iftheminimum-costdisjointpathsethas asmallercostthanthecurrentdisjointpathset,boththe disjointand the local paths are updated and a ‘PathInfo’
messagecontainingthe newlocalpath setis transmitted tothesetofneighbors.Thisprocess continuesuntilthere arenomoreupdatesinthedisjointpathsets.
Afterdetermining thedisjointpaths,each sensornode determines its required neighbors, which include the neighborsthatdisjointpathsusetheedgesbetween.After determiningtherequiredneighbors,eachnode adjustsits transmissionpowerto reach itsfarthest neighboraccord- ingtotheresultingtopology.
Algorithm 3.3 Finding Disjoint Paths to Supernodes (MinDisSet).
Input: T andk Output: D
1: D←∅;
2: if
|
T|
>kthen3: Q←
{
q⊂ T| |
q|
=k}
;4: c←∞;
5: qmin←∅;
6: forallq∈Q do
7: if qconsistsof disjointpathsthen
8: if Cost(q)<cthen
9: c←Cost(q);
10: qmin←q;
11: end if
12: end if
13: endfor
14: D←qmin;
15: end if
3.5. Connectivityrestorationphase
We start the connectivity restoration procedure only when k-vertex supernode connectivity is broken due to node failure. Thus, the first step after a node failure is tocheck whetherthenetworkisstillk-vertexsupernode- connectedornot.As thisisacostly operation[40],ADPV employsasimpledistributedgreedyheuristicwithnofalse positives. That is, if ADPV postulates the network is k- vertexsupernode-connected,thenthenetworkisdefinitely connected. However, the network can still be connected even if ADPV claims it is not. Therefore, ADPV ensures strongk-vertexsupernodeconnectivity.
When a node failure occurs, ADPV ensures all the node’sneighborsinitiateafailuremessagetoinformothers aboutthefailure. Uponreceivingafailuremessage,asen- sornoderemovesallpathsincludingthefailednodefrom its restoration set. Sincefrequent transmission powerad- justmentisdifficulttorealizeinpractice,weemployperi- odicaltransmissionpowercontrol,andduringeach period wecheckwhetheranyfailednodesexistonanyofthedis- jointpaths.Ifafailednodedisconnectsadisjointpath,the restoration process takes place.Notethat this eventdoes not necessarily imply k-vertex supernode disconnectivity, yet because ADPV takes early action it never allows the connectivity to break. After deciding k-vertex supernode connectivity must be restored, ADPV applies a two-step process:updatingthelifetimesoftherestorationpathsand computingminimum-costdisjointpathsfromtherestora- tionset.
In the first step, path lifetimes in the restoration set areupdatedviamessagestransmittedalongthepathfrom the source node tothe destination supernode. Eachnode redirects a received messageto the next hopin thepath and returnsa message that containsupdated lifetime in- formationofthesensornodesbackalongthatpath.Inthe second step, minimum-cost disjoint paths are computed using the previously discussed disjoint-path-selection al- gorithm, Algorithm3.3. An overview of the connectivity- checkingandconnectivity-restorationproceduresaregiven inAlgorithm3.4.
x y
A z Supernode Sensornode
Residual Energy: 100
Consumpon: 1.2 Residual Energy: 100 Consumpon: 2
Residual Energy: 100 Consumpon: 1
x x->A x->y->A
Disjoint Paths y y->A y->x->A
z z->y->A z->x->A
a
x
A z
Residual Energy: 40 Consumpon: 1.2
Residual Energy: 50 Consumpon: 5
x x->A x->z->A
Disjoint Paths
z z->A z->x->A
b
x
A
Residual Energy: 28 Consumpon: 1
x x->A Disjoint Paths
c
Fig. 4. Sample connectivity restoration for k = 2 .
Algorithm3.4 ConnectivityRestorationinADPV.
Input: k, R,D Output: D
1: FailedNodes←∅;
2: forall receivednodefailuremessage
δ
do3: forallPathr ∈R do
4: if r contains
δ
.FailedNodethen5: R←R− r;
6: end if
7: endfor
8: FailedNodes←FailedNodes∪
δ
.FailedNode9: ifcertaintimeelapsed sincelastperiod then
10: for allPathp ∈D do
11: if
(
p∩FailedNodes)
=∅then12: UpdateCosts(R);
13: D←MinDisSet(R,k);
14: break;
15: endif
16: end for
17: FailedNodes←∅;
18: endif
19: end for
For instance, continuing from the example given in Section2.2,fork=2,ADPVoptimizesthetopologyshown inFig.1,asinFig.4(a).In thistopology,allinitial energy levelsareequal.Assumingthedatagenerationrateisuni- formfor all nodes,thepower consumption ofnodes x, y andzare 1.2,2and1,respectively. Withthispower con- sumption, node ydies first(100/2=50 slater), both node xandnodezloseoneoftheirdisjointpathsandthenet- workbecomesone-vertex,butnot two-vertex,supernode- connected.TheADPValgorithmrestoresconnectivity,asin Fig. 4(b),by adjustingthe transmissionrange ofz,which introduces alinkfromnodez tosupernodeA.Because of its increasing power consumption, node z happens to be thesecond dyingnode (50/5=10 slater)andthusnode x losestwo-vertexsupernodeconnectivityoncemore.How- ever, because it has no alternative routes, it adjusts its transmissionpoweragainandworksasaconnectednode, as in Fig. 4(c), for the rest of its life (28/1=28 more seconds). For this network, the two-vertex supernode- connectedlifetimeisbroken whennodezdies.Therefore,
ni
Fig. 5. One-hop neighbors of node i .
the two-vertex supernode-connected lifetimeof this net- workequals60sandtheone-vertexsupernode-connected lifetimeequals88s.
Lemma1. Theconnectivity restoration processof ADPVen- suresk-vertexsupernodeconnectivity.
Proof. Bydefinition,thenetworkgetsk-vertexsupernode- connected ifeach sensor in the network is connected to atleast one supernode withk-vertex disjoint paths. This translatesintothedisjointpathsetofeachsensornodeof beingsize k, andifthere exist more thank paths in the restorationset,ADPVchoosesadisjointsetandensuresk- vertexsupernodeconnectivity.
Lemma2. Inthe restoration pathset,there areatmost paths,whereisthemaximumdegreeofasensornode.
Proof. Wearegoingtoprovethisbycontradiction.Asdis- cussedinSection3.5,eachsensornodekeepsamaximum setpackofsomesizeintheirrestorationsets,sothateach pathinthesetispairwisedisjointwiththeothers.Let denotethemaximumdegreeofa nodeandassume there exists a node, say node i, that has more than paths inits restoration set.Since there are more than paths that are usingat most neighbors,according tothe pi- geonholeprinciple, there existtwo restoration paths that usethesameneighbor.LetFig.5representnode iandits one-hopneighbors.Iftheneighborthattwopaths havein commonisasupernode,thennodeiwillhaveexactlythe same two paths in its restoration set, which is not pos- sible,because the MSP procedure calculatesthe union of theselected pathsto guaranteediversity. Ifthatneighbor
is a sensornode, those paths will not be disjoint, which violatestheMSPdefinition.Therefore,noneighbor,neither sensornode nor supernode,can havetwo paths incom- mon, andthe number ofelements in the restoration set cannotexceed.
3.6.Runningtimeanalysis
Wecomputethelifetimeofarestorationpathviames- sagestransmittedalongthepath fromthesourcenode to thedestinationsupernode. Eachnoderedirects areceived messageto the next hop inthe pathand returnsa mes- sage that containsthe sensors’updated lifetimeinforma- tionbackalongthesamepath.Therefore,foreachrestora- tion path,the number ofmessages equals two times the lengthofthepath.Weassumethatpathlengthisbounded by a constant, say l,following previous studies [23]. No- tice that the number of restoration paths is less than or equalto , where is the maximumdegree of a node.
Then, there are at most l × messages in total, and thus,themessagecomplexityisO()ateachconnectivity- restorationphase.Intheworstcase,foreachsensornode, connectivity restoration is carried out for O() times,as the restoration path set embodies at most l × nodes.
Therefore, at each sensor node, total message complex- ity becomes O(2) for connectivity restoration. For path informationcollection,ADPV hasthe messagecomplexity of O(n), which also equals that ofDPV [23]. Therefore, thetotal message complexitybecomes O(2) +O(n) = O(n).
TheADPValgorithmconsumescomputationalpowerin the initialization phase for disjoint and restoration path constructionandduringtheconnectivityrestorationphase for determining new disjoint paths from the restoration set.Duringtheinitializationphase,whensensornodesre- ceivea‘PathInfo’message,theycalculatetheunionofthe local path information and the received paths in the in- comingmessage.Therunningtimecomplexityofthisstep dependson thenumberofpaths (p)in thelocalpathin- formationtable. InADPV, since themaximum numberof pathsthatcan bestoredina sensornode’spathinforma- tion table isset to a constant value, both calculating the unionofthe two pathinformationtablesand sortingthe pathsaccordingtotheircoststakeconstanttime.
In the initialization phase, there are two more pro- cedures that consumes processing power: maximum set packinganddisjoint-path-selectionalgorithms.Thegreedy heuristic forMSP,shown inAlgorithm 3.1,is usedtwice:
onceforconstructingtherestorationpathandagainforse- lectingthelocalpathinformationtable.Asdiscussedinthe secondlemma,thenumberofrestoration pathsislimited bythe maximumdegreeofnode (), andthenumberof pathsinthelocalpath informationtable isaconstant (l).
Therefore,the MSP algorithmconsists ofnumerous itera- tions, each consistingoftwo steps:i) selecting themini- mumintersectingpathandii)removingthepathsthatin- tersect withit. In thelatter step, thealgorithm traverses all path pairs and determines the minimum intersecting one. The activity of removing the intersecting paths also traversesthe set once more. Considering set size is rep- resented by s, the running time complexity of the MSP
algorithm equals O(s2+s). Therefore, the MSP running time complexityineach stepisO(2++l2+l), which canbereducedtoO(2).
To calculate the minimum disjoint set, Algorithm 3.3 enumeratesallsubsetsofsizekandfindsthesetwiththe minimumcost. EnumeratingallthesesubsetstakesO(pk), whereprepresentsthe numberofpathsinthe givenset.
Since the input of the minimum disjoint set procedure is the localpath information table, which hasa constant number ofelements, therunning time complexity ofthe minimumdisjointsetcalculationisalsoaconstant.
Considering that the messagetransmission complexity of ADPVis O(n) andthedominatingstep (MSP) isexe- cutedonceforeveryincomingmessage,therunningtime complexityofthetotalinitializationphaseisO(n3).
Fortherestorationphases,asdiscussedabove,themax- imum number of restoration phases a node can execute isO(l),andineach phasetherearetwo operations:up- datingpathcosts,whichonlyusesmessagetransmissions, andcalculatingtheminimumdisjointsetfromtherestora- tion set. Since the maximum number ofelements inthe restorationsetis,therunningtimecomplexityfordeter- miningtheminimum-costdisjointpathsfromtherestora- tion set will be O(k), and the total running time com- plexityoftherestorationphaseswillbeO(k+1).
Sincen>>,andthecommonlyacceptedvaluesofk are2and3[22],theADPVrunningtimecomplexityequals O(n3).
3.7. ExpectednumberofrestorationsinADPV
In this section we discusstheoretical expectationsre- sulting from theADPV algorithm andanalyze how many times ADPV can restore k-vertex supernode connectivity fora givennode. SinceADPV can restore suchconnectiv- itywhen thereare atleastk pathsin therestoration set, we willdetermine the expected number ofnode failures beforeanodecannotrestore itsconnectivity.Letndenote the number of sensor nodesin the network and assume the sensor node batteries deplete uniformlyin any order withthesame probability
ρ
.The parametersused inthissectionaregiveninTable1.
Considering that the number of sensor nodes in the restoration setequals r× l,theexpectednumberof sen- sornodesthat diebeforeoneoftheser× lsensornodes diesequals:
n
r× l. (5)
Forexample, ifthereare 100 nodesin theentire net- work and20takepartintherestorationset,thentheex- pected number of node failures before one ofthe nodes in therestoration setfailsequals five.Whena node ona pathdies,thenthatpathwillnolongerbevalidandthere- fore will be removed from the restoration sets available.
As a result, with a node failure, the number of restora- tion pathswilldiminishby one.Therefore,whenthefirst node on a restoration set dies, r− 1 paths, which con- sist of (r− 1)× l sensornodes, will remain. At the same time, thenumberofremainingsensornodesintheentire
networkwillequal:
n− n
r× l. (6)
Continuing from the previous example, 100− 5=95 sensor nodeswill remain in theentire network afterthe firstnodeintherestorationsetdies.Theremainingsensor nodesaftertheithrestorationpathremovalcanbegener- alizedasfollows:
ni+1=ni− ni
(
r− i)
× l, (7)whichalsoequals:
ni+1=ni×
1− 1
(
r− i)
× l(8) andwhichcanalsobewrittenasaproductof:
ni+1=n×
i j=0
1− 1
(
r− j)
× l . (9)Since ADPV can restore k-vertex supernode connectivity whenthereare atleastk pathsintherestorationset,the numberofsensornodeswhenk-vertexsupernodeconnec- tivity ofthe givennode cannot be restoredequals nr−k+1 andcanbecalculatedas:
nr−k+1=n×
r−k
j=0
1− 1
(
r− j)
× l . (10)Thenbychangingtheparametertot=r− j, nr−k+1=n×
r
t=k
1− 1 t× l
. (11)Accordingtotheaboveformula,thenumberofsuccess- ful restorations will be proportional to the sensor node count. Also, withthe increasing average path length, the numberofremainingsensorsincreases,whichinturnde- creases the possibility of successful restorations. There- fore, choosing paths with smaller path lengths may be preferable.
4. Experimentsandresults
In thissection we report ourmeasurements regarding lifetime and other metrics for the DPV and ADPV algo- rithms andtry to evaluate ADPV’s success. For thiseval- uation, we implemented ADPV using an extended ver- sionof acustom simulator,which hasalsobeenused for evaluating the DPV algorithm. We added a time dimen- sionandabatterymodelintotheexistingframework and thusprovidedanenvironmentthatcouldevaluatenetwork lifetime.
4.1. Experimentalsetup
In our experiments, we assumed that sensor nodes and supernodes are uniformly and randomly deployed in an area of 600 m x 600 m and that the initial maximum transmission range Rmax of the sensor nodes is set at 100 m. We repeated our experiments for
{
100,150,...,500}
sensornode, for k=2,3(as theseareTable 2
Simulation parameters.
Deployment area 600 m x 600 m
Initial transmission range of sensor nodes: R max 100 m Number of sensor nodes: N [10 0 . . . 50 0]
Number of supernodes: M 5% and 10% of N
Degree of disjoint connectivity: k 2 and 3
Packet loss rate 10%
commonly accepted k values), and for a supernode ratio (sr) of5%and10% overtheregion.Finally,we assumeda packetlossrateof10% foreach messagetransmission.As aresult,wehad9× 2× 2experimentalinstances,andon each weexecuted both algorithms 20timesandreported the averages. Oursimulation parameters are summarized inTable2.
4.2.Results
InFig.6,wecomparethenodefailuretoleranceofDPV and ADPV. For each algorithm, we measure performance in terms of the fraction of dead sensor nodes when the networkgets (i) supernodedisconnected and(ii)k-vertex supernode disconnected. If there exists a path (single or multi-hop)betweenasensornodeandanyoneofthesu- pernodes, then the sensor node is said to be connected.
Ifevery (alive)sensornode inthe networkhask disjoint paths to theset ofsupernodes, then the network iscon- sideredask-vertexsupernode-connected.Withthesemea- surements we determine the maximum number of node failures that can occur before supernode connectivity is broken. Here,we observethe moststrikingresult, andat thesametime,evidenceofthisstudy’smotivationregard- ingtheinstabilityofstaticalgorithms andeffectivenessof ADPV for keeping the network supernode-connected. As seeninthefigure,evenbeforethefailureof5%ofthesen- sornodes,thenetwork’ssupernodeconnectivityisbroken whenweemploytheDPValgorithm.Thisresultlimitsus- ingDPVasafault-tolerantalternative.Ontheotherhand, ADPVsuccessfullykeepsthenetworksupernode-connected uptofailureofabout95%ofthesensornodes.
In the figure, we observethat when the network be- comes denser, ADPV keeps it supernode-connectivity for longer. This result can be attributed to ADPV becoming moreeffectiveatfindingalternative routesdueto thein- creasingnumberofsensornodes.Forinstance,inoneex- treme,when we examine the results ofa 500-node net- work for k=2 and sr=10%, as shown in Fig. 6(a), we see that the network is still supernode-connected up to a failure of 95% of the sensor nodes. In the other ex- treme, where the number of initially deployed sensor nodes equals 100, ADPV sustains supernode connectivity uptothefailure of20%ofthesensornodes.Looking into eachofthesub-figures,whentheinitialnumberofsensor nodesisbetween250and300,wenoticethat ADPVsuc- ceedsinkeepingsupernodeconnectivityevenaftertheac- tivesensornodesarehalved.Consideringallexperimental instances,onaverage,ADPVmaintainssupernode connec- tivityuptoafailureof52%ofsensornodesfork=2,and 55% of sensor nodes for k=3. Since the optimized net- work topologiesfor k=3 contain moreconnections, it is
0,00%
10,00%
20,00%
30,00%
40,00%
50,00%
60,00%
70,00%
80,00%
90,00%
100,00%
100 150 200 250 300 350 400 450 500
Node Failure Tolerance (%)
Number of Sensor Nodes ADPV Connected
ADPV 2-Connected DPV Connected DPV 2-Connected
(a) k=2, sr=10%
0,00%
10,00%
20,00%
30,00%
40,00%
50,00%
60,00%
70,00%
100 150 200 250 300 350 400 450 500
Node Failure Tolerance (%)
Number of Sensor Nodes ADPV Connected
ADPV 2-Connected DPV Connected DPV 2-Connected
(b) k=2, sr=5%
0,00%
10,00%
20,00%
30,00%
40,00%
50,00%
60,00%
70,00%
80,00%
90,00%
100,00%
100 150 200 250 300 350 400 450 500
Node Failure Tolerance (%)
Number of Sensor Nodes ADPV Connected
ADPV 3-Connected DPV Connected DPV 3-Connected
(c) k=3, sr=10%
0,00%
10,00%
20,00%
30,00%
40,00%
50,00%
60,00%
70,00%
100 150 200 250 300 350 400 450 500
Node Failure Tolerance (%)
Number of Sensor Nodes ADPV Connected
ADPV 3-Connected DPV Connected DPV 3-Connected
(d) k=3, sr=5%
Fig. 6. Percentage of failed sensor nodes when the network becomes (i) supernode disconnected and (ii) k -vertex supernode disconnected.
expectedthatthosenetworkswillhaveahighertolerance fornodefailures.Ontheotherhand,moreconnectionswill consume more battery power, which will affect network lifetime.Wethereforealsoexaminelifetimemeasurements ofthenetworksforthesamesetofscenarios.
Further, aswe can observe in Fig. 6, network topolo- giesgeneratedby theDPValgorithmbecome k-vertexsu- pernodedisconnectedafterthefailureofatmost1%ofthe sensornodes.Eventhoughtheinitialoptimizedtopologies generatedby DPVarek-vertexsupernode-connected,after the failure of a few sensor nodes, the remaining topolo- giesbecome atmost(k− 1)-vertex supernode-connected.
The ADPV algorithm, on the other hand, maintains two- vertex supernode connectivity up to a failure of 32% of sensornodes,andthree-vertex supernodeconnectivityup toa failureof21% ofsensornodes,onaverageamong all experimentalinstances.
InFig.7, we comparethe lifetimeresultsofthe same setofexperimentalinstanceswiththoseofFig.6.Apromi- nent aspectof ADPV isconsidering the sensor nodes’re- mainingenergylevels;asaresult,energydepletionoccurs lessfrequently. Thisfactor, when coupledwiththe adap-
tivenatureofthealgorithm,resultsinlongernetworklife- times.
As we observe in Fig. 7, in terms of one-vertex supernode-connected lifetimes, on average, ADPV results in atwo-foldincrease withrespect toDPV. Forthe same experimental instances, ADPV provides respectively 65%
and 46% longer two-vertex and three-vertex supernode- connectedlifetimesthanDPV.Asweobserveinthefigures, network density has almost no effect on the lifetimes of thetopologiesgeneratedbyDPV.Ontheotherhand,ADPV successfullyprolongs thenetwork lifetimealmost propor- tionallytothenetworkdensityforallk=1,2,3.
We observe inFig. 7(c)and(d) that if thenumber of sensornodesdropsbelowacertainthreshold(inourcase, 150sensornodesina600mx600mareawhensr=10%, and200sensornodeswhensr=5%),itis hardtorestore three-vertex supernode connectivity.Thisfinding suggests that aminimum numberofsensornodesforeveryk and sr value is necessary to restore k-vertex supernode con- nectivity. Figs. 6 and 7, respectively compare node fail- ure tolerance andnetwork lifetimefordifferentvalues of sr=5%,10%andk=2,3.Accordingtotheresults,withthe
15 20 25 30 35 40 45 50 55 60 65
100 150 200 250 300 350 400 450 500
Lifeme (s)
Number of Sensor Nodes ADPV Connected
ADPV 2-Connected DPV Connected DPV 2-Connected
(a) k=2, sr=10%
15 20 25 30 35 40 45
100 150 200 250 300 350 400 450 500
Lifeme (s)
Number of Sensor Nodes ADPV Connected
ADPV 2-Connected DPV Connected DPV 2-Connected
(b) k=2, sr=5%
15 20 25 30 35 40 45 50
100 150 200 250 300 350 400 450 500
Lifeme (s)
Number of Sensor Nodes ADPV Connected
ADPV 3-Connected DPV Connected DPV 3-Connected
(c) k=3, sr=10%
15 17 19 21 23 25 27 29 31 33 35
100 150 200 250 300 350 400 450 500
Lifeme (s)
Number of Sensor Nodes ADPV Connected
ADPV 3-Connected DPV Connected DPV 3-Connected
(d) k=3, sr=5%
Fig. 7. Lifetime comparison of the DPV and ADPV algorithms.
15 20 25 30 35 40 45 50 55 60
150 200 250 300 350 400 450 500
emitefiL(s)
Number of Sensor Nodes ADPV Connected
ADPV 4-Connected DPV Connected DPV 4-Connected
(a) Lifetime comparison
0.00%
10.00%
20.00%
30.00%
40.00%
50.00%
60.00%
70.00%
80.00%
90.00%
100.00%
150 200 250 300 350 400 450 500
NodeeruliaFecnareloT(%)
Number of Sensor Nodes ADPV Connected
ADPV 4-Connected DPV Connected DPV 4-Connected
(b) Node failure tolerance comparison
Fig. 8. Lifetime and node failure tolerance of DPV and ADPV algorithms for k = 4 .
0 5000 10000 15000 20000 25000 30000
100 150 200 250 300 350 400 450 500
Number of Message Transmissions
Number of Sensor Nodes
ADPV DPV
(a) k=2, sr=10%
0 5000 10000 15000 20000 25000 30000 35000 40000
100 150 200 250 300 350 400 450 500
Number of Message Transmissions
Number of Sensor Nodes
ADPV DPV
(b) k=2, sr=5%
0 5000 10000 15000 20000 25000 30000 35000 40000 45000
100 150 200 250 300 350 400 450 500
Number of Message Transmissions
Number of Sensor Nodes
ADPV
DPV
(c) k=3, sr=10%
0 10000 20000 30000 40000 50000 60000
100 150 200 250 300 350 400 450 500
Number of Message Transmissions
Number of Sensor Nodes
ADPV
DPV
(d) k=3, sr=5%
Fig. 9. Number of message transmissions in DPV and ADPV algorithms.
increasing number of supernodes, lifetime also increases, however,therelationbetweentheincreaseinthenumber of supernodesand the increase in the lifetime is sublin- ear,andtherefore,weexpectthat theincreaseinthelife- time becomesinsubstantial asthenumber ofsupernodes exceedsacertain threshold.Wealsonoticethat,withthe increasing k value, more disjoint paths are required and thismakesprovidingalternativeroutesharder.InFig.8(a) and(b),we compareDPV andADPV algorithms fork=4 intermsofnetworklifetimeandnodefailuretolerance,re- spectively.AsseeninFig.8(a),ADPVsuccessfullyprolongs bothone-vertexandfour-vertexsupernode-connectedlife- timesofthenetwork.Also,inFig.8(b),we seethat ADPV canpreservefour-vertexsupernode-connectivityup tothe failure of 50% of the sensor nodes on dense networks, whichinturn,achievesalmost a two-foldincrease inthe four-vertexsupernode-connectedlifetime.
Anotherimportantmetricwemeasureduringouranal- ysis is the number of message transmissions. Message transmissionisanimportantmetricbecausewemustnot onlyconsiderpowerconsumptionintheresultingtopolo- giesbutalsoconsiderthepowerrequiredtogeneratethose topologies, which can be viewed as a fixed cost of ob- taining the final topologies. If this cost is high, then the power efficiencyof the resulting topology mightbecome meaningless. In Fig. 9, we compare the number of DPV andADPV messagetransmissions. Tosimulate theworst-
0 50 100 150 200 250 300 350 400
100 150 200 250 300 350 400 450 500
Number of Connecvity Restoraons
Number of Sensor Nodes
for k=3
for k=2
Fig. 10. Number of connectivity restorations.
case scenario of ADPV, we set the waiting periodin the restoration phase to zero, which means that every node failure that affectsdisjointpathswilltrigger arestoration phase for that node. According to the results, for k=2, ADPV makes atleast 2.25timesand atmostthree times the messagetransmissions than DPV does, andfork=3, ADPVmakesatleastthreetimesandatmost3.5timesthe messagetransmissionsofDPV.As seeninthesub-figures,