wireless sensor networks routingtechniques in wireless sensor

IEEE Wireless Communications • December 20046 1536-1284/04/$20.00 © 2004 IEEE

This research was sup-ported in part by theICUBE initiative of IowaState University, Ames,and the Hashemite Uni-versity, Zarqa, Jordan.

1 In this article, we con-sider routing toward a BSonly.

WIRELESS SENSOR NETWORKS

INTRODUCTIONDue to recent technological advances, the manu-facturing of small and low-cost sensors hasbecome technically and economically feasible.These sensors measure ambient conditions inthe environment surrounding them and thentransform these measurements into signals thatcan be processed to reveal some characteristicsabout phenomena located in the area aroundthese sensors. A large number of these sensorscan be networked in many applications thatrequire unattended operations, hence producinga wireless sensor network (WSN). In fact, theapplications of WSNs are quite numerous. Forexample, WSNs have profound effects on mili-tary and civil applications such as target fieldimaging, intrusion detection, weather monitor-ing, security and tactical surveillance, distributedcomputing, detecting ambient conditions such astemperature, movement, sound, light, or the

presence of certain objects, inventory control,and disaster management. Deployment of a sen-sor network in these applications can be in ran-dom fashion (e.g., dropped from an airplane in adisaster management application) or manual(e.g., fire alarm sensors in a facility or sensorsplanted underground for precision agriculture).Creating a network of these sensors can assistrescue operations by locating survivors, identify-ing risky areas, and making the rescue teammore aware of the overall situation in a disasterarea.

Typically, WSNs contain hundreds or thou-sands of these sensor nodes, and these sensorshave the ability to communicate either amongeach other or directly to an external base station(BS). A greater number of sensors allows forsensing over larger geographical regions withgreater accuracy. Figure 1 shows a schematicdiagram of sensor node components. Basically,each sensor node comprises sensing, processing,transmission, mobilizer, position finding system,and power units (some of these components areoptional, like the mobilizer). The same figureshows the communication architecture of aWSN. Sensor nodes are usually scattered in asensor field, which is an area where the sensornodes are deployed. Sensor nodes coordinateamong themselves to produce high-quality infor-mation about the physical environment. Eachsensor node bases its decisions on its mission,the information it currently has, and its knowl-edge of its computing, communication, and ener-gy resources. Each of these scattered sensornodes has the capability to collect and routedata either to other sensors or back to an exter-nal BS(s).1 A BS may be a fixed or mobile nodecapable of connecting the sensor network to anexisting communications infrastructure or to theInternet where a user can have access to thereported data.

In the past few years, intensive research thataddresses the potential of collaboration amongsensors in data gathering and processing, andcoordination and management of the sensingactivity was conducted. In most applications,sensor nodes are constrained in energy supplyand communication bandwidth. Thus, innovativetechniques to eliminate energy inefficiencies thatshorten the lifetime of the network and efficient

JAMAL N. AL-KARAKI, THE HASHEMITE UNIVERSITY

AHMED E. KAMAL, IOWA STATE UNIVERSITY

ABSTRACTWireless sensor networks consist of small

nodes with sensing, computation, and wirelesscommunications capabilities. Many routing,power management, and data dissemination pro-tocols have been specifically designed for WSNswhere energy awareness is an essential designissue. Routing protocols in WSNs might differdepending on the application and network archi-tecture. In this article we present a survey ofstate-of-the-art routing techniques in WSNs. Wefirst outline the design challenges for routingprotocols in WSNs followed by a comprehensivesurvey of routing techniques. Overall, the rout-ing techniques are classified into three cate-gories based on the underlying networkstructure: flit, hierarchical, and location-basedrouting. Furthermore, these protocols can beclassified into multipath-based, query-based,negotiation-based, QoS-based, and coherent-based depending on the protocol operation. Westudy the design trade-offs between energy andcommunication overhead savings in every rout-ing paradigm. We also highlight the advantagesand performance issues of each routing tech-nique. The article concludes with possible futureresearch areas.

ROUTING TECHNIQUES INWIRELESS SENSOR NETWORKS: A SURVEY

WSNs consist ofsmall nodes withsensing, computation,and wireless communicationscapabilities. Manyprotocols have beenspecifically designedfor WSNs whereenergy awareness is an essentialdesign issue.

Authorized licensed use limited to: UNIVERSITAT POLITÈCNICA DE CATALUNYA. Downloaded on December 21, 2009 at 03:54 from IEEE Xplore. Restrictions apply.

IEEE Wireless Communications • December 2004 7

use of the limited bandwidth are highly required.Such constraints combined with a typical deploy-ment of large number of sensor nodes posemany challenges to the design and managementof WSNs and necessitate energy-awareness at alllayers of the networking protocol stack. Forexample, at the network layer, it is highly desir-able to find methods for energy-efficient routediscovery and relaying of data from the sensornodes to the BS so that the lifetime of the net-work is maximized.

Routing in WSNs is very challenging due tothe inherent characteristics that distinguish thesenetworks from other wireless networks likemobile ad hoc networks or cellular networks.First, due to the relatively large number of sen-sor nodes, it is not possible to build a globaladdressing scheme for the deployment of a largenumber of sensor nodes as the overhead of IDmaintenance is high. Thus, traditional IP-basedprotocols may not be applied to WSNs. Further-more, sensor nodes that are deployed in an adhoc manner need to be self-organizing as the adhoc deployment of these nodes requires the sys-tem to form connections and cope with the resul-tant nodal distribution, especially as theoperation of sensor networks is unattended. InWSNs, sometimes getting the data is moreimportant than knowing the IDs of which nodessent the data. Second, in contrast to typical com-munication networks, almost all applications ofsensor networks require the fbw of sensed datafrom multiple sources to a particular BS. This,however, does not prevent the flow of data to bein other forms (e.g., multicast or peer to peer).Third, sensor nodes are tightly constrained interms of energy, processing, and storage capaci-ties. Thus, they require careful resource manage-ment. Fourth, in most application scenarios,nodes in WSNs are generally stationary afterdeployment except for maybe a few mobilenodes. Nodes in other traditional wireless net-works are free to move, which results in unpre-dictable and frequent topological changes.However, in some applications, some sensornodes may be allowed to move and change theirlocation (although with very low mobility). Fifth,sensor networks are application-specific (i.e.,design requirements of a sensor network changewith application). For example, the challengingproblem of low-latency precision tactical surveil-lance is different from that of a periodic weathermonitoring task. Sixth, position awareness ofsensor nodes is important since data collection isnormally based on the location. Currently, it isnot feasible to use Global Positioning System(GPS) hardware for this purpose. Methods basedon triangulation [1], for example, allow sensornodes to approximate their position using radiostrength from a few known points. It is found in[1] that algorithms based on triangulation ormultilateration can work quite well under condi-tions where only very few nodes know their posi-tions a priori (e.g., using GPS hardware). Still, itis favorable to have GPS-free solutions [2] forthe location problem in WSNs. Finally, data col-lected by many sensors in WSNs is typicallybased on common phenomena, so there is a highprobability that this data has some redundancy.Such redundancy needs to be exploited by the

routing protocols to improve energy and band-width utilization. Usually, WSNs are data-centricnetworks in the sense that data is requestedbased on certain attributes (i.e., attribute-basedaddressing). An attribute-based address is com-posed of a set of attribute-value pair query. Forexample, if the query is something like [tempera-ture > 60°F], sensor nodes that sense tempera-ture > 60°F only need to respond and reporttheir readings.

Due to such differences, many new algo-rithms have been proposed for the routing prob-lem in WSNs. These routing mechanisms havetaken into consideration the inherent features ofWSNs along with the application and architec-ture requirements. The task of finding and main-taining routes in WSNs is nontrivial since energyrestrictions and sudden changes in node status(e.g., failure) cause frequent and unpredictabletopological changes. To minimize energy con-sumption, routing techniques proposed in the lit-erature for WSNs employ some well-knownrouting tactics as well as tactics special to WSNs,such as data aggregation and in-network pro-cessing, clustering, different node role assign-ment, and data-centric methods. Almost all ofthe routing protocols can be classified accordingto the network structure as flit, hierarchical, orlocation-based. Furthermore, these protocols canbe classified into multipath-based, query-based,negotiation-based, quality of service (QoS)-based, and coherent-based depending on theprotocol operation. In flat networks all nodes playthe same role, while hierarchical protocols aimto cluster the nodes so that cluster heads can dosome aggregation and reduction of data in orderto save energy. Location-based protocols utilizeposition information to relay the data to thedesired regions rather than the whole network.The last category includes routing approachesbased on protocol operation, which vary accord-ing to the approach used in the protocol. In thisarticle we explore these routing techniques inWSNs that have been developed in recent yearsand develop a classification for these protocols.

n Figure 1. The components of a sensor node.

Processing unit

Position finding system Mobilizer

TargetSensor node

BS

User

Internet

ProcessorStorage

Sensing unit

Sensor

Transmissionunit

Powergenerator

ADC Transceiver

Power unit


IEEE Wireless Communications • December 20048

Then we discuss each of the routing protocolsunder this classification. Our objective is to pro-vide deeper understanding of the current routingprotocols in WSNs and identify some openresearch issues that can be further pursued.

Although there are some previous efforts onsurveying the characteristics, applications, andcommunication protocols in WSNs [3, 4], thescope of the survey presented in this article isdistinguished from these surveys in many aspects.The surveys in [3, 4] addressed several designissues and techniques for WSNs describing thephysical constraints on sensor nodes, applica-tions, architectural attributes, and the protocolsproposed in all layers of the network stack.However, these surveys were not devoted torouting only. Due to the importance of routingin WSNs and the availability of a significantbody of literature on this topic, a detailed surveybecomes necessary and useful at this stage. Ourwork is a dedicated study of the network layer,describing and categorizing the differentapproaches to data routing. In addition, we sum-marize routing challenges and design issues thatmay affect the performance of routing protocolsin WSNs. The rest of this article is organized asfollows. We discuss routing challenges anddesign issues in WSNs. A classification and com-prehensive survey of routing techniques in WSNsis presented. A summary of future researchdirections on routing in WSNs is discussed. Wethen conclude with final remarks.

ROUTING CHALLENGES ANDDESIGN ISSUES IN WSNS

Despite the innumerable applications of WSNs,these networks have several restrictions, such aslimited energy supply, limited computing power,and limited bandwidth of the wireless links con-necting sensor nodes. One of the main designgoals of WSNs is to carry out data communica-tion while trying to prolong the lifetime of thenetwork and prevent connectivity degradation byemploying aggressive energy management tech-niques. The design of routing protocols in WSNsis influenced by many challenging factors. Thesefactors must be overcome before efficient com-munication can be achieved in WSNs. In the fol-lowing, we summarize some of the routingchallenges and design issues that affect the rout-ing process in WSNs.

Node deployment: Node deployment in WSNsis application-dependent and can be either man-ual (deterministic) or randomized. In manualdeployment, the sensors are manually placedand data is routed through predetermined paths.However, in random node deployment, the sen-sor nodes are scattered randomly, creating an adhoc routing infrastructure. If the resultant distri-bution of nodes is not uniform, optimal cluster-ing becomes necessary to allow connectivity andenable energy-efficient network operation. Inter-sensor communication is normally within shorttransmission ranges due to energy and band-width limitations. Therefore, it is most likely thata route will consist of multiple wireless hops.

Energy consumption without losing accuracy:Sensor nodes can use up their limited supply of

energy performing computations and transmit-ting information in a wireless environment. Assuch, energy-conserving forms of communicationand computation are essential. Sensor node life-time shows a strong dependence on battery life-time [5]. In a multihop WSN, each node plays adual role as data sender and data router. Themalfunctioning of some sensor nodes due topower failure can cause significant topologicalchanges, and might require rerouting of packetsand reorganization of the network.

Data reporting method: Data reporting inWSNs is application-dependent and also dependson the time criticality of the data. Data reportingcan be categorized as either time-driven, event-driven, query-driven, or a hybrid of all thesemethods. The time-driven delivery method issuitable for applications that require periodicdata monitoring. As such, sensor nodes will peri-odically switch on their sensors and transmitters,sense the environment, and transmit the data ofinterest at constant periodic time intervals. Inevent-driven and query-driven methods, sensornodes react immediately to sudden and drasticchanges in the value of a sensed attribute due tothe occurrence of a certain event, or respond toa query generated by the BS or another node inthe network. As such, these are well suited totime-critical applications. A combination of theprevious methods is also possible. The routingprotocol is highly influenced by the data report-ing method in terms of energy consumption androute calculations.

Node/link heterogeneity: In many studies, allsensor nodes were assumed to be homogeneous(i.e., have equal capacity in terms of computa-tion, communication, and power). However,depending on the application a sensor node canhave a different role or capability. The existenceof a heterogeneous set of sensors raises manytechnical issues related to data routing. Forexample, some applications might require adiverse mixture of sensors for monitoring tem-perature, pressure, and humidity of the sur-rounding environment, detecting motion viaacoustic signatures, and capturing images orvideo tracking of moving objects. Either thesespecial sensors can be deployed independentlyor the different functionalities can be included inthe same sensor nodes. Even data reading andreporting can be generated from these sensors atdifferent rates, subject to diverse QoS con-straints, and can follow multiple data reportingmodels. For example, hierarchical protocols des-ignate a cluster head node different from thenormal sensors. These cluster heads can be cho-sen from the deployed sensors or be more pow-erful than other sensor nodes in terms of energy,bandwidth, and memory. Hence, the burden oftransmission to the BS is handled by the set ofcluster heads.

Fault tolerance: Some sensor nodes may failor be blocked due to lack of power, physicaldamage, or environmental interference. The fail-ure of sensor nodes should not affect the overalltask of the sensor network. If many nodes fail,medium access control (MAC) and routing pro-tocols must accommodate formation of newlinks and routes to the data collection BSs. Thismay require actively adjusting transmit powers

One of the maindesign goals ofWSNs is to carry outdata communicationwhile trying to prolong the lifetimeof the network andprevent connectivitydegradation byemploying aggressive energymanagement techniques.



and signaling rates on the existing links to reduceenergy consumption, or rerouting packetsthrough regions of the network where moreenergy is available. Therefore, multiple levels ofredundancy may be needed in a fault-tolerantsensor network.

Scalability: The number of sensor nodesdeployed in the sensing area may be on theorder of hundreds or thousands, or more. Anyrouting scheme must be able to work with thishuge number of sensor nodes. In addition, sen-sor network routing protocols should be scalableenough to respond to events in the environment.Until an event occurs, most sensors can remainin the sleep state, with data from the few remain-ing sensors providing coarse quality.

Network dynamics: In many studies, sensornodes are assumed fixed. However, in manyapplications both the BS or sensor nodes can bemobile [6]. As such, routing messages from or tomoving nodes is more challenging since routeand topology stability become important issues,in addition to energy, bandwidth, and so forth.Moreover, the phenomenon can be mobile (e.g.,a target detection/ tracking application). On theother hand, sensing fixed events allows the net-work to work in a reactive mode (i.e., generatingtraffic when reporting), while dynamic events inmost applications require periodic reporting tothe BS.

Transmission media: In a multihop sensornetwork, communicating nodes are linked by awireless medium. The traditional problems asso-ciated with a wireless channel (e.g., fading, higherror rate) may also affect the operation of thesensor network. In general, the required band-width of sensor data will be low, on the order of1–100 kb/s. Related to the transmission media isthe design of MAC. One approach to MACdesign for sensor networks is to use time-divisionmultiple access (TDMA)-based protocols thatconserve more energy than contention-basedprotocols like carrier sense multiple access(CSMA) (e.g., IEEE 802.11). Bluetooth technol-ogy [7] can also be used.

Connectivity: High node density in sensornetworks precludes them from being completelyisolated from each other. Therefore, sensornodes are expected to be highly connected. This,however, may not prevent the network topologyfrom being variable and the network size fromshrinking due to sensor node failures. In addi-tion, connectivity depends on the possibly ran-dom distribution of nodes.

Coverage: In WSNs, each sensor node obtainsa certain view of the environment. A given sen-sor’s view of the environment is limited in bothrange and accuracy; it can only cover a limitedphysical area of the environment. Hence, areacoverage is also an important design parameterin WSNs.

Data aggregation: Since sensor nodes maygenerate significant redundant data, similarpackets from multiple nodes can be aggregatedto reduce the number of transmissions. Dataaggregation is the combination of data from dif-ferent sources according to a certain aggregationfunction (e.g., duplicate suppression, minima,maxima, and average). This technique has beenused to achieve energy efficiency and data trans-

fer optimization in a number of routing proto-cols. Signal processing methods can also be usedfor data aggregation. In this case, it is referredto as data fusion where a node is capable of pro-ducing a more accurate output signal by usingsome techniques such as beamforming to com-bine the incoming signals and reducing the noisein these signals.

Quality of service: In some applications, datashould be delivered within a certain period oftime from the moment it is sensed, or it will beuseless. Therefore, bounded latency for datadelivery is another condition for time-con-strained applications. However, in many applica-tions, conservation of energy, which is directlyrelated to network lifetime, is considered rela-tively more important than the quality of datasent. As energy is depleted, the network may berequired to reduce the quality of results in orderto reduce energy dissipation in the nodes andhence lengthen the total network lifetime.Hence, energy-aware routing protocols arerequired to capture this requirement.

ROUTING PROTOCOLS IN WSNS

In this section we survey the state-of-the-artrouting protocols for WSNs. In general, routingin WSNs can be divided into flat-based routing,hierarchical-based routing, and location-basedrouting depending on the network structure. Inflat-based routing, all nodes are typicallyassigned equal roles or functionality. In hierar-chical-based routing, nodes will play differentroles in the network. In location-based routing,sensor nodes’ positions are exploited to routedata in the network. A routing protocol is con-sidered adaptive if certain system parameterscan be controlled in order to adapt to currentnetwork conditions and available energy levels.Furthermore, these protocols can be classifiedinto multipath-based, query-based, and negotia-tion-based, QoS-based, or coherent-based routingtechniques depending on the protocol operation.In addition to the above, routing protocols canbe classified into three categories, proactive,reactive, and hybrid, depending on how thesource finds a route to the destination. In proac-tive protocols, all routes are computed beforethey are really needed, while in reactive proto-cols, routes are computed on demand. Hybridprotocols use a combination of these two ideas.When sensor nodes are static, it is preferable tohave table-driven routing protocols rather thanreactive protocols. A significant amount of ener-gy is used in route discovery and setup of reac-tive protocols. Another class of routing protocolsis called cooperative. In cooperative routing,nodes send data to a central node where datacan be aggregated and may be subject to furtherprocessing, hence reducing route cost in terms ofenergy use. Many other protocols rely on timingand position information. We also shed somelight on these types of protocols in this article.In order to streamline this survey, we use a clas-sification according to the network structure andprotocol operation (routing criteria). The classi-fication is shown in Fig. 2 where numbers in thefuture indicate the references.

In the rest of this section we present a

In WSNs, each sensor node obtains

a certain view of theenvironment. A givensensor’s view of the

environment is limited in both range

and accuracy; it can only cover a

limited physical areaof the environment.

Hence, area coverage is also an

important designparameter in WSNs.



detailed overview of the main routing paradigmsin WSNs. We start with network-structure-basedprotocols.

NETWORK-STRUCTURE-BASED PROTOCOLSThe underlying network structure can play a sig-nificant role in the operation of the routing pro-tocol in WSNs. In this section we survey in detailmost of the protocols that fall into this category.

Flat Routing — The first category of routing proto-cols are the multihop flat routing protocols. Inflat networks, each node typically plays the samerole and sensor nodes collaborate to perform thesensing task. Due to the large number of suchnodes, it is not feasible to assign a global identi-fier to each node. This consideration has led todata-centric routing, where the BS sends queriesto certain regions and waits for data from thesensors located in the selected regions. Sincedata is being requested through queries,attribute-based naming is necessary to specifythe properties of data. Early work on data cen-tric routing (e.g., SPIN and directed diffusion[8]) were shown to save energy through datanegotiation and elimination of redundant data.These two protocols motivated the design ofmany other protocols that follow a similar con-cept. In the rest of this subsection, we summa-rize these protocols, and highlight theiradvantages and performance issues.

Sensor Protocols for Information via Negoti-ation: Heinzelman et al. in [9, 10] proposed afamily of adaptive protocols called Sensor Proto-cols for Information via Negotiation (SPIN) thatdisseminate all the information at each node toevery node in the network assuming that allnodes in the network are potential BSs. Thisenables a user to query any node and get therequired information immediately. These proto-cols make use of the property that nodes in closeproximity have similar data, and hence there is aneed to only distribute the data other nodes donot posses. The SPIN family of protocols usesdata negotiation and resource-adaptive algo-rithms. Nodes running SPIN assign a high-levelname to completely describe their collected data(called meta-data) and perform metadata negoti-

ations before any data is transmitted. Thisensures that there is no redundant data sentthroughout the network. The semantics of themeta-data format is application-specific and notspecified in SPIN. For example, sensors mightuse their unique IDs to report meta-data if theycover a certain known region. In addition, SPINhas access to the current energy level of thenode and adapts the protocol it is running basedon how much energy is remaining. These proto-cols work in a time-driven fashion and distributethe information all over the network, even whena user does not request any data.

The SPIN family is designed to address thedeficiencies of classic flooding by negotiationand resource adaptation. The SPIN family ofprotocols is designed based on two basic ideas:

1) Sensor nodes operate more efficiently andconserve energy by sending data that describethe sensor data instead of sending all the data;for example, image and sensor nodes must moni-tor the changes in their energy resources.

2) Conventional protocols like flooding orgossiping-based routing protocols [11] wasteenergy and bandwidth when sending extra andunnecessary copies of data by sensors coveringoverlapping areas. The drawbacks of floodinginclude implosion, which is caused by duplicatemessages sent to the same node, overlap whentwo nodes sensing the same region send similarpackets to the same neighbor, and resourceblindness in consuming large amounts of energywithout consideration for energy constraints.Gossiping avoids the problem of implosion byjust selecting a random node to which to sendthe packet rather than broadcasting the packetblindly. However, this causes delays in propaga-tion of data through the nodes.

SPIN’s meta-data negotiation solves the clas-sic problems of flooding, thus achieving a lot ofenergy efficiency. SPIN is a three-stage protocolas sensor nodes use three types of messages,ADV, REQ, and DATA, to communicate. ADVis used to advertise new data, REQ to requestdata, and DATA is the actual message itself.The protocol starts when a SPIN node obtainsnew data it is willing to share. It does so bybroadcasting an ADV message containing meta-

nnnn Figure 2. Routing protocols in WSNs: a taxonomy.

Flatnetworkrouting

2,3,7,1314,15,16,18

39,41,49

1,8,9,12,1719,22,23,35

31,26,4825,33,42

46,47

Hierarchicalnetworkrouting

Location-based

routing

Network structure

Negotiation-based

routing

3,72,10,26,28

29,34 2,20,27

Multipath-based

routing

Query-based

routing

11,44

QoS-based

routing

11,2,33

Coherent-based

routing

Protocol operation

Routing protocols in WSNs



data. If a neighbor is interested in the data, itsends a REQ message for the DATA and theDATA is sent to this neighbor node. The neigh-bor sensor node then repeats this process withits neighbors. As a result, the entire sensor areawill receive a copy of the data.

The SPIN family of protocols includes manyprotocols. The main two are called SPIN-1 andSPIN-2; they incorporate negotiation beforetransmitting data in order to ensure that onlyuseful information will be transferred. Also, eachnode has its own resource manager that keepstrack of resource consumption and is polled bythe nodes before data transmission. The SPIN-1protocol is a three-stage protocol, as describedabove. An extension to SPIN-1 is SPIN-2, whichincorporates a threshold-based resource aware-ness mechanism in addition to negotiation.When energy in the nodes is abundant, SPIN-2communicates using the three-stage protocol ofSPIN1. However, when the energy in a nodestarts approaching a low threshold, it reduces itsparticipation in the protocol; that is, it partici-pates only when it believes it can complete allthe other stages of the protocol without goingbelow the low energy threshold. In conclusion,SPIN-1 and SPIN-2 are simple protocols thatefficiently disseminate data while maintaining noper-neighbor state. These protocols are well suit-ed to an environment where the sensors aremobile because they base their forwarding deci-sions on local neighborhood information. Otherprotocols of the SPIN family are (please refer to[3, 7] for more details):• SPIN-BC: This protocol is designed for broad-

cast channels.• SPIN-PP: This protocol is designed for point-

to-point communication (i.e., hop-by-hoprouting).

• SPIN-EC: This protocol works similar toSPIN-PP, but with an energy heuristic addedto it.

• SPIN-RL: When a channel is lossy, a protocolcalled SPIN-RL is used where adjustments areadded to the SPIN-PP protocol to account forthe lossy channel.One of the advantages of SPIN is that topo-

logical changes are localized since each nodeneed know only its single-hop neighbors. SPINprovides more energy savings than flooding, andmetadata negotiation almost halves the redun-dant data. However, SPIN’s data advertisementmechanism cannot guarantee delivery of data.To see this, consider the application of intrusiondetection where data should be reliably reportedover periodic intervals, and assume that nodesinterested in the data are located far away fromthe source node, and the nodes between sourceand destination nodes are not interested in thatdata; such data will not be delivered to the desti-nation at all.

Directed diffusion: In [12], C. Intanagonwi-wat et al. proposed a popular data aggregationparadigm for WSNs called directed diffusion.Directed diffusion is a data-centric (DC) andapplication-aware paradigm in the sense that alldata generated by sensor nodes is named byattribute-value pairs. The main idea of the DCparadigm is to combine the data coming fromdifferent sources en route (in-network aggrega-

tion) by eliminating redundancy, minimizing thenumber of transmissions, thus saving networkenergy and prolonging its lifetime. Unlike tradi-tional end-to-end routing, DC routing findsroutes from multiple sources to a single destina-tion that allows in-network consolidation ofredundant data.

In directed diffusion, sensors measure eventsand create gradients of information in theirrespective neighborhoods. The BS requests databy broadcasting interests. An interest describes atask required to be done by the network. Aninterest diffuses through the network hop byhop, and is broadcast by each node to its neigh-bors. As the interest is propagated throughoutthe network, gradients are set up to draw datasatisfying the query toward the requesting node(i.e., a BS may query for data by disseminatinginterests and intermediate nodes propagatethese interests). Each sensor that receives theinterest sets up a gradient toward the sensornodes from which it receives the interest. Thisprocess continues until gradients are set up fromthe sources back to the BS. More generally, agradient specifies an attribute value and a direc-tion. The strength of the gradient may be differ-ent toward different neighbors, resulting indifferent amounts of information flow. At thisstage, loops are not checked, but are removed ata later stage. Figure 3 shows an example of theworking of directed diffusion (sending interests,building gradients, and data dissemination).When interests fit gradients, paths of informa-tion flow are formed from multiple paths, andthen the best paths are reinforced to preventfurther flooding according to a local rule. Inorder to reduce communication costs, data isaggregated on the way. The goal is to find agood aggregation tree that gets the data fromsource nodes to the BS. The BS periodicallyrefreshes and resends the interest when it startsto receive data from the source(s). This is neces-sary because interests are not reliably transmit-ted throughout the network.

All sensor nodes in a directed-diffusion-basednetwork are application-aware, which enablesdiffusion to achieve energy savings by selectingempirically good paths, and by caching and pro-cessing data in the network. Caching canincrease the efficiency, robustness, and scalabili-ty of coordination between sensor nodes, whichis the essence of the data diffusion paradigm.Other usage of directed diffusion is to sponta-neously propagate an important event to somesections of the sensor network. Such a type ofinformation retrieval is well suited only to persis-tent queries where requesting nodes are notexpecting data that satisfy a query for a durationof time. This makes it unsuitable for one-timequeries, as it is not worth setting up gradientsfor queries that use the path only once.

The performance of data aggregation meth-ods used in the directed diffusion paradigm isaffected by a number of factors, including thepositions of the source nodes in the network, thenumber of sources, and the communication net-work topology. In order to investigate these fac-tors, two models of source placement (shown inFig. 4) were studied in [12]. These models arecalled the event radius (ER) model and the ran-

The goal is to find agood aggregationtree that gets thedata from sourcenodes to the BS.

The BS periodicallyrefreshes and

resends the interestwhen it starts to

receive data fromthe source(s). This is

necessary becauseinterests are not

reliably transmittedthroughout the

network.



dom sources (RS) model. In the ER model, asingle point in the network area is defined as thelocation of an event. This may correspond to avehicle or some other phenomenon beingtracked by the sensor nodes. All nodes within adistance S (called the sensing range) of thisevent that are not BSs are considered to be datasources. The average number of sources isapproximately πS2n in a unit area network withn sensor nodes. In the RS model, k of the nodesthat are not BSs are randomly selected to besources. Unlike the ER model, the sources arenot necessarily clustered near each other. Inboth models of source placement, for a givenenergy budget, a greater number of sources canbe connected to the BS. However, each one per-forms better in terms of energy consumptiondepending on the application. In conclusion, theenergy savings with aggregation used in directeddiffusion can be transformed to provide a greaterdegree of robustness with respect to dynamics inthe sensed phenomena.

Directed diffusion differs from SPIN in twoaspects. First, directed diffusion issues dataqueries on demand as the BS sends queries tothe sensor nodes by flooding some tasks. InSPIN, however, sensors advertise the availabilityof data, allowing interested nodes to query thatdata. Second, all communication in directed dif-fusion is neighbor to neighbor with each nodehaving the capability to perform data aggrega-tion and caching. Unlike SPIN, there is no needto maintain global network topology in directeddiffusion. However, directed diffusion may notbe applied to applications (e.g., environmentalmonitoring) that require continuous data deliv-ery to the BS. This is because the query-drivenon-demand data model may not help in thisregard. Moreover, matching data to queriesmight require some extra overhead at the sensornodes.

Rumor routing: Rumor routing [13] is a vari-ation of directed diffusion and is mainly intend-ed for applications where geographic routing isnot feasible. In general, directed diffusion usesflooding to inject the query to the entire net-work when there is no geographic criterion to

diffuse tasks. However, in some cases there isonly a small amount of data requested from thenodes; thus, the use of flooding is unnecessary.An alternative approach is to flood the events ifthe number of events is small and the numberof queries is large. The key idea is to route thequeries to the nodes that have observed a par-ticular event rather than flooding the entire net-work to retrieve information about the occurringevents. In order to flood events through the net-work, the rumor routing algorithm employslong-lived packets called agents. When a nodedetects an event, it adds the event to its localtable, called an events table, and generates anagent. Agents travel the network in order topropagate information about local events to dis-tant nodes. When a node generates a query foran event, the nodes that know the route mayrespond to the query by inspecting its eventtable. Hence, there is no need to flood thewhole network, which reduces the communica-tion cost. On the other hand, rumor routingmaintains only one path between source anddestination as opposed to directed diffusionwhere data can be routed through multiplepaths at low rates. Simulation results showedthat rumor routing can achieve significant ener-gy savings compared to event flooding and canalso handle a node’s failure. However, rumorrouting performs well only when the number ofevents is small. For a large number of events,the cost of maintaining agents and event tablesin each node becomes infeasible if there is notenough interest in these events from the BS.Moreover, the overhead associated with rumorrouting is controlled by different parametersused in the algorithm such as time to live (TTL)pertaining to queries and agents. Since thenodes become aware of events through theevent agents, the heuristic for defining the routeof an event agent highly affects the performanceof next-hop selection in rumor routing.

Minimum Cost Forwarding Algorithm: TheMinimum Cost Forwarding Algorithm (MCFA)[8] exploits the fact that the direction of routingis always known (i.e., toward the fixed externalBS). Hence, a sensor node need not have a

nnnn Figure 3. An example of interest diffusion in a sensor network.

(a) Propagate interest

Source Sink

Source Sink

(b) Set up gradients

(c) Send data and path reinforcement

Source Sink

The energy savingswith aggregationused in the directeddiffusion can betransformed to provide a greaterdegree of robustnesswith respect todynamics in thesensed phenomena.



unique ID nor maintain a routing table. Instead,each node maintains the least cost estimate fromitself to the BS. Each message to be forwardedby the sensor node is broadcast to its neighbors.When a node receives the message, it checks if itis on the least cost path between the source sen-sor node and the BS. If this is the case, itrebroadcasts the message to its neighbors. Thisprocess repeats until the BS is reached.

In MCFA, each node should know the leastcost path estimate from itself to the BS. This isobtained as follows. The BS broadcasts a mes-sage with the cost set to zero, while every nodeinitially sets its least cost to the BS to infinity(∞). Each node, upon receiving the broadcastmessage originated at the BS, checks to see ifthe estimate in the message plus the link onwhich it is received is less than the current esti-mate. If yes, the current estimate and the esti-mate in the broadcast message are updated. Ifthe received broadcast message is updated, it isresent; otherwise, it is purged and nothing fur-ther is done. However, the previous proceduremay result in some nodes having multipleupdates, and those nodes far away from the BSwill get more updates from those closer to theBS. To avoid this, MCFA was modified to runa backoff algorithm at the setup phase. Thebackoff algorithm dictates that a node will notsend the updated message until a * lc timeunits have elapsed from the time at which themessage is updated, where a is a constant andlc is the link cost at which the message wasreceived.

Gradient-based routing: Schurgers et al. [14]proposed another variant of directed diffusion,called gradient-based routing (GBR). The keyidea in GBR is to memorize the number of hopswhen the interest is diffused through the wholenetwork. As such, each node can calculate aparameter called the height of the node, which isthe minimum number of hops to reach the BS.The difference between a node’s height and thatof its neighbor is considered the gradient on thatlink. A packet is forwarded on a link with thelargest gradient. GBR uses some auxiliary tech-niques such as data aggregation and trafficspreading in order to uniformly divide the trafficover the network. When multiple paths passthrough a node, which acts as a relay node, that

relay node may combine data according to a cer-tain function. In GBR, three different data dis-semination techniques have been discussed:• A stochastic scheme, where a node picks one

gradient at random when there are two ormore next hops that have the same gradient

• An energy-based scheme, where a nodeincreases its height when its energy dropsbelow a certain threshold so that other sensorsare discouraged from sending data to thatnode

• A stream-based scheme, where new streamsare not routed through nodes that are cur-rently part of the path of other streamsThe main objective of these schemes is to

obtain balanced distribution of the traffic in thenetwork, thus increasing the network lifetime.Simulation results of GBR showed that GBRoutperforms directed diffusion in terms of totalcommunication energy.

Information-driven sensor querying and con-strained anisotropic diffusion routing: Tworouting techniques, information-driven sensorquerying (IDSQ) and constrained anisotropicdiffusion routing (CADR), were proposed in[15]. CADR aims to be a general form of direct-ed diffusion. The key idea is to query sensorsand route data in the network such that informa-tion gain is maximized while latency and band-width are minimized. CADR diffuses queries byusing a set of information criteria to select whichsensors can get the data. This is achieved by acti-vating only the sensors that are close to a partic-ular event and dynamically adjusting data routes.The main difference from directed diffusion isthe consideration of information gain in additionto communication cost. In CADR, each nodeevaluates an information/cost objective androutes data based on the local information/costgradient and end-user requirements. Estimationtheory was used to model information utility. InIDSQ, the querying node can determine whichnode can provide the most useful informationwith the additional advantage of balancing theenergy cost. However, IDSQ does not specifical-ly define how the query and information arerouted between sensors and the BS. Therefore,IDSQ can be seen as a complementary optimiza-tion procedure. Simulation results showed thatthese approaches are more energy-efficient than

nnnn Figure 4. Two models used in a data-centric routing paradigm such as directed diffusion: a) event radiusmodel; b) random source model.

(a)

Source nodeSink nodeSink

(b)

Sink

The key idea in GBRis to memorize the

number of hopswhen the interest isdiffused through the

whole network. As such, each node

can calculate aparameter called theheight of the node,

which is the minimum number

of hops to reach the BS.



directed diffusion where queries are diffused inan isotropic fashion and reach nearest neighborsfirst.

COUGAR: Another data-centric protocolcalled COUGAR [16] views the network as ahuge distributed database system. The key ideais to use declarative queries in order to abstractquery processing from the network layer func-tions such as selection of relevant sensors and soon. COUGAR utilizes in-network data aggrega-tion to obtain more energy savings. The abstrac-tion is supported through an additional querylayer that lies between the network and applica-tion layers. COUGAR incorporates an architec-ture for the sensor database system where sensornodes select a leader node to perform aggrega-tion and transmit the data to the BS. The BS isresponsible for generating a query plan thatspecifies the necessary information about thedata flow and in-network computation for theincoming query, and sends it to the relevantnodes. The query plan also describes how toselect a leader for the query. The architectureprovides in-network computation ability that canprovide energy efficiency in situations when thegenerated data is huge. COUGAR provides anetwork-layer-independent method for dataquery. However, COUGAR has some draw-backs. First, the addition of a query layer oneach sensor node may add extra overhead interms of energy consumption and memory stor-age. Second, to obtain successful in-networkdata computation, synchronization among nodesis required (not all data are received at the sametime from incoming sources) before sending thedata to the leader node. Third, the leader nodesshould be dynamically maintained to preventthem from being hotspots (failure-prone).

ACQUIRE: In [17], Sadagopan et al. pro-posed a technique for querying sensor networkscalled Active Qwery Forwarding in Sensor Net-works (ACQUIRE). Similar to COUGAR,ACQUIRE views the network as a distributeddatabase where complex queries can be furtherdivided into several subqueries. The operation ofACQUIRE can be described as follows. The BSnode sends a query, which is then forwarded byeach node receiving the query. During this, eachnode tries to respond to the query partially byusing its precached information and then for-wards it to another sensor node. If the pre-cached information is not up-to-date, the nodesgather information from their neighbors within alookahead of d hops. Once the query is resolvedcompletely, it is sent back through either thereverse or shortest path to the BS. Hence,ACQUIRE can deal with complex queries byallowing many nodes to send responses. Notethat directed diffusion may not be used for com-plex queries due to energy considerations asdirected diffusion also uses a flooding-basedquery mechanism for continuous and aggregatequeries. On the other hand, ACQUIRE can pro-vide efficient querying by adjusting the value ofthe lookahead parameter d. When d is equal tonetwork diameter, ACQUIRE behaves similar toflooding. However, the query has to travel morehops if d is too small. A mathematical modelingwas used to find an optimal value of the parame-ter d for a grid of sensors where each node has

four immediate neighbors. However, there is novalidation of results through simulation. Toselect the next node for forwarding the query,ACQUIRE either picks it randomly or the selec-tion is based on maximum potential query satis-faction. Recall that either selection of the nextnode is based on information gain (CADR andIDSQ) or the query is forwarded to a node thatknows the path to the searched event (rumorrouting).

Energy-Aware Routing: The objective of theEnergy-Aware Routing protocol [18], a destina-tion-initiated reactive protocol, is to increase thenetwork lifetime. Although this protocol is simi-lar to directed diffusion, it differs in the sensethat it maintains a set of paths instead of main-taining or enforcing one optimal path at higherrates. These paths are maintained and chosen bymeans of a certain probability. The value of thisprobability depends on how low the energy con-sumption is that each path can achieve. By hav-ing paths chosen at different times, the energy ofany single path will not deplete quickly. This canachieve longer network lifetime as energy is dis-sipated more equally among all nodes. Networksurvivability is the main metric of this protocol.The protocol assumes that each node is address-able through class-based addressing that includesthe locations and types of the nodes. The proto-col initiates a connection through localizedflooding, which is used to discover all routesbetween a source/ destination pair and theircosts, thus building up the routing tables. High-cost paths are discarded, and a forwarding tableis built by choosing neighboring nodes in a man-ner that is proportional to their cost. Then for-warding tables are used to send data to thedestination with a probability inversely propor-tional to the node cost. Localized flooding isperformed by the destination node to keep thepaths alive. Compared to directed diffusion, thisprotocol provides an overall improvement of21.5 percent energy saving and a 44 percentincrease in network lifetime. However, theapproach requires gathering location informa-tion and setting up the addressing mechanismfor the nodes, which complicate route setupcompared to directed diffusion.

Routing protocols with random walks: Theobjective of the random-walks-based routingtechnique [19] is to achieve load balancing in astatistical sense by making use of multipath rout-ing in WSNs. This technique considers onlylarge-scale networks where nodes have very lim-ited mobility. In this protocol, it is assumed thatsensor nodes can be turned on or off at randomtimes. Furthermore, each node has a uniqueidentifier but no location information is needed.Nodes were arranged such that each node fallsexactly on one crossing point of a regular grid ona plane, but the topology can be irregular. Tofind a route from a source to its destination, thelocation information or lattice coordination isobtained by computing distances between nodesusing the distributed asynchronous version of thewell-known Bellman-Ford algorithm. An inter-mediate node would select as the next hop theneighboring node that is closer to the destina-tion according to a computed probability. Bycarefully manipulating this probability, some

The objective of random walks basedrouting technique isto achieve load balancing in a statistical sense andby making use ofmulti-path routing in WSNs. This technique considersonly large scale networks wherenodes have very limited mobility.



kind of load balancing can be obtained in thenetwork. The routing algorithm is simple asnodes are required to maintain little state infor-mation. Moreover, different routes are chosen atdifferent times even for the same pair of sourceand destination nodes. However, the main con-cern about this protocol is that the topology ofthe network may not be practical.

Hierarchical Routing — Hierarchical or cluster-based routing methods, originally proposed inwireline networks, are well-known techniqueswith special advantages related to scalability andefficient communication. As such, the concept ofhierarchical routing is also utilized to performenergy-efficient routing in WSNs. In a hierarchi-cal architecture, higher-energy nodes can beused to process and send the information, whilelow-energy nodes can be used to perform thesensing in the proximity of the target. The cre-ation of clusters and assigning special tasks tocluster heads can greatly contribute to overallsystem scalability, lifetime, and energy efficiency.Hierarchical routing is an efficient way to lowerenergy consumption within a cluster, performingdata aggregation and fusion in order to decreasethe number of transmitted messages to the BS.Hierarchical routing is mainly two-layer routingwhere one layer is used to select cluster headsand the other for routing. However, most tech-niques in this category are not about routing, butrather “who and when to send or process/ aggre-gate” the information, channel allocation, and soon, which can be orthogonal to the multihoprouting function.

LEACH protocol: Heinzelman, et al. [5] intro-duced a hierarchical clustering algorithm forsensor networks, called Low Energy AdaptiveClustering Hierarchy (LEACH). LEACH is acluster-based protocol, which includes distribut-ed cluster formation. LEACH randomly selects afew sensor nodes as cluster heads (CHs) androtates this role to evenly distribute the energyload among the sensors in the network. InLEACH, the CH nodes compress data arrivingfrom nodes that belong to the respective cluster,and send an aggregated packet to the BS inorder to reduce the amount of information thatmust be transmitted to the BS. LEACH uses aTDMA/code-division multiple access (CDMA)MAC to reduce intercluster and intracluster col-lisions. However, data collection is centralizedand performed periodically. Therefore, this pro-tocol is most appropriate when there is a needfor constant monitoring by the sensor network.A user may not need all the data immediately.Hence, periodic data transmissions are unneces-sary, and may drain the limited energy of thesensor nodes. After a given interval of time, ran-domized rotation of the role of CH is conductedso that uniform energy dissipation in the sensornetwork is obtained. The authors found, basedon their simulation model, that only 5 percent ofthe nodes need to act as CHs.

The operation of LEACH is separated intotwo phases, the setup phase and the steady statephase. In the setup phase, the clusters are orga-nized and CHs are selected. In the steady statephase, the actual data transfer to the BS takesplace. The duration of the steady state phase is

longer than the duration of the setup phase inorder to minimize overhead. During the setupphase, a predetermined fraction of nodes, p,elect themselves as CHs as follows. A sensornode chooses a random number, r, between 0and 1. If this random number is less than athreshold value, T(n), the node becomes a CHfor the current round. The threshold value is cal-culated based on an equation that incorporatesthe desired percentage to become a CH, the cur-rent round, and the set of nodes that have notbeen selected as a CH in the last (1/P) rounds,denoted G. It is given by

where G is the set of nodes that are involved inthe CH election. All elected CHs broadcast anadvertisement message to the rest of the nodesin the network that they are the new CHs. Allthe non-CH nodes, after receiving this advertise-ment, decide on the cluster to which they wantto belong. This decision is based on the signalstrength of the advertisement. The non-CHnodes inform the appropriate CHs that they willbe a member of the cluster. After receiving allthe messages from the nodes that would like tobe included in the cluster and based on the num-ber of nodes in the cluster, the CH node createsa TDMA schedule and assigns each node a timeslot when it can transmit. This schedule is broad-cast to all the nodes in the cluster.

During the steady state phase, the sensornodes can begin sensing and transmitting data tothe CHs. The CH node, after receiving all thedata, aggregates it before sending it to the BS.After a certain time, which is determined a pri-ori, the network goes back into the setup phaseagain and enters another round of selecting newCHs. Each cluster communicates using differentCDMA codes to reduce interference from nodesbelonging to other clusters.

Although LEACH is able to increase the net-work lifetime, there are still a number of issuesabout the assumptions used in this protocol.LEACH assumes that all nodes can transmitwith enough power to reach the BS if neededand that each node has computational power tosupport different MAC protocols. Therefore, itis not applicable to networks deployed in largeregions. It also assumes that nodes always havedata to send, and nodes located close to eachother have correlated data. It is not obvious howthe number of predetermined CHs (p) is goingto be uniformly distributed through the network.Therefore, there is the possibility that the elect-ed CHs will be concentrated in one part of thenetwork; hence, some nodes will not have anyCHs in their vicinity. Furthermore, the idea ofdynamic clustering brings extra overhead (headchanges, advertisements, etc.), which may dimin-ish the gain in energy consumption. Finally, theprotocol assumes that all nodes begin with thesame amount of energy capacity in each electionround, assuming that being a CH consumesapproximately the same amount of energy foreach node. The protocol should be extended toaccount for non-uniform energy nodes (i.e., usean energy-based threshold). An extension toLEACH, LEACH with negotiation, was pro-

T np

p r pn G( )

( mod( / )),=

−∈

1 1if

The operation ofLEACH is separated

into two phases, thesetup phase and thesteady state phase.In the setup phase,

the clusters are organized and CHsare selected. In thesteady state phase,

the actual data trans-fer to the base

station takes place.



posed in [5]. The main theme of the proposedextension is to precede data transfers with high-level negotiation using meta-data descriptors asin the SPIN protocol discussed earlier. Thisensures that only data that provides new infor-mation is transmitted to the CHs before beingtransmitted to the BS. Table 1 compares SPIN,LEACH, and directed diffusion according to dif-ferent parameters. It is noted from the table thatdirected diffusion shows a promising approachfor energy-efficient routing in WSNs due to theuse of in-network processing.

Power-Efficient Gathering in Sensor Infor-mation Systems: In [20], an enhancement overthe LEACH protocol was proposed. The proto-col, called Power-Efficient Gathering in SensorInformation Systems (PEGASIS), is a near opti-mal chain-based protocol. The basic idea of theprotocol is that in order to extend network life-time, nodes need only communicate with theirclosest neighbors, and they take turns in commu-nicating with the BS. When the round of allnodes communicating with the BS ends, a newround starts, and so on. This reduces the powerrequired to transmit data per round as the powerdraining is spread uniformly over all nodes.Hence, PEGASIS has two main objectives. First,increase the lifetime of each node by using col-laborative techniques. Second, allow only localcoordination between nodes that are closetogether so that the bandwidth consumed incommunication is reduced. Unlike LEACH,PEGASIS avoids cluster formation and uses onlyone node in a chain to transmit to the BS insteadof multiple nodes.

To locate the closest neighbor node inPEGASIS, each node uses the signal strength tomeasure the distance to all neighboring nodesand then adjusts the signal strength so that onlyone node can be heard. The chain in PEGASISwill consist of those nodes that are closest toeach other and form a path to the BS. Theaggregated form of the data will be sent to theBS by any node in the chain, and the nodes inthe chain will take turns sending to the BS. Thechain construction is performed in a greedyfashion. Simulation results showed that PEGA-SIS is able to increase the lifetime of the net-work to twice that under the LEACH protocol.Such performance gain is achieved through theelimination of the overhead caused by dynamiccluster formation in LEACH, and decreasingthe number of transmissions and reception byusing data aggregation. Although the clusteringoverhead is avoided, PEGASIS still requires

dynamic topology adjustment since a sensornode needs to know about the energy status ofits neighbors in order to know where to route itsdata. Such topology adjustment can introducesignificant overhead, especially for highly uti-lized networks. Moreover, PEGASIS assumesthat each sensor node is able to communicatewith the BS directly. In practical cases, sensornodes use multihop communication to reach theBS. Also, PEGASIS assumes that all nodesmaintain a complete database of the location ofall other nodes in the network. The method bywhich the node locations are obtained is notoutlined. In addition, PEGASIS assumes that allsensor nodes have the same level of energy andare likely to die at the same time. Note also thatPEGASIS introduces excessive delay for distantnodes on the chain. In addition, the single lead-er can become a bottleneck. Finally, although inmost scenarios sensors will be fixed or immobileas assumed in PEGASIS, some sensors may beallowed to move and hence affect the protocolfunctionality.

An extension to PEGASIS, called Hierarchi-cal PEGASIS, was introduced in [2] with theobjective of decreasing the delay incurred forpackets during transmission to the BS. For thispurpose, simultaneous transmissions of data arestudied in order to avoid collisions throughapproaches that incorporate signal coding andspatial transmissions. In the latter, only spatiallyseparated nodes are allowed to transmit at thesame time. The chain-based protocol withCDMA-capable nodes constructs a chain ofnodes that forms a tree-like hierarchy, and eachselected node at a particular level transmits datato a node in the upper level of the hierarchy.This method ensures data transmitting in paral-lel and reduces delay significantly. Such a hierar-chical extension has been shown to performbetter than the regular PEGASIS scheme by afactor of about 60.

Threshold-Sensitive Energy Efficient Proto-cols: Two hierarchical routing protocols calledThreshold-Sensitive Energy Efficient SensorNetwork Protocol (TEEN) and Adaptive Period-ic TEEN (APTEEN) are proposed in [21, 22].These protocols were proposed for time-criticalapplications. In TEEN, sensor nodes sense themedium continuously, but data transmission isdone less frequently. A CH sensor sends itsmembers a hard threshold, which is the thresh-old value of the sensed attribute, and a softthreshold, which is a small change in the value ofthe sensed attribute that triggers the node toswitch on its transmitter and transmit. Thus, thehard threshold tries to reduce the number oftransmissions by allowing the nodes to transmitonly when the sensed attribute is in the range ofinterest. The soft threshold further reduces thenumber of transmissions that might otherwiseoccur when there is little or no change in thesensed attribute. A smaller value of the softthreshold gives a more accurate picture of thenetwork, at the expense of increased energy con-sumption. Thus, the user can control the trade-off between energy efficiency and data accuracy.When CHs are to change (Fig. 5a), new valuesfor the above parameters are broadcast. Themain drawback of this scheme is that if the

n Table 1. Comparison between SPIN LEACH anddirected diffusion.

DirectedSPIN LEACH diffusion

Optimal route No No Yes

Network lifetime Good Very good Good

Resource Yes Yes Yesawareness

Use of meta-data Yes No Yes

PEGASIS assumesthat all sensor nodeshave the same levelof energy and theyare likely to die atthe same time. Notealso that PEGASISintroduces excessivedelay for distantnode on the chain.In addition, the single leader canbecome a bottleneck.



thresholds are not received, the nodes will nevercommunicate, and the user will not get any datafrom the network at all.

The nodes sense their environment continu-ously. The first time a parameter from theattribute set reaches its hard threshold value, thenode switches its transmitter on and sends thesensed data. The sensed value is stored in aninternal variable called sensed value (SV). Thenodes will transmit data in the current clusterperiod only when the following conditions aretrue:• The current value of the sensed attribute is

greater than the hard threshold.• The current value of the sensed attribute dif-

fers from SV by an amount equal to or greaterthan the soft threshold.Important features of TEEN include its suit-

ability for time-critical sensing applications.Also, since message transmission consumes moreenergy than data sensing, the energy consump-tion in this scheme is less than in proactive net-works. The soft threshold can be varied. Atevery cluster change time, fresh parameters arebroadcast, so the user can change them asrequired.

APTEEN, on the other hand, is a hybrid pro-tocol that changes the periodicity or thresholdvalues used in the TEEN protocol according touser needs and the application type. InAPTEEN, the CHs broadcast the followingparameters (Fig. 5b):• Attributes (A): a set of physical parameters

about which the user is interested in obtaininginformation

• Thresholds: consists of the hard threshold(HT) and soft threshold (ST)

• Schedule: a TDMA schedule, assigning a slotto each node

• Count time (CT): the maximum time periodbetween two successive reports sent by a nodeThe node senses the environment continuous-

ly, and only those nodes that sense a data valueat or beyond HT transmit. Once a node senses avalue beyond HT, it transmits data only whenthe value of that attribute changes by an amountequal to or greater than ST. If a node does notsend data for a time period equal to CT, it isforced to sense and retransmit the data. ATDMA schedule is used, and each node in thecluster is assigned a transmission slot. Hence,APTEEN uses a modified TDMA schedule toimplement the hybrid network. The main fea-tures of the APTEEN scheme include the fol-lowing. It combines both proactive and reactivepolicies. It offers a lot of flexibility by allowing

the user to set the CT interval, and the thresholdvalues for energy consumption can be controlledby changing the CT as well as the threshold val-ues. The main drawback of the scheme is theadditional complexity required to implement thethreshold functions and CT. Simulation of TEENand APTEEN has shown that these two proto-cols outperform LEACH. The experiments havedemonstrated that APTEEN’s performance issomewhere between LEACH and TEEN interms of energy dissipation and network lifetime.TEEN gives the best performance since itdecreases the number of transmissions. Themain drawbacks of the two approaches are theoverhead and complexity associated with form-ing clusters at multiple levels, the method ofimplementing threshold-based functions, andhow to deal with attribute-based naming ofqueries.

Small minimum energy communication net-work (MECN): In [23], a protocol is proposedthat computes an energy-efficient subnetwork,the minimum energy communication network(MECN), for a certain sensor network utilizinglow-power GPS. MECN identifies a relay regionfor every node. The relay region consists ofnodes in a surrounding area where transmittingthrough those nodes is more energy-efficientthan direct transmission. The enclosure of anode i is created by taking the union of all relayregions node i can reach. The main idea ofMECN is to find a subnetwork that will havefewer nodes and require less power for transmis-sion between any two particular nodes. In thisway, global minimum power paths are foundwithout considering all the nodes in the network.This is performed using a localized search foreach node considering its relay region. MECN isself-reconfiguring and thus can dynamicallyadapt to node failure or the deployment of newsensors. The small MECN (SMECN) [24] is anextension to MECN. In MECN, it is assumedthat every node can transmit to every othernode, which is not possible every time. InSMECN possible obstacles between any pair ofnodes are considered. However, the network isstill assumed to be fully connected as in the caseof MECN. The subnetwork constructed bySMECN for minimum energy relaying is prov-ably smaller (in terms of number of edges) thanthe one constructed in MECN. Hence, the sub-network (i.e., subgraph G ′) constructed bySMECN is smaller than the one constructed byMECN if the broadcast region is circular aroundthe broadcasting node for a given power setting.Subgraph G′ of graph G, which represents the

nnnn Figure 5. Time line for the operation of a) TEEN and b) APTEEN.

(a)

Parameters

Time

Clusterhead receivesmessage

Cluster changetime

Attribute > threshold

(b)

TDMA scheduleand parameters

Time

Frame timeCluster formationCluster change

time

Slot fornode i

Important features of TEEN include itssuitability for time-

critical sensing applications. Also,

since message transmission

consumes more energy than data

sensing, the energyconsumption in this

scheme is less than in proactive

networks.



sensor network, minimizes the energy usage sat-isfying the following conditions:• The number of edges in G′ is less than in G

while containing all nodes in G.• The energy required to transmit data from a

node to all its neighbors in subgraph G’ is lessthan the energy required to transmit to all itsneighbors in graph G. Assume that r = (u, u1,…, v) is a path between u and v that spans k –1 intermediate nodes u1, … uk–1. The totalpower consumption of one path like r is givenby

where u = u0 and v = uk, and the powerrequired to transmit data under this protocol is

p(u,v) = t.d(u,v)n

for some appropriate constant t, n is the pathloss exponent of outdoor radio propagationmodels n ≥ 2, and d(u,v) is the distance betweenu and v. It is assumed that a reception at thereceiver takes a constant amount of powerdenoted c . The subnetwork computed bySMECN helps in sending messages on mini-mum-energy paths. However, the proposed algo-rithm is local in the sense that it does notactually find the minimum-energy path, it justconstructs a subnetwork in which it is guaran-teed to exist. Moreover, the subnetwork con-structed by SMECN makes it more likely thatthe path used is one that requires less energyconsumption. In addition, finding a subnetworkwith a smaller number of edges introduces moreoverhead in the algorithm.

Self-organizing protocol: Subramanian et al.[25] describes a self-organizing protocol (SOP)and an application taxonomy that was used tobuild architecture to support heterogeneoussensors. Furthermore, these sensors can bemobile or stationary. Some sensors probe theenvironment and forward the data to a designat-ed set of nodes that act as routers. Router nodesare stationary and form the backbone for com-munication. Collected data are forwardedthrough the routers to the more powerful BSnodes. Each sensing node should be able toreach a router in order to be part of the net-work. A routing architecture that requiresaddressing of each sensor node has been pro-posed. Sensing nodes are identifible through theaddress of the router node to which they areconnected. The routing architecture is hierarchi-cal where groups of nodes are formed andmerge when needed. The Local Markov Loops(LML) algorithm, which performs a randomwalk on spanning trees of a graph, was used tosupport fault tolerance and as a means of broad-casting. Such an approach is similar to the ideaof a virtual grid used in some other protocolsdiscussed later under location-based routingprotocols. In this approach, sensor nodes can beaddressed individually in the routing architec-ture; hence, it is suitable for applications wherecommunication to a particular node is required.Furthermore, this algorithm incurs a small costfor maintaining routing tables and keeping abalanced routing hierarchy. It was also found

that the energy consumed for broadcasting amessage is less than that consumed in the SPINprotocol. This protocol, however, is not an on-demand protocol, especially in the organizationphase of the algorithm, and thus introducesextra overhead. Another issue is related to theformation of a hierarchy. It could happen thatthere are many cuts in the network, and hencethe probability of applying reorganization phaseincreases, which is an expensive operation.

Sensor aggregates routing: In [26], a set ofalgorithms for constructing and maintaining sen-sor aggregates were proposed. The objective isto collectively monitor target activity in a certainenvironment (target tracking applications). Asensor aggregate comprises those nodes in a net-work that satisfy a grouping predicate for a col-laborative processing task. The parameters ofthe predicate depend on the task and its resourcerequirements. The formation of appropriate sen-sor aggregates were discussed in [26] in terms ofallocating resources to sensing and communica-tion tasks. Sensors in a sensor field are dividedinto clusters according to their sensed signalstrength, so there is only one peak per cluster.Then local cluster leaders are elected. One peakmay represent one target, multiple targets, or notarget if the peak is generated by noise sources.To elect a leader, information exchangesbetween neighboring sensors are necessary. If asensor, after exchanging packets with all its one-hop neighbors, finds that it is higher than all itsone-hop neighbors on the signal field landscape,it declares itself a leader. This leader-basedtracking algorithm assumes that the unique lead-er knows the geographical region of the collabo-ration.

Three algorithms were proposed in [26]. Firstwas a lightweight protocol, Distributed Aggre-gate Management (DAM), for forming sensoraggregates for a target monitoring task. The pro-tocol comprises a decision predicate P for eachnode to decide if it should participate in anaggregate and a message exchange scheme Mabout how the grouping predicate is applied tonodes. A node determines if it belongs to anaggregate based on the result of applying thepredicate to the data of the node as well asinformation from other nodes. Aggregates areformed when the process eventually converges.Second, Energy-Based Activity Monitoring(EBAM) estimates the energy level at each nodeby computing the signal impact area, combininga weighted form of the detected target energy ateach impacted sensor, assuming that each targetsensor has equal or constant energy level. Thethird algorithm, Expectation-Maximization LikeActivity Monitoring (EMLAM), removes theconstant and equal target energy level assump-tion. EMLAM estimates the target positions andsignal energy using received signals, and uses theresulting estimates to predict how signals fromthe targets may be mixed at each sensor. Thisprocess is iterated until the estimate is sufficient-ly good.

The distributed track initiation managementscheme, combined with the leader-based track-ing algorithm described in [26], forms a scalablesystem. The system works well in tracking multi-ple targets when the targets are not interfering,

C r p u u ci ii

k( ) ( ( , ) )= ++

=

−∑ 1

0

1

The subnetwork constructed bySMECN makes itmore likely that thepath used is onethat requires lessenergy consumption.In addition, finding asub-network with asmaller number ofedges introducesmore overhead inthe algorithm.



and it can recover from intertarget interferenceonce the targets move apart.

Virtual grid architecture routing: An energy-efficient routing paradigm is proposed in [27]that utilizes data aggregation and in-networkprocessing to maximize the network lifetime.Due to the node stationarity and extremely lowmobility in many applications in WSNs, a rea-sonable approach is to arrange nodes in a fixedtopology, as briefly mentioned in [28]. A GPS-free approach [2] is used to build clusters thatare fixed, equal, adjacent, and nonoverlappingwith symmetric shapes. In [27], square clusterswere used to obtain a fixed rectilinear virtualtopology. Inside each zone, a node is optimallyselected to act as CH. Data aggregation is per-formed at two levels: local and then global. Theset of CHs, also called local aggregators (LAs),perform local aggregation, while a subset ofthese LAs are used to perform global aggrega-tion. However, the determination of an optimalselection of global aggregation points, calledmaster aggregators (MAs), is NP-hard. Figure 6illustrates an example of fixed zoning and theresulting virtual grid architecture (VGA) used toperform two-level data aggregation. Note thatthe location of the BS is not necessarily at theextreme corner of the grid; it can be located atany arbitrary place.

Two solution strategies for the routing withdata aggregation problem are presented in [27]:an exact algorithm using an integer linear pro-gram (ILP) formulation, and some near-optimalbut simple and efficient approximate algorithms:a genetics-algorithm-based heuristic, a k-meansheuristic, and a greedy-based heuristic. In [29],another efficient heuristic, the Clustering-BasedAggregation Heuristic (CBAH), was also pro-posed to minimize energy consumption in thenetwork and hence prolong the network lifetime.The objective of all algorithms is to select anumber of MAs out of the LAs that maximizenetwork lifetime. For a realistic scenario, it isassumed in [27] that LA nodes form possiblyoverlapping groups. Members of each group sen-sie the same phenomenon; hence, their readingsare correlated. However, each LA node thatexists in the overlapping region will send data toits associated MA for each of the groups towhich it belongs. It was noted in [29] that theproblem of assigning MAs to LAs in CBAH issimilar to the classical bin packing problem, amajor difference being that neither the identitiesnor the amount of power each MA will be usingfor different LAs are known. In CBAH, the setof MAs are selected based on incremental filingof some bins with capacities. Besides being fastand scalable to large sensor networks, theapproximate algorithms in [27, 29] produceresults not far from the optimal solution.

Hierarchical power-aware routing: In [30],hierarchical power-aware routing was proposed.The protocol divides the network into groups ofsensors. Each group of sensors in geographicproximity are clustered together as a zone, andeach zone is treated as an entity. To performrouting, each zone is allowed to decide how itwill route a message hierarchically across theother zones such that the battery lives of thenodes in the system are maximized. Message are

routed along the path that has the maximumover all the minimum of the remaining power,called the max-min path. The motivation is thatusing nodes with high residual power may bemore expensive than the path with the minimalpower consumption. An approximation algo-rithm, called the max-min zPmin algorithm, wasproposed in [30]. The crux of the algorithm isbased on the trade-off between minimizing thetotal power consumption and maximizing theminimal residual power of the network. Hence,the algorithm tries to enhance a max-min pathby limiting its power consumption as follows.First, the algorithm finds the path with the leastpower consumption (Pmin) by using the Dijkstraalgorithm. Second, the algorithm finds a paththat maximizes the minimal residual power inthe network. The proposed algorithm tries tooptimize both solution criteria. This is achievedby relaxing the minimal power consumption forthe message to be equal to zPmin with parameterz ≥ 1 to restrict the power consumption for send-ing one message to zPmin. The algorithm con-sumes at most zPmin while maximizing theminimal residual power fraction.

Another algorithm that relies on max-minzPmin, called zone-based routing, is also pro-posed in [30]. Zone-base routing is a hierarchicalapproach where the area covered by the (sensor)

nnnn Figure 6. Regular shape tessellation applied to the network area.

Base station

Sensor node Local aggregator node

Master aggregator node



network is divided into a small number of zones.To send a message across the entire area, a glob-al path from zone to zone is found. The sensorsin a zone autonomously direct local routing andparticipate in estimating the zone power level.Each message is routed across the zones usinginformation about the zone power estimates. Aglobal controller for message routing is assignedthe role of managing the zones. This may be thenode with the highest power. If the network canbe divided into a relatively small number ofzones, the scale for the global routing algorithmis reduced. The global information required tosend each message across is summarized by thepower level estimate of each zone. A zone graphwas used to represent connected neighboringzone vertices if the current zone can go to thenext neighboring zone in that direction. Eachzone vertex has a power level of 1. Each zonedirection vertex is labeled by its estimated powerlevel computed by a procedure, which is a modi-fied Bellman-Ford algorithm. Moreover, twoalgorithms were outlined for local and globalpath selection using the zone graph.

Two-Tier Data Dissemination: An approachin [6], called Two-Tier Data Dissemination(TTDD), provides data delivery to multiplemobile BS. In TTDD, each data source proac-tively builds a grid structure that is used to dis-seminate data to the mobile sinks by assumingthat sensor nodes are stationary and location-aware. In TTDD, sensor nodes are stationaryand location-aware, whereas sinks may changetheir locations dynamically. Once an eventoccurs, sensors surrounding it process the signal,and one of them becomes the source to generatedata reports. Sensor nodes are aware of theirmission, which will not change frequently. Tobuild the grid structure, a data source choosesitself as the start crossing point of the grid, andsends a data announcement message to each ofits four adjacent crossing points using simplegreedy geographical forwarding. When the mes-sage reaches the node closest to the crossingpoint (specified in the message), it will stop.

During this process, each intermediate nodestores the source information and further for-wards the message to its adjacent crossing pointsexcept the one from which the message comes.This process continues until the message stops atthe border of the network. The nodes that storethe source information are chosen as dissemina-tion points. After this process, the grid structureis obtained. Using the grid, a BS can flood aquery, which will be forwarded to the nearestdissemination point in the local cell to receivedata. Then the query is forwarded along otherdissemination points upstream to the source.The requested data then flows down in thereverse path to the sink. Trajectory forwarding isemployed as the BS moves in the sensor field.Although TTDD is an efficient routing approach,there are some concerns about how the algo-rithm obtains location information, which isrequired to set up the grid structure. The lengthof a forwarding path in TTDD is larger than thelength of the shortest path. The authors ofTTDD believe that the suboptimality in the pathlength is worth the gain in scalability. Finally,how TTDD would perform if mobile sensornodes are allowed to move in the network is stillan open question. Comparison results betweenTTDD and directed diffusion showed thatTTDD can achieve longer lifetimes and shorterdata delivery delays. However, the overheadassociated with maintaining and recalculatingthe grid as network topology changes may behigh. Furthermore, TTDD assumed the avail-ability of a very accurate positioning system thatis not yet available for WSNs.

The above mentioned flat and hierarchicalprotocols are different in many aspects. At thispoint, we compare the different routingapproaches for flat and hierarchical sensor net-works as shown in Table 2.

Location-Based Routing Protocols — In this kind ofrouting, sensor nodes are addressed by means oftheir locations. The distance between neighbor-ing nodes can be estimated on the basis of

n Table 2. Hierarchical vs. flat topologies routing.

Hierarchical routing Flat routing

Reservation-based scheduling Contention-based scheduling

Collisions avoided Collision overhead present

Reduced duty cycle due to periodic sleeping Variable duty cycle by controlling sleep time of nodes

Data aggregation by clusterhead Node on multihop path aggregates incoming data from neighbors

Simple but non-optimal routing Routing can be made optimal but with an added complexity.

Requires global and local synchronization Links formed on the fly without synchronization

Overhead of cluster formation throughout the network Routes formed only in regions that have data for transmission

Lower latency as multiple hops network formed by Latency in waking up intermediate nodescluster- heads always available and setting up the multipath

Energy dissipation is uniform Energy dissipation depends on traffic patterns

Energy dissipation cannot be controlled Energy dissipation adapts to traffic pattern

Fair channel allocation Fairness not guaranteed



incoming signal strengths. Relative coordinatesof neighboring nodes can be obtained byexchanging such information between neighbors[1, 2, 31]. Alternatively, the location of nodesmay be available directly by communicating witha satellite using GPS if nodes are equipped witha small low-power GPS receiver [28]. To saveenergy, some location-based schemes demandthat nodes should go to sleep if there is no activ-ity. More energy savings can be obtained by hav-ing as many sleeping nodes in the network aspossible. The problem of designing sleep periodschedules for each node in a localized mannerwas addressed in [32, 28]. In the rest of this sec-tion, we review most of the location- or geo-graphic-based routing protocols.

Geographic Adaptive Fidelity: GAF [28] isan energy-aware location-based routing algo-rithm designed primarily for mobile ad hoc net-works, but may be applicable to sensor networksas well. The network area is first divided intofixed zones and form a virtual grid. Inside eachzone, nodes collaborate with each other to playdifferent roles. For example, nodes will electone sensor node to stay awake for a certainperiod of time, and then the rest go to sleep.This node is responsible for monitoring andreporting data to the BS on behalf of the nodesin the zone. Hence, GAF conserves energy byturning off unnecessary nodes in the networkwithout affecting the level of routing fidelity.Each node uses its GPS-indicated location toassociate itself with a point in the virtual grid.Nodes associated with the same point on thegrid are considered equivalent in terms of thecost of packet routing. Such equivalence isexploited in keeping some nodes located in aparticular grid area in sleeping state in order tosave energy. Thus, GAF can substantiallyincrease the network lifetime as the number ofnodes increases. There are three states definedin GAF: discovery, for determining the neigh-bors in the grid; active, reflecting participationin routing; and sleep, when the radio is turnedoff. In order to handle mobility, each node inthe grid estimates its time of leaving the gridand sends this to its neighbors. The sleepingneighbors adjust their sleeping time accordinglyin order to keep routing fidelity. Before theleaving time of the active node expires, sleepingnodes wake up and one of them becomes active.GAF is implemented both for nonmobility(GAF-basic) and mobility (GAF-mobility adap-tation) of nodes. Figure 7 shows an example offixed zoning that can be used in sensor networkssimilar to that proposed in [28]. The fixed clus-ters in [28] are selected to be equal and square.The selection of the square size is dependent onthe required transmitting power and communi-cation direction. Vertical and horizontal com-munication is guaranteed to happen if the signaltravels a distance of a = r/2√2, chosen such thatany two sensor nodes in adjacent vertical orhorizontal clusters can communicate directly.For diagonal communication to happen, the sig-nal has to span a distance of b = r/2√2. Theissue is how to schedule roles for the nodes toact as CHs. A CH can ask the sensor nodes inits cluster to switch on and start gathering dataif it senses an object. Then the CH is responsi-

ble for receiving raw data from other nodes inits cluster and forwarding it to the BS. Theauthors in [28] assumed that sensor nodes canknow their locations using GPS cards, which isinconceivable with current technology. GAFstrives to keep the network connected by keep-ing a representative node always in active modefor each region on its virtual grid. Simulationresults show that GAF performs at least as wellas a normal ad hoc routing protocol in terms oflatency and packet loss, and increases the life-time of the network by saving energy. AlthoughGAF is a location-based protocol, it may also beconsidered a hierarchical protocol, where theclusters are based on geographic location. Foreach particular grid area, a representative nodeacts as the leader to transmit the data to othernodes. The leader node, however, does not doany aggregation or fusion as in the case of otherhierarchical protocols discussed earlier.

Geographic and Energy Aware Routing: Yuet al. [33] discussed the use of geographic infor-mation while disseminating queries to appropri-ate regions since data queries often includegeographic attributes. The protocol, Geographicand Energy Aware Routing (GEAR), uses ener-gy-aware and geographically informed neighborselection heuristics to route a packet toward thedestination region. The key idea is to restrict thenumber of interests in directed diffusion by onlyconsidering a certain region rather than sendingthe interests to the whole network. By doingthis, GEAR can conserve more energy thandirected diffusion.

Each node in GEAR keeps an estimatedcost and a learning cost of reaching the destina-tion through its neighbors. The estimated costis a combination of residual energy and dis-tance to destination. The learned cost is arefinement of the estimated cost that accountsfor routing around holes in the network. A holeoccurs when a node does not have any closerneighbor to the target region than itself. Ifthere are no holes, the estimated cost is equalto the learned cost. The learned cost is propa-

n Figure 7. An example of zoning in sensor net-works.

Local aggregator (LA)

a

a

r

b

To save energy,some location based

schemes demandthat nodes should go

to sleep if there isno activity. More

energy savings canbe obtained by

having as manysleeping nodes in the network

as possible.



gated one hop back every time a packet reachesthe destination so that route setup for the nextpacket will be adjusted. There are two phasesin the algorithm:• Forwarding packets toward the target region:

Upon receiving a packet, a node checks itsneighbors to see if there is one neighbor thatis closer to the target region than itself. Ifthere are more than one, the nearest neighborto the target region is selected as the nexthop. If they are all further than the nodeitself, this means there is a hole. In this case,one of the neighbors is picked to forward thepacket based on the learning cost function.This choice can then be updated according tothe convergence of the learned cost during thedelivery of packets

• Forwarding the packets within the region: Ifthe packet has reached the region, it can bediffused in that region by either recursive geo-graphic forwarding or restricted flooding.Restricted flooding is good when the sensorsare not densely deployed. In high-density net-works, recursive geographic forwarding ismore energy-efficient than restricted flooding.In that case, the region is divided into four suregions and four copies of the packet are cre-ated. This splitting and forwarding processcontinues until regions with only one node areleft.In [33], GEAR was compared to a similar

non-energy-aware routing protocol, GPSR [34],which is one of the earlier methods in geograph-ic routing and uses planar graphs to solve theproblem of holes. In GPSR, the packets followthe perimeter of the planar graph to find theirroute. Although the GPSR approach reduces thenumber of states a node should keep, it wasdesigned for general mobile ad hoc networks,and requires a location service to map locationsand node identifiers. GEAR not only reducesenergy consumption for route setup, but alsoperforms better than GPSR in terms of packetdelivery. The simulation results show that foruneven traffic distribution, GEAR delivers 70–80percent more packets than GPSR. For uniformtraffic pairs GEAR delivers 25–35 percent morepackets than GPSR.

MFR, DIR, and GEDIR: Stojmenovic andLin [35] described and discussed basic localizedrouting algorithms. These protocols deal withbasic distance, progress, and direction-basedmethods. The key issues are forward and back-ward directions. A source node or any interme-diate node will select one of its neighborsaccording to a certain criterion. The routingmethods that belong to this category are MostForward within Radius (MFR), GeographicDistance Routing (GEDIR) that is a variant ofgreedy algorithms, the two-hop greedy method,alternate greedy method, and DIR (a compassrouting method). GEDIR is a greedy algorithmthat always moves the packet to the neighbor ofthe current vertex whose distance to the desti-nation is minimized. The algorithm fails whenthe packet crosses the same edge twice in suc-cession. In most cases, the MFR and greedymethods have the same path to the destination.In the DIR method, the best neighbor has theclosest direction (i.e., angle) toward the desti-

nation. That is, the neighbor with the minimumangular distance from the imaginary line joiningthe current node and the destination is select-ed. In MFR, the best neighbor A will minimizethe dot product DA— .DS—, where S, D are thesource and destination nodes, respectively, andSD— represents the Euclidian distance betweenthe two nodes S, D. Alternatively, one can max-imize the dot product SD—.SA—. Each methodstops forwarding the message at a node forwhich the best choice is to return the messageback to a previous node. GEDIR and MFRsare loop-free, while DIR may create loopsunless past traffic is memorized or a time-stamp is enforced [35].

A comparison study [35] between these algo-rithms showed that the three basic algorithmshad comparable performance in terms of deliv-ery rate and average dilation. Moreover, simula-tions revealed that the nodes in MFR and greedymethods select the same forwarding neighbor inmore than 99 percent of cases, and the entireselected paths were identical in most cases.

The Greedy Other Adaptive Face Routing: In[36], a geometric ad hoc routing algorithm com-bining greedy and face routing was proposed.We will now briefly review the key points ofGreedy Other Adaptive Face Routing(GOAFR). The greedy algorithm of GOAFRalways picks the neighbor closest to a node to benext for routing. However, it can easily be stuckat some local minimum (i.e., no neighbor is clos-er to a node than the current node). Other FaceRouting (OFR) is a variant of Face Routing(FR). The FR algorithm [35] is the first thatguarantees success if the source and destinationare connected. However, the worst case cost ofFR is proportional to the size of the network interms of number of nodes. The first algorithmthat can compete with the best route in theworst case is Adaptive Face Routing (AFR).Moreover, by a lower bound argument, AFR isshown to be asymptotically worst-case optimal.But AFR is not average-case efficient. OFR uti-lizes the face structure of planar graphs suchthat the message is routed from node s to node tby traversing a series of face boundaries. Theaim is to find the best node on the boundary(i.e., the closest node to the destination t) byusing geometric planes. When finished, the algo-rithm returns to s the best node on the bound-ary. The simple greedy algorithm behaves well indense networks, but fails for very simple configu-rations, as was shown in [36]. It was shown thatGOAFR can achieve both worst-case optimalityand average-case efficiency. Based on the simu-lation results of GOAFR, there are several waysto further improve the average-case perfor-mance. It was also shown that GOAFR outper-forms other prominent algorithms, such as GPSRand AFR.

SPAN: Another position-based algorithmcalled SPAN [32] selects some nodes as coor-dinators based on their positions. The coordi-nators form a network backbone used toforward messages. A node should become acoordinator if two neighbors of a non-coordi-nator node cannot reach each other directly orvia one or two coordinators (three-hop reacha-bility). New and existing coordinators are not

The simulationresults show that for an uneven trafficdistribution, GEARdelivers 70 percentto 80 percent morepackets than GPSR.For uniform trafficpairs GEAR delivers25 percent to 35percent more packets than GPSR.



necessarily neighbors in [32], which in effectmakes the design less energy-efficient becauseof the need to maintain the positions of two-or three-hop neighbors in the complicatedSPAN algorithm.

ROUTING PROTOCOLS BASED ONPROTOCOL OPERATION

In this section we review routing protocols withdifferent routing functionality. It should benoted that some of these protocols may fallunder one or more of the above routing cate-gories.

Multipath Routing Protocols — In this subsection westudy routing protocols that use multiple pathsrather than a single path in order to enhancenetwork performance. The fault tolerance(resilience) of a protocol is measured by thelikelihood that an alternate path exists betweena source and a destination when the primarypath fails. This can be increased by maintainingmultiple paths between the source and destina-tion at the expense of increased energy con-sumption and traffic generation. These alternatepaths are kept alive by sending periodic mes-sages. Hence, network reliability can beincreased at the expense of increased overheadin maintaining the alternate paths.

The authors in [37] proposed an algorithmthat routes data through a path whose nodeshave the largest residual energy. The path ischanged whenever a better path is discovered.The primary path will be used until its energyfalls below the energy of the backup path, atwhich time the backup path is used. Using thisapproach, the nodes in the primary path will notdeplete their energy resources through continualuse of the same route, hence achieving longerlife. However, the path switching cost was notquantified in the article.

The authors of [38] proposed the use of a setof suboptimal paths occasionally to increase thelifetime of the network. These paths are chosenby means of a probability that depends on howlow the energy consumption of each path is.

The path with the largest residual energywhen used to route data in a network may bevery energy-expensive too, so there is a trade-off between minimizing the total power con-sumed and the residual energy of the network.The authors in [30] proposed an algorithm inwhich the residual energy of the route isrelaxed a bit in order to select a more energy-efficient path.

In [39], multipath routing was used toenhance the reliability of WSNs. The proposedscheme is useful for delivering data in unreliableenvironments. It is known that network reliabili-ty can be increased by providing several pathsfrom source to destination and sending the samepacket on each path. However, using this tech-nique, traffic will increase significantly. Hence,there is a trade-off between the amount of traf-fic and the reliability of the network. This trade-off is studied in [39] using a redundancy functionthat is dependent on the multipath degree andfailing probabilities of the available paths. Theidea is to split the original data packet into sub-

packets and then send each subpacket throughone of the available multipaths. It has beenfound that even if some of these subpackets arelost, the original message can still be recon-structed. According to their algorithm, it hasalso been found that for a given maximum nodefailure probability, using a higher multipathdegree than a certain optimal value will increasethe total probability of failure.

Directed diffusion [12] is a good candidatefor robust multipath routing and delivery. Basedon the directed diffusion paradigm, a multipathrouting scheme that finds several partially dis-joint paths is studied in [40] (alternate routes arenot node disjoint, i.e., routes are partially over-lapped). It has been found that the use of multi-path routing provides a viable alternative forenergy-efficient recovery from failures in WSNs.The motivation for using these braided paths isto keep the cost of maintaining the multipathslow. The costs of alternate paths are comparableto the primary path because they tend to bemuch closer to the primary path.

Query-Based Routing — In this kind of routing, thedestination nodes propagate a query for data(sensing task) from a node through the network,and a node with this data sends the data thatmatches the query back to the node that initiat-ed the query. Usually these queries are describedin natural language or high-level query lan-guages. For example, client C1 may submit aquery to node N1 and ask: Are there moving vehi-cles in battle space region 1? All the nodes havetables consisting of the sensing task queries theyreceive, and send data that matches these taskswhen they receive it. Directed diffusion [12]described earlier is an example of this type ofrouting. In directed diffusion, the BS node sendsout interest messages to sensors. As the interestis propagated throughout the sensor network,the gradients from the source back to the BS areset up. When the source has data for the inter-est, the source sends the data along the interest’sgradient path. To lower energy consumption,data aggregation (e.g., duplicate suppression) isperformed en route.

The rumor routing protocol [41] uses a set oflong-lived agents to create paths that are direct-ed toward the events they encounter. Wheneveran agent crosses a path leading to an event it hasnot encountered yet, it creates a path state thatleads to the event. When agents come acrossshorter paths or more efficient paths, they opti-mize the paths in routing tables accordingly.Each node maintains a list of its neighbors andan events table that is updated whenever newevents are encountered. Each node can also gen-erate an agent in a probabilistic fashion. Eachagent contains an events table that is synchro-nized with every node it visits. The agent has alifetime of a certain number of hops, after whichit dies. A node will not generate a query unlessit learns a route to the required event. If there isno route available, the node transmits a query ina random direction. Then the node waits toknow if the query reached the destination for acertain amount of time, after which the nodefloods the network if no response is receivedfrom the destination.

It was shown thatGOAFR can achieve

both worst-case opti-mality and average-

case efficiency.Based on the

simulation results ofGOAFR, there are

several ways to further improve the

average-case performance. It was

also shown thatGOAFR outperforms

other prominentalgorithms, such as

GPSR or AFR.



Negotiation-Based Routing Protocols — These proto-cols use high-level data descriptors in order toeliminate redundant data transmissions throughnegotiation. Communication decisions are alsomade based on the resources available to them.The SPIN family protocols [9] discussed earlierand the protocols in [10] are examples of nego-tiation-based routing protocols. The motivationis that the use of flooding to disseminate datawill produce implosion and overlap between thesent data, so nodes will receive duplicate copiesof the same data. This operation consumesmore energy and processing by sending thesame data by different sensors. The SPIN proto-cols are designed to disseminate the data of onesensor to all other sensors, assuming these sen-sors are potential BSs. Hence, the main idea ofnegotiation-based routing in WSNs is to sup-press duplicate information and prevent redun-dant data from being sent to the next sensor orthe BS by conducting a series of negotiationmessages before the real data transmissionbegins.

QoS-based Routing — In QoS-based routing proto-cols, the network has to balance between energyconsumption and data quality. In particular, thenetwork has to satisfy certain QoS metrics(delay, energy, bandwidth, etc.) when deliveringdata to the BS.

Sequential Assignment Routing (SAR) pro-posed in [42] is one of the first routing proto-cols for WSNs to introduce the notion of QoSinto routing decisions. A routing decision inSAR is dependent on three factors: energyresources, QoS on each path, and the prioritylevel of each packet. To avoid single route fail-ure, a multipath approach and localized pathrestoration schemes are used. To create multi-ple paths from a source node, a tree rooted atthe source node to the destination nodes (i.e.,the set of BSs) is built. The paths of the tree arebuilt while avoiding nodes with low energy orQoS guarantees. At the end of this process,each sensor node will be part of a multipathtree. As such, SAR is a table-driven multipathprotocol that aims to achieve energy efficiencyand fault tolerance. In essence, SAR calculatesa weighted QoS metric as the product of theadditive QoS metric and a weight coefficientassociated with the priority level of the packet.The objective of SAR is to minimize the aver-age weighted QoS metric throughout the life-time of the network. If topology changes due tonode failures, path recomputation is needed. Asa preventive measure, a periodic recomputationof paths is triggered by the BS to account forany changes in topology. A handshake proce-dure based on a local path restoration schemebetween neighboring nodes is used to recoverfrom a failure. Failure recovery is done byenforcing routing table consistency betweenupstream and downstream nodes on each path.Simulation results showed that SAR offers lesspower consumption than the minimum energymetric algorithm, which focuses only the energyconsumption of each packet without consideringits priority. SAR maintains multiple paths fromnodes to BS. Although this ensures fault toler-ance and easy recovery, the protocol suffers

from the overhead of maintaining the tables andstates at each sensor node, especially when thenumber of nodes is huge.

Another QoS routing protocol for WSNs thatprovides soft real-time end-to-end guaranteeswas introduced in [43]. The protocol requireseach node to maintain information about itsneighbors and uses geographic forwarding tofind the paths. In addition, SPEED strives toensure a certain speed for each packet in thenetwork so that each application can estimatethe end-to-end delay for the packets by dividingthe distance to the BS by the speed of the pack-et before making an admission decision. More-over, SPEED can provide congestion avoidancewhen the network is congested. The routingmodule in SPEED is called Stateless GeographicNondeterministic Forwarding (SNFG) and workswith four other modules at the network layer.Delay estimation at each node is basically madeby calculating the elapsed time before an ACK isreceived from a neighbor as a response to atransmitted data packet. By looking at the delayvalues, SNGF selects the node that meets thespeed requirement. If it fails, the relay ratio ofthe node is checked, calculated by looking at themiss ratios of the neighbors of a node (the nodesthat could not provide the desired speed) and isfed to the SNGF module. When compared toDSR and AOVD, SPEED performs better interms of end-to-end delay and miss ratio. More-over, the total transmission energy is less due tothe simplicity of the routing algorithm; controlpacket overhead is less. However, SPEED doesnot consider any further energy metric in itsrouting protocol. Therefore, for more realisticunderstanding of SPEED’s energy consumption,there is a need to compare it to a routing proto-col that is energy-aware.

Coherent and Noncoherent Processing — Data process-ing is a major component in the operation ofwireless sensor networks. Hence, routing tech-niques employ different data processing tech-niques. In general, sensor nodes will cooperatewith each other in processing different dataflooded in the network area. Two examples ofdata processing techniques proposed in WSNsare coherent and noncoherent data-processing-based routing [42]. In noncoherent data process-ing routing, nodes will locally process the rawdata before it is sent to other nodes for furtherprocessing. The nodes that perform further pro-cessing are called aggregators. In coherent rout-ing, the data is forwarded to aggregators afterminimum processing. The minimum processingtypically includes tasks like timestamping andduplicate suppressio. To perform energy-effi-cient routing, coherent processing is normallyselected.

Noncoherent functions have fairly low datatraffic loading. On the other hand, since coher-ent processing generates long data streams,energy efficiency must be achieved by path opti-mality. In noncoherent processing, data process-ing incurs three phases:• Target detection, data collection, and prepro-

cessing• Membership declaration• Central node election

The main idea ofnegotiation-basedrouting in WSNs isto suppress duplicateinformation and prevent redundantdata from being sentto the next sensor orthe basestation byconducting a seriesof negotiation messages before the real data transmission begins.



During phase 1, a target is detected, its data col-lected and preprocessed. When a node decidesto participate in a cooperative function, it willenter phase 2 and declare this intention to allneighbors. This should be done as soon as possi-ble so that each sensor has a local understandingof the network topology. Phase 3 is the electionof the central node. Since the central node isselected to perform more sophisticated informa-tion processing, it must have sufficient energyreserves and computational capability.

In [42], single and multiple winner algorithmswere proposed for noncoherent and coherentprocessing, respectively. In the single winneralgorithm (SWE), a single aggregator node iselected for complex processing. The election of anode is based on the energy reserves and com-putational capability of that node. By the end ofthe SWE process, a minimum-hop spanning treewill completely cover the network. In the multi-ple winner algorithm (MWE), a simple extensionto SWE is proposed. When all nodes are sourcesand send their data to the central aggregatornode, a large amount of energy will be con-sumed; hence, this process has a high cost. Oneway to lower the energy cost is to limit the num-ber of sources that can send data to the centralaggregator node. Instead of keeping a record ofonly the best candidate node (master aggregatornode), each node will keep a record of up to nnodes of those candidates. At the end of theMWE process, each sensor in the network has aset of minimum-energy paths to each sourcenode (SN). After that, SWE is used to find thenode that yields the minimum energy consump-tion. This node can then serve as the centralnode for coherent processing. In general, theMWE process has longer delay, higher overhead,and lower scalability than that for noncoherentprocessing networks.

We observed that there are some hybrid pro-tocols that fit under more than one category. Wesummarize recent results on data routing inWSNs in Table 3. The table shows how differentrouting protocols ft under different categoriesand also compares different routing techniquesaccording to many metrics.

ROUTING IN WSNS: FUTURE DIRECTIONS

The future vision of WSNs is to embed numer-ous distributed devices to monitor and interactwith physical world phenomena, and to exploitspatially and temporally dense sensing and actu-ation capabilities of those sensing devices. Thesenodes coordinate among themselves to create anetwork that performs higher-level tasks.

Although extensive efforts have been exertedso far on the routing problem in WSNs, thereare still some challenges that confront effectivesolutions to the routing problem. First, there istight coupling between sensor nodes and thephysical world. Sensors are embedded in unat-tended places or systems. This is different fromtraditional Internet, PDA, and mobility applica-tions that interface primarily and directly withhuman users. Second, sensors are characterizedby a small footprint, and as such nodes presentstringent energy constraints since they areequipped with small finite energy sources. This

is also different from traditional fixed butreusable resources. Third, communications isthe primary consumer of energy in this environ-ment where sending a bit over 10 or 100 m con-sumes as much energy as thousands to millionsof operations (known as R4 signal energydropoff) [44].

Although the performance of these protocolsis promising in terms of energy efficiency, fur-ther research is needed to address issues such asQoS posed by video and imaging sensors andreal-time applications. Energy-aware QoS rout-ing in sensor networks will ensure guaranteedbandwidth (or delay) through the duration ofconnection as well as provide the use of themost energy efficient path. Another interestingissue for routing protocols is the considerationof node mobility. Most current protocols assumethat the sensor nodes and BS are stationary.However, there might be situations such as bat-tle environments where the BS and possibly thesensors need to be mobile. In such cases, fre-quent update of the position of the commandnode and sensor nodes and propagation of thatinformation through the network may excessivelydrain the energy of nodes. New routing algo-rithms are needed in order to handle the over-head of mobility and topology changes in suchan energy-constrained environment. Futuretrends in routing techniques in WSNs focus ondifferent directions; all share the common objec-tive of prolonging network lifetime. We summa-rize some of these directions and give somepertinent references as follows:

•Exploit redundancy: Typically a large num-ber of sensor nodes are implanted inside orbeside the phenomenon. Since sensor nodes areprone to failure, fault tolerance techniques comeinto the picture to keep the network operatingand performing its tasks. Routing techniquesthat explicitly employ fault tolerance techniquesin an efficient manner are still under investiga-tion (e.g., [39]).

•Tiered architectures (mix of form/energyfactors): Hierarchical routing is an old techniqueto enhance scalability and efficiency of the rout-ing protocol. However, novel techniques of net-work clustering that maximize network lifetimeare also a hot area of research in WSNs (e.g.,[45]).

•Exploit spatial diversity and density of sen-sor/actuator nodes: Nodes will span a networkarea that might be large enough to provide spa-tial communication between sensor nodes.Achieving energy-efficient communication in thisdensely populated environment deserves furtherinvestigation. Dense deployment of sensor nodesshould allow the network to adapt to an unpre-dictable environment.

•Achieve desired global behavior with adap-tive localized algorithms (i.e., do not rely onglobal interaction or information): However, in adynamic environment, this is hard to model (e.g.,[12]).

•Leverage data processing inside the networkand exploit computation near data sources toreduce communication (i.e., perform in-networkdistributed processing): WSNs are organizedaround naming data, not nodes’ identities. Sincewe have large collections of distributed ele-

The future vision ofWSNs is to embed

numerous distributeddevices to monitor

and interact withphysical world

phenomena, and toexploit spatially and

temporally densesensing and

actuation capabilitiesof those sensing

devices.



ments, localized algorithms that achieve system-wide properties in terms of local processing ofdata before it is sent to the destination are stillneeded. Nodes in the network will store nameddata and make it available for processing. Thereis a high need to create efficient processingpoints in the network (e.g., duplicate suppres-sion, aggregation, correlation of data). How toefficiently and optimally find those points is stillan open research issue (e.g., [27]).

•Time and location synchronization: Energy-efficient techniques for associating time and spa-tial coordinates with data to supportcollaborative processing are also required [1].

•Localization: Sensor nodes are randomlydeployed into an unplanned infrastructure. Theproblem of estimating spatial coordinates of the

node is referred to as localization. GPS cannotbe used in WSNs as GPS can work only out-doors and not in the presence of any obstruc-tion. Moreover, GPS receivers are expensive andunsuitable for the construction of small cheapsensor nodes. Hence, there is a need to developother means of establishing a coordinate systemwithout relying on an existing infrastructure.Most of the proposed localization techniquestoday depend on recursive trilateration/multilat-eration techniques (e.g., [46]), which would notprovide enough accuracy in WSNs.

•Self-configuration and reconfiguration areessential to the lifetime of unattended systems ina dynamic and energy constrained environment.This is important for keeping the network upand running. As nodes die and leave the net-

n Table 3. Classification and comparison of routing protocols in wireless sensor networks.

Classifi- Mobility Position Power Negotiation- Data aggre- Local- QoS State comp- Scalab- Multi- Query-cation awareness usage based gation ization lexity ility path based

SPIN Flat Poss. No Ltd. Yes Yes No No Low Ltd. Yes Yes

Direct Flat Ltd. No Ltd. Yes Yes Yes No Low Ltd. Yes Yesdiffusion

Rumor Flat Very Ltd. No N/A No Yes No No Low Good No Yesrouting

GBR Flat Ltd. No N/A No Yes No No Low Ltd. No Yes

MCFA Flat No No N/A No No No No Low Good No No

CADR Flat No No Ltd. No Yes No No Low Ltd. No No

COUGAR Flat No No Ltd. No Yes No No Low Ltd. No Yes

ACQUIRE Flat Ltd. No N/A No Yes No No Low Ltd. No Yes

EAR Flat Ltd. No N/A No No No Low Ltd. No Yes

LEACH Hierarchical Fixed BS No Max. No Yes Yes No CHs Good No No

TEEN & Hierarchical Fixed BS No Max. No Yes Yes No CHs Good No NoAPTEEN

PEGASIS Hierarchical Fixed BS No Max. No No Yes No Low Good No No

MECN & Hierarchical No No Max. No No No No Low Low No NoSMECN

OP Hierarchical No No N/A No No No No Low Low No No

HPAR Hierarchical No No N/A No No No No Low Good No No

VGA Hierarchical No No N/A Yes Yes Yes No CHs Good Yes No

Sensor Hierarchical Ltd. No N/A No Yes No No Low Good No Poss.aggregate

TTDD Hierarchical Yes Yes Ltd. No No No No Mod. Low Poss. Poss.

GAF Location Ltd. No Ltd. No No No No Low Good No No

GEAR Location Ltd. No Ltd. No No No No Low Ltd. No No

SPAN Location Ltd. No N/A Yes No No No Low Ltd. No No

MFR, Location No No N/A No No No No Low Ltd. No NoGEDIR

GOAFR Location No No N/A No No No Low Good No No

SAR Location No No N/A Yes Yes No Yes Mod. Ltd. No Yes

SPEED QoS No No N/A No No No Yes Mod. Ltd. No Yes



work, update and reconfiguration mechanismsshould take place. A feature that is important inevery routing protocol is to adapt to topologychanges very quickly and to maintain the net-work functions (e.g., [9]).

•Secure routing: Current routing protocolsoptimize for the limited capabilities of nodesand the application-specific nature of networks,but do not consider security. Although theseprotocols have not been designed with securityas a goal, it is important to analyze their securityproperties. One aspect of sensor networks thatcomplicates the design of a secure routing proto-col is in-network aggregation. In WSNs, in-net-work processing makes end-to-end securitymechanisms harder to deploy because intermedi-ate nodes need direct access to the contents ofthe messages (e.g., [47, 48]).

Other possible future research for routingprotocols includes the integration of sensor net-works with wired networks (i.e., the Internet).Most applications in security and environmentalmonitoring require the data collected from sen-sor nodes to be transmitted to a server so thatfurther analysis can be done. On the other hand,the requests from the user should be made tothe BS through the Internet. Since the routingrequirements of each environment are different,further research is necessary for handling thesekinds of situations.

CONCLUSIONS

Routing in sensor networks is a new area ofresearch, with a limited but rapidly growing setof research results. In this article we present acomprehensive survey of routing techniques inwireless sensor networks that have been present-ed in the literature. They have the commonobjective of trying to extend the lifetime of thesensor network while not compromising datadelivery.

Overall, the routing techniques are classifiedbased on the network structure into three cate-gories: flat, hierarchical, and location-basedrouting protocols. Furthermore, these protocolsare classified into multipath-based, query-based,negotiation-based, and QoS-based routing tech-niques depending on protocol operation. Wealso highlight the design trade-offs betweenenergy and communication overhead savings insome of the routing paradigm, as well as theadvantages and disadvantages of each routingtechnique. Although many of these routing tech-niques look promising, there are still many chal-lenges that need to be solved in sensor networks.We highlight those challenges and pinpointfuture research directions in this regard.

REFERENCES[1] N. Bulusu, J. Heidemann, and D. Estrin, “GPS-less Low

Cost Out Door Localization for Very Small Devices,”Tech. rep. 00729, Comp. Sci. Dept., USC, Apr. 2000.

[2] A. Savvides, C.-C. Han, and M. Srivastava, “DynamicFine-Grained Localization in Ad-Hoc Networks of Sen-sors,” Proc. 7th ACM MobiCom, July 2001, pp. 166–79.

[3] I. Akyildiz et al., “A Survey on Sensor Networks,” IEEECommun. Mag., vol. 40, no. 8, Aug. 2002, pp. 102–14.

[4] S. Tilak, N. Abu-Ghazaleh, W. Heinzelman, “A Taxono-my of Wireless Micro-sensor Network Models,” ACMSIGMOBILE Mobile Comp. Commun. Rev., vol. 6, no. 2,Apr. 2002, pp. 28–36.

[5] W. Heinzelman, A. Chandrakasan and H. Balakrishnan,“Energy-Efficient Communication Protocol for WirelessMicrosensor Networks,” Proc. 33rd Hawaii Int’l. Conf.Sys. Sci., Jan. 2000.

[6] F. Ye et al., “A Two-Tier Data Dissemination Model forLarge-Scale Wireless Sensor Networks,” Proc. ACM/IEEEMOBICOM, 2002.

[7] http://www.ieee802.org/15/[8] F. Ye et al., “A Scalable Solution to Minimum Cost Forward-

ing in Large Sensor Networks,” Proc. 10th Int’l. Conf.Comp. Commun. and Networks, 2001, pp. 304–09.

[9] W. Heinzelman, J. Kulik, and H. Balakrishnan, “AdaptiveProtocols for Information Dissemination in WirelessSensor Networks,” Proc. 5th ACM/IEEE Mobicom, Seat-tle, WA, Aug. 1999. pp. 174–85.

[10] J. Kulik, W. R. Heinzelman, and H. Balakrishnan,“Negotiation-Based Protocols for Disseminating Infor-mation in Wireless Sensor Networks,” Wireless Net-works, vol. 8, 2002, pp. 169–85.

[11] S. Hedetniemi and A. Liestman, “A Survey of Gossip-ing and brocadcasting in Communication Networks,”IEEE Network, vol. 18, no. 4, 1988, pp. 319–49.

[12] C. Intanagonwiwat, R. Govindan, and D. Estrin,“Directed Diffusion: a Scalable and Robust Communica-tion Paradigm for Sensor Networks,” Proc. ACM Mobi-Com 2000, Boston, MA, 2000, pp. 56–67.

[13] D. Braginsky and D. Estrin, “Rumor Routing Algorithmfor Sensor Networks,” Proc. 1st Wksp. Sensor Networksand Apps., Atlanta, GA, Oct. 2002.

[14] C. Schurgers and M.B. Srivastava, “Energy EfficientRouting in Wireless Sensor Networks,” MILCOM Proc.Commun. for Network-Centric Ops.: Creating the Info.Force, McLean, VA, 2001.

[15] M. Chu, H. Haussecker, and F. Zhao, “Scalable Infor-mation Driven Sensor Querying and Routing for Ad HocHeterogeneous Sensor Networks,” Int’l. J. High Perf.Comp. Apps., vol. 16, no. 3, Aug. 2002.

[16] Y. Yao and J. Gehrke, “The Cougar Approach to In-network Query Processing in Sensor Networks,” SIG-MOD Record, Sept. 2002.

[17] N. Sadagopan et al., “The ACQUIRE Mechanism forEfficient Querying in Sensor Networks,” Proc. 1st Int’l.Wksp. Sensor Network Protocol and Apps,. Anchorage,AK, May 2003.

[18] R. C. Shah and J. Rabaey, “Energy Aware Routing forLow Energy Ad Hoc Sensor Networks,” IEEE WCNC,Orlando, FL, Mar. 17–21, 2002.

[19] S. Servetto and G. Barrenechea, “Constrained RandomWalks on Random Graphs: Routing Algorithms forLarge Scale Wireless Sensor Networks,” Proc. 1st ACMInt’l. Wksp. Wireless Sensor Networks and Apps.,Atlanta, GA, 2002.

[20] S. Lindsey and C. Raghavendra, “PEGASIS: Power-Effi-cient Gathering in Sensor Information Systems,” IEEEAerospace Conf. Proc., 2002, vol. 3, 9–16, pp. 1125–30.

[21] A. Manjeshwar and D. P. Agarwal, “TEEN: a Routing Pro-tocol for Enhanced Efficiency in Wireless Sensor Networks,”1st Int’l. Wksp. on Parallel and Distrib. Comp. Issues inWireless Networks and Mobile Comp., April 2001.

[22] A. Manjeshwar and D. P. Agarwal, “APTEEN: A HybridProtocol for Efficient Routing and Comprehensive Infor-mation Retrieval in Wireless Sensor Networks,” Proc.Int’l. Parallel and Distrib. Proc. Symp., pp. 195–202.

[23] V. Rodoplu and T. H. Meng, “Minimum Energy MobileWireless Networks,” IEEE JSAC, vol. 17, no. 8, Aug.1999, pp. 1333–44.

[24] L. Li, and J. Y. Halpern, “Minimum-Energy Mobile Wire-less Networks Revisited,” IEEE ICC 2001, vol. 1, pp. 278–83.

[25] L. Subramanian and R. H. Katz, “An Architecture forBuilding Self Configurable Systems,” Proc. IEEE/ACMWksp. Mobile Ad Hoc Net. and Comp., Boston, MA,Aug. 2000.

[26] Q. Fang, F. Zhao, and L. Guibas, “Lightweight Sensing andCommunication Protocols for Target Enumeration andAggregation,” Proc. 4th ACM MOBIHOC, 2003, pp. 165–76.

[27] J. N. Al-Karaki et al., “Data Aggregation in WirelessSensor Networks — Exact and Approximate Algo-rithms,” Proc. IEEE Wksp. High Perf. Switching andRouting 2004, Phoenix, AZ, Apr. 18–21, 2004.

[28] Y. Xu, J. Heidemann, and D. Estrin, “Geography-informed Energy Conservation for Ad-hoc Routing,”Proc. 7th Annual ACM/IEEE Int’l. Conf. Mobile Comp.and Net., 2001, pp. 70–84.

[29] J. N. Al-Karaki, and A. E. Kamal, “On the CorrelatedData Gathering Problem in Wireless Sensor Networks,”to appear in the Proc. 9th IEEE Symp. Comp. and Com-mun., Alexandria, Egypt, July 2004.

One aspect of sensornetworks that

complicates thedesign of a secure

routing protocol is in-network aggregation.

In WSNs, in-networkprocessing makes

end-to-end securitymechanisms harderto deploy because

intermediate nodesneed direct access to the contents of

the messages.



[30] Q. Li, J. Aslam and D. Rus, “Hierarchical Power-AwareRouting in Sensor Networks,” Proc. DIMACS Wksp. Per-vasive Net., May, 2001.

[31] S. Capkun, M. Hamdi, and J. Hubaux, “GPS-free Posi-tioning in Mobile Ad-hoc Networks,” Proc. 34th AnnualHawaii Int’l. Conf. Sys. Sci., 2001 pp. 3481–90.

[32] B. Chen et al., “SPAN: an Energy-efficient CoordinationAlgorithm for Topology Maintenance in Ad Hoc Wire-less Networks,” Wireless Networks, vol. 8, no. 5, Sept.2002, pp. 481–94.

[33] Y. Yu, D. Estrin, and R. Govindan, “Geographical andEnergy-Aware Routing: A Recursive Data DisseminationProtocol for Wireless Sensor Networks,” UCLA Comp.Sci. Dept. tech. rep., UCLA-CSD TR-010023, May 2001.

[34] B. Karp and H. T. Kung, “GPSR: Greedy PerimeterStateless Routing for Wireless Sensor Networks,” Proc.MobiCom 2000, Boston, MA, Aug. 2000.

[35] I. Stojmenovic and X. Lin, “GEDIR: Loop-Free LocationBased Routing in Wireless Networks,” Int’l. Conf. Paralleland Distrib. Comp. and Sys., Boston, MA, Nov. 3–6, 1999.

[36] F. Kuhn, R. Wattenhofer, and A. Zollinger, “Worst-CaseOptimal and Average-Case Efficient Geometric Ad HocRouting,” Proc. 4th ACM Int’l. Conf. Mobile Comp. andNet., 2003, pp. 267–78.

[37] J.-H. Chang and L. Tassiulas, “Maximum Lifetime Rout-ing in Wireless Sensor Networks,” Proc. Adv. Telecom-mun. and Info. Distrib. Research Prog., College Park,MD, Mar. 2000.

[38] C. Rahul and J. Rabaey, “Energy Aware Routing forLow Energy Ad Hoc Sensor Networks,” IEEE WCNC, vol.1, Mar. 17–21, 2002, Orlando, FL, pp. 350–55.

[39] S. Dulman et al., “Trade-Off between Traffic Overheadand Reliability in Multipath Routing for Wireless SensorNetworks,” WCNC Wksp., New Orleans, LA, Mar. 2003.

[40] D. Ganesan et al., “Highly Resilient, Energy-EfficientMultipath Routing in Wireless Sensor Networks,” ACMSIGMOBILE Mobile Comp. Commun. Rev., vol. 5, no. 4,2001, pp. 11–25.

[41] D. Braginsky and D. Estrin, “Rumor Routing AlgorithmFor Sensor Networks,” Int’l. Conf. Distrib. Comp. Sys.,Nov. 2001.

[42] K. Sohrabi and J. Pottie, “Protocols for Self-Organiza-tion of a Wireless Sensor Network,” IEEE Pers. Com-mun., vol. 7, no. 5, 2000, pp. 16–27.

[43] T. He et al., “SPEED: A Stateless Protocol for Real-timeCommunication in Sensor Networks,” Proc. Int’l. Conf.Distrib. Comp. Sys., Providence, RI, May 2003.

[44] D. Goodman, Wireless Personal Communications Sys-tems, Reading, MA: Addison-Wesley, 1997.

[45] S. Bandyopadhyay and E. Coyle, “An Energy Efficient Hierarchical Clustering Algorithm for WirelessSensor Networks,” Proc. INFOCOM 2003, vol. 3, pp.1713–23.

[46] N. Bulusu et al., “Scalable Coordination for WirelessSensor Networks: Self-Configuring Localization Sys-tems,” Proc. 6th Int’l. Symp. Commun. Theory andApps., Ambleside, U.K., July 2001

[47] A. Perrig et al., “SPINS: Security Protocols for Sensor Net-works,” Wireless Networks, vol. 8, 2000, pp. 521–34.

[48] C. Karlof and D. Wagner, “Secure Routing in WirelessSensor Networks: Attacks and Countermeasures,” AdHoc Networks, vol. 1, 2003, pp. 293–315.

ADDITIONAL READING[1] S. Hedetniemi, S. Hedetniemi, and A. Liestman, “A Sur-

vey of Gossiping and Broadcasting in CommunicationNetworks,” Networks, vol. 18, 1988.

BIOGRAPHIESJAMAL N. AL-KARAKI [M] ([email protected]) is an assistantprofessor in the Electrical and Computer EngineeringDepartment at the Hashemite University, Zarqa, Jordan. Heobtained his Ph.D. in computer engineering from IowaState University in 2004. He received his B.Sc. and M.Sc.degrees in electrical/computer engineering from JordanUniversity of Science and Technology in 1993 and 1995,respectively. His research interests lie in protocols andarchitectures for wireless and mobile networks, particularlymobile ad hoc networks and wireless sensor networks. Heis also interested in fault-tolerant computing and parallelprocessing. He has published more than 20 technicalpapers in these areas.

AHMED E. KAMAL [SM] ([email protected]) received a B.Sc.(distinction with honors) and an M.Sc. both from CairoUniversity, Egypt, and an M.A.Sc. and a Ph.D. both fromthe University of Toronto, Canada, all in electrical engineer-ing, in 1978, 1980, 1982, and 1986, respectively. He is cur-rently a professor of electrical and computer engineering atIowa State University. His research interests include opticalnetworks, wireless and sensor networks, performance eval-uation, and QoS in the Internet.

We presented a comprehensive survey of routingtechniques in wireless sensor networks which havebeen presented inthe literature. Theyhave the commonobjective of trying toextend the lifetimeof the sensor net-work, while not compromising data delivery.


wireless sensor networks routingtechniques in wireless sensor

Documents