A survey on cross-layer solutions for wireless sensor networks

Review

A survey on cross-layer solutions for wireless sensor networks

Lucas D.P. Mendes, Joel J.P.C. Rodrigues !

Instituto de Telecomunicac- ~oes, University of Beira Interior, Rua Marques D’Avila e Bolama, 6201-001 Covilh ~a, Portugal

a r t i c l e i n f o

Article history:Received 7 July 2010Received in revised form11 November 2010Accepted 18 November 2010

Keywords:Cross-layerSurveyTypes of wireless sensor networks (WSNs)

a b s t r a c t

Ever since wireless sensor networks (WSNs) have emerged, different optimizations have been proposedto overcome their constraints. Furthermore, the proposal of new applications forWSNs have also creatednew challenges to be addressed. Cross-layer approaches have proven to be themost efficient optimizationtechniques for these problems, since they are able to take the behavior of the protocols at each layer intoconsideration. Thus, this surveyproposes to identify the keyproblemsofWSNs and gather available cross-layer solutions for them that have been proposed so far, in order to provide insights on the identification ofopen issues and provide guidelines for future proposals.

& 2010 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12. Drivers for cross-layer approaches. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33. Considered technologies at each layer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

3.1. Application layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43.2. Network layer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53.3. Medium access control layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73.4. Physical layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93.5. Multiple-layer standards. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

3.5.1. IEEE 802.15.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93.5.2. IEEE 802.11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93.5.3. IEEE 802.15.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93.5.4. Bluetooth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

4. Open issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95. Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

1. Introduction

Wireless sensor networks (WSNs) have presented interestingand challenging issues since their creation. These networks havebeen drawing even more attention lately, mainly because of theplethora of different applications that have been emerging. Thus,they are the aim of many optimization proposals, and the cross-layer approaches have proved to achieve better optimizationresults than their layered counterparts. As several proposals have

the same objectives, their efficiency comparison becomes impor-tant, and also as they consider different approaches, it is possible todiscover which ones result in more gains to networks, avoidingcomplicated computation and excessive overhead. Furthermore,the mission of the WSN must be considered by the optimizationproposal since, for instance, throughput maximization is not asimportant for an electrocardiogram (ECG) application as it is forvideo transmission. Finally, a look at the past of sensor networkscan reveal the issues as they appeared, and also provide insights forthe future challenges.

The first sensor networks have not considered wireless com-munication (Kohno et al., 1999). Moreover, wired systems havetheir own wired power supply, so energy consumption was not anissue at that time. The main concerns at the second-half of the

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/jnca

Journal of Network and Computer Applications

1084-8045/$ - see front matter & 2010 Elsevier Ltd. All rights reserved.doi:10.1016/j.jnca.2010.11.009

! Corresponding author. Tel.: +351 275319891; fax: +351 275319899.E-mail addresses: [email protected] (L.D. Mendes),

[email protected] (J.J.P.C. Rodrigues).

Please cite this article as: Mendes LDP, J.P.C. Rodrigues JJPC. A survey on cross-layer solutions for wireless sensor networks. J NetworkComput Appl (2010), doi:10.1016/j.jnca.2010.11.009

Journal of Network and Computer Applications ] (]]]]) ]]]–]]]

www.elsevier.com/locate/jnca

dx.doi.org/10.1016/j.jnca.2010.11.009

mailto:[email protected]

mailto:[email protected]

dx.doi.org/10.1016/j.jnca.2010.11.009

dx.doi.org/10.1016/j.jnca.2010.11.009

1990s were on the manufacturing and packaging of sensors andactuators, ways of providing power supply to arrays of devices,internal and external communications (first steps towardswirelesssensors), and signal processing (which is still an important line ofresearch on the area) (Ko, 1996).

Nowadays, it is well known that the advances in radio technol-ogy have achieved a low power consumption level that allowed thecreation of very small battery-powered communication devices.When joinedwith low-poweredmicroprocessors connected to anykind of tiny sensors (e.g., light intensity, temperature, 3D move-ment, heartbeat), a small battery-powered sensor, capable ofprocessing data and communicating with other wireless devicesis created. Furthermore, this new device allows the measurementof some phenomena that occur not only in a single point, but in awide area. Finally, the sensed data can be communicated to one ormore devices called sink nodes, which make the data available to agateway, and further, to the user. Also, the user can change param-eters of the sensor nodes (e.g., measurement report intervals,sensor nodes topology) through the sink nodes. This scenario char-acterizes a WSN (Baronti et al., 2007), as depicted in Fig. 1.

However, WSNs can be deployed in different environments andfor different purposes. Hence, they have received some particulardenominations in the literature. For instance, sensor networksresponsible for transmitting video and/or audio captured in anarea,usually for surveillance purposes, may be called wireless multi-media sensor network (WMSN) or surveillance wireless sensornetwork (SWSN). Sensor networks deployed inside factories arecalled industrial wireless sensor networks (IWSNs) in the litera-ture. When used for healthcare, they can be labeled wireless bodyarea networks (WBANs). Furthermore,when the sensors are able totake actions according to predetermined conditions, the networkmay be called wireless sensor and actuator network (WSAN orSANET). And finally, if the sensors are not in fixed positions, theycan be labeledmobile sensor networks (MSNs) if they cannot reachhigh speeds, or vehicular sensor networks (VSNs) if they are able tomove at greater speeds. These characterizations ofWSNs are usefulto determine the different set of challenges that a sensor networkmay face. Thus, for instance, before reading about a solutiondesigned for an IWSN, one can infer that interference from heavymachinery is going to be considered.

Nevertheless, some problemsmay arise in any type ofWSN. Forinstance, although the components that assemble wireless sensorshave been developed to save energy, sensor nodes consumption isstill one of the main concerns (Akyildiz et al., 2007). The lifetime ofa network must be enough for the mission of the WSN to beaccomplished, thus protocols that waste this resource can make anetwork useless. Also, inefficient routing of packets can lead towaste of energy. A protocol that needs many routing advertise-ments will make use of sensors energy to send them, reducing thenetwork lifetime (Choi et al., 2005). Moreover, the collaboration

of nodes can help to reduce energy consumption by avoidingretransmissionswhenapacket doesnot reach its destination (Lianget al., 2007). Another problem, especially for WMSNs, is quality ofservice (QoS). Video packets cannot experience long delays, or theobjective of the network may be compromised (Chen et al., 2007).Furthermore, wireless links are not reliable, especially for sensorssince they are built with radios to save energy, and not for longrange transmission. Thus, error control techniques are necessary,however they must also be energy efficient (Yu et al., 2007).Moreover, the capacity of the network is a parameter of interest.New sensorsmay be needed to extend the sensing area of a runningWSNapplication.However, the impact of newnodes in thenetworkmust be studied or the performance of the whole network maybe degraded (Liang and Liang, 2009). And like any wireless system,there is also the security problem. Cryptography can protect com-munication from eavesdropping, however the tradeoff betweendata protection and energy consumption must be studied (Misic,2008). And finally, in order to enable WSNs to use contentionlessmedium access and coordinated sleep schedules, they need to besynchronized (Ye et al., 2005). All the solutions for the aforemen-tioned problems conflict directly with energy saving, since thetechniques to solve them require resources from one or morenodes. Thus, every proposed solution need to be energy efficient.

Cross-layer design states that parameters of two or more layerscan be retrieved and/or changed in order to achieve an optimizationobjective. The concept of cross-layering has been first proposed forTCP/IP networks, when wireless links were deployed (Srivastava andMotani, 2005). Since the TCP/IP stack has been proposed for wiredconnections, there was a loss of performance when wireless technol-ogy became part of existing networks. Lately, cross-layering is a fieldthat has beenattractingmore attention inWSNs research and it is stillin its early development in this type of networks since it has not beendeployed on many testbeds or networks yet (Misra et al., 2008).However, different solutions have already been proposed in theliterature, and at least in numerical frameworks or simulations, theyhave proven to achieve better performance gains than their layeredcounterparts. Commongoalsof cross-layer optimizations inWSNsarereduction of energy consumption (Kulkarni et al., 2006), efficientrouting (Choi et al., 2005), QoS provisioning (Yuan et al., 2006b), andoptimal scheduling (Shu and Krunz, 2009), as can be verifiedthroughout this work. Some of the most used protocols on cross-layer design and new protocol proposals are going to be presented,along with the optimizations they provide, in order to gather a smalldatabase on what has been proposed so far.

The goal of this survey is to show how cross-layer design hasbeen adopted onwireless sensor networks (WSNs) to improve theirperformance by discussing the state-of-the-art literature on thesubject. Also, open issues on cross-layering will be identified tofacilitate the work of researchers towards further improvementsapplied to WSNs. Although Akyildiz et al. (2007) have assembled

Fig. 1. Example of a wireless sensor network architecture.

L.D.P. Mendes, J.J.P.C. Rodrigues / Journal of Network and Computer Applications ] (]]]]) ]]]–]]]2


dx.doi.org/10.1016/j.jnca.2010.11.009

a comprehensive survey on wireless multimedia sensor networksand Yick et al. (2008) have done one for general WSNs, the aim ofthis paper is to gather and discuss cross-layer techniques proposedfor all kinds of currently used WSNs.

The remainder of this paper is organized as follows. Section 2provides an insight on the problems frequently addressed by cross-layer approaches. Theprotocols and techniquesused on the studiedsolutions and new proposals are discussed in Section 3. Finally, inSection 4 the conclusions are drawn.

2. Drivers for cross-layer approaches

This section proposes to discuss the main problems addressedby cross-layer design forWSNs, pointing out themain one—energyconsumption—and its relation to the other presented problems.

The most important advantage of wireless sensors is that theycan be placed anywhere and measure phenomena that wiredsensors cannot. To do so, wireless sensors need a source of energy,small enough to allow them to be placed in tight places, and eveninside the human body. Thus, WSNs are power constrained, whichmakes sensors energy an invaluable resource. After the energydepletion of one ormore nodes of aWSN, the application theyweresupposed to feed becomes compromised. Hence, the design ofWSNs must be thoroughly thought in order to be energy-efficient.

According toWang et al. (2008a),WSNs aremission-driven,whichmeans that measurements of the behavior of a phenomenon will becarried out in an enclosed area. This fact can lead to redund-ancy since sensors placed near each other will make measurementswithclose values. Thus, transmitting correlated, or even repeateddatais wasting energy. Moreover, sensor nodes will not transmit orforward packets all the time. Thus, keeping its functionalities turnedon all the time is a waste of energy as well (Ha et al., 2006a,b).

Furthermore, the commonmetric to determine the lifetime of aWSN is the energy depletion of the first node of theWSN. However,this metric may not be precise when the running applicationtolerates a determined number of packet losses or a determineddelay (Ozgovde and Ersoy, 2009). Thus, the lifetime of the networkshouldbe application-specific, andhence eachdifferent applicationmust be studied to determine if the first node energy depletion isenough to characterize the lifetime of the WSN. If it is not, newmetrics should be proposed.

The used routing protocol is tightly coupled with the WSNlifetime. A routing protocol might require the flooding of controlpackets to determine the routes, which induces an initial wasteof energy. Also, topology changes are very likely to occur whenconsidering WSNs, especially because of nodes that leave thenetwork due to energy depletion. Furthermore, constant controlmessages exchange may be necessary to keep information aboutroutes, adding transmission overhead and consuming sensorsenergy. Thus, from these problems it is possible to see that routingprotocols for WSNs must be energy-aware and energy-efficient,adding the least overhead possible to avoid reducing the networklifetime to unacceptable thresholds (Zhang and Zhang, 2009).

WSNs are also used to measure mission-critical data (e.g., videoand/or audio for surveillance, health monitoring data for criticalpatients). Thus, transmission of data through these networks mustcope with strict delay constraints and provide a suitable aggregatethroughput. It is worth noting that these QoS requirements conflictdirectly with energy consumption issues since they may requirethe use of more transmission power to reduce the occurrence ofchannel errors and amore frequent access to themedium (Melodiaand Akyildiz, 2010). Hence, the tradeoff between QoS and energyconsumption must be thoroughly studied.

Since wireless media are used to transmit data, packets aresubject to interference from other transmission, resulting in errors.

This issue is closely related to energy constraints, since packeterrors will cause retransmission, requiring more energy from thesensor. Moreover, retransmission will affect delay and datathroughput, also affecting QoS. The use of error control techniquescan prevent retransmission, however at the cost of introducingsome transmission overhead, and thus increasing the sensorsenergy consumption. Hence, it can be seen that transmissionerrors, QoS constraints, and energy consumption are closelyrelated, requiring cross-layer solutions to address these problemssimultaneously (Vuran and Akyildiz, 2009). Also, the effects of theuse of automatic repeat request (ARQ), forward error correction(FEC), and their combination on the discussed parameters shouldbe studied to find an optimal solution for WSNs.

The number of sensors in the network can affect the perfor-mance of thewhole network (Olariu and Xu, 2005). In aWSNmadeof only a few sensors, more energy will be used to reach the othernodes, thus reducing the network lifetime. Also, the farthestsensors may become isolated from the rest of the network whenthe relay node they depend on has no battery left. On the other side,in aWSNwith toomany sensors, coverage andenergy consumptionto reach other nodes will not be an issue (Habib, 2007). However,the forwarding of messages may cause congestion and an energy-efficient routing becomes complex to be calculated. Hence, thecapacity planning of aWSN is an important design parameter to beconsidered for efficient use of its scarce resources.

Given the broadcast nature of the wireless medium, sensornodes in WSNs become a potential target for attacks (Chen et al.,2010).Maliciousnodes can join the network andeavesdrop in orderto damage transmissions or steal information. Some attacks havebeen identified in the literature and also some attack detection andprotection techniques have been proposed.

One common type of security breach is the spoofing attack. It isnot difficult to find out a node identification in a wireless network.Just by capturing packets on the air may allow a node to discovera valid node address. Hence, the malicious node can change itsaddress to the valid captured one and start transmitting false datainto the network (Chen et al., 2010). In WSNs, the Sybil attack ismore damaging. In this case, an attacker assumesmultiple networkidentification and starts broadcasting messages, thus disruptingrouting and preventing the network from working properly.

Another type of attack is the wormhole attack. Its goal is tocapture traffic from one part of the network and retransmit it inanother. The traffic can be captured by a malicious node andtransmitted to another through a different communication link(e.g., wired link, different wireless band or technology). Then, thesecond malicious node replicates the captured traffic to itsneighbors in order to interfere with routing protocols.

Cryptography can also be used for packet exchange safety.However, cryptographic keys must be exchanged beforehand toenable nodes to encrypt their messages and also more processingresources will be necessary to encrypt and decrypt data. Thisprocess consumes precious energy from the sensors (Misic et al.,2006a), making the tradeoff between network security and energyconsumption clear.

The sensors hardware clock reference can directly affect anapplication operation. For instance, a WSN for target detection isuseless if it cannot register both the position and the detectiontime of an event (Cheng et al., 2009a). This shows once more theapplication layer dependence on the physical layer. However, thecomplexity of the used synchronization protocol can directly affectthe network lifetime (Ganeriwal et al., 2009). Hence, the networktimeprotocol (NTP), used for synchronizationon the Internet, couldnot be used in WSNs, and then three (non-cross-layer) approacheshave been defined for the proposal of new synchronizationprotocols—sender to receiver synchronization (SRS), receiver toreceiver synchronization (RRS), and flooding. The flooding time

L.D.P. Mendes, J.J.P.C. Rodrigues / Journal of Network and Computer Applications ] (]]]]) ]]]–]]] 3


dx.doi.org/10.1016/j.jnca.2010.11.009

synchronization protocol has been proposed according to thelatter category, and it can be inferred that this technique is notenergy-efficient. Furthermore, the timing-sync protocol for sensornetworks (TPSN) and the tiny synchronization/mini synchroniza-tion (TS/MS) protocols that belong to the SRS category require thatsensor nodes receive and transmit once to achieve synchronization.Although it is more energy-efficient than flooding, where duplicatetransmissionsmay occur, it still wastes some energy. Finally, in theRRS category, the reference broadcast synchronization (RBS),the pairwise broadcast synchronization (PBS), the network-widepair selection algorithm, and the group-wise pair selection algo-rithm (GPS) have been proposed. In these proposals, some sensorstake advantage of the broadcast medium to overhear the synchro-nization transmission between other nodes, saving transmissionresources (Cheng et al., 2009a). Nevertheless, none of the men-tioned are cross-layer proposals, acting on the synchronizationregardless of the application. Thus, a tradeoff between synchroni-zation accuracy and sensors energy consumption can be achieved ifthe required application quality of service is taken into account.

From the discussion presented in this section, it can be seen thatall limitations of WSNs are connected. Also, routing, QoS con-straints, security, and time synchronization conflict directly withsensors energy consumption, requiring a comprehensive study toidentify the existing tradeoffs. Thus, in order to tackle the problemsinterdependence, a cross-layer solution is needed. Clearly, theseproposals are more complex than the non-cross-layer ones.However, energy consumptionmust be reduced at all costs in thesenetworks, or else theywill not operate for the extendedperiods theyare supposed to. The ideal cross-layer solution—but also the mostcomplex—would involve parameters from all layers of the stacksince all of them affect energy consumption to some degree. Hence,the increase in the sensors design complexity is inevitable andinversely proportional to the sensors energy capacity.

3. Considered technologies at each layer

Differently from the conventional networks layers, WSNsconsider only the following layers (Baronti et al., 2007): application(APP), network (NWK or NET), medium access control (MAC), andphysical (PHY). Although there is no transport layer, since it iscomplex and it would waste sensors energy (Misra et al., 2008),some WSN protocols have been designed for congestion controland reliable end-to-end communication (functions of the transportlayer). Some of them are presented below, in the Network layersubsection. Nevertheless, some authors defend the definition of atransport layer on WSNs (Misra et al., 2008), keeping the samelayers stack as defined by TCP/IP. The considered stack can be seenin Fig. 2with twoexamples of cross-layer interactions. For instance,

the QoS requirements at the application layer can be informedto the MAC layer in order to achieve better scheduling for therunning application, and the channel state information (CSI) can befed to the network layer so the routing protocol can avoid pathsincluding channels in a bad state.

Table 1 summarizes the technologies, protocols, and schemesconsidered by the cross-layer proposals in the same order ascommented in the following subsections. It is worth noting thatsome cross-layer proposals are mathematical frameworks thatconsider a set of parameters at each layer, however without usingany specific technology or scheme at the considered layer. Thus,a cross-layer proposal may have a technology or scheme repre-sented at only one layer in Table 1.

3.1. Application layer

MPEG and H.26 are the most used compression standardsfor video. The compression they provide allows more video to bestored and less bandwidth for video transmission is required(Chen et al., 2006b). Chen et al. (2006b) have proven that the IEEE802.15.3 technology is not suitable for variable bitrate (VBR) videotransmission. Hence, they have proposed the energy diffservapplication-aware scheduling (EDAS). This cross-layer algorithmincreases the quality of the received video by increasing thefraction of decodable frames to more than 85%, while the IEEE802.15.3 achieves a worst case of 65%. Moreover, energy consump-tion is reduced up to 30% by compressing video with lower qualitywhere QoS admits it, thus transmitting less data, and transmittinghigher video quality where QoS constraints are stricter. Also, EDASincludes an admission controlmechanism to provide the necessaryQoS to all the video flows in transmission.

Yuan et al. (2006b) have proposed a cross-layer solution toincrease the lifetime of clustered WSNs. The QoS parameters aretaken into account by the adaptive code position modulationscheme (ACPMS) at the cluster heads, adapting the transmissionaccording to the computed BER. Intra- and inter-clusterQoS perfor-mance is assessed through simulation, proving that QoS is providedin termsof deliveredpacket ratio and end-to-enddelay and that thefirst node death is delayed at least for 60 rounds, where the roundsare the moments when the clusters are rebuilt.

Chen et al. (2007) have proposed another solution to saveenergy in WSNs transmitting H.26L videos. The proposed cross-layer routing scheme called directional geographical routing (DGR)breaks a video transmission in several flows that are transmittedthroughdifferent routes in order to avoid energy depletion of nodesin a single route. Also, FEC is used to protect the multiple videoflows on their path to the sink node, helping to guarantee the entirevideo arrival.

Disregarding the type of data generated by the sensors, but stillconsidering the application layer, data source coding (DSC) is atechnique used to encode application data in order to reduce theamount of transmissions in a network. For instance, sensorsdeployed close to each other may measure values that do notdiffer much. Hence, each sensor encodes its measurement usingfewer bits than it regularly would, making it impossible to retrievethe measurement value with information from only one node.However, when all measurements reach the sink node, it cancombine the sensors encoded data to retrieve the values measuredby all the sensors. Thus, fewer bits are transmitted and energyis saved. Wang et al. (2008a, 2009) have proposed cross-layersolutions to transmit data using DSC combined with multipletransmission ratemedium access protocols. InWang et al. (2008a),multi-level rate routing (MLRR) is proposed, while in Wang et al.(2009) quadrature phase-shift keying (QPSK) and quadratureamplitude modulation (QAM) are considered. Both proposals areFig. 2. Considered layers stack with two examples of cross-layer interactions.



dx.doi.org/10.1016/j.jnca.2010.11.009

able to calculate the best transmission rate to achieve low packeterror rates.

Data fusion (also knownas data aggregation) is a techniqueusedto reducemeasurement errors. In this case, data are not encoded toreduce the number of transmitted bits. Instead, data received fromseveral sensors are compared to search for errors. Thus, a reliableapproximation of themeasured phenomena is achieved. Liang et al.(2007) have proposed a cross-layer solution to enhance data fusionand the network lifetime. They have used sensors channel stateinformation (CSI) to schedule transmissions only to the sensorsexperiencing better channels, and they have also used this infor-mation to weight the reliability of a sensor measurement whenperforming data fusion. Hence, preventing sensors experiencingbad channels from transmitting, the bit error rate could be reducedup to 1000 timeswhen compared to themaximum ratio combiningfusion algorithm, thus wasting less energy.

Data aggregation has also been used by the GRASS protocol,proposed by Al-Karaki et al. (2009). In the analyzed scenario,clusters are formed and the cluster heads (CHs) aggregates the datacoming from the sensors in the cluster. Then, the CHs are respon-sible to find a route to the base station thatmaximizes the networklifetime. Results have shown a gain of at least 35% on the lifetime

when compared to the directed diffusion protocol. Also, to avoidlimiting the network lifetime to the CHs lifetime, new CHs areelected periodically inside each cluster.

3.2. Network layer

Routing protocols attract special attention by their impact onthe network lifetime. One of the most quoted architecture byrouting proposals for clustered WSNs is LEACH (low-energyadaptive clustering hierarchy) (Heinzelman et al., 2002). In thisproposal, the network is divided in clusters, and cluster heads (CHs)are randomly chosen inside a cluster every round, guaranteeingthat all the nodes in the cluster have the opportunity of being theCH and thus sharing the energy consumption among all nodes. Theproposal has proven that for the same amount of sensors initialenergy, the proposed protocol can transmit 40% more data thanwhen using LEACH.

A cross-layer routing protocol for WSNs proposed by Choi et al.(2005), the cell based energy density-aware (CEDA) protocol,considers the division of the WSN in cells, extending LEACH tomultihop routing inside cells. The routing tables include one more

Table 1Main technologies, protocols, and schemes considered in cross-layer proposals.

Proposal Authors Technologies and/or schemes at each considered layer

Application Network MAC Physical

EDAS Chen et al. MPEG-4 – EDAS –– Yuan et al. – Cluster-based routing ACPMS ACPMSDGR Chen et al. H.26L / FEC DGR – –MLRR Wang et al. DSC MLRR Rate assignment CSI/REI– Wang et al. DSC – – QPSK/QAM– Liang et al. Data fusion – TDMA/TDD 16-QAM/CSIGRASS Karaki et al. – GRASS – –CEDA Choi et al. – CEDA 802.11-based –EDDD Chen et al. RT-/BE-traffic EDDD S-MAC –PCCP Wang et al. – PCCP CSMA-like –GMREE Sanchez et al. – GMREE – –A-MAC Nam et al. – AODV A-MAC –APRMAC Mitchell et al. – One-hop neighbors discovery APRMAC –– Cheng et al. Data aggregation – TDMA Transmission power controlCAGIF Zhang and Zhang – CAGIF CDMA –XLCU Melodia and Akyildiz – GIF TH-IR-UWB UWB/dynamic channel coding– Shi et al. – Minimum-energy routing – UWB– Shi and Hou – – – UWB/Transmission power control– Liang and Liang – – – Virtual MIMO

Yuan et al. – – CSMA/TDMA Virtual MIMOMisic et al. – – CSMA/CA Duty cycles

CAEM Lin and Kwok – LEACH CSMA/CD Duty cycles/CSISS-Trees Ha et al. – – CSMA Duty cycles– Misic – – CSMA/CA BPSK/O-QPSK/Duty cyclesEX-MAC Hong and Kim – – EX-MAC –ENCO Demirkol and Ersoy – – Contention-based Duty cycles– Kwon et al. – based/ARQ TDMA-power control Transmission– Mao et al. – – TDMA-based -– Madan et al. – – TDMA –– Cui et al. – – TDMA M-QAM– Su and Zhang – – TDMA CSI/M-QAM/battery discharge dynamics– Shi and Fapojuwo – – TDMA FSK– Phan et al. – – TDMA –– Wang et al. – – TDMA-based –– Shu and Krunz – – Hybrid TDMA/CDMA –CLPC Messier et al. – – ALOHA/ARQ CLPCDOMRA Miao et al. – Two-hop knowledge ALOHA CSIDQBAN Otal et al. – – DQBAN –DPLM Chen and Zhao – – – CSI/REITM-MAC Ren and Liang – – TM-MAC UWBLEMMA Macedo et al. – – LEMMA –TC-MAC Kim and Park – – TC-MAC Duty cycles– Chehri et al. – – TH-IR-UWB Transmission power control/UWB/Duty cycles



dx.doi.org/10.1016/j.jnca.2010.11.009

variable—the energy level from the neighboring nodes. Then, theenergy level information is used as a weight factor when routingdata, avoiding the nodes with less remaining energy. Hence, therouting protocol has proven that only 5% of the nodes have theirenergy depleted when all the nodes are dead using the AODVprotocol.

The directed diffusion routing protocol has been extended byChen et al. (2006a), since its use has not shown to be adequate forWSNs. Their proposal, the energy-efficient differentiated directeddiffusion (EDDD), is able to differentiate between real-time (RT)and best effort (BE) flows. Hence, RT flows are routed consideringtheir end-to-end time constraints, while BE flows are routed in anenergy-efficientmanner. Also, route-repairing techniques improvethe ones used by directed diffusion, reducing the time to reroute RTflows in case of node failure. Their results have shown that thenetwork lifetime is increasedwhen compared to directed diffusion.

Wang et al. (2007) have proposed a cross-layer priority-basedcongestion control protocol (PCCP) for WSNs. Although congestioncontrol is a function of the transport layer, this work addresses thecongestion problem from the network layer. The proposal includestwo new queues between network andMAC layers—one for trafficoriginated at that node and another for the traffic routed throughthe node. By verifying the inter-arrival packet times in the routedtraffic queue, PCCP is able to detect congestion. Then, implicitcongestionnotification (ICN) is used to inform the congestion to theother nodes. Finally, each node reduces its traffic routing queue-scheduling rate to control the detected congestion. The resultspresented in the work show that the normalized network through-put is stabilized above 90% and the fluxes are fairly scheduled.

The energy-efficient geographic multicast routing (GMREE) hasbeen proposed by Sanchez et al. (2007) to reduce the energy con-sumption in SANETs. For that, they have created the cost overprogress ratio, a metric that informs the cost (energy amount) ofreaching a neighbor and the progress it provides towards a destina-tion. Thus, multicast trees are formed according to that metric withthe objective of reducing energy consumption. Results show thatGMREE is able to reduce energy consumption inmulticast SANETs by10 times when compared to the IPowerProgress algorithm.

In the cross-layer design proposed by Nam et al. (2007), theirnew adaptive MAC (A-MAC) requires a change in the used routingprotocol. New metrics that depend on the duty cycle (the ratebetween a sensor active and sleep times) influence the routingdecisions. The authors have chosen the ad hoc on-demand distancevector (AODV) routing protocol for their tests. A-MAC changes thenodes duty cycle dynamically to achieve a predetermined networklifetime. Thus, AODV routes the packets according to the adoptedmetrics in order to verify the behavior of end-to-end delay andnetwork lifetime. Their results have shown that the commonrouting metric (minimum hop count) is more energy-efficientwhen combined to A-MAC, however the delays are higher. Theothermetrics implemented onAODV can reduce delay, but they areless energy-efficient, showing that there is a tradeoff betweenenergy consumption and delay.

Mitchell et al. (2010) have proposed the aerial platform basedrouting and medium access control (APRMAC). In this protocol,each sensor nodeneeds to discover their one-hop neighbor throughsending hello packets. Then, all sensors send this information to anaerial node in LOS with all the sensors. The aerial platform isresponsible for calculating the routes from all the sensors to theirnearest sink nodes, and then it schedules medium access time forall the nodes. Thus, each node knows when it can transmit, when itis supposed to listen, and when it can enter sleep mode, saving atleast 5% more sensors energy than TRAMA. Figure 3 shows anexample of an APRMAC scenario.

Cheng et al. (2009b) have argued that shortest path orminimumenergy routing solutions have not considered the bandwidth

constraints of WSNs. Thus, they have proposed a cross-layerrouting protocol that is able to adjust the time division multipleaccess (TDMA) link rates according to their utilization. It has beenproven in theirwork that the lifetime of the network is increased by40% when compared to shortest path routing and the capacity ofthe links is respected.

A new routing metric has been defined by Zhang and Zhang(2009) for their cross-layer channel-aware geographic informedforwarding (CAGIF) protocol. The efficient-advancement metric(EAM)determines the optimal relay node position to reach the nexthop towards the destination. They have considered CDMA asmedium access method, also taking interference into account.Through the use of the EAM and three positions (the current node,the source, and the destination), CAGIF provides routing with atleast 50% less energy consumption and up to 4 times less packetcontrol overhead when compared to GIF.

Melodia and Akyildiz (2010) have proposed a cross-layercontrol unit (XLCU) to provide QoS for WMSNs. They have consid-ered geographical forwarding to control the flows and hop-by-hopQoS requirements. Also, the XLCU controls the scheduling and rateassignment of the time hopping impulse radio UWB (TH-IR-UWB),preventing collisions and idle listening. Thus, throughput andreliability are increased and delay is reduced, according to theresults.

A centralized nonlinear programming problem for optimalrouting, scheduling, and power control has been defined by Shiet al. (2006). For large networks, a heuristic algorithm for theoptimization problemwas proposed. In a later proposal by Shi andHou (2008), they have included rate assignment to the optimiza-tion problem and then derived closed-form expressions to calcu-late the capacity of the WSN.

Liang and Liang (2009) have proposed a maximum spanningtree searching (MASTS) algorithm to determine the best set ofnodeswhen creating virtualmultiple-inmultiple-out (MIMO) linksin order to connect the routes determined by the routing protocol.Although a specific routing protocol has not been considered,implying that the nodes already knew the routes, the authorshave proven that the bit error rates experienced by the users isreduced up to 10 times when compared to regular virtual MIMOcommunication.

Yuan et al. (2006a) have also consideredMIMO communicationin their cross-layer proposal. They have considered the LEACHprotocol as base for multicluster routing. The cluster heads (CHs)

Fig. 3. Example of an APRMAC scenario (Mitchell et al., 2010).



dx.doi.org/10.1016/j.jnca.2010.11.009

are responsible for choosing the nodes that will be participating inthe MIMO communication to the next cluster head. This is done byachieving the minimal acceptable packet error probability, accord-ing to the end-to-end QoS parameters. Thus, avoiding packetretransmission minimizes the overall energy consumption.

3.3. Medium access control layer

Cross-layer designs that do not consider an optimizationinvolving themediumaccess control layer are rare in the literature.The large number of proposals in this subsection proves thisstatement. These proposals will be briefly discussed, emphasizingthe considered technologies used for optimization.

Carrier sensemultiple accesswith collisionavoidance (CSMA/CA)is the medium access technique used by IEEE 802.15.4. Misic et al.(2006b) have considered this technology in their cross-layer activitymanagement scheme. The sensors duty cycles (sleep and listen/transmit time) are calculated in order to achieve a determined eventsensing reliability. This is done according to the packet loss prob-ability, calculated with physical layer measurements,and packet collision probability, packet blocking probability, andservice time probability distribution, measured at the mediumaccess layer. Two scenarios were considered for duty cyclecalculations—centralized and distributed. For the centralized sce-nario, the sink node determines the duty cycles and transmits themto the other sensor nodes. This method achieves the requiredreliability, but it can deplete the coordinator node energy faster.On the distributed case, the coordinator only sends the number ofalive nodes to the sensors, and then each of them can calculate itsown duty cycle to keep the reliability above the predefined limit.Thus, the coordinator node consumes less energy and the networkevent sensing reliability is kept at the same level.

Lin and Kwok (2006) have proposed CAEM, a cross-layer energymanagement approach considering carrier sense multiple accesswith collision detection (CSMA-CD) as channel access method.All sensors continuously sense the channel state. If one of them hasa packet to be sent, it is done if the channel is in a good state. If not,the packet is buffered, the sensor enters sleep mode, and a fairscheduling and queuing algorithm avoids starvation in case thechannel stays in the bad state for a long time. Thus, the lifetime ofthe network is increased up to 40% when compared to pure LEACHthrough the knowledge of the channel state.

According to Ha et al. (2006a), pure CSMA is preferred overcontentionlessmethods, e.g., code divisionmultiple access (CDMA)and time division multiple access (TDMA) because it is simpler tobe implemented, and over CSMA/CA because the latter addsoverhead to transmission. Hence, Ha et al. (2006a) have proposedthe sense-sleep trees (SS-Trees) where trees of sensor nodes areorganized to facilitate the sensors sleep scheduling. Then, trans-mitting nodes access themediumusing CSMA, and through the useof downstream and upstream phases, the number of packetcollisions is reduced. Thus, they have achieved reduced energyconsumption while meeting the surveillance requirements.

Another advantage of CSMA/CA over TDMA is its capability ofhandling communication among a large number of sensors (Misic,2007). Based on this, Misic (2007) has modeled the lifetime of amultilevel clustered IEEE802.15.4WSN. The effect of the number ofsensors in each cluster on the network lifetime is analyzed by usingthemodel, validated through simulations. Finally, results show thatadjusting the number of nodes in each cluster can increase thelifetime of the network.

Proposed to reduce end-to-end latency when duty-cyclesare considered for energy consumption reduction, express MAC(EX-MAC), designed by Hong and Kim (2009), considers CSMA/CAas multiple access method. Whenmultiple sensors sense an event,

event data reservation (EDR) is used to schedule transmissionsalong the path to the sink in order to reduce end-to-end delay.EX-MAC has also proven to be more energy-efficient than previousmultiple access proposals, achieving up to 15 times less energyconsumption than X-MAC.

A new medium access metric, the estimated number of con-tenders (ENCO), has been proposed by Demirkol and Ersoy (2009).They have considered CSMA as the multiple access method andthey have discussed the effect of the number of nodes contendingfor the channel on the end-to-end delay. As a result, they havemanaged to reduce contention delay by calculating and feeding thesensors with each ENCO value before the network operation.Results have shown that the end-to-end delay and energy con-sumption are reduced when compared to the traditional conten-tion method.

Despite the commented advantages of CSMA over TDMA,authors of cross-layer proposals also consider TDMA because ofits simpler deterministic analysis and transmission guarantees.Kwon et al. (2006) have proposed a cross-layer routing, automaticrepeat request (ARQ), and power control scheme consideringTDMAmethod at theMAC layer. The centralized algorithm verifiesthe channel state at every link. Then, it candefine the best routes foreach information stream, also adapting the retransmission limits toincrease the probability of successful packet delivery. Hence, atradeoff between the packet delivery probability and the networklifetime is respected.

TDMA has also been considered in the cross-layer proposal byMaoet al. (2007). In this proposal, the authors schedule all availabletransmission slots in an optimal manner by using a combination oftwo algorithms—particle swarm optimization (PSO) and hybridalgorithm (HA). The results shown in the work prove that theproposed combined algorithm reduces the energy consumption forTDMA slots allocation by 17% when compared to max degree firstcoloring algorithm and by 5% when compared to the node basedscheduling algorithm.

Madan et al. (2006) have considered TDMA in their cross-layerproposal as well. They have proposed their own routing schemeand they have also considered a generic multilevel modulation astransmission method of the interfering nodes. Then, their cross-layer algorithm calculates the optimal routes to balance thenetwork load, the best slot scheduling for TDMA, and the optimalmodulation scheme in order to reduce interference and transmitusing lower power. Thus, results have shown that theWSN lifetimeis increased by 7 times when compared to optimal TDMA.

The joint optimization of network, MAC, and physical layers hasbeen proposed by Cui et al. (2007). At the physical layer, M-aryquadrature amplitude modulation (MQAM) is considered. Themodulation adapts to the channel conditions by increasing itstransmission rate (also increasing BER) when the channel is in agood state and decreasing it to bemore robust when the channel isin a bad state. The MAC layer schedules more TDMA time slots tothe links with less transmission rate to keep the throughput, andthe network layer reroutes traffic to the links with better channelstates. By solving the mathematical optimization problem con-sidering these parameters, the proposal is able to reduce sensorsenergy consumption.

Su and Zhang (2009) have proposed a cross-layer battery-awareTDMA MAC protocol for WBANs. They have studied batterydynamics, concluding that the energy of a battery takes longerto be depleted if its use is subject to idle times. Thus, their proposalschedules transmissions keeping a node from transmitting for thelongest possible time, however respecting QoS constraints. Also,adaptive modulation is considered to reduce packet losses due tothe channel varying characteristics with time. Their analyticalmodel has been validated through simulations considering an ECGsensors application. Their tests have proven that the number of



dx.doi.org/10.1016/j.jnca.2010.11.009

packets transmitted in the network is increased by up to 50 timeswhen compared to IEEE 802.15.4 and by up to 33 times whencompared to Bluetooth, while QoS constraints are respected forboth cases.

The cross-layer nonlinear optimization model proposed by Shiand Fapojuwo (2010) considers TDMA as medium access method.The mathematical framework collects information on the trafficload, the retransmission and modulation scheme, and the bit errorrates to calculate the optimal transmission powers and the flowrate used for each link. By the analysis of the trend of thecalculations with the nonlinear optimization model, an algorithmfor minimum delay scheduling has been proposed. The algorithmhas been tested on linear, grid, and random topologies, proving tobe energy-efficient.

Node admission and reduction of the energy consumption inWMSNs are the goals of the solution proposed by Phan et al. (2009).They have considered that sensor nodes access the mediumthrough the TDMAmethod and that their speed ranges from staticup to low mobility. They have formulated a mixed-integer linearprogramming (MILP) problem for network lifetime maximizationwith node admission. Then, they have shown the effect of the nodestransmissionbandwidths on thenumber of admittednodes and thenetwork lifetime.

A TDMA-based MAC protocol is part of the cross-layer modelproposed by Wang et al. (2008b). Power control, link access, androuting are considered in their mathematical model to satisfy theKarish–Kuhn–Tucker (KKT) optimality conditions. The presentedresults have shown that the derived equations provide an upperbound for the network lifetime, validated through simulations.

Shu and Krunz (2009) have proposed a cross-layer solutionconsidering hybrid TDMA/CDMA medium access method. First,intercluster interference is avoided by allocating different super-frames (sets of TDMA frames) to interfering clusters, mediumaccess known as spatial TDMA (STDMA). Thus, when nodes trans-mit in a cluster, adjacent interfering clusterswill not transmit at thesame time. Second, CDMA scheduling is allocated to sets of sensorsinside each cluster according to a power and time control (PTC) inorder to reduce power consumption inside the clusters. Throughthe combination of those techniques, the network lifetime isincreased, according to their results.

Some cross-layer proposals have also considered ALOHA asmedium access method. This has been the medium access methodstudied by Messier et al. (2008), combined with an infinitelypersistent ARQ. They have proposed a cross-layer power control(CLPC) algorithm that is capable of calculating the best receivedpower for each node (and thus optimal transmission power can becalculated) taking the multiple access interference (MAI) intoaccount. Simulation results have shown energy savings of up to74% when compared to noise CLPC.

ALOHA medium access method has been considered by Miaoet al. (2009) in their cross-layer proposal as well. In order to derivethe equations for decentralized optimization for multichannelrandom access (DOMRA), they have also considered that somenodes are out of the range of others, differently fromother channel-aware ALOHA proposals considering that all nodes interfere witheach other. Through the collection of traffic information fromneighbors and instantaneous CSI measurement, transmissionpower is allocated to every sensor node in order to reduceinterference and increase network utility and transmissionfairness.

The distributed queuing body area network (DQBAN) is a cross-layer solution proposed by Otal et al. (2009). Transmission of apacket is scheduled according to a fuzzy-rule based on the QoSconstraints of the packet and the current traffic. Thus, a packet issent through slotted ALOHA medium access method when thetraffic is light, a reservation protocol is used if the traffic is heavy,

and a polling scheme is used if delay constraints are tight andcannot afford collisions. Their proposal has shown to reduceWBAN energy consumption up to 100,000 times and increasethe packet delivery rate by up to 42.75% when compared todistributed queuing without energy policy, and decrease meanpackets delay by 8% when compared to distributed queuing with-out scheduler.

Chen and Zhao (2007) have identified two important physicallayer parameters that should be used by MAC protocols—thechannel state information (CSI) and the residual energy informa-tion (REI). The proposed medium access protocol schedules trans-mission to the sensors that experience a good channel when thesensor nodes have full energy, andwhen part of the sensors energyhas been used, the scheduling prioritizes the sensors with moreenergy left in order to even the nodes energy. Thus, the energyconsumption of the WSN is reduced by 8 times when compared tothe pure conservative scheme.

Ren and Liang (2008) have proposed a new cross-layer mediumaccess method—the throughput maximized MAC (TM-MAC). It isable to divide ultra-wideband piconets in sets and to calculate theoptimal scheduling and transmission rate in order to minimizeinterference. Results have shown that throughput is increased byup to 22% and that packets transmission time is reduced by up to32% when compared to the IEEE 802.15.3 performance.

The latency-energy minimization medium access (LEMMA) is across-layer proposal by Macedo et al. (2009) that jointly optimizesrouting in sporadic-traffic WSNs. Transmission of higher powertones is used to avoid scheduling of sensors at the same time slot ofthe TDMA method. When a tone from a node is received, no othersensor attempts to occupy the time slot. This has proven to reducetransmission interference in the considered lognormal shadowingscenario. Thus, it has been shown that delay is reduced by 12 timesand the network throughput is increased by up to 28% whencompared to LMAC.

Some proposals that make use of listen and sleep schemes(duty-cycles) cause an increase in the end-to-end latency and areduction of the network throughput. Furthermore, the highnumber of nodes in WSNs may result in congestion near the sinknodes (Kim and Park, 2009). To solve these problems, Kim and Park(2009) have proposed the transport controlledMAC (TC-MAC). Thelisten periods are used formulti-hop channel reservation, reducingthe end-to-end latency since packets will be forwarded as soon asthey reach the next hop. Also, congestion is controlled through abackpressuremechanism,where a specialmessage is sent locally toprevent close nodes from contending for the channel, and anexplicit congestion notificationwith a trafficmonitor that providesfairness. Their results have proven that energy consumption isreduced when compared to other cross-layer MAC proposals,reaching up to a reduction of 6 times the energy consumptionachieved by IEEE 802.11.

Duty-cycles have been also considered in the cross-layerproposal by Misic et al. (2006a). First, a desired sensing reliability(number of packets per second for reliable event sensing at thesink) is set. Then, the proposed algorithms are responsible for duty-cycle scheduling in a multi-cluster WSN (Bluetooth scatternet) inorder to avoid congestion and achieve the minimal sensingreliability. Their simulation results have shown that the reliabilitylevel is sustained, the packet loss ratio is decreased, sensors energyis saved, and congestion is reduced at the bridge nodes (responsiblefor inter-cluster communication).

Chehri et al. (2009) have studied the effects of rate and powerallocation for sensor nodes on the same route. They have noticedthat individually optimizing the routes, it becomes harder to meetend-to-end delay constraints. Thus, they have proposed a designcapable of adapting its modulation scheme and transmissionpower in order to reduce the energy consumption of the nodes



dx.doi.org/10.1016/j.jnca.2010.11.009

on the route. They have achieved up to 51% energy consumptionreduction compared to the fixed rate and power scheme.

3.4. Physical layer

Ultra-wideband (UWB) is one of the technologies considered forthe physical layer in cross-layer proposals for WSNs. This technol-ogy for short-range high-data rate transmission presents lowcomplexity, low cost, and robustness against multipath fadingand jamming (Chehri et al., 2009). Shi et al. (2006), Shi and Hou(2008), Chehri et al. (2009), and Melodia and Akyildiz (2010) haveconsidered UWB in their optimization problems explainedpreviously.

A cross-layer adaptive code position modulation (ACPM)scheme has been proposed by Yuan et al. (2005). The authors havedefined a new layer—the packet scheduling layer, defined as a layercapable of calculating the delay and packet loss ratio requirements.From the delay requirement, the optimal modulation scheme canbe chosen, and from the packet loss ratio, the physical layer candetermine the maximum BER a packet may experience whentransmitted. Thus, the transmission power for the decided mod-ulation is adjusted in order to meet the requirements.

Sensors are small devices equipped with only one antenna fortransmission and reception. However, MIMO communication,which requires a larger set of antennas per device, can be virtuallyachieved through the cooperation of nodes. This communicationmode is also able to reduce energy consumption (Liang and Liang,2009). Two cross-layer proposals have considered virtual MIMOcommunications—one by Liang and Liang (2009) and another byYuan et al. (2006a). They have been commented previously in theNetwork Layer subsection.

Multilevel modulation has been considered in some cross-layerproposals. For instance, the previously discussed optimizationframework proposed by Madan et al. (2006) considers a genericmultilevelmodulationwith constellation size greater than or equalto four. Extending this proposal, Cui et al. (2007) have proposedanother optimization framework, also discussed previously, butconsidering MQAM.

Tian and Ekici (2007) have proposed themultihop taskmappingand scheduling (MTMS) in theirwork. This algorithm seeks the bestapplication task mapping at each sensor node. Through dynamicvoltage scaling (DVS), minimal power is used to schedule the tasks(and thus they take more time to be served), however respectingthe delay constraints. Thus, nodes energy consumption is reducedby up to 52% when compared to the distributed computingarchitecture (DCA).

3.5. Multiple-layer standards

3.5.1. IEEE 802.15.4The IEEE 802.15.4 standard (IEEE 802.15.4, 2003b) (IEEE

802.15.4, 2006) has been created for low-rate wireless personalarea networks (Misic et al., 2006b), making it suitable for WSNs.It defines specifications for medium access control and physicallayer. There are four possible transmission schemes. Three of themoperate in the 868/915 MHz frequency range, one with directsequence spread spectrum (DSSS) and binary phase-shift keying(BPSK) modulation, the other with DSSS and offset quadraturephase-shift keying (O-QPSK), and the last with parallel sequencespread spectrum (PSSS), BPSK and amplitude shift keying (ASK)modulations. The last scheme operates in the 2450 MHz frequencyrangewithDSSS andO-QPSKmodulation. Regarding theMAC layer,the twopossible topologies rely on a PANcoordinator node towork,using CSMA/CA as medium access method. In peer-to-peer com-munication, nodes can communicate with others in range directly,

but in star topology they need to communicate through the coordi-nator node (Misic et al., 2006b). Misic (2007) and Misic et al.(2006b) have considered this standard in their cross-layer propo-sals explained in a previous subsection.

3.5.2. IEEE 802.11Also based on CSMA medium access method, the IEEE 802.11

(IEEE 802.11, 2007) defines 4 PHY schemes. The first operates in the2.4 GHz frequency rangeusing frequencyhopping spread spectrum(FHSS). The second uses the same frequency range with DSSS.The third scheme uses communication through infrared (IR) withwavelengths from 850 to 950 nm. The fourth uses orthogonalfrequency division multiplexing at the 5 GHz band. Also, there aretwo data rate extensions defined in the standard. At theMAC layer,it uses request to send (RTS) and clear to send (CTS) frames tocontend for the medium (when operating in distributed coordina-tion function—DCF-mode). This is suitable for large networks, butit wastes more energy on idle listening when traffic is light (Kimand Park, 2009). Thus, Kim and Park (2009) have proposed anextension to the IEEE 802.11 MAC to avoid this waste of energy.This is carried out by the assignment of listen periods and changingnodes to sleep mode when they are not supposed to listen, asexplained previously.

3.5.3. IEEE 802.15.3The IEEE 802.15.3 standard was proposed for high-rate applica-

tions in wireless personal area networks (WPANs) (IEEE 802.15.3,2003a), which is suitable for video and audio transmissionon WSNs. It specifies the operation of WPANs at the 2.4 GHzwith QPSK, differential quadrature phase-shift keying (DQPSK),16-quadrature amplitude modulation (QAM), 32-QAM, and 64-QAM. The MAC specification defines the use of a hybrid CSMA/CAand TDMA. Chen et al. (2006b) have considered a modification ofthese medium access methods, the pseudo-static TDMA (pTDMA).This medium access method prevents collisions since it is conten-tion-free, nevertheless it affects fairness since transmitting nodesand idle nodes have transmission time allocated to them equally.Thus, Chen et al. (2006b) have proposed a cross-layer solution forthis problem, assigning QoS priorities to the different data flowsand providing medium access based on the defined priorities.More details on this solution have been given in a previoussubsection.

3.5.4. BluetoothBluetooth has also been considered for WSNs. This technology

uses frequency hopping spread spectrum technique, reducing theinterference from other networks operating at the same frequency.It can achieve up to 3 Mbps and up to 100 m of transmission range,making Bluetooth suitable for medium-rate WPANs. Also, thistechnology organizes networks in piconets (clusters with up toseven active devices and onepiconetmaster), andmediumaccess isdone by TDMA/time-division duplex (TDD) (Misic et al., 2006a). Aswith IEEE 802.15.3, collisions are prevented, but energy can besaved if nodes go to sleep state when other nodes are assigned totransmit. However, event reliability, measured as the number ofpackets per second for reliable event detection at the sink node, is arequirement that conflicts with nodes sleep duration, and thus atradeoff must be sustained. In order to study these effects ofreliability, Misic et al. (2006a) have proposed a study for BluetoothWSNs, explained previously.

4. Open issues

It is possible to infer that QoS and energy saving are the mostimportant aspects of video transmission over WSNs. Furthermore,



dx.doi.org/10.1016/j.jnca.2010.11.009

it can be seen that regular routing can quickly deplete sensorsenergy in the video stream path. Thus, it is clear that QoS provi-sioning, routing, and energy consumption are tightly connected.Video transmission still has to be further investigated in order tocreate optimal energy-efficient and QoS-aware routing protocolssuitable not only for IEEE 802.15.3, but also for other technologies.Furthermore, only VBR video and FEC techniques have been testedso far, and even though it is known that ARQ can be harmful forQoS delay requirements, it is worth to run some tests on videotransmission over WSNs to analyze its effects. Moreover, DSC hasbeen tested only without considering real data transmission.It would be interesting to combine the transmission of video andDSC to test if data can be decoded at the sink node while savingenergy by reducing the number of bits transmitted. This solutionwould probably require advanced video processing/coding tech-niques. Also, it is clear that data fusion cannot be combined withvideo transmission since each sensor will gather different data.However, data fusion algorithms could benefit from routingprotocols by forwarding themeasured data to the nearest commonnode. Then, the fusion could be carried out, avoiding the transmis-sion of several measurements data to a sink node, and thusreducing energy consumption. Finally, it can be seen that only afew proposals involving the application layer have been actuallydeployed. There is still much room for improvements solving theproblems that arise during the deployment of cross-layerproposals.

Furthermore, it can be concluded that new routing protocols aredifficult to be created. All aforementioned solutions involvingrouting protocols are extensions of other previous protocols. Thus,other known routing protocols can become part of cross-layerbandwidth-aware optimizations, creating the opportunity formany contributions. Moreover, the commonly used metrics forrouting (e.g., shortest path) can be replaced by others that includeenergy consumption, interference and/or channel state. VirtualMIMO communication can affect routing since it can make farthernodes become neighbor ones. Thus, the combination of virtualMIMO and routing protocols must be further investigated to verifyits advantages and drawbacks. Also, none of the cross-layersolutions has been deployed on testbeds. Comparisons betweenthe simulated or calculated WSN performances and resultsachieved in testbeds can be a source for further routing protocoldevelopment. Also, the most efficient routing hierarchy hasnot been identified. Some solutions suppose it is the clusteredhierarchy, others use single-hop approaches, and also flat hier-archies can be found. Thus, each hierarchy considered needs to beassessed in terms of the network objective in order to identifywhich one results in higher throughput, less end-to-end delay,and so on.

From the presented proposals, it is clear that the most usedmedium access control methods are CSMA and TDMA. However,these access methods are considered in different scenarios, assum-ing the chosen one will perform better for the considered WSN.Literature lacks of a comparison of thesemost usedMAC protocols.Also, the scheduling of sleep, listen and transmit times has showedto be of great importance to save sensors energy. Nevertheless,it can also severely interfere on QoS constraints. Thus, theseschedules must be carefully studied in each scenario, and then ageneral rule should be proposed. Furthermore, the use of CSI helpsto avoidwasting network resources.Mediumaccess is scheduled tothe sensors experiencing a good channel state (opportunisticscheduling), however this can lead to unfairness. Thus, mechan-isms that make use of CSI must also include methods to improvescheduling fairness. Moreover, CSI can be used to change modula-tion schemes and for transmission power control. Modulationschemes have different error resiliencies, thus more robustschemes can be used when a sensor experiences a bad channel,

and also transmission power can be increased to reach other nodeswithmore reliability. However, more interference can reduce linkscapacity and increase packets error probability. Hence, the tradeoffbetween self-benefit and network benefit must be analyzed whenchanges in modulation and power control are used.

It is also possible infer that UWB is a promising technology forWSNs due to its high achievable transmission rate and its robust-ness. Moreover, it can be combined to CSI and REI, as discussed inprevious subsections, to improve energy efficiency. Virtual MIMOhas been considered as well, and this technique has proven to saveenergy by the exploitation of nodes cooperation. It could also beachievedby regularMIMO techniques, but sensor nodes are power-and size-constrained, and thus placing more antennas on sensornodes becomes infeasible. Furthermore, multilevel modulationenables sensors to adapt to the channel quality dynamically, alsoleading to energy saving. Moreover, DVS has been used to savesensors energy by reducing data processing energy consumption,when QoS constraints allow a small increase in processing delays.Finally, we are able to conclude that physical layer considerationsare the key for energy consumption reduction. Thus, all sensorsphysical capabilities available (virtual MIMO, CSI, REI, multilevelmodulation, and DVS) must be exploited in order to reach optimalnetwork energy consumption in real networks.

Scenarios considered in the solutions comprise different num-bers of sensors in the WSN. It is clear that scalability can severelyaffect the operation of routing protocols, also leading to congestion.Little work has been done on the limits of proposals operation.It is also worth noting that scalability is different depending on theconsidered scenario. For body area networks, a scalable solutionmight be able to comprise up to 20 sensors, while in a disaster areamonitoring solution, up to many hundreds of sensors might benecessary. Thus, the operation boundaries of the solutions shouldbe assessed, and they must cope with the considered applicationscenario.

Furthermore, only a few proposals consider sensors mobility.It has been argued that the proposal of new routing protocols is notan easy task, and by considering mobility more complexity isexpected. While lowmobility is experimented by tracked assets ina warehouse, a vehicular traffic sensoring solution needs to copewith high mobility. This is certainly one of the larger challengescross-layer research will have to deal with.

One of the disadvantages of cross-layer design is its lack ofinteroperability with protocols running on commercial solutions.A few solutions have considered interoperating with existingsystems, and thus it can be seen that this will remain an openissue for some time, specially because newprotocols are frequentlyproposedwithout regarding interoperation.More efforts should bedirected towards the creation of architectures, or a single archi-tecture, that can interoperate with the existing protocol stack.After this definition, protocols could be proposed following itsrules, guaranteeing that different cross-layer solutions couldcoexist in systems operating with non-cross-layer applications.Moreover, as shown by cross-layer optimization framework pro-posals, the more layers are considered, the better the optimizationapproach. Thus, it can be inferred that by considering all the layersshould result in the best cross-layer solutions. However, thecomplexity of the solutions also increasewith the number of layersconsidered. Hence, the proposed architecture must handle all thelayers. Another assumptionmade by optimization proposals is thatparameters of thewhole networkare accessible by theoptimizer. Inpractice, this would increase the control transmission, yieldingoverheads. On the other hand, distributed solutions cannot achieveoptimal parameters selection. Thus, the tradeoff between over-heads and the optimality of the solution needs to be studied, and ifan interoperable architecture is to be proposed, it must also copewith parameters exchange between sensor nodes.



dx.doi.org/10.1016/j.jnca.2010.11.009

5. Conclusions

The great number of cross-layer approaches that address thechallenges presented by the new applications of WSNs proves thatthere is still need for further optimization of these networks, andthat cross-layering is efficient to accomplish that. Thus, in thissurvey most of the recent research on this field has been gatheredand discussed. Proposals have shown that there are differentcategories of WSNs, and that each of them has their own set ofproblems to be addressed. Furthermore, well-known problemshave been discussed and some available cross-layer solutions havebeen briefly presented. Then, the layers and used technologies havebeen discussed, also presenting cross-layer approaches that areexamples of the used technologies. Moreover, open issues havebeen identified through the insights providedwhen the approacheswere gathered. At the application layer, QoS provisioning is themain challenge. For the MAC layer, duty-cycles scheduling hasproven to save energy with different designs for each mediumaccess method. And at the physical layer, some effects of the use ofmultilevel modulation and the consideration of parameters mea-sured at this layer, e.g., CSI and REI, by other layers has been shown.Furthermore, some of the standards considered in cross-layerapproaches have also been presented.

Some directions for future research are given next. From theapplication layer, it has been seen that strict QoS parameters mayneed to be respected depending on the running application.However, as it can be counterproductive for the WSN lifetime,further investigation needs to be carried out to verify the behaviorof the used multimedia coder (MPEG, H.263, H.264) on the sensorsenergy consumption. Generally, the work that considers these QoSrequirements and coders only consider packet error rates at thephysical level. Thus, asmore complex channelmodels are availablein researchwork that does not consider the application layer, theseproposals could be combined to result in a more detailed designthat considers the effects of the physical characteristics on theapplication. Also, none of the discussed work involving theapplication layer has been deployed, and thus only analyticaland simulation results are available. This is an interesting challengeto be addressed in the future. Moreover, routing comprises themost challenging set of issues on WSNs. There is no consensus onthe best approach for routing since some proposals considerclustered networks, others consider flat routing (no hierarchy),and also some consider centralized routing. The number of sensorsin the WSN can clearly affect the performance of each approach,requiring scalability tests that have not been done in the literature.Furthermore, sensors mobility are rarely considered. The highcomplexity of the proposed routing protocols forWSNs is increasedby considering mobility, demanding a great amount of effort fromresearchers to fill this gap. Also concerning the network layer,MIMO effects on routing protocols have been only super-ficially verified, and spatial diversity has not been explored yet.Besides, cross-layer proposals do not agree on the most efficientmedium access method for WSNs. Several MAC protocols havebeen created based on TDMA and CSMA, but the literature lacks ofthe comparison between them. In addition, boundaries betweenuplink and downlink phases, optimal frame sizes, and synchroni-zation methods are still open issues. Finally, to properly addressthe efficiency of the different modulation and transmission tech-nologies considered in cross-layer design, an extensive comparisonbetween them is needed. Clearly, as they are closely related tothe channel properties and energy consumption profile, theydirectly affect the parameters at the upper layers. Hence, there isstill much to be done in order to achieve a comprehensive cross-layer design that addresses the issues at every layer of the stack inan energy-efficient manner, but also considering the applicationrequirements.

Acknowledgments

This work has been partially supported by Instituto de Tele-comunicac- ~oes, Next Generation Networks and Applications Group(NetGNA), Portugal, in the framework of the BodySens Project.

References

Akyildiz IF, Melodia T, Chowdhury KR. A survey on wireless multimedia sensornetworks. Computer Networks 2007;51(4):921–60.

Al-Karaki JN, Ul-Mustafa R, Kamal AE. Data aggregation and routing in wirelesssensor networks: optimal and heuristic algorithms. Computer Networks2009;53(7):945–60.

Baronti P, Pillai P, Chook V, Chessa S, Gotta A, Hu YF. Wireless sensor networks: asurvey on the state of the art and the 802.15.4 and ZigBee standards. ComputerCommunications 2007;30(7):1655–95.

Chehri A, Fortier P, Tardif P-M. Cross-layer link adaptation design for UWB-basedsensor networks. Computer Communications 2009;32(13–14):1568–75.

ChenM, Kwon T, Choi Y. Energy-efficient differentiated directed diffusion (EDDD) inwireless sensor networks. Computer Communications 2006a;29(2):231–45.

Chen M, Leung VCM, Mao S, Yuan Y. Directional geographical routing for real-timevideo communications inwireless sensor networks. Computer Communications2007;30(17):3368–83.

Chen X, Xiao Y, Cai Y, Lu J, Zhou Z. An energy diffserv and application-aware MACscheduling for VBR streaming video in the ieee 802.15.3 high-rate wirelesspersonal area networks. Computer Communications 2006b;29(17):3516–26.

Chen Y, Yang J, TrappeW,Martin RP. Detecting and localizing identity-based attacksin wireless and sensor networks. IEEE Transactions on Vehicular Technology2010;59(5):2418–34.

ChenY, ZhaoQ.An integrated approach to energy-awaremediumaccess forwirelesssensor networks. IEEE Transactions on Signal Processing 2007;55(7):3429–44.

Cheng K, Lui K, Wu Y, Tam V. A distributed multihop time synchronization protocolfor wireless sensor networks using pairwise broadcast synchronization. IEEETransactions on Wireless Communications 2009a;8(4):1764–72.

Cheng M, Gong X, Cai L. Joint routing and link rate allocation under bandwidth andenergy constraints in sensor networks. IEEE Transactions on Wireless Commu-nications 2009b;8(7):3770–9.

Choi JY, Kim HS, Baek I, Kwon WH. Cell based energy density aware routing: a newprotocol for improving the lifetime of wireless sensor networks. ComputerCommunications 2005;28(11):1293–302.

Cui S, Madan R, Goldsmith AJ, Lall S. Cross-layer energy and delay optimization insmall-scale sensor networks. IEEE Transactions on Wireless Communications2007;6(10):3688–99.

Demirkol I, Ersoy C. Energy and delay optimized contention for wireless sensornetworks. Computer Networks 2009;53(12):2106–19.

Ganeriwal S, Tsigkogiannis I, Shim H, Tsiatsis V, Srivastava MB, Ganesan D.Estimating clock uncertainty for efficient duty-cycling in sensor networks.IEEE/ACM Transactions on Networking 2009;17(3):843–56.

Ha RW, Ho P-H, Shen XS. Cross-layer application-specific wireless sensor networkdesign with single-channel CSMA MAC over sense-sleep trees. ComputerCommunications 2006a;29(17):3425–44.

Ha RW, Ho P-H, Shen XS, Zhang J. Sleep scheduling for wireless sensor networks vianetwork flow model. Computer Communications 2006b;29(13–14):2469–81.

Habib SJ. Modeling and simulating coverage in sensor networks. ComputerCommunications 2007;30(5):1029–35.

HeinzelmanWB, Chandrakasan AP, BalakrishnanH. An application-specific protocolarchitecture for wireless microsensor networks. IEEE Transactions on WirelessCommunications 2002;1(4):660–70.

Hong S-H, Kim H-K. A multi-hop reservation method for end-to-end latencyperformance improvement in asynchronous MAC-based wireless sensor net-works. IEEE Transactions on Consumer Electronics 2009;55(3):1214–20.

IEEE 802.11, Part 11: wireless LANmedium access control (MAC) and physical layer(PHY) specifications. IEEE Computer Society; June 2007.

IEEE 802.15.3, Part 15.3: wireless medium access control (MAC) and physical layer(PHY) Specifications for high rate wireless personal area networks (WPANs).IEEE Computer Society; September 2003.

IEEE 802.15.4, Part 15.4: wireless medium access control (MAC) and physical layer(PHY) specifications for low-rate wireless personal area networks (LR-WPANs).IEEE Computer Society; October 2003.

IEEE 802.15.4, Part 15.4: wireless medium access control (MAC) and physical layer(PHY) specifications for low-rate wireless personal area networks (WPANs).IEEE Computer Society; September 2006.

Kim J, Park KH. An energy-efficient, transport-controlled mac protocol for wirelesssensor networks. Computer Networks 2009;53(11):1879–902.

KoWH. The future of sensor and actuator systems. Sensors andActuators A: Physical1996;56(1–2):193–7.

Kohno M, Matsunaga M, Anzai Y. An adaptive sensor network system for complexenvironments. Robotics and Autonomous Systems 1999;28(2–3):115–25.

Kulkarni S, Iyer A, Rosenberg C. An address-light, integrated mac and routingprotocol for wireless sensor networks an address-light, integrated MAC androuting protocol for wireless sensor networks. IEEE/ACM Transactions onNetworking 2006;14(4):793–806.



dx.doi.org/10.1016/j.jnca.2010.11.009

Kwon H, Kim TH, Choi S, Lee BG. A cross-layer strategy for energy-efficient reliabledelivery in wireless sensor networks. IEEE Transactions on Wireless Commu-nications 2006;5(12):3689–99.

Liang J, Liang Q. Channel selection in virtual MIMO wireless sensor networks. IEEETransactions on Vehicular Technology 2009;58(5):2249–57.

Liang Q, Yuan D, Wang Y, Chen H-H. A cross-layer transmission scheduling schemefor wireless sensor networks. Computer Communications 2007;30(14–15):2987–94.

Lin X-H, Kwok Y-K. CAEM: a channel adaptive approach to energy management forwireless sensor networks. Computer Communications 2006;29(17):3343–53.

Macedo M, Grilo A, Nunes M. Distributed latency-energy minimization andinterference avoidance in TDMAwireless sensor networks. Computer Networks2009;53(5):569–82.

Madan R, Cui S, Lal S, Goldsmith A. Cross-layer design for lifetime maximization ininterference-limited wireless sensor networks. IEEE Transactions on WirelessCommunications 2006;5(11):3142–52.

Mao J,WuZ,WuX. A TDMA scheduling scheme formany-to-one communications inwireless sensor networks. Computer Communications 2007;30(4):863–72.

Melodia T, Akyildiz IF. Cross-layer QoS-aware communication for ultra wide bandwireless multimedia sensor networks. IEEE Journal on Selected Areas inCommunications 2010;28(5):653–63.

Messier G, Hartwell J, Davies R. A sensor network cross-layer power controlalgorithm that incorporates multiple-access interference. IEEE Transactionson Wireless Communications 2008;7(8):2877–83.

Miao G, Li GY, Swami A. Decentralized optimization for multichannel randomaccess. IEEE Transactions on Communications 2009;57(10):3012–23.

Misic J. Algorithm for equalization of cluster lifetimes in a multi-level beaconenabled 802.15.4 sensor network. Computer Networks 2007;51(11):3252–64.

Misic J. Traffic and energy consumption of an IEEE 802.15.4 network in the presenceof authenticated, ECC Diffie-Hellman ephemeral key exchange. ComputerNetworks 2008;52(11):2227–36.

Misic J, Reddy GR, Misic VB. Activity scheduling based on cross-layer information inBluetooth sensor networks. Computer Communications 2006a;29(17):3385–96.

Misic J, Shafi S, Misic VB. Cross-layer activity management in an 802-15.4 sensornetwork. IEEE Communications Magazine 2006b;44(1):131–6.

Misra S, Reisslein M, Xue G. A survey of multimedia streaming in wireless sensornetworks. IEEE Communications Surveys & Tutorials 2008;10(4):18–39.

Mitchell PD, Qiu J, Li H, Grace D. Use of aerial platforms for energy efficient mediumaccess control in wireless sensor networks. Computer Communications2010;33(4):500–12.

NamY, Kwon T, Lee H, Jung H, Choi Y. Guaranteeing the network lifetime inwirelesssensor networks: a MAC layer approach. Computer Communications2007;30(13):2532–45.

Olariu S, Xu Q. A simple and robust virtual infrastructure for massively deployedwireless sensor networks. Computer Communications 2005;28(13):1505–16.

Otal B, Alonso L, Verikoukis C. Highly reliable energy-saving MAC for wireless bodysensor networks in healthcare systems. IEEE Journal on Selected Areas inCommunications 2009;27(4):553–65.

Ozgovde A, Ersoy C. WCOT: a utility based lifetime metric for wireless sensornetworks. Computer Communications 2009;32(2):409–18.

Phan KT, Fan R, Jiang H, Vorobyov SA, Tellambura C. Network lifetimemaximizationwithnode admission inwirelessmultimedia sensor networks. IEEE Transactionson Vehicular Technology 2009;58(7):3640–6.

Ren Q, Liang Q. Throughput and energy-efficiency-aware protocol for ultrawide-band communication in wireless sensor networks: a cross-layer approach. IEEETransactions on Mobile Computing 2008;7(6):805–16.

Sanchez JA, Ruiz PM, Stojmenovic I. Energy-efficient geographic multicast routingfor sensor and actuator networks. Computer Communications 2007;30(13):2519–31.

Shi L, Fapojuwo AO. TDMA scheduling with optimized energy efficiency andminimum delay in clustered wireless sensor networks. IEEE Transactions onMobile Computing 2010;9(7):927–40.

Shi Y, Hou YT. On the capacity of UWB-based wireless sensor networks. ComputerNetworks 2008;52(14):2797–804.

Shi Y, Hou YT, Sherali HD, Midkiff SF. Optimal routing for UWB-based sensornetworks. IEEE Journal on Selected Areas in Communications 2006;24(4):857–63.

Shu T, Krunz M. Energy-efficient power/rate control and scheduling in hybridTDMA/CDMA wireless sensor networks. Computer Networks 2009;53(9):1395–408.

Srivastava V, Motani M. Cross-layer design: a survey and the road ahead. IEEECommunications Magazine 2005;43(12):112–9.

Su H, Zhang X. Battery-dynamics driven TDMA MAC protocols for wireless body-area monitoring networks in healthcare applications. IEEE Journal on SelectedAreas in Communications 2009;27(4):424–34.

Tian Y, Ekici E. Cross-layer collaborative in-network processing inmultihopwirelesssensor networks. IEEE Transactions on Mobile Computing 2007;6(3):297–310.

Vuran MC, Akyildiz IF. Error control in wireless sensor networks: a cross layeranalysis. IEEE/ACM Transactions on Networking 2009;17(4):1186–99.

Wang C, Li B, Sohraby K, Daneshmand M, Hu Y. Upstream congestion control inwireless sensor networks through cross-layer optimization. IEEE Journal onSelected Areas in Communications 2007;25(4):786–95.

WangH, Peng D,WangW, Sharif H, hwa ChenH. Cross-layer routing optimization inmultirate wireless sensor networks for distributed source coding basedapplications. IEEE Transactions on Wireless Communications 2008a;7(10):3999–4009.

WangH, Yang Y,M a, He J,Wang X. Network lifetimemaximizationwith cross-layerdesign in wireless sensor networks. IEEE Transactions on Wireless Commu-nications 2008b;7(10):3759–68.

WangW, PengD,WangH, Sharif H, ChenH-H. Cross-layermultirate interactionwithdistributed source coding in wireless sensor networks. IEEE Transactions onWireless Communications 2009;8(2):787–95.

Ye Q, Zhang Y, Cheng L. A study on the optimal time synchronization accuracy inwireless sensor networks. Computer Networks 2005;48(4):549–66.

Yick J, Mukherjee B, Ghosal D.Wireless sensor network survey. Computer Networks2008;52(12):2292–330.

Yu G, Zhang Z, Qiu P. Efficient ARQ protocols for exploiting cooperative relaying inwireless sensor networks. Computer Communications 2007;30(14–15):2765–73.

Yuan Y, He Z, Chen M. Virtual MIMO-based cross-layer design for wireless sensornetworks. IEEE Transactions on Vehicular Technology 2006a;55(3):856–64.

Yuan Y, Yang Z, He J, Chen W. An adaptive code position modulation scheme forwireless sensor networks. IEEE Communication Letters 2005;9(6):481–3.

Yuan Y, Yang Z, He Z, He J. An integrated energy awarewireless transmission systemfor qos provisioning in wireless sensor network. Computer Communications2006b;29(2):162–72.

Zhang L, Zhang Y. Energy-efficient cross-layer protocol of channel-aware geo-graphic-informed forwarding in wireless sensor networks. IEEE Transactions onVehicular Technology 2009;58(6):3041–52.



dx.doi.org/10.1016/j.jnca.2010.11.009

Documents

A survey on cross-layer solutions for wireless sensor networks