Research Article

International Journal of DistributedSensor Networks2017, Vol. 13(4)� The Author(s) 2017DOI: 10.1177/1550147717703966journals.sagepub.com/home/ijdsn

Survey on coexistence ofheterogeneous wireless networks in2.4 GHz and TV white spaces

Duzhong Zhang1,2, Quan Liu1, Lin Chen2 and Wenjun Xu1

AbstractGiven the unprecedentedly dense deployment of wireless networks due to the proliferation of inexpensive, widely avail-able wireless devices in the limited spectrum bands, many of these wireless networks should operate at least partiallyoverlapped or even on the same spectrum band. Incompatible communication pattern of different networks with differ-ent applications (e.g. Internet of Things, wireless sensor networks and individual wireless communication) would causeserious wireless networks coexistence problem. For the representative, two bands, 2.4 GHz industrial, scientific andmedical band and the TV white spaces, effective signalling protocols, as well as radio resource allocation mechanismscould achieve an efficient and fair utilization of spectrum. In this survey, we report and analyse the recent technicaladvance and development of coexistence protocols. Aiming at tracing the latest developments in this field, we attemptto deliver a comprehensive coverage on existing literatures with a proper technical depth to introduce the design ideaand philosophy and analyse the pros and cons of each surveyed coexistence solution. We also discuss a number ofimportant and relevant research challenges that have not been addressed in the existing literatures and that deserve fur-ther research attention and investigation.

KeywordsCognitive radio networks, heterogeneous wireless networks, coexistence, survey, 2.4 GHz and TV white spaces

Date received: 24 May 2016; accepted: 6 March 2017

Academic Editor: Riccardo Colella

Introduction

The last two decades have witnessed an unprecedentedsuccess of wireless networks due to the proliferation ofinexpensive, widely available wireless devices, resultingin a significant densification of deployed wireless net-works, which have been applied from individual bodymonitoring to satellites communication.

However, given the limited spectrum resources,many of these wireless networks should operate at leastpartially overlapped or even on the same spectrumband. Such phenomenon, termed as networks coexis-tence, is pronounced through a lot of spectrum bands,especially in 2.4 GHz industrial, scientific and medical(ISM) band.1 As multiple wireless communication net-work standards (e.g. Bluetooth, ZigBee and WiFi) are

operating on this band for different applications(Internet of Things (IoT), wireless sensor networks(WSNs), individual wireless communication etc.), thewireless networks communication in this band hasmore complexities than other bands.2,3

1School of Information Engineering, Wuhan University of Technology,

Wuhan, China2Laboratoire de Recherche en Informatique (LRI-CNRS UMR 8623),

Universite Paris-Sud, Orsay, France

Corresponding author:

Quan Liu, School of Information Engineering, Wuhan University of

Technology, Jianhu Campus, 122 Luoshi Road, Wuhan 430070, Hubei,

China.

Email: quanliu@whut.edu.cn

Creative Commons CC-BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License

(http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without

further permission provided the original work is attributed as specified on the SAGE and Open Access pages (http://www.uk.sagepub.com/aboutus/

openaccess.htm).

On the contrary, some authorized spectrum bandsare only used in limited geographical areas or periodsof time, which lead a lot of spectrum resources wasted.In order to reutilize these vacant spectrum resources,cognitive radio (CR), as a promising technology, hasemerged to allow secondary users (SUs) unlicensed net-works to access underutilized spectrum bands which isauthorized to the primary users (PUs).4,5 So that, CRtechnology has many benefits on wireless networks’data transmission under complex environment byimproving spectrum utilization, which gives out numer-ous novel ideas (e.g. CR for IoT, CR in vehicle net-works and CR WSN).6–8 Due to the ongoing researchof CR networks, new CR network standards arerequired to propose on TV White Spaces (TVWSs) forexperiments. Incompatible communication pattern ofheterogeneous CR network standards has also causedspectrum competing, and new CR networks coexistencechallenges which are different with traditional coexis-tence scenario have been appeared.9,10

Therefore, in order to introduce heterogeneous net-works coexistence precisely, the wireless networks coexis-tence mechanisms of 2.4 GHz and TVWSs would bestudied as representatives in this article. Due to the het-erogeneous natures of the coexisting networks, the coexis-tence task is particularly challenging. Specifically, theheterogeneity may contain one or more following aspects:

� Heterogeneity in wireless technologies at the phys-ical layer: Coexisting networks may employ dif-ferent wireless technologies, such as IEEE 802.11(WiFi),11 IEEE 802.15.1 (Bluetooth)12 and IEEE802.15.4 (ZigBee),13 all of which share the same2.4 GHz ISM band. Due to their different physi-cal (PHY) layers, each two networks of themcannot decode the packets of each other directly.

� Heterogeneity in medium access mechanisms:Coexisting networks may use different mediumaccess mechanisms, for example, WiFi and someZigBee networks are based on Carrier SenseMultiple Access (CSMA) mechanism, while theother types of ZigBee networks use TimeDivision Multiple Access (TDMA)-basedmedium access pattern. They are by nature con-flicting with each other.

� Heterogeneity in network protocols: Coexistingnetworks may use different network protocolswith different packet formats. As a result, theycannot communicate with each other directly asthe packets are not interpretable by others.

� Heterogeneity in power level: Coexisting net-works may transmit their data packets at differ-ent power levels, for example, the maximumtransmission power of a WiFi node is 20 dBmcompared with that of a ZigBee node which isonly 0 dBm. The heterogeneity in power level

may cause the hidden terminal problem, which,as analysed later in our survey, may significantlydegrade the system performance.

� Heterogeneity in applications and Quality ofServices (QoS) requirement: The applicationsrunning in coexisting networks may be heteroge-neous. A ZigBee network which is utilized forWSN usually has low bitrate application trafficwith high stability guarantee, while a WiFi net-work may support real-time multimedia applica-tions. Consequently, the QoS requirement ofthese networks are by nature heterogeneous.

We now illustrate the above heterogeneity using theexamples of the 2.4-GHz ISM band and TVWSs:

� 2.4 GHz ISM band: The coexistence problem inthis spectrum band is mainly between ZigBeeand WiFi networks due to their ubiquitousdeployments nowadays. The coexistence has adevastated effect on the performance of ZigBeeif the traffic of the WiFi network becomes dense.The reason is that the transmission power ofZigBee nodes is order of magnitude lower thanthat of WiFi nodes, making it difficult for thelatter to be informed of the existence of the for-mer. Even in some particular cases where theWiFi nodes are aware of the existence of theZigBee nodes since they use different wirelesstechnologies and communication protocols, theycannot coordinate between each other to regu-late the access to the spectrum.

� TVWSs: Due to the development of digital televi-sion technology, analogue TV broadcast signalsare massively transferred to digital pattern. Then,vacant TV spectrum bands are announced to beexperimental bandplan ‘TV White Spaces’ byFCC (Federal Communications Commission) fornovel CR networks working.9 Examples includeIEEE 802.22 wireless regional area networks(WRAN),14 IEEE 802.11af (WiFi over TVWSs)15

and ECMA 392 (wireless personal area network(WPAN) over TVWSs).16 The heterogeneity men-tioned above is also present in this scenario suchas the incompatible medium access control(MAC)/PHY designs of coexisting networks andthe heterogeneous QoS requirements.

Formally, we use the term heterogeneous networkscoexistence to denote the problem of designing distribu-ted coordination and/or resource allocation mechan-isms to achieve harmonious and efficient share of thespectrum resources among the coexisting networks. Thecoexistence mechanisms should

� Enable communication and informationexchange among heterogeneous networks;

2 International Journal of Distributed Sensor Networks

� Ensure efficient and fair resource allocationamong networks; in some cases, this means strik-ing a balance between system-level optimalityand fairness among each individual network.

To tackle the problem of heterogeneous networkscoexistence, we identify the following researchchallenges:

� Designing coordination and signalling mechan-isms: Given the heterogeneity in wireless technol-ogies and communication protocols, thecoexisting networks cannot exchange coordina-tion information among them. However, with-out coordination, it would be too hard toachieve harmonious coexistence among them.Hence, one fundamental step is to design effec-tive coordination and signalling protocols thatcan enable communication and informationexchange among heterogeneous coexistingnetworks.

� Solving the hidden terminal problem: The hiddenterminal problem in heterogeneous networks ismore complex than the original version inCSMA networks. Specifically, we distinguishtwo scenarios of the problem:

- Asymmetrical scenario: Consider two net-works 1 and 2. The transmission power ofthe nodes in network 2 is higher than that ofnodes in network 1, as illustrated by Figure1(a). Due to such power asymmetry, node 1in network 2 cannot be aware of the trans-missions of the nodes in network 1; thus,data collision cannot be avoided and causingthe hidden terminal problem.

- Symmetrical scenario: Again, consider twonetworks 1 and 2. In the symmetrical case,the transmission power of the nodes in net-work 1 is the same (or comparable) as that ofnodes in network 2, as illustrated by Figure1(b). Node 1 is transmitter and node 2 isreceiver without power emitting in both net-works. In this case, the communication ofnode 2 in network 1 may be interfered sincesignal emitters in two networks cannot senseeach other. Hence, despite the symmetry intransmission power, the hidden terminalproblem still exists because the two networksmay have different communication protocolsand different PHY interfaces,17 making theclassic CSMA/CA (Carrier sense multipleaccess with collision avoidance) mechanisminapplicable in this scenario.

� Designing efficient resource allocation mechan-isms: Another important issue is the design of

efficient resource allocation mechanisms, meet-ing the heterogeneous QoS requirements of thecoexisting networks. Ideally, the radio resourceshould be allocated among the networks propor-tionally to their QoS requirements. How toachieve such an equilibrium in a distributed wayis the key design problem here.

� Coexisting with PUs: In TVWSs, the impact ofPUs should also be taken into considerationwhen designing coexistence mechanisms andprotocols. In this regard, relevant problemsinclude how to identify a primary signal, how todistinguish the primary signal from secondarysignals and how to coordinate the access to spec-trum holes left by the PUs among the secondarynetworks.

Due to the fundamental importance of the heteroge-neous network coexistence problems in the future wire-less communication systems and the particular designchallenges brought by the heterogeneity analysed previ-ously, we devote this survey to reporting and analysingthe recent technical advance and development of coex-istence protocols. Aiming at tracing the latest develop-ments in this field, we attempt to deliver acomprehensive coverage on the existing literatures witha proper technical depth to introduce the design idea

Figure 1. Illustration of the hidden terminal problem: (a)asymmetrical scenario and (b) symmetrical scenario.

Zhang et al. 3

and philosophy and analyse the pros and cons of eachsurveyed coexistence solution. We complete the surveyby pointing out a number of important and relevantresearch challenges that have not been addressed in theexisting literatures and that deserve further researchattention and investigation.

The remaining sections are organized as follows.Section ‘Overview and taxonomy of existing coexis-tence solutions’ points out the existing main coexistencesolutions on 2.4 GHz and TVWSs. And introducesthem in different classifications. Section ‘Survey on2.4 GHz coexistence’ provides the main coexistence sce-narios and research progresses in 2.4 GHz. Section‘Survey on TVWSs coexistence’ provides main coexis-tence scenarios and research progresses in TVWSs.Section ‘Conclusion and challenges’ concludes the coex-istence solutions mentioned in this article with futureresearch directions which have not been addressed.

Overview and taxonomy of existingcoexistence solutions

In this section, we present a high-level overview on theexisting work on the heterogeneous network coexis-tence. Specifically, we present a comprehensive taxon-omy that classifies the existing coexistence mechanismsusing a set of typical criteria, as illustrated in Figure 2.Through the taxonomy presented in this section, weaim at establishing an overall picture on the existingcoexistence solutions and the related design ideas andphilosophies, which can further stimulate novel solu-tions in this field.

Classification by spectrum bands

Perhaps the most natural classification criterion of net-works coexistence is their operating spectrum bands.Using this criterion, the representative solutions couldmainly be divided into two spectrum bands, the2.4 GHz ISM band, mainly from 2400 to 2484.5 MHz,and the TVWSs, mainly on very high frequency/ultra

high frequency (VHF/UHF) bands ranging from 54 to806 MHz in the case of USA. The reasons that thesetwo bands have attracted both academic and industrialresearch attentions are as follows:

� The ISM spectrum bands are radio bandsreserved internationally for the use of radio fre-quency (RF) energy for industrial, scientific andmedical purposes other than telecommunica-tions. Due to its license-free nature, many com-munication equipments are designed to operateon this spectrum band, among which two repre-sentative ones are WiFi11 and ZigBee.18 Giventheir ubiquitous presence in daily life and per-sonal applications, designing coexistencemechanisms between them becomes a pressingtopic.

� The FCC has recently authorized license exemp-tion for operations of CR devices in TV bandsleft available by the transition from analogue todigital TV systems.9 Given the overcrowding ofthe ISM band and a number of desirable signalpropagation properties of TV bands, a largenumber of emerging standards based on CRhave targeted them as their operation spec-trums.14–16,19 Heterogeneity by nature, these net-works are expected to coexist in the samespectrum band, thus calling for efficient coexis-tence protocols among them.

In what follows, a summary of coexistence solutionsin these two spectrum bands would be stated.

Coexistence solutions on 2.4 GHz ISM band. Given the ubi-quitousness of WiFi and ZigBee technologies and theheterogeneity between them, most of the existing solu-tions are focused on the coexistence of either WiFi net-works or WiFi and ZigBee networks. Somerepresentative solutions are listed in the following forillustration. A detailed survey is presented in section‘Survey on 2.4 GHz coexistence’:

� WiFi–WiFi coexistence: Since there are onlythree independent channels which can be usedby different WiFi networks simultaneously,when there are more WiFi networks, they arerequired to coordinate among them to share thespectrum. In this regard, Zhang and Shin20

study the partial spectrum overlapping betweenWiFi networks and propose a distributedmechanism called adaptive subcarrier nulling(ASN) to enable coexistence among partiallyoverlapped WiFi networks by sharing over-lapped spectrum. And in the work by Hanet al.,21 Zhang’s group introduces a coexistenceFigure 2. Taxonomy of existing coexistence solutions.

4 International Journal of Distributed Sensor Networks

mechanism between IEEE 802.11ac and IEEE802.11a for settling single interference caused byasymmetric spectrum bandwidths. A new fine-grained spectrum sharing (FSS) mechanism isproposed to minimize networks’ channel band-widths and reallocate them into appropriatespectrum bands to decrease heterogeneous net-works’ spectrum overlapping. An adaptive chan-nel bonding (ACB) protocol is introduced byHuang et al.22 using alterable bandwidthapproach to mitigate inefficiency and unfairnessin heterogeneous coexistence scenario. Anotherresearch strand, represented by the Flashbackmechanism developed in the work by Cidonet al.,23 consists of establishing in-band controlchannels to exchange control informationamong coexisting networks on top of datapackets.

� WiFi–ZigBee coexistence: There exists a signifi-cant asymmetry in terms of transmission powerbetween WiFi and ZigBee nodes. Given suchasymmetry, a coexistence mechanism, as devel-oped in the work by Zhou et al.,24 is to makeWiFi signals distinguishable to ZigBee nodesthrough detecting periodic beacons of WiFi net-works, and in the work by Yan et al.,25 a wide-band antenna is utilized on reprogrammedZigBee nodes for decoding WiFi signals andrearranging ZigBee networks’ communicationchannels, while in the work by Hao et al.,26 peri-odic beacons of WiFi are applied to calibrateZigBee networks’ synchronization. Anothersolution, developed by Liang et al.,27 increasesthe robustness of ZigBee signals to make themsurvivable under WiFi interference. While in theworks by Yubo et al.28 and Yang et al.,29

multiple-input multiple-output (MIMO) systemis employed to alleviate WiFi interference inZigBee data. The third research direction on thistopic is to design coordination mechanismsamong WiFi and ZigBee nodes. In this regard,Zhang and Shin30–32 designed two types of solu-tions: they developed a coordination mechanismcalled cooperative carrier signalling (CCS) orcooperative busy tone (CBT) that uses asymme-trical bandwidth between WiFi and ZigBee net-works and puts a high power level busy signal onadjacent ZigBee channel to inform WiFi devices;in the work by Zhang and Shin,33 a Gap Sense(GSense) mechanism is devised that puts a high-power preamble before TDMA networks’ (e.g.ZigBee) packets to inform CSMA networks (e.g.WiFi) devices and uses gaps between beacons inpreamble to exchange control information; and aDynamic medium access control (DynMAC)protocol is presented by Correia et al.34 for the

coexistence problem in TDMA tree topologyWSNs. By utilizing CR technology, nodes in treeare reconfigurated into different time slots, andappropriate channels are evaluated by DynMACand allocated to each node in order to avoidCSMA networks interferences.

� Coexistence between WiFi and high-power non-WiFi sources: This type is a special one as itfocuses on the WiFi networks coexisting withhigh-power non-WiFi sources which can emitthe cross-technology interference (CTI) on2.4 GHz band. Due to rare attention, there isonly one solution presented by Gollakotaet al.,35 a MIMO system is applied to WiFi net-work to enhance the robustness of WiFi signalsand mitigate the CTI.

Coexistence solutions in TVWSs. As mentioned previously,many emerging CR networks are designed to operateon TVWSs band with a number of self-coexistencemechanisms depicted in the related standards.Specifically, both IEEE 802.22 and ECMA 392 provideself-coexistence mechanisms:14,16 a two-degree coexis-tence mechanism is designed in IEEE 802.22 (networkscoexistence and cells coexistence) to guarantee that dis-tinct channels are chosen between neighbour base sta-tions (BSs) and no conflicts exist in terminals’communication; coexistence beacons are employed inECMA 392 to allow adjacent devices channel sharing.In IEEE 802.16h,36 two types of coexistence mechan-isms, uncoordinated mechanism and coordinatedmechanism, are proposed in heterogeneous and homo-geneous scenarios, respectively. Besides the solutions inexisting standards, a number of novel coexistencemechanisms have been proposed for specific scenariosand applications. Zhao and colleagues37,38 proposed afair MAC protocol to solve coexistence problembetween CR networks and between CR networks andPUs using a three-state model. Bian et al.17 introducedSpectrum Sharing for Heterogeneous Coexistence(SHARE) mechanism to mitigate hidden terminalproblem and Symbiotic Heterogeneous CoexistenceArchitecture (SHARE) to establish coordinationbetween heterogeneous CR networks.39,40 But a calcu-lating fault in the work by Bian and colleagues39,40 hasbeen found, so Zhang et al.41 provided a betterecology-based spectrum sharing mechanism.

Classification by domains

However, coexistence mechanisms for the heteroge-neous wireless networks are distinctive with each other,but they could also be classified into different domainswhich their coexistence approaches work in. Then, wewould like to introduce some representative solutionsunder time, frequency and code domains in this part.

Zhang et al. 5

Coexistence in time domain. In this class of coexistencesolutions, each coexisting network gets a time share ofthe total radio resource. A coordination protocol isrequired to negotiate how time is shared among coexist-ing networks, so that the collision can be avoided. Arepresentative coexistence solution in time domain isthe SHARE mechanism developed by Bian et al.17

SHARE divides communication time slots into differ-ent types in order to carry the frames of different net-works. An algorithm is developed to allocate time slotsamong coexisting networks with the objective of achiev-ing fairness among them. And in the work by Amjadet al.,42 heterogeneous CR networks’ data transmissionperiods are split into equal small time slots and allo-cated in appropriate channels by game-based algo-rithm, so that data collisions between CR networkscould be minimized and coexistence is realized.

Coexistence in frequency domain. Another coexistencesolution is to allow coexisting networks to coexist inthe frequency domain. Specifically, they are required tocoordinate their operation frequencies in order toswitch to non-overlapping channels. A coexistencemechanism in the frequency domain is provided byZhang and Shin20 where ASN mechanism is developedto divide a channel into several subbands and to regu-late the coexistence on the basis of subbands. While inthe work by Sun et al.,43 a novel spectrum sharingmechanism is proposed for coordinating licensed usersand CR sensor networks’ spectrum shares and datarate through utilizing Stackelberg game theory in orderto improve victim nodes’ overall throughput.

Coexistence in code domain. This type of solutions isbased on code domain, and by utilizing specific codes,a side channel could be constructed for transmittingcontrol messages on top of data packets. As an exam-ple, Wu et al.44 proposed to create a side control chan-nel through special interference codes on top of datacommunication.

Classification by mechanisms

Depending on whether the decision and the coordina-tion mechanisms are performed by a centralized entityor in a distributed fashion, the coexistence solutionscan be categorized into centralized and distributedapproaches.

Centralized coexistence approach. In centralizedapproaches, the coexistence decision is performed by acentral entity, which is then sent to each coexisting net-work. An example of centralized approach is that

defined in IEEE 802.19.1,19 where a mediator is desig-nated to make the coexistence decision.

Distributed coexistence approach. In distributed coexis-tence approaches, on the contrary, the decision is madein a decentralized fashion without relying on a centralentity. An example of distributed approaches is GapSense,33 where the coexistence is coordinated distribu-tively via the preambles of data packets.

Classification by control channel requirements

In some coexistence solutions, one or several controlchannels are required to convey the control and coordi-nation information. In other coexistence approaches,the coordination is achieved without a control channel.

Control channel is essential to MAC protocol andcoexistence mechanism. Consequently, the existence ofcontrol channel can divide coexistence solutions intotwo part.

Coexistence approaches requiring a control channel. In manycases, a control channel is essential to the MAC proto-col and can significantly facilitate many basic networkfunctionalities such as coexistence. A coexistence con-trol channel is specified in IEEE 802.16h,36 which isdedicated to the synchronization, spectrum detectionand inter-system coordination among secondary cogni-tive users.

Control channel free. In many practical scenarios, a con-trol channel is impractical or too expensive to main-tain. In this case, coexistence is negotiated andcoordinated without a control channel. As an example,the BuzzBuzz protocol devised by Liang et al.27 doesnot require any form of control channel. Instead, itincreases the robustness of the ZigBee packets via repe-tition and error correction to make ZigBee networkscoexist with WiFi networks.

To conclude this section, we would like to emphasisthat the above classifications are by no means mutuallyexclusive on to another. On the contrary, one coexis-tence solution can belong to multiple categories in theclassification or support several operation modes in dif-ferent categories. In the following two sections, we pro-vide a comprehensive survey on the existing coexistencesolutions following the classification based on the oper-ation frequency of the coexisting networks. Specifically,section ‘Survey on 2.4 GHz coexistence’ is dedicated tothe coexistence solutions on 2.4 GHz ISM band andsection ‘Survey on TVWSs coexistence’ is focused onTVWSs.

6 International Journal of Distributed Sensor Networks

Survey on 2.4 GHz coexistence

In this section, a comprehensive survey on 2.4 GHzcoexistence solutions is presented. As introduced in theprevious section, three main coexistence scenarios,WiFi-WiFi coexistence, WiFi-ZigBee coexistence andthe coexistence between WiFi and high-power non-WiFi sources, are particularly pronounced in this spec-trum band. In the following, we organize our surveybased on these three scenarios.

WiFi–WiFi coexistence

We start with WiFi–WiFi coexistence. Figure 3 illus-trates the default spectrum operation setting of mostWiFi networks today. Specifically, 14 channels of22 MHz each are available on the 2.4-GHz ISM band,resulting a total bandwidth of 83.5 MHz. Among the14 channels, there are only three non-overlapping chan-nels, namely, the channels 1, 6 and 11 according to theIEEE 802.11 standard. Given the high density of thedeployment of WiFi hot-spots today, it is inevitablethat a WiFi network shares part or all of its operatingspectrum band with neighbour WiFi networks.Consequently, designing coexistence mechanismsamong WiFi networks becomes an urgent researchproblem. The problem of WiFi–WiFi coexistencebecomes even more pressing as IEEE 802.11ac45 fur-ther envisions to increase the channel bandwidth to 80and 160 MHz.

Coexistence via partial spectrum sharing. Current WiFi net-works rely on CSMA/CA to coordinate transmissionson the same channel. However, the CSMA/CA cannotcoordinate the transmissions of different WiFi net-works operating on partially overlapping channels. Inresponse to the problem of coexistence of WiFi net-works on partially overlapping spectrum channels,Zhang and Shin20 developed a solution called adaptivesubcarrier nulling.

The main purpose of ASN is to mitigate collisionsdue to partial overlapping by enabling wireless localarea networks (WLANs) sharing overlapped spectrum.To motivate the design of ASN, a comprehensive mea-surement about partial overlapping interference inIEEE802.11b/g networks is conducted.20 Particularly,due to orthogonal frequency division multiplexing(OFDM) PHY layer of IEEE802.11g, partial

overlapping may cause more destructive damage todata transmission in IEEE802.11g than inIEEE802.11b since full overlapping and partial over-lapping may both lead to all packets loss inIEEE802.11g network. Specifically, this damage maycause two types of problem as shown in Figure 4, ’A’,’B’, ’C’ represent different networks nodes’ transmis-sion channel width, and (a), (b), (c) in Figure 4 are dif-ferent overlapping scenarios:

� The first type, illustrated in Figure 4(a), is calledpartial channel blocking. As CSMA/CA mechan-ism is employed by IEEE 802.11, when part ofchannel is occupied by co-located narrow bandWLANs, the entire channel should be suspended.

� The second, illustrated in Figure 4(b) and (c), iscalled channel starvation. In these two scenarios,node A can transmit only if nodes B and C areboth idle. It is rare that both B and C completetransmission at the same time and lose conten-tion with A. So, A would remain starvation withlarge probability.

ASN is proposed to address the above problem. InASN, a WiFi channel is split into several subbands. Asshown in Figure 3, the minimum overlapping spectrumbandwidth between adjacent channels is 5 MHz in802.11 g, each channel of 20 MHz is thus split intofour subbands. When mutual interference is detected, anode can adjust its bandwidth on the basis on sub-bands so as to share the spectrum with other nodes. InASN, since the transmission spectrum may change,receivers cannot detect and decode packets as in origi-nal IEEE 802.11 networks. Hence, a suite of mechan-isms is proposed in ASN in this context, includingpacket detection, synchronization and decoding.Moreover, adjacent subbands may generate interfer-ence even though they are orthogonal. To mitigateinterference between subbands, narrow guard bandsare introduced in ASN. Specifically, two adjacent sub-bands should be separated by three subcarriers.

In summary, ASN is easy to implement in existingnetwork devices without any other extra equipmentand can increase the overall throughput of heteroge-neous networks and decreases collisions among themby facilitating spectrum sharing. However, due to adja-cent channel interference, guard bands should be

Figure 3. WiFi–ZigBee spectrum.Figure 4. Heterogeneous channel width and partialoverlapping interference in IEEE 802.11 networks.

Zhang et al. 7

configured at the subband level, which may decreasethe overall spectrum efficiency.

Coexistence via control channel. However, an additionalin-band control channel mechanism Flashback is pro-posed by Cidon et al.23 to coordinate heterogeneousWiFi networks coexistence.

In current WiFi networks communication, ordinaryextra control channel construction is unappropriate toapply, since the ISM band is overcrowded, no extraspectrum can be equipped as control channel. Forexample, a 2-MHz control channel is employed in aWiFi channel, and an extra 1.875-MHz guard bandmust be added to eliminate the mutual interferencebetween control channel and data plane. Hence, con-trol channel would occupy 25% bandwidth of onechannel in a 20-MHz WiFi channel at least, collapse ofthroughput cannot be avoided. This extra cost is toohigh for WiFi networks communication. While anothersolution of this problem, orthogonal frequency divisionmultiple access (OFDMA) protocol, can realize thedata and control packets transmitting at the same timeon different frequency without extra spectrum require-ment. But OFDMA protocol requires tight clock syn-chronization among all network nodes which is notsuitable for WiFi network. And other solutions likeRTS (request to send) and CTS (clear to send) packetsare also unappropriate for WiFi network due to extraconsumption of transmission duration which lead todecrease in throughput too. So, using in-band controlchannel approach without consumption on throughputis the best idea to solve this problem.

To realize this idea, there are two challenges, linkmargin and flashes decoding. First is link margin;before its explanation, we should present the conceptof flash at first. As shown in Figure 5, flash is a type ofsimple high-power sinusoid and has same frequencyand duration to a OFDM subcarrier since WiFi net-works have the OFDM PHY layer. In essential, flash isa kind of noise to normal data. However, link marginis a kind of statistic data to estimate current channelquality. It is the difference between the instantaneouschannel signal-to-noise ratio (SNR) and the minimalSNR which must be guaranteed to decode packet.Therefore, link margin is designed to measure the inter-ference from flash to ordinary data transmission. Theflash emitting should not surpass the link margin con-straint so that flashes and ordinary data both can bedecoded correctly in receiver, and in-band controlchannel can be constructed. Second is flashes decoding,which is easy to be understood. As the literal meaning,flashes are only a simple sinusoid signal which is impos-sible to bring complex message in a single signal, andthis challenge is to design a method to transmit anddecode the control messages in flashes.

Consequently, to address these problems and chal-lenges, Flashback technique is proposed. First, flashestransmission is important. By applying subcarriers ofOFDM PHY, each flash can be transmitted in oneOFDM symbol, which only occupied 1/64 of channeland last for 4 ms in WiFi network. Hence, compared towhole WiFi packet transmission, the transmission offlashes is rare, which would not cause important impacton data transmission in WiFi network. And for oneperiod of data packet transmission, flashes may appearat different subcarrier in different time slot, so thesedynamical transmission pattern can be abstracted intoa time–frequency coordinate matrix, which is conveni-ent for locating and encoding flashes of Flashback, sothat control messages can be encoded in these matrix.Second, for each flash decoding, erasure mechanism isdesigned. For each packet receiving, flashes and pack-ets are received simultaneously by detecting flashesfirst, and the error bit which is flashed would be erasedsince correcting it would cost much more than erasure.Due to the burst feature of flash, the duration ofdecoding flash is much shorter than data packet trans-mission, receiver can easily use algorithm to decodewhole packet without the erased error bits, so controlmessage and data can be received completely. Third,due to asynchronous WiFi devices, ordinary time–frequency grid cannot be applied to encode and decodecontrol messages directly. Hence, a distance detectingmechanism is employed. It is easy to understand thismechanism, for instance, two flashes can be sent at thesame time at subcarriers 2 and 27, so that the distancebetween them is 25, which can be the code of controlmessages. Using distances, control messages can be senton an average 175 Kbps bitrate, meanwhile, an 8-bitcyclic redundancy check (CRC) is also added into

Figure 5. Flash on top of data transmission.

8 International Journal of Distributed Sensor Networks

control packets for error correction. Finally, the linkmargin is very easy to be guaranteed using an algo-rithm to calculate the appropriate transmitting powerbefore each flash transmission can eliminate this prob-lem perfectly.

In general, by detecting control messages, coordina-tion between WiFi nodes can be completed well, theinterference can be mitigated without any extra costand an extra control channel is constructed to transmitinformation. But this solution only focuses on nodescoexistence in same WiFi network, which is not suit-able for heterogeneous IEEE 802.11 standard networkscoexistence.

WiFi–ZigBee coexistence

In this subsection, we focus on WiFi–ZigBee coexis-tence. The coexistence between WiFi and ZigBee net-works has attracted significant research attentionrecently due to the following reasons: first, due to ubi-quitous employment of WiFi and ZigBee networks indaily life, coexistence problem between them becomes apressing problem, and hence, WiFi–ZigBee coexistencesolutions have a practical impact; second, WiFi–ZigBeecoexistence represents a very typical coexistencebetween heterogeneous networks, and consequently,research on this particular scenario can be easilyextended to a broader coexistence context.

The major design challenges of coexistence protocolsbetween WiFi and ZigBee networks are twofold: theasymmetry in transmission power and channel band-width, recaptured as follows:

� Asymmetry in transmission power: The transmis-sion power of ZigBee devices is 20 dBm lowerthan that of WiFi devices. Consequently, ZigBeedevices cannot be properly sensed by WiFidevices. In contrast, WiFi signal may severelyinterfere ZigBee receivers.

� Asymmetry in channel bandwidth: Each ZigBeechannel occupies 5 MHz, while each WiFi chan-nel occupies 22 MHz, which overlaps fourZigBee channels.

In the following, we present major existing WiFi–ZigBee coexistence mechanisms and how they addressthe above design challenge.

Exploring WiFi white spaces. Huang et al.46 propose acoexistence approach termed WISE (WhIte Space-aware framE adaptation) that exploits the white spacesbetween WiFi frames. The motivation of WISE is thatthe WiFi traffic is highly bursty and thus leaves signifi-cant amount of white spaces between WiFi frames thatcan be exploited by ZigBee transmission.

To address the coexistence problem using the whitespaces, Huang et al. presented an experiment to mea-sure the interference caused by WiFi nodes based onwhich they identified three types of terminal problem,hidden terminal problem, exposed terminal problem andblind terminal problem, which are shown in Figure 6and summarized as follows:

� Hidden terminal problem: The ZigBee receiver isin the interference range of WiFi nodes but thesender is outside the range.

Figure 6. Terminal problems in WiFi–ZigBee scenario:(a) hidden terminal, (b) exposed terminal and (c) blind terminal.

Zhang et al. 9

� Exposed terminal problem: The ZigBee sender isin the interference range of WiFi nodes, whilethe receiver is not.

� Blind terminal problem: The ZigBee sender andreceiver are both in the interference range ofWiFi nodes but WiFi nodes cannot sense them.

Both hidden terminal problem and blind terminalproblem can lead to data collision. WISE addresses theblind terminal problem. WISE consists of two maincomponents, white space modelling and frameadaptation:

� The first component predicts the white spacesduration. According to experiment, WiFi framesform clusters with inter-frame durations typicallyless than 1 ms within clusters. ZigBee packetheaders being more than 544 ms at least, inter-frame durations within clusters cannot be usedby ZigBee nodes. The real white spaces whichcan be used are intervals between clusters. Giventhat the WiFi traffic arrival process exists self-similarity, Pareto model is applied to fit the arri-val process of WiFi frame clusters and to predictWiFi white spaces duration.

� The second component calculates the size ofZigBee frames based on the outcome of the firstcomponent. The ZigBee frames are then dividedinto sub-frames which are suitable for transmis-sion in white spaces without collision. SessionIDs and delimiter mechanism are also appliedfor frame reconstruction and correction.

To summarize, WISE resolves the blind terminalproblem with no extra consumption and cost at ZigBeedevices. However, this approach needs to suspendZigBee transmissions when WiFi traffic arrives, whichmay decrease ZigBee throughput. It is also unsuitablefor TDMA packets and delay-sensitive applicationswhere the Pareto model is only an estimation.

Among other solutions that detect WiFi signals anduse WiFi white spaces, Zhou et al.24 developed a system‘ZiFi’ which utilizes ZigBee radios to detect WiFi sig-nals from noise signals. As WiFi networks broadcastbeacon signals periodically, WiFi access points (APs)can be discovered by sensing these signals. But due tothe heterogeneity of WiFi and ZigBee networks, ZigBeedevices cannot decode WiFi’s signals. To address thisproblem, the authors developed a digital signal process-ing (DSP) algorithm Common Multiple Folding (CMF)to amplify unknown periodic signals in received signalstrength (RSS) sample using folding technique, firstused in pulsar searching on large radio telescope. Aconstant false alarm rate (CFAR) detector is designedto detect WiFi beacons in the results of CMF. Theadvantages of ZiFi include high accuracy, low overhead

and low delay. However, the mitigation of interferencebetween ZigBee and WiFi is not addressed.

Enhancing robustness of ZigBee packets. Due to powerasymmetry between WiFi and ZigBee, ZigBee packetssuffer from the interference from WiFi signals. Lianget al.27 designed a novel approach called BuzzBuzz toenhance the robustness of ZigBee packets such thatthey can be successfully decoded at the receiver sideeven under the presence of WiFi signals.

To guide the design of BuzzBuzz, the authors con-ducted a series of experiments to measure the interfer-ence between WiFi and ZigBee on bit level. Theexperimented scenarios can be divided into two types,which can be analogized to symmetric and asymmetricscenario as shown in Figure 1. The following observa-tion is drawn:

� In the symmetric scenario, the bit errors ofZigBee packets due to collision are distributed atthe beginning of ZigBee packets since CSMAmechanism can work and WiFi transmissionsare suspended shortly after detecting ZigBeetransmissions.

� In the asymmetric scenario, which fails to senseof ZigBee signals, WiFi transmission cannotstop when ZigBee data are transmitted.Consequently, bit errors are distributed acrossthe whole ZigBee packet.

Based on the above observation, BuzzBuzz containstwo components, each addressing a scenario:

� In the symmetric scenario, the starting bits ofZigBee packets are corrupted by WiFi signals,thus leading to CRC check failure at the receiverside, despite the remaining bits are received cor-rectly. If there exists some approach that canmake ZigBee packets pass CRC check, datatransmission can be completed successfully.Motivated by this argument, the Multi-Headers(MH) mechanism is proposed, which puts sev-eral headers into the beginning of data part inZigBee frame such that the remaining headerscan pass CRC check when some headers are col-lided by WiFi signals.

� In the asymmetric scenario, due to the even dis-tribution of bit errors across the whole packet,the MH mechanism cannot work any more. Theforward error correction (FEC) approach isintroduced to recover the ZigBee packetsusing Reed–Solomon (RS) code, a block-basedlinear error-correction code widely used in digi-tal communication. The authors developed a

10 International Journal of Distributed Sensor Networks

full-featured TinyOS-compatible RS library(TinyRS) to decode the ZigBee packets.

In conclusion, the advantages of BuzzBuzz includethe low cost, the ease of using and the low consump-tion. The major disadvantage is the overhead createdby the MH and the FEC components in terms of packetsize and extra processing overhead.

Enabling WiFi–ZigBee coordination. Another researchstrand on WiFi–ZigBee coexistence is to devise coordi-nation protocols between them.30–33 The main chal-lenges of the coordination mechanism design are theasymmetry in transmission power, operation band-width and PHY layer protocols. Moreover, for ZigBeenetworks in CSMA mode, the fastest clear channelassessment (CCA) operation takes 128 ms and the rx/txswitching time 192 ms, while the CCA duration of WiFinetworks lasts 20 or 9 ms. As a result, WiFi nodes maypre-empt transmission channel when ZigBee switchesfrom receive mode to transmit mode and may causecollision.

To address these design challenges, Zhang andShin30–32 provided an approach called CooperativeBusy Tone, whose core idea is illustrated in Figure 7:

� A ZigBee signaller is deployed to solve the prob-lem of WiFi–ZigBee power asymmetry. The sig-naller is a separate ZigBee node, which is close tothe WiFi network or has more power than ordi-nary ZigBee nodes. It is designed to emit busytone to inform WiFi nodes when the ZigBee

network has data to transmit. Note that busytone can be sensed without decoding.

� As busy tone can interfere ZigBee network’scommunication too if they are working in thesame channel, and channel bandwidth asymme-try (one WiFi channel bandwidth covers fourZigBee channels) is utilized to solve the problem.A frequency flipping mechanism is developedsuch that when ZigBee nodes are transmittingdata, the signaller will transmit busy tone onadjacent channel.

� To avoid collisions caused by asymmetric CCAduration, two different approaches are furtherdeveloped for ZigBee CSMA and TDMAmodes:- In CSMA mode, ZigBee devices transmit

data after sensing and contention. In CBTmechanism, every data transmission must beinitiated by the signaller, who broadcasts anotification message (referred to as CTS) onoriginal channel and then switches to adja-cent channel to start emitting busy tone.Meanwhile, ZigBee transmitter contends forchannel access normally. Under suchmechanism, WiFi nodes can sense busy tonebefore ZigBee transmission. When ZigBeedevices complete communication, an ACKpacket is transmitted. When hearing theACK packet, the signaller stops broadcast-ing busy tone.

- In TDMA mode, each data transmission isalso initiated by the signaller, who performsCCA for at most Km times before transmis-sion. If the Kj th CCA is idle, the signallerhops to adjacent channel to broadcast busytone until the end of the transmission.Hence, busy tone broadcasting is(Km 2 Kj)CCA earlier than data transmis-sion. In this case, WiFi nodes can also sensebusy tone before data transmission andavoid collision.

To summarize, the CBT mechanism can coordinatecoexistence between ZigBee and WiFi networks effec-tively and avoids collisions. Such advantage comes withthe price of deploying an extra ZigBee channel and anextra device (the signaller). Moreover, fairness isanother concern when long duration busy tone isbroadcast.

Zhang and Shin33 have also developed anothermechanism called Gap Sense to coordinate heteroge-neous wireless networks using a preamble constructionand detection algorithm. The core idea of GSense is tosend a number of energy pulses in the preamble of eachpacket and encode information using the time gapbetween two consecutive pulses. Heterogeneous net-works then utilize such control channel to exchange

Figure 7. CBT illustration: Zt, Zr, St and Wt are ZigBeetransmitter, ZigBee receiver, signaller and WiFi transmitter,respectively.

Zhang et al. 11

control information and coordinate with each other.The disadvantages of GSense are as follows. First, thepreambles extend the duration of data packets anddecrease the network throughput. Second, the sensingand interference ranges of different networks may beasymmetric, meaning that the pulses need to be ampli-fied to be sensed by other networks, leading to extraenergy consumption. Third, for TDMA-based net-works, it is expensive or even impractical to monitorchannels continuously.

Coexistence between WiFi and high-power non-WiFisources

Besides WiFi–WiFi and WiFi–ZigBee coexistence ana-lysed previously, CTI from high-power non-WiFisources has recently become a major problem whichmay have detrimental effect on WiFi performance dueto the deployment of such sources including surveil-lance cameras, baby monitors, microwave ovens, digitaland analogue cordless phones and outdoor microwavelinks. A traditional solution to CTI is to increase therobustness of the WiFi transmission by reducing itstransmission rate. However, this approach cannot workwith high-power interference. An alternative approachis to let the interfered WiFi communication hop tonon-interfered channels, but this channel hopping–based mechanism is less effective nowadays with moreand more devices being deployed in the 2.4-GHz ISMband, making it more and more difficult to find avail-able channels without interference.

In response to the above problem, Gollakota et al.35

proposed a novel system called TIMO (TechnologyIndependent Multi-Output) using the MIMO capabilityinherent to 802.11n to mitigate high-power CTI.

Specifically, TIMO uses multiple interfaces in onedevice to decode a signal of interest, even when thechannel from other concurrent transmissions isunknown. The core idea of TIMO is presented as fol-lows via a simple illustrative example. Consider a pairof two-antenna WiFi nodes that want to communicatein the presence of a high-power unknown interferer.Let s(t) be the signal of interest and i(t) the interferencesignal. The receiver node receives the following signalson its two antennas

y1ðtÞ= hi � iðtÞ+ hs � sðtÞ ð1Þ

y2ðtÞ= h0i � iðtÞ+ h0s � sðtÞ ð2Þ

where hi and h0i are the channel functions from theinterferer to the receiver and are unknown to the recei-ver, hs and h0s are the channel functions from the senderto the receiver and are known to the receiver and i(t)and s(t) are the signal of interference and the signal of

interest. If the receiver can calculate the ratio hi=h0t¼D b,

termed as the interferer’s channel ratio, then it can solves(t) from the above two equations. Note that the ideacan also be extended to the generic case with multi-interfaces devices.

The main flowchart in a TIMO receiver is shown inFigure 8, which uses equations (1) and (2) to calculatethe signal of interest s(t). A TIMO receiver consists ofthree main components:

� An algorithm for computing the interferer’schannel ratio in an OFDM subcarrier withoutany prior knowledge. Note that the interferer’schannel ratio is calculated for each OFDMsubcarrier.

� A decoder that allows the receiver to decode thesignal of interest, given the interferer’s channelratio in every OFDM subcarrier. As the inter-symbol interference (ISI) in some wide-bandCTI, which can lead to low accuracy of decod-ing, an inverse filter approach is also employedin receiver to eliminate the ISI.

� An iteration mechanism that reduces the noise inthe computation of channel ratios, hence increas-ing SNR. By iterating computed signal of interestof the last time slot, the signal of interest can beestimated more accurately and can converselyincrease the accuracy of interference channel ratio.

In summary, TIMO leverages the MIMO capabilityinherent to 802.11n to mitigate high-power CTI. Theprice paid is the loss of performance gain (in terms ofthroughput) brought by MIMO.

Survey on TVWSs coexistence

In this section, we present a comprehensive survey onheterogeneous networks coexistence in TVWSs, whichhas become a pressing research problem because manyCR networks operate on TVWSs with PUs (typicallyTV signals). Due to specialities of CR networks (e.g.the presence of PUs), the coexistence problem inTVWSs is more complex in certain aspect than that in2.4 GHz presented in section ‘Survey on 2.4 GHz coex-istence’. This section is structured in the following way.We first present CR network standards in TVWSsdeveloped recently and the coexistence mechanisms in

Figure 8. Flowchart of TIMO.

12 International Journal of Distributed Sensor Networks

the standards. We then survey the state-of-the-art coex-istence solutions proposed by the research community.

Coexistence mechanisms in CR standards

We start with CR network standards coexistence.Emerging CR network standards on this band includeIEEE 802.19.1,19 IEEE 802.22,14 IEEE 802.11af,15

ECMA 392,16 and IEEE 802.16h36. These emergingstandards have heterogeneous spectrum bandwidth,transmitting power and PHY and MAC layer. In gen-eral, the coexistence scenario in these standards is com-pletely asymmetric. Hence, most of them have theirown coexistence mechanism. But due to the focuses ofstandards are different, the coexistence mechanismtypes and targets are different.

IEEE 802.22. We first review the coexistence of homo-geneous networks in the CR standards. There are twostandards addressing the homogeneous coexistenceissue, IEEE 802.22 and ECMA 392. This subsection isfocused on IEEE 802.22.

There are mainly three challenges in devising homo-geneous network coexistence mechanisms:

� Presence of PUs: A particularity of CR networksis the presence of PUs. How to distinguish a pri-mary signal and a secondary signal and how toreact accordingly should be taken into consider-ation in the design of coexistence mechanisms;

� Multi-channel access and allocation: CR net-works are by nature multi-channel environ-ments. Coexistence mechanisms should specify

how to coordinate cells’ channel selection and/orshare the channel;

� Inter-network information exchange: Since differ-ent networks may operate on different channels,how to exchange coordination information amongcoexisting wireless networks should be addressed.

To address the above design challenges, a coexis-tence protocol is developed in IEEE 802.22 called coex-istence beacon protocol (CBP). CBP has two types ofhomogeneous coexistence approaches:

� The first approach is spectrum etiquette whichmanages the coexistence of different IEEE802.22 networks operating on different channels.Specifically, as illustrated in Figure 9, spectrumetiquette is responsible for channel selection andcoordination. The main objective is to ensurethat the operation channel and the first backupchannel of a BS is different and orthogonal tothose of its neighbour BSs and PUs. To realizethe objective, a self-coexistence window (SCW) isdeployed before each frame to exchange controlinformation.

� The second approach is on-demand framecontention, which manages the coexistence ofcells operating on the same channel. In thisapproach, superframe control headers (SCH) areemployed to decide each frame allocation insame channel. When hearing SCH, cells candecode the channel allocation information. Onlythe cell to whom the channel is allocated trans-mits data on the allocated channel while othersstay silent.

Figure 9. Spectrum etiquette.

Zhang et al. 13

ECMA 392. ECMA 392 has another homogeneous net-work coexistence mechanism. In ECMA 392, due tothe deployment of beacon period (BP) mechanism inframe construction, frames of various devices in samenetwork can be allocated appropriately. Each BP con-tains several different signal windows, which is designedfor the contention and reservation, so different devicescan emit their beacons in BP to compete the framesallocation after BP. Using this mechanism, intra-cellcollisions can be eliminated. For inter-cell coexistence,three types of self-coexistence scenario are specified inECMA 392:

� Coexistence of two master–slaver networks;� Coexistence of two peer-to-peer networks;� Coexistence of a master–slaver network and a

peer-to-peer network.

To realize coexistence in these three types of coexis-tence scenario, a lot of coexistence mechanisms aredeveloped which can be summarized as two types:

� The first one is BP merging that merges the BPsof different networks based on a set of strategies,so that different networks in different coexistencescenarios can use the same BP to allocate theirframes and thus avoid collision.

� The second one is changing coexistence scenario,for all cell networks, to keep only one master celland let the other cells be in the slave mode. Bythis mechanism, all cells’ frame schedules in dif-ferent networks can be managed by the masterto coordinate the coexistence.

IEEE 802.16h. We now turn to heterogeneous networkcoexistence, which is by nature more challenging thanhomogeneous network coexistence. In this subsection,we survey the coexistence mechanism in IEEE 802.16h.IEEE 802.16h specifies two types of coexistencemechanisms, coordinated and uncoordinated mechan-isms, targeted to coexistence scenarios with recognizeddevices and non-recognizable devices, respectively:

� Coordinated mechanism: The coordinatedmechanism mainly manages coexistence of IEEE802.16h devices. The first step is to synchronizethe devices’ MAC frames and to separate basestation from its subscriber station transmission.The second step is to allocate a dedicated chan-nel as a control channel to exchange coexistenceinformation using dynamic channel selection(DCS) and adaptive channel selection (ACS).Finally, by utilizing coexistence frame, coordi-nated scheduling and fairness algorithm are usedto share the channel among different networks.

� Uncoordinated mechanism: The uncoordinatedmechanism is focused on networks with non-recognizable devices, including PUs and SUs.Specifically, a dynamic frequency selection (DFS)protocol is developed to manage the coexistencewith PUs by searching the channels free of PUsand switching SUs to such idle channels. TheDCS protocol is used to coordinate coexistenceof SUs by finding the best operating channel tobe shared with other SUs.

The main advantage of the coexistence mechanismin IEEE 802.16h is its generic nature which is applicablein a wide range of scenarios in TVWSs. The disadvan-tage is its high complexity which makes it implementa-ble only in high-speed CPUs.

IEEE 802.19.1. IEEE 802.19.1 is a generic coexistenceprotocol that plays the role of a mediator to coordinatecoexistence of heterogeneous CR networks.

The architecture of IEEE 802.19.1 is illustrated inFigure 10. IEEE 802.19.1 has three main components,coexistence manager (CM), coexistence enabler (CE)and coexistence discovery and information sever (CDIS):

� CM is the core component in the IEEE 802.19.1systems because it is responsible for coexistencedecision-making and discovers and com-municates with other CMs. Coexistencedecision-making is the process of exchanging thecoexistence information, sending correspondentrequests, commands or control messages toother CEs.

� CE connects CMs and TV band devices (TVBD)or TV band networks. It has two functionalities.The first one is to provide coexistence informa-tion from TVBD network or device to CMs. Thesecond one is to transfer the control messages,

Figure 10. IEEE 802.19.1 system architecture.

14 International Journal of Distributed Sensor Networks

commands and requests from CMs to TVBDnetwork or device.

� The function of CDIS is to allow CMs to obtaininformation about other CMs and that relatedto TVWSs coexistence and PUs collected byCDIS or from TVWSs database, which storesPU-occupied channel list. The OperatorManagement Entity in Figure 10 maintains theoperator-related information and provides sup-port for CMs.

The advantage of IEEE 802.19.1 is its generality thatcan be used in a wide range of CR scenarios. However,it is still too complex to be implemented in practicaldevices and may not scale.

Other CR coexistence mechanisms

In this subsection, we survey other novel coexistencesolutions proposed by the research community. Thesesolutions are mainly concentrated on the coordinationamong secondary networks and the coexistence withprimary networks, which are surveyed as follows.

As pointed out in section ‘Introduction’, one of thedesign challenges of heterogeneous network coexistencemechanism is the asymmetry in transmission power,which may cause the hidden terminal problem. The hid-den terminal problem is particularly challenging if thenodes involved belong to two heterogeneous networksin terms of network protocols.

In this regard, Bian et al.17 developed a novel proto-col to solve the hidden terminal problem between aTime-division multiplexing (TDM)-based CR networkand a CSMA-based CR network in TVWSs, a typicalcoexistence scenario.

The coexistence scenario is illustrated in Figure 11.The TDM-based network has long-duration super-frames and quite period (QP). The CSMA-based net-work has short time slots and CSMA frames have shortsensing operation time, so that they can fit in the QP ofthe TDM superframes. In the work by Bian et al.,17 acoexistence protocol Spectrum sHARing for heteroge-neous coexistencE (SHARE) is proposed to address thehidden terminal problem in the following two cases:

� Avoiding collision at CSMA receivers: Byapproximating the packet arrival time ofCSMA-based networks using Poisson distribu-tion,31,46 Bian et al.17 develop the collision avoid-ance algorithm to avoid collision at CSMAreceivers. Specifically, each TDM transmitterestimates the CSMA packet transmission framelength and then configures the QP terminationtime to keep silence for one more CSMA packettransmission or terminate QP for TDM packetstransmission. To ensure the fairness in terms of

channel access time between the TDM-basednetwork and the CSMA network, a weight-fair-ness maintenance mechanism is designed inSHARE to adjust the channel access time of thetwo networks by adding or removing additionaltime slots in future QPs.

� Avoiding collision at TDM receivers: A beacontransmission mechanism is provided in SHAREto avoid collisions at TDM receivers.Specifically, SHARE lets TDM receivers broad-cast short beacon signals at the beginning of eachtime slot to inform CSMA nodes to suspenddata transmission. Since the TDM frame dura-tion is fixed, the estimation of transmitting dura-tion for CSMA nodes is feasible. Theconfiguration of BPs in SHARE also takes intoaccount the channel conditions via estimation.

Bian and colleagues39,40 presented an ecology-inspired framework, Symbiotic Heterogeneouscoexistence ARchitecturE (SHARE), to coordinate het-erogeneous network coexistence. The core idea comesfrom the similarity of the interaction between heteroge-neous networks and that between ecology species inecosystems (symbiotic relation). Specifically, two typesof biological algorithms are developed under the sym-biotic framework to address the coexistence problem:

� First, an ecology-inspired spectrum share alloca-tion algorithm is proposed in which by analogiz-ing classical Lotka–Volterra (L-V) predator-preymodel47

dNi

dt= riNi 1�

Ni +Pj 6¼i

ai;jNj

Ki

0B@

1CA ð3Þ

where Ni represents the population size of predator,and Ki is the maximized population size of species i.While ai,j indicates mutual impacts between species iand j, ri indexes intrinsic rate of increase. The spec-trum bandwidths which are needed by CR networkscould be transformed as predators’ population, andtotal spectrum shared by heterogeneous networks issimilar to prey’s amount. Hence, a fair spectrumcompetition mechanism is constructed.� Second, a foraging-based channel selection algo-

rithm is introduced to coordinate coexistence.

Figure 11. Superframe structure of TDM networks.

Zhang et al. 15

Based on the analogies between animals’ fora-ging behaviours and channel selection process,the algorithm enables CR network agents toselect appropriate channels to reach an equili-brium point of the whole system.

However, Amjad et al.42 provided a novel coexis-tence protocol for heterogeneous CR WSNs with gametheory on frequency and time dimensions.

As shown in Figure 12, suppose three heterogeneousCR networks (CRN 1, CRN 2 and CRN 3) coexistencewith three channels for communication. Then, coexis-tence scenarios could become

� Because the quality of channel 1 is the best, self-ish channel selection mechanism in these net-works would cause congestion as shown inFigure 12(a).

� Even heterogeneous CR networks employ ordi-nary fair spectrum resources allocation mechan-ism, and results are also hard to avoid datatransmission collisions perfectly as shown inFigure 12(b).

� In this case, Amjad et al.42 split different CR net-work frames into unified time slots and introducethe pure and mixed strategy Nash equilibria gamesolution which is called correlated equilibrium onthe evaluation of relationship between unifieddata frames and channels selection, so that fairand efficient communication resources distribu-tion could be realized as shown in Figure 12.

A particularity that makes the coexistence of CRnetworks challenging is the presence of PUs. In this

regard, Zhao and colleagues37,38 developed a suite ofsolutions to address the coexistence problem betweenCR networks under the presence of PUs. More specifi-cally, a fairness-oriented media access control (FMAC)protocol is proposed that mainly contains twocomponents:

� A cooperative spectrum sensing mechanism isproposed to estimate the location of the PUunder the assumption that the location of PUs isfixed and already known by cognitive nodes. Insuch context, the distance between the primarytransmitter and an SU can be estimated throughsignal strength received by the SU receiver.Mathematically, it is well known that if threeCR users measure the distance between themand the PU, the location of the PU can be calcu-lated as shown in Figure 13. Then, the primary

Figure 12. Channel access pattern of CR Networks: (a) selfish behaviour from CR networks for best quality channel (channel 1)will always result in a collision, (b) fair distribution of spectrum resource when CR networks mix their choice of channels and (c) fairand efficient resource distribution with correlated equilibrium.

Figure 13. Estimation of PU location: (a) two CR users are notenough and (b) three CR users are sufficient.

16 International Journal of Distributed Sensor Networks

signal can be distinguished from a secondarysignal.

� Once the primary signal is distinguished, a three-state sensing model is proposed to classify thesensed channel occupation states, as shown inequation (4)

ri =ni; H0

xp + ni; H1

xs + ni; H2

8<: ð4Þ

where xs is the emitted signal strength of an SU, xpis the emitted signal strength of a PU and ni is zero-mean additive white Gaussian noise (AWGN). Thethree-state model classifies the channel states intoH0 (idle), H1 (occupied by a PU) and H2 (occupiedby a SU). If a PU is present, the SU transmissionshould be refrained. Otherwise, the secondary net-work can compete for the occupation of channelwith other secondary networks.

The solutions proposed by Zhao and colleagues37,38

consist of MAC protocols for coexisting CR networks.The major drawbacks are the assumption that the PUs’locations are fixed and the prior knowledge on the PUlocations, which limit their application.

Conclusion and challenges

In previous sections, we have surveyed the latest devel-opments of heterogeneous networks coexistence bystructuring the existing work on the 2.4-GHz band andthe TVWSs. Despite the large body of heterogeneousnetwork coexistence solutions proposed in the litera-ture, especially in the past few years, there are still someresearch challenges left unaddressed, which may attractfuture research attention.

Coexistence of networks with different QoSconstraints

One of the challenges for heterogeneous networks coex-istence is how to coordinate heterogeneous networkswith heterogeneous QoS constraints. Most of existingsolutions on coexistence problem only focus on hetero-geneous networks coexistence without any specific QoSconsideration for the coexisting networks. The work ofBian et al.17 partially addresses this challenge using aweighted fairness maintenance algorithm to divide trans-mission time into two portions WiFi and ZigBee net-works to meet their different QoS requirements.However, the ratio between the time allocated to theheterogeneous networks is fixed and does not take intoaccount real-time QoS constraints. Therefore, we think

that research efforts should be devoted to filling thisgap by developing dynamic and adaptive QoS-awarecoexistence protocols.

Moreover, different applications would lead differ-ent QoS requirements (i.e. WSNs in complex industryenvironment need more reliable communication thanhigh-speed data transmission), which could notapprove overall system throughput performancerequirements of existing coexistence mechanisms. Inthis regard, it seems necessary to integrate the heteroge-neous QoS constraints of the coexisting network intothe performance metric in the design of coexistenceprotocols in order to strike a desired balance betweenmaximizing the overall throughput and accommodat-ing specific QoS constraints.

Coexistence under presence of PUs

As mentioned in the previous section, in TVWSs, a par-ticularity that makes the coexistence of CR networkschallenging is the presence of PUs, whose transmissionmay not necessarily follow a predictable pattern.Hence, coordination among heterogeneous secondarynetworks under the presence of PUs consists of animportance design. We point out that although thePUs issue is mentioned in a number of CR coexistencestandards (e.g. TVWSs database in IEEE 802.19.119),the problem of how to react when PUs arrive has notbeen specified. For example, in IEEE 802.19.1, it ismentioned that a PU list in TVWSs database can pro-vide PU information to CMs to manage spectrum hop-ping of the coexisting networks, but the problem ofcoexistence under the presence of PUs is not specified.

Zhao and colleagues37,38 develop a solution to thisproblem. However, their solution requires that the loca-tions of PUs are prior known to the coexisting networksand are fixed. Such assumptions may be too strongalthough impractical in many cases. Even in the caseswhere the assumptions hold, their solution may not beeffective or even fail to work if PUs are geographicallyclose to cognitive nodes, which makes it difficult to dis-tinguish the primary signal from secondary signal.

Given the above argument, a primary building blockenabling coexistence of CR networks with the presenceof PUs is the detection of primary signals. Once the pri-mary signal is detected, distributed mechanisms shouldbe devised to let cognitive nodes react and coordinatethe spectrum access among them. We illustrate the chal-lenge in this phase via the following example. Two CRnetworks (A and B) operate on a band of 30 MHz. Aoccupies first 20 MHz and B occupies last 20 MHz withan overlapped 10-MHz band in the middle part thatthey share. The event that a PU arrives at first 10 MHzwill not only trigger A to relocate its spectrum but also

Zhang et al. 17

probably impact B who might need to adapt its opera-tion band.

From two-network to multi-network coexistence

Finally, the coexistence of multiple networks consists ofanother important future work. We note that most ofexisting coexistence mechanisms are focused on thecanonic two-network coexistence scenario, the simplestcoexistence scenario we can imagine. When there aremultiple coexisting networks, the problem becomesmuch more complex, thus calling for more generic coex-istence solutions.

Specifically, we identify the following three typicalmulti-network coexistence scenarios:

� Homogeneous multi-network coexistence: Thesimplest scenario of multi-network coexistence isthe coexistence of multiple homogeneous net-works, that is, networks using the same technol-ogy operating on a swath of spectrum bands. Inthis scenario, some solutions have been proposedin the existing literature, for example, the spec-trum etiquette mechanism in IEEE 802.22.14

However, these approaches are usually centra-lized mechanisms. The development of distribu-ted coexistence mechanisms consists of animportant direction for future research.

� Heterogeneous multi-network coexistence: In con-trast to the first scenario, the second coexistencescenario is the coexistence of multiple heteroge-neous networks problem, that is, networks usingdifferent technologies operating on a swath ofspectrum bands. Again, solutions based on acentralized coordination mediator as in IEEE802.19.119 may fail to function in distributedenvironments. Moreover, CSMA-based mechan-isms suffer from the problems of heterogeneoustransmitting power and spectrum and are notdirectly applicable in this context.

� Hybrid multi-network coexistence: The third sce-nario is a combination of the previous two.Consequently, the challenges in the previous twoscenarios should be addressed here.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest withrespect to the research, authorship, and/or publication of thisarticle.

Funding

The author(s) disclosed receipt of the following financial sup-port for the research, authorship, and/or publication of thisarticle: This research is supported by the National NaturalScience Foundation of China (Grant Nos 51475343 and

51675389) and International Science & TechnologyCooperation Program, Hubei Technological InnovationSpecial Fund (Grant No. 2016AHB005).

References

1. Yang D, Xu Y and Gidlund M. Wireless coexistence

between IEEE 802.11- and IEEE 802.15.4-based net-works: a survey. Int J Distrib Sens N 2011; 7: 912152.

2. Borges LM, Velez FJ and Lebres AS. Survey on thecharacterization and classification of wireless sensor net-work applications. IEEE Commun Surv Tut 2014; 16:

1860–1890.3. Zhang D, Chen Z, Ren J, et al. Energy harvesting-aided

spectrum sensing and data transmission in heterogeneouscognitive radio sensor network. IEEE T Veh Technol

2017; 66: 831–843.4. Mitola J. Cognitive radio architecture evolution. Proc

IEEE 2009; 97: 626–641.5. Mitola J and Maguire GQ. Cognitive radio: making soft-

ware radios more personal. IEEE Pers Commun 1999; 6:13–18.

6. Paul A, Chilamkurti N, Daniel A, et al. Chapter 5: cog-nitive radio in vehicular network. In: Paul A, Chilam-kurti N and Daniel A (eds) Intelligent vehicular networksand communications. Amsterdam: Elsevier, 2017, pp.113–

129.7. Rawat P, Singh KD and Bonnin JM. Cognitive radio for

M2M and Internet of Things: a survey. Comput Commun

2016; 94: 1–29.8. Wu Y and Cardei M. Multi-channel and cognitive radio

approaches for wireless sensor networks. Comput Com-

mun 2016; 94: 30–45.9. Federal Communications Commission (FCC) 08-260.

Second report and order and memorandum opinion andorder in ET docket nos. 2-380 (additional spectrum for

unlicensed devices below 900 MHz and in the 3 GHzband) and 04-186 (unlicensed operation in the TV broad-cast bands), 14 November 2008, https://apps.fcc.gov/edocs_public/attachmatch/FCC-08-260A1.pdf

10. Gao B, Park JM, Yang Y, et al. A taxonomy of coexis-

tence mechanisms for heterogeneous cognitive radio net-works operating in TV white spaces. IEEE Wirel

Commun 2012; 19(4): 41–48.11. IEEE Std 802.11: IEEE standard for information

technology-telecommunications and information exchangebetween systems-local and metropolitan area networks-specific requirement, 2007, http://ieeexplore.ieee.org/docu-ment/654749/

12. Bluetooth SIG Inc. Bluetooth basics, 2011, www.

bluetooth.com13. IEEE. IEEE 802.15.4b standard: wireless medium access

control and physical layer specification for low rate wire-

less personal area networks. New York: IEEE, 2006.14. IEEE 802.22 Working Group. http://www.ieee802.org/

22/15. IEEE P802.11 Task Group AF. Wireless LAN in the TV

white space, http://www.ieee802.org/11/16. ECMA Standard. ECMA TC48-TG1, standard ECMA-

392: MAC and PHY for operation in TV white space.

Report, ECMA, Geneva, 2009.

18 International Journal of Distributed Sensor Networks

17. Bian K, Park JM, Chen L, et al. Addressing the hiddenterminal problem for heterogeneous coexistence between

TDM and CSMA networks in white space. IEEE T Veh

Technol 2014; 63: 4450–4463.

18. ZigBee Alliance. ZigBee specification. Davis, CA: ZigBeeAlliance, 2008.

19. IEEE 802.19 Task Group 1. Wireless coexistence in the

TV white space, http://www.ieee802.org/19/pub/TG1.html20. Zhang X and Shin K. Adaptive subcarrier nulling:

enabling partial spectrum sharing in wireless LANs. In:

Proceedings of the 19th IEEE international conference on

network protocols (ICNP), Vancouver, AB, Canada, 17–

20 October 2011, pp.311–320. New York: IEEE.21. Han S, Zhang X and Shin KG. Fair and efficient coexis-

tence of heterogeneous channel widths in next-generation

wireless LANs. IEEE T Mobile Comput 2016; 15:2749–2761.

22. Huang P, Yang X and Xiao L. Adaptive channel bond-

ing in multicarrier wireless networks. In: Proceedings of

the international symposium on mobile ad hoc networking

and computing (MobiHoc), Bangalore, India, 29 July–1August 2013, pp.297–300. New York: ACM.

23. Cidon A, Nagaraj K, Katti S, et al. Flashback: decoupled

lightweight wireless control. In: Proceedings of the ACM

SIGCOMM 2012 conference on applications, technologies,

architectures, and protocols for computer communication,Helsinki, 13–17 August 2012, pp.223–234. New York:

ACM.24. Zhou R, Xiong Y, Xing G, et al. ZiFi: wireless LAN dis-

covery via ZigBee interference signatures. In: Proceedings

of the annual international conference on mobile computing

and networking (MOBICOM), Chicago, IL, 20–24 Sep-

tember 2010, pp.49–60. New York: ACM.25. Yan Y, Yang P, Li XY, et al. WizBee: wise ZigBee coex-

istence via interference cancellation with single antenna.

IEEE T Mobile Comput 2015; 14: 2590–2603.26. Hao T, Zhou R, Xing G, et al. WizSync: exploiting Wi-Fi

infrastructure for clock synchronization in wireless sensor

networks. In: Proceedings of the real-time systems sympo-

sium, Vienna, 29 November–2 December 2011, pp.149–

158. New York: IEEE.27. Liang CJM, Priyantha NB, Liu J, et al. Surviving Wi-Fi

interference in low power ZigBee networks. In: Proceed-

ings of the 8th ACM conference on embedded networked

sensor systems (SenSys 2010), Zurich, 3–5 November

2010, pp.309–322. New York: ACM.28. Yubo Y, Panlong Y, Xiangyang L, et al. ZIMO:

building cross-technology MIMO to harmonize zigbee

smog with WiFi flash without intervention. In: Proceed-ings of the 19th annual international conference on mobile

computing & networking (MobiCom ’13), Miami, FL, 30September–4 October 2013, pp.465–476. New York:

ACM.29. Yang P, Yan Y, Li XY, et al. Taming cross-technology

interference for Wi-Fi and ZigBee coexistence networks.

IEEE T Mobile Comput 2016; 15: 1009–1021.30. Zhang X and Shin KG. A case for the coexistence of het-

erogeneous wireless networks. In: Proceedings of the

annual international conference on mobile computing and

networking (MOBICOM), Las Vegas, NV, 19–23 Sep-

tember 2011, pp.1–3. New York: ACM.31. Zhang X and Shin KG. Enabling coexistence of hetero-

geneous wireless systems: case for ZigBee and WiFi. In:

Proceedings of the international symposium on mobile ad

hoc networking and computing (MobiHoc), Paris, 17–19

May 2011. New York: ACM.32. Zhang X and Shin KG. Cooperative carrier signaling:

harmonizing coexisting WPAN and WLAN devices.

IEEE ACM T Network 2013; 21(2): 426–439.33. Zhang X and Shin KG. Gap sense: lightweight coordina-

tion of heterogeneous wireless devices. In: Proceedings of

the IEEE INFOCOM, Turin, 14–19 April 2013, pp.3093–

3101. New York: IEEE.34. Correia LH, Tran TD, Pereira VN, et al. DynMAC: a

resistant MAC protocol to coexistence in wireless sensor

networks. Comput Netw 2015; 76: 1–16.35. Gollakota S, Adib F, Katabi D, et al. Clearing the RF

smog: making 802.11 robust to cross-technology interfer-

ence. In: Proceedings of the ACM SIGCOMM 2011 con-

ference (SIGCOMM’11), Toronto, ON, Canada, 15–19

August 2011, pp.170–181. New York: ACM.36. IEEE Standard. IEEE 802.16 working group. IEEE stan-

dard for local and metropolitan area networks part 16:

air interface for broadband wireless access systems

amendment 2: improved coexistence mechanisms for

licensed-exampt operation. July 2010, http://ieeexplor-

e.ieee.org/document/5538195/37. Zhao Y, Song M and Xin C. FMAC: a fair MAC proto-

col for coexisting cognitive radio networks. In: Proceed-

ings of the IEEE INFOCOM, Turin, 14–19 April 2013,

pp.1474–1482. New York: IEEE.38. Zhao Y, Song M, Xin C, et al. Spectrum sensing based

on three-state model to accomplish all-level fairness for

co-existing multiple cognitive radio networks. In: Pro-

ceedings of the IEEE INFOCOM, Orlando, FL, 25–30

March 2012, pp.1782–1790. New York: IEEE.39. Bian K, Park JMJ, Du X, et al. Ecology-inspired coexis-

tence of heterogeneous wireless networks. In: Proceedings

of the IEEE global communication conference, Atlanta,

GA, 9–13 December 2013. New York: IEEE.40. Bian K, Park JM and Gao B. Ecology-inspired coexis-

tence of heterogeneous cognitive radio networks. In: Bian

K, Park JM and Gao B (eds) Cognitive radio networks.

Berlin: Springer, 2014, pp.117–131.41. Zhang D, Liu Q, Chen L, et al. Ecology-based coexis-

tence mechanism in heterogeneous cognitive radio net-

works. In: Proceedings of the 2015 IEEE global

communications conference (GLOBECOM), San Diego,

CA, 6–10 December 2015, pp.1–6. New York: IEEE.42. Amjad MF, Chatterjee M and Zou CC. Coexistence in

heterogeneous spectrum through distributed correlated

equilibrium in cognitive radio networks. Comput Netw

2016; 98: 109–122.43. Sun S, Chen N, Ran T, et al. A stackelberg game spec-

trum sharing scheme in cognitive radio-based heteroge-

neous wireless sensor networks. Signal Process 2016; 126;

18–26.

Zhang et al. 19

44. Wu K, Tan H, Liu Y, et al. Side channel: bits overinterference. IEEE T Mobile Comput 2012; 11(8):1317–1330.

45. Hiertz G, Denteneer D, Stibor L, et al. The IEEE 802.11universe. IEEE Commun Mag 2010; 48(1): 62–70.

46. Huang J, Xing G, Zhou G, et al. Beyond co-existence:exploiting WiFi white space for ZigBee performance

assurance. In: Proceedings of the international conference

on network protocols (ICNP), Kyoto, Japan, 5–8 Octo-

ber 2010, pp.305–314. New York: IEEE.47. Murray JD. Mathematical biology. I. An introduction

(vol. 40). 3rd ed. Dordrecht: Kluwer Academic Publish-

ers, 2002.

20 International Journal of Distributed Sensor Networks