Spectrum Mobility in Cognitive Radio Networks

IEEE Communications Magazine • June 2012114 0163-6804/12/$25.00 © 2012 IEEE

INTRODUCTION

Rapid development of wireless networking tech-nology has raised a large demand for spectrumband. Whereas the number of devices utilizingunlicensed industrial, scientific, and medical(ISM) band is growing, the allocated spectrumremains the same. This spectrum scarcity prob-lem happens because the current static spectrumallocation policy used by governmental agenciesis unable to accommodate the growing band-width demand. In fact, exclusively allocatedlicensed spectrum bands whose availability variesboth spatially and temporally, are proved to beunderutilized. Meanwhile, a large portion ofspectrum in UHV and VHF range will becomeavailable in the future upon completing the tran-sition to digital TV [1]. This so-called TV whitespace calls large interest because radio frequen-cy of the respective spectrum bands has a num-ber of advantageous characteristics that wouldopen numerous new wireless applications in thefuture.

Cognitive radio (CR) is a key enabling tech-nology in dynamic spectrum access (DSA). CRnetworks (CRNs) can maximize the use of band-width resources without changing well-estab-lished regulation of spectrum allocation. Owingto spectrum awareness, CR enabled devices orsecondary users (SUs) can make use of unoccu-pied licensed spectrum bands opportunisticallywhen the legacy devices or primary users (PUs)are idling. Thus, CRN becomes a compellingsolution to the spectrum scarcity problem.

In [2, 3], spectrum management framework isintroduced, where a CRN system executes fourinterrelated functions: spectrum sensing, spec-trum decision, spectrum sharing, and spectrummobility. In spectrum sensing, CR nodes searchfor spectrum holes in licensed spectrum bandsthat can be used for SU communication. Basedon spectrum sensing results, CR nodes decide onthe best available communication channel. Spec-trum sharing coordinates channel access amongthe CR nodes. Finally, when a PU reclaims alicensed channel temporarily occupied by a SU,spectrum mobility suspends the transmission,vacates the channel, and resumes ongoing com-munication using another vacant channel.

Spectrum mobility is a challenging topic for var-ious research activities related to heterogeneousnetworks and CRNs. In this article, spectrummobility in CRNs is examined with respect to spec-trum handoff. Various spectrum handoff strategiesare reviewed and compared qualitatively. The com-parison shows that even a simple strategy canachieve satisfactory performance for sparse com-munication needs, whereas more advanced adap-tive handoff strategies are required to gain highestbenefit as a trade-off for higher computationeffort. Then, we discuss current issues and researchchallenges in spectrum mobility.

The rest of the article is organized as follows:In the following section, some background con-cepts are overviewed. We introduce design issuesin spectrum mobility. Different spectrum hand-off strategies are classified, technically reviewed,and qualitatively compared. We discuss currentissues and research challenges. Finally, the con-clusions of the article are presented.

PRELIMINARIES

COGNITIVE RADIO NETWORK ARCHITECTUREFrom network architecture standpoint, CRN dif-fers from legacy wireless networks in that it con-siders spatial and temporal spectrumheterogeneity and distinguishes legacy (PU) andopportunistic (SU) users. A general classificationof CRNs is illustrated in Fig. 1 where infra-structure-based and non-infrastructure-basedCRNs [3] are depicted.

Infrastructure-based CRNs have a base CRstation (i.e., a local centralized entity), whichorganizes CR communication in a single-hop

ABSTRACT

Cognitive radio networks (CRNs) offer apromising solution for spectrum scarcity problemby means of dynamic spectrum access. So long asin highly dynamic environments, the secondaryuser (SU) communication is often interrupted,spectrum mobility is a key feature enabling con-tinuous SU data transmission. Namely, SU per-forms spectrum handoff by transferring ongoingcommunication to a vacant channel. This articlediscusses some important features of spectrummobility in CRNs. Qualitative comparison ofvarious handoff strategies is considered withregard to handoff latency. Furthermore, essen-tial design issues and associated research chal-lenges are also addressed.

TOPICS IN RADIO COMMUNICATIONS

Ivan Christian, Sangman Moh, Ilyong Chung, and Jinyi Lee, Chosun University

Spectrum Mobility in Cognitive Radio Networks

MOH LAYOUT_Layout 1 5/23/12 11:42 AM 14

IEEE Communications Magazine • June 2012 115

environment, in much the same way it works incellular systems. As a rule, CR nodes are respon-sible for spectrum sensing and the base stationperforms other functions such as spectrum deci-sion, spectrum sharing, and spectrum mobility.

On the other hand, in non-infrastructure-based CRNs, which are also known as CR adhoc networks (CRAHNs), distributed communi-cation is performed between CR nodes in a mul-tihop environment. Having no centralized entityto organize the network, CR nodes are responsi-ble for performing all spectrum-related functionsby means of collaboration. As a consequence,cooperation between CR nodes is of greatimportance for high performance CRAHNs.

SPECTRUM HANDOFFThe primary objective of spectrum mobility inCRNs is to perform seamless channel switchoverwhile sustaining performance of ongoing SUcommunication. To this end, spectrum mobilityis divided into two processes: spectrum handoffand connection management. Spectrum handoffis the process of transferring ongoing data trans-mission from the current channel to another freechannel. This naturally causes additional latencyto SU communication that eventually affects SUperformance. To compensate the unavoidablehandoff delay, connection management processmanages and adjusts protocol stack parametersdepending on current situation.

There are two PU related events that can trig-ger spectrum handoff in CRNs. First, PU arrivalin the licensed channel necessarily forces SU toperform spectrum handoff. Second, spectrumhandoff can occur because of CR user mobility.As CR users moves spatially, there is a chancethat transmission coverage of the SU overlapswith a PU currently using the same channelband. Being opportunistic users in licensed spec-trum bands, SUs’ activity in legacy networks shallconform to the ground rule: PUs always havehigher priority in using licensed spectrum thanSUs. As a consequence, if SU arrival causesinterference to PU data transmission, the SUshall leave the licensed channel immediately.

In addition, SUs can also perform spectrumhandoff because of link quality degradation. Sincethe radio spectrum of CRNs is predominantlyoccupied by users outside the control of SUs, thequality of the communication channel in CRNsmay in particular vary dynamically over time andspace. Thus, it is important for SUs to monitorand analyze periodically the quality of the chan-nel being used for data transmission. If channelquality tends to degrade in the future, spectrumhandoff is required to maintain the QoS level.

Spectrum handoff can be explained as cyclicprocess as shown in Fig. 2. It consists of twophases [4]: evaluation phase and link mainte-nance phase. In evaluation phase, SU observesthe environment and analyzes whether handofftriggering events occur. Once SU decides to per-form spectrum handoff, it goes into link mainte-nance phase. In this phase, SU first pauses theongoing transmission. Then SU hands over thereclaimed channel to PU and resumes datatransmission session over another free channel.Finally, SU leaves link maintenance phase andresumes the cycle. Note that backup target chan-

nel can be searched either in evaluation or inlink maintenance phase depending on whetherproactive or reactive handoff is used.

PERFORMANCE MEASUREMENTTo measure spectrum mobility performance, sev-eral parameters have been proposed [4, 5]. Thenumber of spectrum handoff is defined as thenumber of handoffs occurring during one sessionof SU data transmission. Link maintenanceprobability is the probability that the communi-cation link is successfully maintained when SUvacates the channel. Handoff latency is definedas the delay resulting from spectrum handoffprocess, and effective data rate is the averageamount of data which is successfully transferredbetween two communicating SUs in each ses-sion. Among these parameters, link maintenanceprobability and handoff latency are key perfor-mance metrics for spectrum mobility.

The correlation between these parameters isexplained as follows: From SU point of view,increasing the number of spectrum handoffsincreases link maintenance probability. However,at the same time, this also reduces effective datarate per connection. On the other hand, from PUpoint of view, handoff latency becomes the mainconcern. The greater the handoff latency to theSU is, the greater the interference to the PU is.

In general, spectrum mobility techniques withhigher link maintenance probability and lowerhandoff latency give better spectrum agility toCRNs. In order to achieve high performance inspectrum mobility, multiple spectrum handoffsshould be avoided if possible.

DESIGN ISSUES INSPECTRUM MOBILITY

This section briefly discusses design issues inspectrum mobility.

PU DETECTIONSensing speed and accuracy in spectrum sensingare two important factors for efficient spectrummobility. In fact, there is a trade-off between thetwo. Fast sensing speed would lead to less accu-

Figure 1. A legacy network together with: a) a cognitive radio network (CRN)and b) a cognitive radio ad hoc network (CRAHN).

(a) (b)

CR node CR basestation

Legacybase station

CR node

CR node

CR node

CR node

CR node

Legacynode

MOH LAYOUT_Layout 1 5/23/12 11:42 AM Page 115

IEEE Communications Magazine • June 2012116

rate sensing output. In addition, recent spectrumsensing techniques are also prone to imperfectsensing results due to radio propagation effects,such as channel fading or shadowing.

To increase sensing speed and accuracy, CRnodes can select other idle CR nodes as partnersto perform cooperative spectrum sensing insteadof local sensing. In [6], a cross-layer protocolcalled ESCAPE (Embedded Spectrally AgileRadio Protocol for Evacuation) using coopera-tive spectrum sensing in CRAHNs was suggest-ed. Specifically, CR nodes are divided intoevacuation groups and each group shares thesame CDMA spreading code. If PU arrival isdetected, warning messages are spread periodi-cally to the entire group.

HANDOFF DECISIONGenerally SUs use spectrum overlay techniquewhere radio signals are transmitted with a powerabove PU noise level. In contrast, SUs withunderlay technique transmit radio signals with apower below PU noise level so that both SU andPU can use the same licensed band at the sametime. Nevertheless, if the aggregate interferenceto the PU exceeds a certain threshold, SU shouldleave the licensed band. Therefore, spectrumhandoff decision is an important issue even inCRNs with spectrum underlay technique.

In [7], fuzzy logic based spectrum handoffalgorithm in multihop underlay CRAHNs is pro-posed. Using fuzzy logic controller, SU will firstadjust the transmitting power of the radio signal.If power adjustment is unlikely to lower theaggregate interference, then SU should do spec-trum handoff.

TARGET CHANNEL SELECTIONFinding a suitable target channel over which aSU can continue data transmission session is themost pressing issue in CR research related tospectrum mobility. In fact, target channel selec-

tion for spectrum handoff is a non-trivial task,because it depends on many factors, such aschannel capacity, channel availability at the timeof handoff, and probability of channel beingavailable in the future. Poor target channelselection can cause multiple spectrum handoffsin a single data transmission session thatdegrades overall performance.

The most common approach to this issue isusing backup channel list (BCL). SU anticipatesspectrum handoff by listing potential target chan-nels into BCL and maintaining it periodicallybetween communicating peers. IEEE 802.22 Wire-less Regional Area Network (WRAN) Standardadopts this approach [8]. Another approach is pre-dicting target channel availability. Using prediction,partial spectrum sensing rather than conventionalfull sensing can be performed to reduce sensingdelay during spectrum handoff [9]. Also, instead ofrelying only on sensing result, SU can get accuratetarget channel selection based on historical spec-trum statistics. In [10], a proactive channel accessalgorithm using both spectrum sensing results andspectrum statistics to determine a handoff targetchannel is suggested. Assuming that PU arrivalpattern is not statistically random (because itdepends on human behavior) PU traffic can bemodeled so that SU can estimate both channelavailability and the length of time that channelswill be available. As a result, intelligent targetchannel selection can be performed.

ROUTING RECOVERYRouting recovery is another important issue inspectrum handoff that requires careful planning.Spectrum handoff is likely to cause route break-ing. Accordingly, SU is required to recover therouting table to maintain network connectivity.In fact, routing recalculation is a costly processin terms of time and resource consumption.Therefore, routing recovery process should beintegrated in spectrum handoff schemes.

Figure 2. Spectrum handoff process.

Handofftriggering

events

Handoffsuccessful

Spectrumsensing

Channelhandover

Target channel error

Pausetransmission

Evaluation phase

Link maintenance phase

EnvironmentResume

transmission

Backup channelsearch

Backup channelsearch

(Proactive)

(Reactive)

Sensing speed and

accuracy in spectrum

sensing are two

important factors for

efficient spectrum

mobility. In fact,

there is a trade-off

between the two.

Fast sensing speed

would lead to less

accurate sensing out-

put. In addition,

recent spectrum

sensing techniques

are also prone to

imperfect sensing

results.



A few works in CR routing have been pro-posed to anticipate spectrum handoff, all ofwhich attempt to keep network connectivity.One approach is to recalculate a new route andto perform spectrum handoff once a new routingtable is ready. Another approach is to avoidrouting recalculation by preparing two redun-dant channels (data channel and backup chan-nel) before starting data transmission. SUmaintains the backup channel periodically sothat the communication link can be immediatelytransferred to the backup channel upon spec-trum handoff event.

SPECTRUM HANDOFF STRATEGYAmong other issues discussed in this section,spectrum handoff strategy is considered as themain issue in spectrum mobility. Since eachmodule involved in spectrum handoff process issubject to imperfect performance, spectrumhandoff strategy plays an important role in com-pensating the deficiency of the related modules.Thus, applying proper handoff strategy to thespecific PU network can give optimal result tospectrum mobility performance.

The next section discusses spectrum handoffstrategies in greater detail.

COMPARISON OFHANDOFF STRATEGIES

Spectrum handoff strategies are characterized byidentifying when spectrum sensing and handoffare performed with regard to handoff triggeringevent occurrence. Spectrum sensing can be per-formed either before or after spectrum handofftriggering events happens [11], and so can behandoff action [10]. The combination of the twoparameters above, gives four spectrum handoffstrategies: non-handoff, pure reactive handoff,pure proactive handoff, and hybrid handoffstrategy. These cases are shown in Fig. 3.

NON-HANDOFF STRATEGYIn non-handoff strategy, SU keeps staying inoriginal channel and being idle until the channelbecomes free again. In other words, SU selectsthe current licensed channel as the next targetchannel. After PU leaves the licensed channel,SU resumes the data transmission again.

The major disadvantage of this approach isthat it causes high waiting latency to SU becausethe delay is as long as PU is active in the corre-sponding channel. In delay sensitive applications,this method would fail to meet QoS require-ments. Also, it is obvious that time is badly wast-ed during SU waiting period.

PURE REACTIVE HANDOFF STRATEGYIn pure reactive handoff strategy, SU appliesreactive spectrum sensing and reactive handoffaction approach. Once a handoff triggeringevent occurs, SU performs spectrum sensing tofind target backup channel. Afterward, link com-munication is transferred to the new target chan-nel. In other words, both target channel selectionand handoff action are performed reactivelyafter a triggering event happens.

The advantage of this approach is that SU

can get an accurate target channel since spec-trum sensing is performed in the most relevantspectrum environment. Nevertheless, it comes ata cost of longer handoff latency due to on-demand spectrum sensing. Since SU performsspectrum sensing after detecting the handoffevent, spectrum sensing becomes the majordelay in the handoff process.

PURE PROACTIVE HANDOFF STRATEGYIn pure proactive handoff strategy, SU usesproactive spectrum sensing and proactive hand-off action approach. SU performs spectrum sens-ing to find a backup target channel before ahandoff triggering event happens. Based on theknowledge of PU traffic model, SU is able topredict PU arrival so that SU evacuates thechannel beforehand. In other words, both targetchannel selection and handoff actions are per-formed proactively before the triggering eventhappens.

There are several advantages in using pureproactive strategy. First, handoff latency can bevery short because everything can be planned inadvance. Second, the possibility of multiplespectrum handoffs can be minimized by consid-ering future target channel usage when select-ing backup target channel. However, thedrawback of this strategy is that backup targetchannel can remain obsolete. There is a chancethat prepared backup channel is already occu-pied by other users at handoff time. In addi-tion, accurate PU traffic model also becomes akey factor in this strategy. Poor predictioncaused by inaccurate PU traffic model maybadly degrade the overall spectrum mobilityperformance.

HYBRID HANDOFF STRATEGYHybrid handoff strategy combines pure reactiveand pure proactive strategy by applying proactivespectrum sensing and reactive handoff action.Target channel selection is prepared beforehandor during SU data transmission while spectrumhandoff is performed after a handoff triggeringevent happens. In other words, target channelselection is performed proactively and handoffaction is performed reactively.

Hybrid handoff strategy is a reasonable com-promise between pure reactive and pure proac-tive strategy. Faster spectrum handoff time canbe achieved as spectrum sensing time is not per-formed during the handoff process. However,target channel can stay obsolescent as it does inpure proactive approach.

Comparison between various spectrum hand-off strategies is summarized in Table 1.

DISCUSSIONIn terms of handoff latency, pure proactivehandoff strategy has superior performance, fol-lowed by hybrid handoff, pure reactive handoff,and non-handoff strategy, respectively. This canbe explained by examining handoff latency factordiagram shown in Fig. 4.

Handoff phase is defined as a period of timefrom the occurrence of a handoff triggeringevent to the time when SU can resume its datatransmission. In principle, handoff latency isdirectly proportional to the number of sequen-

Since each module

involved in spectrum

handoff process is

subject to imperfect

performance, spec-

trum handoff strate-

gy plays an

important role in

compensating the

deficiency of the

related modules.

Thus, applying prop-

er handoff strategy

to the specific PU

network can give

optimal result to

spectrum mobility

performance.



Figure 3. Spectrum handoff strategies: a) non-handoff; b) pure reactive handoff; c) pure proactive handoff;and d) hybrid handoff.

SU detects PU A arrival,then pauses transmission

PU A startstransmission

SU resumestransmission

SU waits for PU A

(a)

(b)

SU PU A

Waiting latency

Handoff latency

SU

Time

Channel 1

PU BChannel 2

SU detects PU A arrival,then pauses transmission

SU

SU

PU A

Time

Channel 1

SU finds backup channel,spectrum handoff

PU A starts transmission,SU resumes transmission

PU BChannel 2

(c)

SU preparesbackup channel

SU

SU

PU A

Time

Channel 1

SU predicts PU A arrival,then pauses transmission


Spectrumhandoff

PU BChannel 2

Handoff latency

(d)

SU preparesbackup channel

SU

SU

PU A

Time

Channel 1

SU detects PU arrival,then pauses transmission


Spectrumhandoff

PU BChannel 2

Handoff latency

In the case of well-

modeled PU net-

works, pure

proactive strategy

would be the best.

On the other hand,

pure reactive hand-

off strategy and

hybrid handoff strat-

egy are good for

CRNs in general PU

networks where PU

arrival is considered

to be random, such

as in battlefield and

natural disaster

evacuation.



tial tasks performed by SU during the handoffphase. Each task requires specific period of timeto finish that is translated into delay to PU side.The delays are accumulated together andbecome handoff latency. Therefore, more tasksperformed by SU in handoff phase lead to longerhandoff latency to the SU, and, in turn, greaterinterference to PU.

In non-handoff strategy, PU data transmis-sion becomes a major delay source to SU. Whilehandoff decision task completion can be estimat-ed, it is not the case for PU activity where itscompletion tends to be random in nature. There-fore, non-handoff strategy suffers from unpre-dictable waiting latency.

For other spectrum handoff models, majordelay sources to SU come from spectrum sens-ing, handoff decision, and channel switchingtask; among which spectrum sensing is consid-ered the most time-consuming. In pure reactivehandoff strategy all three tasks are performedduring the handoff phase. On the contrary, inpure proactive handoff strategy spectrum sensingand handoff decision task are excluded fromhandoff phase. In hybrid handoff strategy spec-trum decision and channel switching tasks areperformed during handoff phase, while spectrumsensing is excluded from handoff phase. As aresult, pure proactive handoff strategy featuresthe fastest response in spectrum mobility, whilepure reactive handoff strategy exhibits relativelyslow response and hybrid handoff strategy hasmoderate response.

It is important to note that the applicationof spectrum handoff strategy depends on uniquecharacteristics of the PU network. Non-handoffstrategy can be suitable for the PU networkwith short data transmission pattern and in thesituation where other licensed spectrum bandsare highly congested. In the case of well-mod-eled PU networks, pure proactive strategywould be the best. On the other hand, purereactive handoff strategy and hybrid handoffstrategy are good for CRNs in general PU net-works where PU arrival is considered to be ran-dom, such as in battlefield and natural disasterevacuation.

CURRENT ISSUES ANDRESEARCH CHALLENGES

This section lists current issues and researchchallenges in the field of spectrum mobility.

CURRENT ISSUES-Parallel Multichannel Transmission — Paral-lel multichannel transmission is used in wirelessnetworks to improve channel capacity. Instead ofusing a single channel, data is transmitted overmultiple channels simultaneously so that thethroughput is maximized. In CRNs, the sametechnique is adopted to minimize spectrumhandoff effect.

This approach is used in the CORVUS(Cognitive Radio Approach for Usage of Vir-tual Unlicensed Spectrum) architecture wherelicensed spectrum bandwidth is divided intofine-grained subchannels [12]. Each SU datalink makes use of multiple scattered subchan-nels to minimize spectrum handoff effects. In[13], a link maintenance protocol with OFDM(Orthogonal Frequency Division Multiplex-ing) modulation scheme in CRNs is proposedto provide a mechanism for exchanging spec-trum information, maintaining backup sub-channels, and switching the subchannels whenrequired.

The CORVUS concept can reduce spectrumhandoff effect and lower the risk of routerecovery. However, its effectiveness is closelyrelated to PU bandwidth usage. The wider thePU bandwidth is, the more effective is thistechnique. In addition, CORVUS requires alarge number of radios that can be impracticalin mobile networks with limited hardware sup-port. Nevertheless, the application ofCORVUS-like concept in mobile networks isworth researching.

Channel Contention During SpectrumHandoff — PU arrival may cause simultaneousspectrum handoffs. So, there is a chance thatSUs will contend with each other to reserve tar-get channels that can decrease spectrum handoffsuccess rate and degrade spectrum mobility per-

Table 1. Comparison of spectrum handoff strategies.

Strategy Non-handoff Pure reactive Pure proactive Hybrid

Main idea Stay and wait * Reactive sensing* Reactive action

* Proactive sensing* Proactive action

* Proactive sensing* Reactive action

Advantages Very low PU interference Accurate targetchannel selection

* Fastest response* Smart target channel selection Fast response

Disadvantages Very high SU interference Slow response * Outdated target channel selection* High computational requirement

Outdated targetchannel selection

Handoff latency Unpredictably high latency Medium latency Very low latency Low latency

Dependency PU activity Spectrum sensing * Backup channel relevancy* Accurate PU traffic model

Backup channelrelevancy

Best suitedenvironment

Short data transmission PUnetwork General PU network Well-modeled PU network General PU network



formance. In fact, most of current works in spec-trum mobility consider a pair of SUs instead ofmultiple SUs.

To date, only few research activities considerchannel contention for multiple SU spectrumhandoffs. In [14], a distributed channel selectionalgorithm is employed, where a SU that needs toperform spectrum handoff selects available chan-nels according to randomized selecting channelsequence. Thus, the target channel contentionproblem can be mitigated in case of simultane-ous spectrum handoff among multiple SUs.

Since this issue is practically important forsuccessful spectrum handoff, the target channelselection algorithm for spectrum handoff has toaccount for channel contention as well.

RESEARCH CHALLENGESAdaptive Spectrum Handoff Strategy —Current works are mostly focused on single spec-trum handoff strategy. Since each spectrumhandoff strategy is best suited for different PUnetworks, a new adaptive spectrum handoff algo-rithm with multiple spectrum handoff strategiesis required. Ideally a SU should know the PUtraffic pattern and apply the most suitable hand-off strategy. When PU traffic pattern changes,SU would notice the change and adapt its hand-off strategy accordingly. Therefore, future spec-trum handoff strategy should consider spectrumlearning factor in the design process.

Cross-Layer Link Maintenance Protocols —Link maintenance is at the heart of spectrummobility. The related design issues are spreadover physical, MAC and network layers. There-fore, cross layer approach between the three lay-ers is required to address this issue efficiently.Most of the research works propose cross-layersolutions, either between physical and MAC lay-ers or between MAC and network layers.Although the proposed solutions are aware ofspectrum handoff, they are ineffective to someextent because some important spectrum hand-off issues are overlooked. For example, routing

algorithms with cross-layer approach betweenMAC and network layers usually do not addressPU detection issue. On the other hand, MACalgorithms with cross-layer approach betweenphysical and MAC layers hardly address theproblem of routing recovery. Therefore, it isnecessary to design a cross-layer link mainte-nance protocol that would efficiently addressspectrum mobility issues in physical, MAC, andnetwork layers to get optimal solution.

Energy Efficiency — In CRAHN, energy efficien-cy becomes a major constraint due to the limitedresources of CR nodes. On the other hand, spec-trum mobility methods usually rely on frequentspectrum information update and spectrum sensingthat take significant power. Therefore, energy effi-cient spectrum mobility is still a challenge.

CONCLUSIONSSpectrum mobility is one of main functionalities inCRNs that gives agility to CR nodes. Design issuesin spectrum mobility include PU detection, hand-off decision, target channel selection, routingrecovery, and spectrum handoff strategy. Amongthese issues, spectrum handoff strategy plays animportant role in spectrum mobility performance.Various handoff strategies are classified into fourmodels: non-handoff, pure reactive, pure proactive,and hybrid handoff strategy. Qualitative compari-son between the four models is made in terms ofhandoff latency performance metric. The analysisof handoff latency factors shows that pure proac-tive handoff strategy has the lowest handoff laten-cy, while hybrid handoff and pure reactive handoffstrategies have moderate and long handoff latency,respectively. These characteristics will determinetheir suitability for various applications in PU net-works. In addition, reducing spectrum handoffeffect using parallel multichannel transmission andmitigating channel contention between multipleSUs during spectrum handoff remain open issuesin spectrum mobility. We also consider adaptivespectrum handoff strategy, cross-layer link mainte-

Figure 4. Handoff latency factors.

PU datatransmission

Handofftriggering

events Waiting latency

Non-handoff

Time

Channelswitching

Handoffdecision

Spectrumsensing

Handoff latency

Pure reactivehandoff

Handoff latency

Pure proactivehandoff

Handoff latency

Hybridhandoff

Ideally a SU should

know the PU traffic

pattern and apply

the most suitable

handoff strategy.

When PU traffic pat-

tern changes, SU

would notice the

change and adapt its

handoff strategy

accordingly. There-

fore, future spectrum

handoff strategy

should consider

spectrum learning

factor in the design

process.



nance protocols, and energy efficiency as researchchallenges in spectrum mobility.

Finally, we would like to share our perspec-tive regarding research directions in spectrummobility. First, there is a tendency to exhibit cog-nitive behavior in dynamic spectrum environ-ment by integrating various learning methods ina spectrum handoff algorithm. Properly appliedlearning techniques would benefit in manyaspects, especially in obtaining spectrum usageinformation in more independent way ratherthan relying on predefined static information.Second, from handoff strategy standpoint, thefocus of spectrum mobility will be shifted from asingle static handoff strategy towards the adap-tive multiple handoff strategies approach. As aresult, spectrum handoff behavior can changedynamically over time from reactive to proactiveapproach to achieve optimal performance indynamic spectrum environment.

ACKNOWLEDGEMENTSThis work was supported by research funds fromChosun University while Dr. Jinyi Lee’s workwas also supported in part by the MKE (TheMinistry of Knowledge Economy), Korea, underthe ITRC (Information Technology ResearchCenter) support program supervised by theNIPA (National IT Industry Promotion Agency)(NIPA-2012-H0301-12-2008). Correspondenceshould be addressed to Dr. Sangman Moh([email protected]).

REFERENCES[1] C.-J. Kim et al., “Dynamic Spectrum Access/Cognitive

Radio Activities in Korea,” Proc. IEEE Symp. New Fron-tiers in Dynamic Spectrum, Apr. 2010.

[2] I. F. Akyildiz, W.-Y. Lee, and K. Chowdhury, “CRAHNs:Cognitive Radio Ad Hoc Networks,” Ad Hoc Networks,vol. 7, No. 5, July 2009, pp. 810–36.

[3] I. F. Akyildiz et al., “A Survey on Spectrum Managementin Cognitive Radio Networks,” IEEE Commun. Mag., vol.46, no. 4, Apr. 2008, pp. 40–48.

[4] Y. Zhang, “Spectrum Handoff in Cognitive Radio Net-works: Opportunistic and Negotiated Situations,” IEEEInt’l. Conf. Commun. (ICC), June 2009.

[5] L.-C. Wang and A. Chen, “On the Performance of Spec-trum Handoff for Link Maintenance in Cognitive Radio,”Proc. Int’l. Symp. Wireless Pervasive Computing(ISWPC), May 2008.

[6] X. Liu and Z. Ding, “ESCAPE: A Channel Evacuation Pro-tocol for Spectrum Agile Networks,” Proc. IEEE Int’l.Symp. New Frontiers in Dynamic Spectrum Access Net-works (DySPAN), Apr. 2007.

[7] L. Giupponi and A. I. Perez-Neira, “Fuzzy-based Spec-trum Handoff in Cognitive Radio Networks,” Proc. Int’l.Conf. Cognitive Radio Oriented Wireless Networks andCommun. (CrownCom), May 2008.

[8] G. Ko et al., “Channel Management in IEEE 802.22WRAN Systems,” IEEE Commun. Mag., vol. 48, no. 9,Sept. 2010, pp. 88–94.

[9] R. Ma, Y. Hsu, and K. Feng, “A POMDP-Based SpectrumHandoff Protocol for Partially Observable CognitiveRadio Networks,” Proc. IEEE Wireless Commun. andNet. Conf. (WCNC), Apr. 2009.

[10] L. Yang, L. Cao, and H. Zheng, “Proactive ChannelAccess in Dynamic Spectrum Networks,” Physical Com-munication (Elsevier), vol. 1, June 2008, pp. 103–11.

[11] L.-C. Wang and C.-W. Wang, “Spectrum Handoff forCognitive Radio Networks: Reactive-Sensing or Proac-tive-Sensing?,” Proc. IEEE Int’l. Conf. Performance,Computing and Commun. (IPCCC), Dec. 2008.

[12] D. Cabric et al., “A Cognitive Radio Approach forUsage of Virtual Unlicensed Spectrum,” Proc. 14th ISTMobile and Wireless Commun. Summit, June 2005.

[13] Q. Shi et al., “Link Maintenance Protocol for CognitiveRadio System with OFDM PHY,” Proc. IEEE Int’l. Symp.New Frontiers in Dynamic Spectrum Access Networks(DySPAN), Apr. 2007.

[14] Y. Song and J. Xie, “Common Hopping Based Proac-tive Spectrum Handoff in Cognitive Radio Ad Hoc Net-works,” Proc. IEEE Global Telecommun. Conf.(GLOBECOM), Dec. 2010.

BIOGRAPHIESIVAN CHRISTIAN ([email protected]) received his B.E.degree in electrical engineering from Bandung Institute ofTechnology, Indonesia in 2005. From 2006 to 2010, heworked at PT Siemens Indonesia, Jakarta, Indonesia. In2010, he was selected as a Korean Government ScholarshipProgram (KGSP) grantee. Currently, he is a graduate stu-dent in the Dept. of Computer Engineering, Chosun Univer-sity, South Korea. His current research interests includecognitive radio networks and cognitive radio ad hoc andsensor networks.

SANGMAN MOH ([email protected]) received the Ph.D.degree in computer engineering from Korea AdvancedInstitute of Science and Technology (KAIST), South Korea in2002. Since late 2002, he has been a faculty member inthe Dept. of Computer Engineering at Chosun University,South Korea. From 1991 to 2002, he had been with Elec-tronics and Telecommunications Research Institute (ETRI),South Korea. His research interests include mobile comput-ing and networking, ad hoc and sensor networks, and cog-nitive radio networks.

ILYONG CHUNG ([email protected]) holds Ph.D. degrees inComputer Science from City University of New York. From1991 to 1994, he was a senior technical staff of Electronicand Telecommunication Research Institute (ETRI), Dajeon,Korea. Since 1994, he has been a Professor in Departmentof Computer Science, Gwangju, Korea. His research inter-ests are in computer networking, security systems and cod-ing theory.

JINYI LEE ([email protected]) received the bachelordegree in mechanical design from Chonbuk University,Korea. Also he received the master and Ph.D. degree fromTohoku University, Japan. He is an Associate Professor ofChosun university, Korea and director of IT-NDT Center. Heis the author or coauthor of thirty patents and over 60 sci-entific papers. His research interest is in development ofmagnetic camera.

The analysis of hand-

off latency factors

shows that pure

proactive handoff

strategy has the low-

est handoff latency,

while hybrid handoff

and pure reactive

handoff strategies

have moderate and

long handoff latency,

respectively. These

characteristics will

determine their suit-

ability for various

applications in PU

networks.


Spectrum Mobility in Cognitive Radio Networks

Documents

Transcript of Spectrum Mobility in Cognitive Radio Networks