Research Article Channel Selection Policy in Multi-SU and

Research ArticleChannel Selection Policy in Multi-SU and Multi-PUCognitive Radio Networks with Energy Harvestingfor Internet of Everything

Feng Hu12 Bing Chen12 Xiangping Zhai12 and Chunsheng Zhu3

1The College of Computer Science and Technology Nanjing University of Aeronautics and Astronautics Nanjing 211106 China2The Collaborative Innovation Center of Novel Software Technology and Industrialization Nanjing 210023 China3The Department of Electrical and Computer Engineering The University of British Columbia Vancouver BC Canada V6T 1Z4

Correspondence should be addressed to Bing Chen cb chinanuaaeducn

Received 22 September 2016 Accepted 17 November 2016

Academic Editor Beniamino Di Martino

Copyright copy 2016 Feng Hu et alThis is an open access article distributed under the Creative Commons Attribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Cognitive radio which will become a fundamental part of the Internet of Everything (IoE) has been identified as a promisingsolution for the spectrum scarcity In a multi-SU and multi-PU cognitive radio network selecting channels is a fundamentalproblem due to the channel competition among secondary users (SUs) and packet collision between SUs and primary users (PUs)In this paper we adopt cooperative sensing method to avoid the packet collision between SUs and PUs and focus on how to collectthe spectrum sensing data of SUs for cooperative sensing In order to reduce the channel competition among SUs we first considerthe hybrid transmissionmodel for single SUwhere a SU can opportunistically access both idle channels operating either theOverlayor the Underlaymodel and the busy channels by using the energy harvesting technology Then we propose a competitive set basedchannel selection policy for multi-SU where all SUs competing for data transmission or energy harvesting in the same channel willform a competitive set Extensive simulations show that the proposed cooperative sensing method and the channel selection policyoutperform previous solutions in terms of false alarm average throughput average waiting time and energy harvesting efficiencyof SUs

1 Introduction

Due to the continuous development of wireless devices andservices our environment is transforming into an Internetof Everything (IoE) [1ndash5] In this IoE paradigm whereeverything and everyone will be connected the bandwidthdemand for limited spectrum has been greatly increasingThe scarcity of the spectrum resources has become a seriousproblem This is mainly due to the traditional static spec-trum allocation policy where a particular portion of thespectrum can be only used by licensed wireless communi-cations systems The impoverishment of available spectrumand the underutilization of licensed spectrum facilitate theappearance of cognitive radio (CR) technology which hasevoked much enthusiasm of many scholars and FederalCommunications Commission (FCC) approved unlicenseduse of licensed spectrum through CR technology [6ndash9]

CR has been regarded as an efficient approach to cope upwith the spectrum shortage and low utilization problems[10ndash12] Therefore the introduction of cognitive radio inIoE environment can provide on-demand spectrum accessamong multiple devices

Dynamic Spectrum Access (DSA) mechanism has beenoffered for spectrum usage There are two major trans-mission models for a secondary user (SU) efficiently usingidle spectrums which are Overlay [13] and Underlay [14]respectively In the Overlay model a SU can exclusively andopportunistically use the licensed spectrum only if a primaryuser (PU) is inactive In other words the SU is not allowedto access the spectrum simultaneously with the PU in orderto prevent colliding with PU transmission In contrast evenif when the PU accesses its spectrum a SU may coexistwith it as long as the interference caused to the PU bythis SU does not degrade its communication quality in the

Hindawi Publishing CorporationMobile Information SystemsVolume 2016 Article ID 6024928 12 pageshttpdxdoiorg10115520166024928

2 Mobile Information Systems

Underlay model However when the PU state changes tobe inactive the transmission power of the SU will be stillbelow the interference threshold constraint in the Underlaymodel Therefore the idle spectrum resources are not fullyutilized and the SU does not achieve optimal performanceOn the other hand when the licensed channels are verybusy the time that SU must wait for an available channelis too long and then it may also significantly reduce theperformance of Overlay model [15] Therefore we need tofind a hybrid transmission model where the advantages ofboth Overlay and Underlay models are combined for thePU state variability so that the performance of SUs can bemaximized

The hybrid transmission model has recently been pro-posed in [16ndash18] In [16] the SU can exchange controlinformation in the Underlay model and transmit data infor-mation in the Overlay model However the decision ofaccessing a model is not based on the sensing results In[17 18] the SU can constantly sense the activity of PUs andtransmit data information in theOverlaymodel when the PUtransmission is not detected Otherwise the SU reduces itstransmission power to access the spectrum in the Underlaymodel However these papers did not take into account thesensing errors and neglected the effect of PU retransmissionon the SU QoS Although these related works indicatedthat the SU can obtain more spectrum access opportunitiesin the hybrid transmission model compared with the twoconventional transmission models the issue of two or moreSUs competing for the same channel has rarely been studiedso far to the best of our knowledge Furthermore the packetcollision probability between SUs and a PU will increase inthe multiple SUs scenario [19] Thus due to the importanceof the collision avoidance in a CR network with multiple SUsand multiple PU channels we need to propose a channelselection policy in the hybrid transmission model to addressthe collision issue

Energy supply is always a critical issue in wireless com-munications In a multi-SU andmulti-PU CR network wheremultiple SUs access multiple PU channels in the hybridtransmission model the SU needs to spend more energy toconstantly detect many channels and switch among multiplechannels Therefore energy efficiency is another importantcriterion in the CR network along with spectrum efficiency[20 21] Furthermore the cost of replacing the battery is oftenexpensive Recently some energy harvesting techniques havebeen introduced in [22ndash24] Such techniques allow devices toharvest natural sourcesrsquo energy such as sun wind acousticand ambient radio frequency (RF) waves Converting elec-tromagnetic waves from ambient RF waves into energy isconsidered to be more suitable and stable for the low energydevices in sensor networks or CR networks compared withother sources [24] Assuming that a SU is equipped withthe RF energy harvesting capability it must not only selectan idle channel to transmit data but also a busy channel toharvest RF energy to obtain enough energy and spectrumusage opportunity Hence a suitable channel selection policyis very important to improve both the spectrum efficiencyand the energy efficiency in CR networks

Inspired by the inherent benefits of the above schemesin this paper we focus on the channel competition amongSUs and packet collision between SUs and PUs in multi-SU and multi-PU CR networks Apart from the existingworks such as adopting the conventional noncooperativespectrum sensing method in the multi-SU CR network [25]and allowing SUs to access idle channels in the Overlayor Underlay model [14] there is no effective solution tothe packet competition among multiple SUs and PUs [26]We adopt the cooperative sensing method and the conceptof competitive set to solve these two problems so that thespectrum sensing accuracy and the throughput of multipleSUs can be improved It is noted that in our study SUs canharvest RF energy from busy channels by using the energyharvesting technology so as to extend their battery life

The main contributions of this paper are as follows

(i) We use the cooperative sensing method to avoid thepacket collision between SUs and PUs In channelsensing phase the SUs that detect the same channelexchange the channel usage information with eachother and make a more accurate decision on thestate of this channel Moreover the packet collisionbetween SUs and PUs can significantly decreasesince the cooperative sensing method can detect theactivity of PUs reoccupy their channels with a largeprobability

(ii) We propose a hybrid transmission model combiningOverlay andUnderlaymodels to fully utilize the avail-able idle spectrum Each SU can opportunisticallyaccess the unoccupied PU channel or underlay part ofits signal into the portion of the channel occupied bythe PU depending on the data queue state and sensingresult or decide whether or not to access a busychannel to harvest RF energy given its energy queueExtensive simulations show that our proposed hybridtransmission model can improve the efficiency ofspectrum usage and the energy harvesting efficiencyof SUs

(iii) With the aim of eliminating the channel competitionamong SUs and reducing their average waiting timeand also decreasing their spectrum handoff delay wepropose a competitive set based channel selectionpolicy In our proposed policy the SUs who form acompetitive set in the same channel randomly obtaininteger labels from zero The SU who obtains the zerolabel has the right to use this channel In particularseveral SUs can obtain multiple labels in differentcompetitive sets Therefore the average waiting timethat SUs spend on switching to other idle channelsor still staying on this channel for data transmissionor accessing a busy channel for energy harvesting islower compared with the random selection policySimulation results show that the proposed channelselection policy is simple and effective on reducing thecollisions among SUs

The rest of this paper is organized as follows The systemmodel is described in Section 2 and the cooperative spectrum

Mobile Information Systems 3

SU T5

SU T6

SU T7

SU T8

SU T3

SU T1

SU T2

SU T4

SU R5

SU R2

SU R1

SU R3

SU R4

SU R7

SU R8

SU R6

Primary user

Secondary user transmitter

Secondary user receiver

Figure 1 A scenario of the multi-SU and multi-PU CR network

sensing method is given in Section 3 In Section 4 wepresent the hybrid transmission model and channel selectionpolicy and then analyze the performance in terms of averagethroughput average waiting time and energy harvestingefficiency of SUsrsquo three aspects The simulation results arelisted in Section 5 Finally we conclude our work in Section 6

2 System Model

21 Multi-SU and Multi-PU CR Networks We consider amulti-SU andmulti-PUCR network with119872 PUs and119873 pairsof SU as depicted in Figure 1 where each PU is allocateda licensed channel (which we call ldquoPU channelrdquo) Similar to[13ndash15] the traffic of each channel is modeled as a two-statecontinuous-time Markov process the spectrum is occupiedby the PU (busy state) and the spectrum is not occupiedby the PU (idle state) For the PUs these two states arereferred to as ON and OFF states respectively Each SUtransmitter and its corresponding SU receiver are withineach otherrsquos transmission range Therefore the existence ofa communication between two SUs depends not only on thedistance between them but also on the time-varying activitiesof the PUs As illustrated in Figure 1 we consider the scenariothat several SUs may access the same channel and one SUmay have more than one channel for selection

As these PUs are in the interference range of some SUsthe channel power gains from the PU transmitter to the PUreceiver SU transmitter to the SU receiver PU transmitter tothe SU receiver and SU transmitter to the PU receiver aredenoted by 119866119901119901 119866119904119904 119866119901119904 and 119866119904119901 respectively We employthe model 119866119894119895 = 119896119889minus120572119894119895 for the channel gain between the 119894thtransmitter and the 119895th receiver where 119896 is an attenuationfactor that represents power variation caused by path loss 119889119894119895denotes the distance between them and120572 is the path loss [27]We assume that the channel power gains and the channel state

information (CSI) are known to each SU and SUs can obtainthe channel availability after spectrum sensing

22 Energy Harvesting Technology RF energy signal cannot only propagate over a distance but also broadcast inall directions [22] However due to the uncertainty oflocation fading and environmental conditions the energysupplied from RF energy may not guarantee QoS in wirelessapplications To ensure the static and stable energy theRF energy signal is transformed to a DC voltage and thenstored into a rechargeable battery [23] It is reasonable todefine the effective zone of the energy harvesting sincethe propagation energy drops off rapidly with the distanceincreases We assume that each SU can only obtain the RFenergy signal from the channels that it can sense Each SUcan harvest RF energy from the busy channels occupied byPUs and store the energy in a rechargeable battery when itstransmitter is equipped with an energy harvesting deviceand the maximum size of battery is 119864max In this paper therechargeable battery is modeled by an ideal linear model[28] where the changes in the energy stored are linearlyrelated to the amounts of energy harvested or spent Since theincreased energy harvested from PU channels can be utilizedfor channel sensing and data transmission the working timeof SUs will be extended

3 Cooperative Spectrum Sensing

Spectrum sensing is the basis of the DSA mechanismFurthermore sensing errors will affect the performance ofSUs transmission and cause packet collision between SUs andPUs In this section we describe our cooperative spectrumsensing method

31 Energy Based Spectrum Sensing Spectrum sensing has tobe performed before data transmission to detect the channelavailability Many signal techniques have been used for theSUs to sense the activity of the PUs [13]The energy detectionmethod not only implements simply but also representsintuitively the proportion of the busy channelsTherefore theenergy detection method is accurate and optimal when theSUs have little or no prior knowledge of the PU signal [29]and we consider it as the spectrum sensing algorithm in ourproposed policy The aim of spectrum sensing is to sense theexistence of signal in licensed spectrumThus under the twohypotheses the signal can be expressed as

1198670 119909 (119905) = 119899 (119905) 1198671 119909 (119905) = 119904 (119905) + 119899 (119905) (1)

where 119899(119905) is an Additive White Gaussian Noise (AWGN)and 119904(119905) is the signal of PU in target channel 1198670 and 1198671 arethe two hypotheses of nonexistence or existence of 119904(119905) From[30] we have known that the probability of detection can bedenoted by 119875119889 with a fixed SNR 120574 in an AWGN channel andit can be written as

119875119889 (120574 120591 120582) = Q(( 1205821205902 minus 120574 minus 1)radic 1205911198911199042120574 + 1) (2)


Channelsensing

Exchangeinformation Decision

Data transmissionEnergy harvesting

CS DTEHTS

Figure 2 An intuitional illustration of the time-slot structure

where 120591 is the sensing duration 120582 is the sensing threshold 119891119904is the sampling frequency 1205902 is the variance of the AWGNand Q(119909) is the tail probability of the normal distributionUnder imperfect sensing there are two types of sensingerrors miss detection and false alarm A false alarm erroroccurs when the SU observes the channel is busy whereasit is actually idle and a miss detection error occurs whenthe SU observes the channel is idle whereas it is actuallybusy Hence the false alarm indicates the waste of spectrumaccess opportunity whereas themiss detection imposes on thepotential interference to PUs The false alarm probability 119875119891andmiss detection probability 119875119898 can be expressed as [31]

119875119891 (120591 120582) = Q(( 1205821205902 minus 1)radic120591119891119904) 119875119898 (120574 120591 120582) = 1 minus Q(1205821205902 minus (1 + 120574)(1 + 120574)radic2120591119891119904)

(3)

where119875119891 and119875119898 are related to the threshold120582 and the sensingtime 120591 Furthermore 119875119898 is also a function of SNR

32 Cooperative Spectrum Sensing Due to the effects ofmultipath fading inside buildings with high penetration lossand local interference the probability of miss detection andfalse alarm will be increased under the conventional non-cooperative spectrum sensing method This phenomenonwill lead to packet collision between SUs and PUs in multi-SU and multi-PU CR networks In order to deal with thisproblem cooperative spectrum sensing has been adopted insome studies [15 17 23] We focus on how to collect thespectrum sensing data of SUs for cooperative sensing andcombine these sensing results to produce the final decisionin this paper

As illustrated in Figure 2 we suppose the multi-SU andmulti-PU CR network with time slotted (TS) that is one TSconsists of two phases which are the channel sensing phase(CS) and the data transmission (DT) or energy harvesting(EH) phase respectively In the first phase SUs sense thePU channels to detect the activity of the PUs and exchangethe channel usage information with other SUs Then eachSU will combine its sensing results with othersrsquo At last twoor more of the same results are considered to be the finaldecision of this channel In particular the channel will beredetected until a decision ismadewhen a channel is detectedby four SUs and two of them believe that the channel is idlewhile the other two are just the opposite The sensing resultis considered to be the final decision when a channel is onlydetected by one SU In the next phase the SU executes RFenergy harvesting or data transmission based on the finaldecision Similar to [32 33] we suppose that sensing duration

PU-T1 PU-R1

PU-TM PU-RM

Busy channel

Idle channel

Dataqueue

queueEnergy

SU-TX

middot middot middot

Figure 3 An illustration of the hybrid transmission model

is small compared with the PU channel traffic state cycle sothat the PU channel traffic state can be considered unchangedduring sensing phases

4 Channel Selection Policy

In this section we first propose a hybrid transmission modelfor single SU and secondly present a channel selection policyfor multi-SU based on the competitive set to alleviate thechannel competition among SUs

41 Hybrid Transmission Model As shown in Figure 3the arriving data is buffered in the data queue of the SUtransmitter 119876119863119894 119894 = 1 2 3 119873 The maximum capacity ofthe data queue is 119876max As mentioned before the RF energyis stored in the energy queue 119876119864119894 119894 = 1 2 3 119873 whosemaximum size is denoted as 119864max

At the beginning when the data arrives at 119894th SU trans-mitter its data queue and energy queue can be representedas 119876119863119894 = 119876119864119894 = 119864max then the SU can perform datatransmission when idle channels are sensed Let 119864(119904) bethe sensing outcome of 119894th SU and we define 120582119874 and 120582119880be the Overlay and the Underlay model energy thresholdrespectively If the channel signal energy is sensed below theOverlay model energy threshold that is 119864(119904) lt 120582119874 the SUwill transmit data with a higher power from its data queuein the Overlay model However if the channel signal energyis above the Overlay model energy threshold but below theUnderlay model energy threshold that is 120582119874 lt 119864(119904) lt 120582119880it means that the PU does not fully occupy this channel andthe SU can access it with PU at the same time by reducing itstransmission power as long as it does not interfere in the PUtransmission that is Underlay transmission model

To make use of the hybrid transmission model each SUtransmitter is assumed to have perfect knowledge of the CSIFor different channels their capacity and utilization rate aredifferent Based on the sensing result of channels each SUcalculates the statistical Overlay and Underlay model energythreshold and updates them according to the two types ofsensing errors that is miss detection and false alarm Whenthe data arrives at a SU transmitter it compares the currentchannel sensing results with the knowledge of CSI to obtainthe occupancy of PU The SU estimates the power of thePU based on the transmission distance and antenna gainwhen the PU does not fully occupy this channel [34] For


Start

Data queueand

energy queuestate

Idle channel sensing(i) Energy detecting

(ii) Transmission modethreshold setting

(iii) Results exchanging

Busy channel sensing(i) Energy detecting

(ii) Harvesting modethreshold setting


Energyharvesting

model

OverlayModel

UnderlayModel

No

N

No

o

Yes

Yes

Yes

QDi = empty

QEi = Emax

QDi = empty

QEi = Emax

QDi = empty

QEi = empty

E(s) lt 120582O

120582O lt E (s) lt 120582U

E (s) gt 120582U

Figure 4 An illustration of the process of selecting the transmission models for each SU

the Overlay model there is no limitation on the transmitpower of SUs and they can transmit data with the initialpower Nevertheless due to the interference caused by SUtowards PU the SUs need to decrease the transmit powerchange the modulation type and adjust the encodedmode toafford a suitable SNR to accommodate the variation of currentchannel in the Underlaymodel

When the data queue of 119894th SU is empty and its energyis used in the previous time-slot that is 119876119863119894 = 119876119864119894 =119864max the SU can harvest RF energy from busy channelsfor increasing energy reserves Hence if a channel is sensedabove the Underlay model energy threshold the SU mayimplement energy harvesting from it

In our proposed hybrid transmission model each SU candetermine to implement either the data transmission or theenergy harvesting depending on the state of data queue andenergy queue Based on the spectrum sensing results andOverlayUnderlaymodel energy thresholds each SU can notonly access a channel alone or with the PU simultaneouslyfor data transmission but also harvest the RF energy fromthe PU occupied channels The SU decides whether or notto stay in the current channel or switch to a new channel fordata transmission or a busy channel for energy harvestingafter sensing channels in the next CS phase The process ofselecting the transmission models for each SU is presented inFigure 4

42 Overview of Channel Selection Policy As the mentionedhybrid transmission model each SU can implement eitherthe data transmission in an idle channel or the energyharvesting from a busy channel However for the multi-SUand multi-PU CR network one of the great challenges ofimplementing multi-SU channel access successfully is the

problem of competition among SUs We explain the detailsof the proposed channel selection policy

421 Channel Selection Policy for Data Transmission TheSU transmitter sends a RTS packet on the channel to itscorresponding SU receiver if an idle channel is detectedThen the SU receiver replies with a CTS packet in the samechannel Notice that the RTSCTS collision may occur whenmore than one pair of SUs contends the same target idlechannel for data transmissions Hence different from theconventional way the SUpair does not access the idle channelimmediately when the CTS packet is successfully received bythe SU transmitter Those SUs who receive CTS packet forma competitive set 119878119868119894 119894 = 1 2 3 119872 which means thatthese SUs are competing to access this PU channel Supposingthat the size of 119878119868119894 is W we randomly assign them integerlabels from zero to119882minus 1 The SU who obtains the zero labelcan transmit data in the DT phase In particular for the SUswho can sense more than one channel they can competefor multiple idle channels and obtain multiple labels whenthe data arrives at their data queue Furthermore the SUcan access the corresponding channel for data transmissionas long as it can obtain the zero label in one competitiveset Similarly when the channel can only be accessed inthe Underlay model for data transmission those SUs whoreceive the CTS packet will form a competitive set 119878119880119894 119894 =1 2 3 119872

The SU that transmits data in the previous time-slot willkeep data transmission until the channel state is changedwhen the sensing outcome of the current channel is 119864(119904) lt120582119874 in the next CS phase All the label values of other SUsare in the same competitive set minus one when the SUwithdraws from the current channelTherefore the SUwhose


Input All PUs channels 119897119899 119899 isin [1119873] including 119897119898 idle channels119898 isin [1119872] and 119901 SUs 119901 isin [1 119875]Output Channel selection for SUs

(1) begin(2) lowast For 119897119898 idle channels lowast(3) if the data queue of SUs 119876119863119901 = then(4) if 119897119898 ge 1 then(5) if the 119902 SUs receive CTS packet successfully then(6) The 119902 SUs form a number of competitive sets 119878119868119894 119894 = 1 119898(7) Randomly assigning them integer labels from zero to 119908 minus 1 119908 le 119902(8) for erery competitive sets 119878119868119894 do(9) The SU who obtains the zero label can access the corresponding channel for data transmission and

withdraw from other competitive sets(10) Label values of other SUs minus one(11) end(12) end(13) end(14) if 119897119898 = 0 then(15) if the 119904 SUs receive CTS packet successfully then(16) if the 119905 SUs transmit the data in the Underlay model and their throughput can be satisfied then(17) The 119905 SUs form a number of competitive sets 119878119880119894 119894 = 1 119899(18) Randomly assigning them integer labels from zero to V minus 1 V le 119905(19) for erery competitive sets 119878119880119894 do(20) The SU who obtains the zero label can reduce its transmitting power and access the corresponding

channel for data transmission and then withdraw from other competitive sets(21) Label values of other SUs minus one(22) end(23) end(24) end(25) end(26) end(27) end

Algorithm 1 The channel selection policy for data transmission

label value is subtracted to be zero can access this channelIf the present channel is detected satisfy 120582119874 lt 119864(119904) lt120582119880 in the next CS phase that is the PU is not completelyoccupied in this channel for data transmission the SUreduces its transmission power to satisfy the interferencepower constraint of PU that it can continue to transmit data intheUnderlaymodel and their corresponding competitive setswill remain in useHowever the transmission of SUwill causeinterference to the communication of PU if the 119864(119904) gt 120582119880Then the data transmission of SU will be stopped in the nextDT phase and the competitive set of this channel will bedissolvedThe algorithm of our channel selection strategy fordata transmission is presented in Algorithm 1

We illustrate the process of randomly assigning integerlabels by Figure 5 First four SUs need to access channelsfor data transmission and they sense the current channelavailability to obtain a list of available channels Secondly theSUs who compete for the same channel form a competitiveset that is the SUs 1 2 3 can use the channel AWe randomlyassign integer labels from zero for these SUs Thirdly the SUswho obtain the zero label can access the channels and theywithdraw from the corresponding competitive sets while thelabel values of other SUs are minus one In particular the SU2 can use channels A or B In the next time-slot the SU 4 canaccess the channel B

422 Channel Selection Policy for Energy Harvesting TheSUs that contend for the same target busy channel forma competitive set 119878119864119895 119895 = 1 2 3 119872 which means thatthese SUs are competing to access the 119895th PU channel forenergy harvestingWe also assign them integer labels and theSU that obtains the zero label can harvest energy in the nextEH phase Then these SUs withdraw from the competitionsets when the data arrives at the data queue of SUs or theirenergy queue is full The algorithm of our channel selectionpolicy for energy harvesting is given in Algorithm 2

In the proposed channel selection policy the SU receiversends a Decode packet to its transmitter when the currenttransmission is complete that is the data has been acceptedsuccessfully The Channel-Switching (CSW) flag is set whenthe SU needs to switch to another channel and then the SUtransmitter and receiver pause their current transmission andperform channel handoff [35 36]

Note that the CSMACA protocol also uses the RTSCTShandshake procedures to ensure that the collision doesnot occur among users and utilizes the exponential-backoffalgorithm to decompose collision that is each node performsa random delay 119905when the collision happens and 119905 obeys the119879(0 sim 119879) on the bottom of the exponential distribution Inour proposed channel selection policy we ensure the usageof idle channels through establishing the competitive sets


0

1 00 0

1 1 12

2 23

3 34

4 4

1

2

3

4

1

2

3

4

User User

A CA B

AB

A CA B

AB

A CA B

AB

Availablechannel

Availablechannel Label

User Availablechannel

AccesschannelLabel User Available

channelAccess

channelLabel

1 00 02

1

2

1

1

0

1 00 0

2

1

1

CA

A CA B

AB

CA

B

The SUs that needto access channels

The SUs sense the usageof current channels andobtain the list ofavailable channels

We randomly assigninteger labels for each SU

Next time-slotThe SU who obtains the zero labelcan access one channel while thelabel values of other SUs minusone

Figure 5 An example of randomly assigning integer labels


(1) begin(2) lowast For 119897119899minus119898 busy channels lowast(3) if the data queue of SUs 119876119863119901 = then(4) if the energy queue of SUs 1198761198631199011015840 = 119876max then(5) The 1199011015840 SUs form a number of competitive sets 119878119864119894 119894 = 1 119899(6) Randomly assigning them integer labels from zero to 119906 minus 1 119906 le 1199011015840(7) for erery competitive sets 119878119864119894 do(8) The SU who obtains the zero laber can access the corresponding busy channel for energy harvesting in the next

EH phase and withdraw from other competitive sets(9) Label values of other SUs minus one(10) end(11) end(12) end(13) end

Algorithm 2 The channel selection policy for energy harvesting

and randomly assigning the integer labels when the collisionoccurs Moreover in order to reduce the average waiting timeof SUs the SUs who access channels withdraw from theircompetitive sets while other SUsrsquo label values are minus one

43 Performance Analysis of the Channel Selection PolicyIn this section we intend to illustrate the spectrum usageperformance of our proposed policy in terms of averagethroughput average waiting time and energy harvestingefficiency of SUsrsquo three aspects

431 Average Throughput of SUs In our proposed channelselection policy SU can transmit data in the Overlay orUnderlay model The service rate of each SU in the hybridmodel is described as 119877ℎ = 119877119900 + 119877119906 and 119877119900 can be denotedby 1198770119900 1198771119900 11987701015840119900 and 11987711015840119900 in the Overlaymodel [37]

1198770119900 = 119861 log2 (1 + 119892119904119875119900119904 ) 1198771119900 = 0


11987701015840119900 = 119861 log2 (1 + 119892119904119875119900119904119892119901119875119901 + 1) 11987711015840119900 = 0

(4)

where1198770119900 represents that the PU does not occupy the channelIn contrast1198771119900 represents that the spectrum is being occupiedby PU 11987701015840119900 and 11987711015840119900 are the service rate of each SU under falsealarm and miss detection respectively Similarly the 119877119906 canbe denoted by 1198770119906 1198771119906 in the Underlaymodel

1198770119906 = 119861 log2 (1 + 119892119904119875119906119904 ) 1198771119906 = 119861 log2 (1 + 119892119904119875119906119904119892119901119875119901 + 1)

(5)

The throughput of SU can be described in terms of the outageas [38]

119879 = 1 minus 119901out (6)

where 119901out is the outage probability The throughput inchannel selection policy 119879ℎ is comprised of 119879119900 and 119879119906respectively 119879119900 is given by

119879119900 = 119901119894 (1 minus 119901119891) (1 minus 119901119900out) + (1 minus 119901119894) 119901119891 (1 minus 1199011199001015840out) (7)

where 119901119894 is the channel idle probability Since the SU trans-mission will cause interference to PU Under false alarm 119901119900outand 1199011199001015840out can be described as

119901119900out = Pr [1198770119900 lt 119877119904] 1199011199001015840out = Pr [11987701015840119900 lt 119877119904]

(8)

where 119877119904 is the required service rate of SU Correspondinglywe can obtain 119879119906 119901119906out and 1199011199061015840out as follows

119879119906 = 119901119894 (1 minus 119901119891) (1 minus 119901119906out) + (1 minus 119901119894) 119901119891 (1 minus 1199011199061015840out) 119901119906out = Pr [1198770119906 lt 119877119904] 1199011199061015840out = Pr [1198771119906 lt 119877119904]

(9)

432 Average Waiting Time of SUs Here we calculate thetime elapsed between each SU which receives the RTS signaland implements data transmission and this elapsed time canreflect the performance of the competitive set The averagewaiting time of SUs will be longer than the conventionalrandom access policy if the design of the competitive set isnot reasonableThus the average waiting time of SUs119879119908 canbe described as

119879119908 = 119879119905 minus 119879RTS (10)

where 119879119905 and 119879RTS are the time-slots that the SU transmitsdata to and receives the RTS signal from respectively

Table 1 Simulation parameters settings

Parameter Value119875119894 08120582119900 03120582119906 07119864max 15119876max 20119877119904 3 bps119875119901 15 dBTime-slot 2msRTS packets length 250 bitCTS packets length 220 bit

433 Energy Harvesting Efficiency We use 119890ℎ to express thepackets of energy that can be harvested by SUs from busychannels and it follows Poisson distribution The energyconsumed by SU for data transmission and spectrum sensingare 119890119905 and 119890119904 respectively We assume that 119890119888 representsanother energy consumption on circuit and 119890119905119903 denotes theresidual energy at the time-slot 119905 Therefore the energyharvesting efficiency is described in terms of the residualenergy in the next time-slot

119890119905+1119903 = min [119890119905119903 + 119890ℎ minus (119890119905 + 119890119904 + 119890119888) 119864max] (11)

5 Simulations

In this section we will provide numerical results to demon-strate the performance of the proposed cooperative sensingmethod and channel selection policy in terms of probabilityof false alarm average throughput average waiting timeand energy harvesting efficiency of SUs Table 1 shows theparameter settings of our simulations and some parametersare valued based on the previous works on CR networks Weconsider a multi-SU and multi-PU CR network with 20 PUs20 available PU channels and 25 pairs of SUs In particularseveral SUs may access the same channel and one SU mayhave more than one channel for selection We set all thespectrum bandwidth to be the same and the packets lengthsof SUs and PUs are fixed in the simulations However theinterference limitation of PUs is differentWe let the path lossconstant 120572 = 2 and attenuation factor 119896 = 05 respectivelyaccording to the empirical values to compute the channelgain Moreover the white Gaussian noise is 8 times 10minus1551 Performance for Probability of False Alarm Figure 6illustrates the probability of false alarm in cooperative sensingand conventional noncooperative sensing method underdifferent numbers of SUs As shown in the figure the prob-ability of false alarm decreases with the increase of detectionprobability For the conventional noncooperative sensingmethod there is no interaction on the detection resultsamong multiple SUs Hence the increase of users has noimpact on the probability of false alarm For a fixed detectionprobability cooperative sensing method can achieve higherdetection accuracy As described in Section 32 two or more

Mobile Information Systems 9Pr

obab

ility

of f

alse

alar

m

001

002

003

004

005

006

007

008

009

Number of SUs1 2 3 4 5 6

Cooperative sensing detection probability 085Noncooperative sensing detection probability 085Cooperative sensing detection probability 09Noncooperative sensing detection probability 09

Figure 6 The probability of false alarm in cooperative sensing andnoncooperative sensing methods under different numbers of SUs

of the same results are considered to be the final decisionof the target channel Therefore the probability of falsealarm decreases significantly when the cooperative SUs are2 or 3 Furthermore the probability of false alarm wherethe detection probability is 085 decreases faster when thenumber of SUs changes from 3 to 4 Thus the cooperativesensing method can improve the accuracy of channel sensingin the case of low detection rate However more than 4cooperative SUs have little effect on the probability of falsealarm

52 Performance for Average Throughput of SUs Figure 7shows the average throughput of SUs in the following trans-mission models our proposed hybrid model existing hybridmodel [39] Overlay-only model and Underlay-only modelunder different numbers of busy channels As can be seenfrom the figure that the average throughput of SUs decreasesin the Overlay-only model with the number of busy channelsincrease It is due to the fact that the high percentage ofbusy channels restricts SUs from transmission in theOverlay-only model However there is a little impact on the averagethroughput of SUs since the SUs can coexist with PUs in theUnderlay-only model It can be observed from the figure thatthe hybridmodel transmission outperforms theOverlay-onlyandUnderlay-onlymodel alone Furthermore we can see thatour proposed hybrid model can achieve higher throughputcompared with the existing hybrid model The reason of thisobservation can be explained as follows With the decreaseof available idle channels the opportunities of SUs for datatransmission become less The collision among SUs becomesmore intense since the existing hybrid model is only basedon the number of SUs and access probability However theconcept of competitive set can improve the utilization of thelimited idle channels

Number of busy channels0 2 4 6 8 10 12 14 16 18 20

OverlayUnderlay

Existing hybridProposed hybrid

Aver

age t

hrou

ghpu

t of S

Us

0

5

10

15

20

25

Figure 7The effect of transmissionmodels on the average through-put of SUs under different numbers of busy channels

53 Performance for Average Waiting Time of SUs Figure 8shows the averagewaiting time of SUs in four differentmodelsunder different numbers of busy channels It is clear that theaverage waiting time of SUs greatly increases in the Overlay-only model with the decrease of available idle channelsIn contrast the average waiting time is not significantlyincreased in the Underlay-only model since the number ofavailable idle channels has little effect on the data transmis-sion of SUs The different interference threshold constraint ofPUs so that the SUs cannot access some channels is in theUnderlay-only model which results in the consequence thatsome SUs need to wait for a long time to access channelsTheaverage waiting time of SUs of our proposed hybrid modelis lower than the existing hybrid model but higher than theUnderlay-only model when lots of channels are occupiedby PUs The explanation for this observation is as followsAs described in Section 4 SUs can continue to transmitdata in the Underlay model when the PU accesses its idlechannel and their current competitive sets will remain in useFurthermore the SUs that compete more than one channelscan wait for accessing opportunities in other competitivesets when the current channel cannot be accessed Thereforethe average waiting time of SUs will be reduced in ourproposed hybrid model However since the hybrid modelwill give priority to whether the channel can be accessed inthe Overlay model when the idle channels become less theaverage waiting time of SUs in the Underlay-only model islower than ours

54 Performance for Energy Harvesting Efficiency of SUsFigure 9 shows the average residual energy of SUs in theconventional CR network existing CR network with energyharvesting [30] and our CR network with energy harvestingversus the simulation time As shown in the figure energyharvesting technology can ensure that enough energy isreserved after long time communication Furthermore our

10 Mobile Information SystemsAv

erag

e wai

ting

time o

f SU

s

15

20

25

30

35

40

45

50


OverlayUnderlay


Figure 8 The effect of transmission models on the average waitingtime of SUs under different numbers of busy channels

Aver

age r

esid

ual e

nerg

y of

SU

s

0

05

1

15

2

25

3

2 4 6 8 10

Simulation time

Conventional CRNExisting CRN with energy harvestingOur CRN with energy harvesting

Figure 9The average residual energy of SUs in the conventional CRnetwork existing CR network with energy harvesting and our CRnetwork with energy harvesting under different simulation time

proposed CR network with energy harvesting outperformsthe existing CR network with energy harvesting This isbecause as described in Section 3 SUs decide to sense the idlechannels for data transmission or busy channels for energyharvesting depending on the state of data queue and energyqueueHence SUs can spend less energy for sensing channelsMeanwhile the SUs may have more opportunities to harvestenergy since the concept of competitive set can reduce thecollision among multiple SUs competing for the same busychannel

6 Conclusion

In this paper aiming at solving the problem of spectrumscarcity in IoE environment we consider a multi-SU andmulti-PU cognitive radio network in which the SUs areequipped with the RF energy harvesting capability In thisnetwork the crucial issues are the channel competitionamong SUs and the packet collision between SUs and PUsWe adopt the cooperative spectrum sensingmethod to reducethe probability of sensing errors and alleviate the interferenceto PUs In order to solve the problem of channel competitionamong SUs we first propose a hybrid transmission model forsingle SU Each SU can either implement data transmissionin an idle channel or energy harvesting from a busy channelgiven its data queue and energy queue state and sensingresult Additionally we present a channel selection policy formulti-SU based on competitive set Our proposed policy canachieve higher throughput compared with the conventionalrandom policy Furthermore the collision will never bedetected by themselves and may last for a quite long timewhen several SUs collide with each other in the conventionalrandom policy Hence the channel competition among SUswill largely limit the performance of conventional randompolicyWhile SUs will detect the collision in the CS phase andstop transmission in the next DTEH phase to avoid longerineffective transmission in our proposed policy Simulationsshow that the proposed cooperative sensing method andchannel selection policy outperform previous solutions interms of probability of false alarm average throughputaverage waiting time and energy harvesting efficiency of SUs

Competing Interests

The authors declare that there are no competing interestsregarding the publication of this paper

Acknowledgments

This work was supported in part by the National NaturalScience Foundation of China under Grant 61672283 theFunding of Jiangsu Innovation Program for Graduate Educa-tion under Grant KYLX15 0325 the Fundamental ResearchFunds for the Central Universities under Grant NS2015094the Natural Science Foundation of Jiangsu Province underGrant BK20140835 and the Postdoctoral Foundation ofJiangsu Province under Grant 1401018B

References

[1] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of things a survey on enabling tech-nologies protocols and applicationsrdquo IEEE CommunicationsSurveys amp Tutorials vol 17 no 4 pp 2347ndash2376 2015

[2] J Jin J Gubbi S Marusic and M Palaniswami ldquoAn informa-tion framework for creating a smart city through internet ofthingsrdquo IEEE Internet of Things Journal vol 1 no 2 pp 112ndash1212014

[3] C Zhu V C M Leung L Shu and E C H Ngai ldquoGreeninternet of things for smart worldrdquo IEEE Access vol 3 pp 2151ndash2162 2015


[4] Z Sheng C Mahapatra C Zhu and V C M Leung ldquoRecentadvances in industrial wireless sensor networks toward efficientmanagement in IoTrdquo IEEE Access vol 3 pp 622ndash637 2015

[5] Z Sheng C Zhu and V C M Leung ldquoSurfing the internet-of-things lightweight access and control of wireless sensornetworks using industrial low power protocolsrdquo EAI EndorsedTransactions on Industrial Networks and Intelligent Systems vol14 no 1 article e2 pp 1ndash11 2014

[6] W Li C Zhu V C M Leung L T Yang and Y Ma ldquoPer-formance comparison of cognitive radio sensor networks forindustrial IoT with different deployment patternsrdquo IEEE Sys-tems Journal 2015

[7] W Li V Leung C Zhu and Y Ma ldquoScheduling and routingmethods for cognitive radio sensor networks in regular topol-ogyrdquo Wireless Communications and Mobile Computing vol 16no 1 pp 47ndash58 2016

[8] J Li H Zhao J Wei et al ldquoSender-jump receiver-wait ablind rendezvous algorithm for distributed cognitive radionetworksrdquo in Proceedings of the IEEE International Symposiumon Personal Indoor andMobile Radio Communications (PIMRCrsquo16) Valencia Spain September 2016

[9] Federal Comunications Commission ldquoUnlicensed operation inthe TV broadcast bandsrdquo Rep ET Docket no 08-260 2008

[10] Y C Liang K C Chen G Y Li and P Mahonen ldquoCognitiveradio networking and communications an overviewrdquo IEEETransactions on Vehicular Technology vol 60 no 7 pp 3386ndash3407 2011

[11] E Z Tragos S Zeadally A G Fragkiadakis and V A SirisldquoSpectrum assignment in cognitive radio networks a compre-hensive surveyrdquo IEEE Communications Surveys amp Tutorials vol15 no 3 pp 1108ndash1135 2013

[12] X Zhai L Zheng and C W Tan ldquoEnergy-infeasibility tradeoffin cognitive radio networks price-driven spectrum access algo-rithmsrdquo IEEE Journal on Selected Areas in Communications vol32 no 3 pp 528ndash538 2014

[13] Y Yilmaz Z Guo and X Wang ldquoSequential joint spectrumsensing and channel estimation for dynamic spectrum accessrdquoIEEE Journal on Selected Areas in Communications vol 32 no11 pp 2000ndash2012 2014

[14] N Khambekar C M Spooner and V Chaudhary ldquoOn improv-ing serviceability with quantified dynamic spectrum accessrdquo inProceedings of the IEEE International Symposium on DynamicSpectrum Access Networks (DySPAN rsquo14) pp 553ndash564 McLeanVa USA April 2014

[15] TMCChuH Phan andH J Zepernick ldquoHybrid interweave-underlay spectrum access for cognitive cooperative radio net-worksrdquo IEEE Transactions on Communications vol 62 no 7 pp2183ndash2197 2014

[16] V Chakravarthy X Li R Zhou Z Wu and M Temple ldquoNoveloverlayunderlay cognitive radio waveforms using sd-smseframework to enhance spectrum efficiency-part II analysis infading channelsrdquo IEEE Transactions on Communications vol58 no 6 pp 1868ndash1876 2010

[17] A K Karmokar S Senthuran and A Anpalagan ldquoPhysicallayer-optimal and cross-layer channel access policies for hybridoverlay-underlay cognitive radio networksrdquo IET Communica-tions vol 8 no 15 pp 2666ndash2675 2014

[18] J Zou H Xiong D Wang and C W Chen ldquoOptimal powerallocation for hybrid overlayunderlay spectrum sharing inmultiband cognitive radio networksrdquo IEEE Transactions onVehicular Technology vol 62 no 4 pp 1827ndash1837 2013

[19] H Cho and G Hwang ldquoAn optimized random channel accesspolicy in cognitive radio networks under packet collisionrequirement for primary usersrdquo IEEE Transactions on WirelessCommunications vol 12 no 12 pp 6382ndash6391 2013

[20] S Xie and Y Wang ldquoConstruction of tree network with limiteddelivery latency in homogeneous wireless sensor networksrdquoWireless Personal Communications vol 78 no 1 pp 231ndash2462014

[21] J Shen H Tan J Wang J Wang and S Lee ldquoA novel routingprotocol providing good transmission reliability in underwatersensor networksrdquo Journal of Internet Technology vol 16 no 1pp 171ndash178 2015

[22] S Sudevalayam and P Kulkarni ldquoEnergy harvesting sensornodes survey and implicationsrdquo IEEE Communications Surveysamp Tutorials vol 13 no 3 pp 443ndash461 2011

[23] Pratibha K H Li and K C Teh ldquoEnergy-harvesting cognitiveradio systems cooperating for spectrum sensing and utiliza-tionrdquo in Proceedings of the IEEEGlobal Communications Confer-ence (GLOBECOM rsquo15) San Diego Calif USA December 2015

[24] L Mohjazi M Dianati G K Karagiannidis S Muhaidatand M Al-Qutayri ldquoRf-powered cognitive radio networkstechnical challenges and limitationsrdquo IEEE CommunicationsMagazine vol 53 no 4 pp 94ndash100 2015

[25] S Hu Y D Yao and Z Yang ldquoCognitivemedium access controlprotocols for secondary users sharing a common channelwith time division multiple access primary usersrdquo WirelessCommunications and Mobile Computing vol 14 no 2 pp 284ndash296 2014

[26] H A B Salameh and M F El-Attar ldquoCooperative OFDM-based virtual clustering scheme for distributed coordination incognitive radio networksrdquo IEEE Transactions on Vehicular Tech-nology vol 64 no 8 pp 3624ndash3632 2015

[27] S P Herath and N Rajatheva ldquoAnalysis of equal gain com-bining in energy detection for cognitive radio over Nakagamichannelsrdquo inProceedings of the IEEEGlobal TelecommunicationsConference (GLOBECOM rsquo08) pp 1ndash5 New Orleans La USANovember 2008

[28] X Lu P Wang D Niyato D I Kim and Z Han ldquoWirelessnetworks with RF energy harvesting a contemporary surveyrdquoIEEE Communications Surveys amp Tutorials vol 17 no 2 pp757ndash789 2015

[29] S Wang Y Wang J P Coon and A Doufexi ldquoEnergy-efficientspectrum sensing and access for cognitive radio networksrdquoIEEE Transactions on Vehicular Technology vol 61 no 2 pp906ndash912 2012

[30] S Park H Kim and D Hong ldquoCognitive radio networks withenergy harvestingrdquo IEEE Transactions onWireless Communica-tions vol 12 no 3 pp 1386ndash1397 2013

[31] T Yucek and H Arslan ldquoA survey of spectrum sensing algo-rithms for cognitive radio applicationsrdquo IEEE CommunicationsSurveys amp Tutorials vol 11 no 1 pp 116ndash130 2009

[32] W B Chien C K Yang and Y H Huang ldquoEnergy-savingcooperative spectrum sensing processor for cognitive radiosystemrdquo IEEE Transactions on Circuits and Systems I RegularPapers vol 58 no 4 pp 711ndash723 2011

[33] Y Zou Y D Yao and B Zheng ldquoCooperative relay techniquesfor cognitive radio systems spectrum sensing and secondaryuser transmissionsrdquo IEEE Communications Magazine vol 50no 4 pp 98ndash103 2012

[34] P J Smith P A Dmochowski H A Suraweera and M ShafildquoThe effects of limited channel knowledge on cognitive radio


system capacityrdquo IEEE Transactions on Vehicular Technologyvol 62 no 2 pp 927ndash933 2013

[35] B Wang Z Ji K J R Liu and T C Clancy ldquoPrimary-prioritizedmarkov approach for dynamic spectrum allocationrdquoin Proceedings of the IEEE International Symposium on NewFrontiers in Dynamic Spectrum Access Networks (DySPAN rsquo07)pp 1854ndash1865 Dublin Ireland April 2007

[36] Y Song and J Xie ldquoProSpect a proactive spectrum handoffframework for cognitive radio ad hoc networks without com-mon control channelrdquo IEEE Transactions onMobile Computingvol 11 no 7 pp 1127ndash1139 2012

[37] M G Khoshkholgh K Navaie and H Yanikomeroglu ldquoAccessstrategies for spectrum sharing in fading environment overlayunderlay and mixedrdquo IEEE Transactions on Mobile Computingvol 9 no 12 pp 1780ndash1793 2010

[38] Y Wang P Ren F Gao and Z Su ldquoA hybrid underlayoverlaytransmission mode for cognitive radio networks with statisticalquality-of-service provisioningrdquo IEEE Transactions on WirelessCommunications vol 13 no 3 pp 1482ndash1498 2014

[39] S Gmira A Kobbane and E Sabir ldquoA new optimal hybridspectrum access in cognitive radio overlay-underlay moderdquoin Proceedings of the International Conference on Wireless Net-works andMobile Communications (WINCOM rsquo15) MarrakechMorocco October 2015

Submit your manuscripts athttpwwwhindawicom

Computer Games Technology

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Distributed Sensor Networks


Advances in

FuzzySystems

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014


ReconfigurableComputing

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014


Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

Artificial Intelligence

HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014

Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications


Hindawi Publishing Corporation

httpwwwhindawicom Volume 2014

Advances in

Multimedia


Biomedical Imaging


ArtificialNeural Systems

Advances in


RoboticsJournal of



Computational Intelligence and Neuroscience

Industrial EngineeringJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Human-ComputerInteraction

Advances in

Computer EngineeringAdvances in



Underlay model However when the PU state changes tobe inactive the transmission power of the SU will be stillbelow the interference threshold constraint in the Underlaymodel Therefore the idle spectrum resources are not fullyutilized and the SU does not achieve optimal performanceOn the other hand when the licensed channels are verybusy the time that SU must wait for an available channelis too long and then it may also significantly reduce theperformance of Overlay model [15] Therefore we need tofind a hybrid transmission model where the advantages ofboth Overlay and Underlay models are combined for thePU state variability so that the performance of SUs can bemaximized

The hybrid transmission model has recently been pro-posed in [16ndash18] In [16] the SU can exchange controlinformation in the Underlay model and transmit data infor-mation in the Overlay model However the decision ofaccessing a model is not based on the sensing results In[17 18] the SU can constantly sense the activity of PUs andtransmit data information in theOverlaymodel when the PUtransmission is not detected Otherwise the SU reduces itstransmission power to access the spectrum in the Underlaymodel However these papers did not take into account thesensing errors and neglected the effect of PU retransmissionon the SU QoS Although these related works indicatedthat the SU can obtain more spectrum access opportunitiesin the hybrid transmission model compared with the twoconventional transmission models the issue of two or moreSUs competing for the same channel has rarely been studiedso far to the best of our knowledge Furthermore the packetcollision probability between SUs and a PU will increase inthe multiple SUs scenario [19] Thus due to the importanceof the collision avoidance in a CR network with multiple SUsand multiple PU channels we need to propose a channelselection policy in the hybrid transmission model to addressthe collision issue

Energy supply is always a critical issue in wireless com-munications In a multi-SU andmulti-PU CR network wheremultiple SUs access multiple PU channels in the hybridtransmission model the SU needs to spend more energy toconstantly detect many channels and switch among multiplechannels Therefore energy efficiency is another importantcriterion in the CR network along with spectrum efficiency[20 21] Furthermore the cost of replacing the battery is oftenexpensive Recently some energy harvesting techniques havebeen introduced in [22ndash24] Such techniques allow devices toharvest natural sourcesrsquo energy such as sun wind acousticand ambient radio frequency (RF) waves Converting elec-tromagnetic waves from ambient RF waves into energy isconsidered to be more suitable and stable for the low energydevices in sensor networks or CR networks compared withother sources [24] Assuming that a SU is equipped withthe RF energy harvesting capability it must not only selectan idle channel to transmit data but also a busy channel toharvest RF energy to obtain enough energy and spectrumusage opportunity Hence a suitable channel selection policyis very important to improve both the spectrum efficiencyand the energy efficiency in CR networks

Inspired by the inherent benefits of the above schemesin this paper we focus on the channel competition amongSUs and packet collision between SUs and PUs in multi-SU and multi-PU CR networks Apart from the existingworks such as adopting the conventional noncooperativespectrum sensing method in the multi-SU CR network [25]and allowing SUs to access idle channels in the Overlayor Underlay model [14] there is no effective solution tothe packet competition among multiple SUs and PUs [26]We adopt the cooperative sensing method and the conceptof competitive set to solve these two problems so that thespectrum sensing accuracy and the throughput of multipleSUs can be improved It is noted that in our study SUs canharvest RF energy from busy channels by using the energyharvesting technology so as to extend their battery life

The main contributions of this paper are as follows

(i) We use the cooperative sensing method to avoid thepacket collision between SUs and PUs In channelsensing phase the SUs that detect the same channelexchange the channel usage information with eachother and make a more accurate decision on thestate of this channel Moreover the packet collisionbetween SUs and PUs can significantly decreasesince the cooperative sensing method can detect theactivity of PUs reoccupy their channels with a largeprobability

(ii) We propose a hybrid transmission model combiningOverlay andUnderlaymodels to fully utilize the avail-able idle spectrum Each SU can opportunisticallyaccess the unoccupied PU channel or underlay part ofits signal into the portion of the channel occupied bythe PU depending on the data queue state and sensingresult or decide whether or not to access a busychannel to harvest RF energy given its energy queueExtensive simulations show that our proposed hybridtransmission model can improve the efficiency ofspectrum usage and the energy harvesting efficiencyof SUs

(iii) With the aim of eliminating the channel competitionamong SUs and reducing their average waiting timeand also decreasing their spectrum handoff delay wepropose a competitive set based channel selectionpolicy In our proposed policy the SUs who form acompetitive set in the same channel randomly obtaininteger labels from zero The SU who obtains the zerolabel has the right to use this channel In particularseveral SUs can obtain multiple labels in differentcompetitive sets Therefore the average waiting timethat SUs spend on switching to other idle channelsor still staying on this channel for data transmissionor accessing a busy channel for energy harvesting islower compared with the random selection policySimulation results show that the proposed channelselection policy is simple and effective on reducing thecollisions among SUs

The rest of this paper is organized as follows The systemmodel is described in Section 2 and the cooperative spectrum


SU T5

SU T6

SU T7

SU T8

SU T3

SU T1

SU T2

SU T4

SU R5

SU R2

SU R1

SU R3

SU R4

SU R7

SU R8

SU R6

Primary user





2 System Model








1198670 119909 (119905) = 119899 (119905) 1198671 119909 (119905) = 119904 (119905) + 119899 (119905) (1)




Channelsensing



CS DTEHTS




(3)




PU-T1 PU-R1

PU-TM PU-RM

Busy channel

Idle channel

Dataqueue

queueEnergy

SU-TX










Start

Data queueand

energy queuestate







Energyharvesting

model

OverlayModel

UnderlayModel

No

N

No

o

Yes

Yes

Yes

QDi = empty

QEi = Emax

QDi = empty

QEi = Emax

QDi = empty

QEi = empty

E(s) lt 120582O

120582O lt E (s) lt 120582U

E (s) gt 120582U





















0

1 00 0

1 1 12

2 23

3 34

4 4

1

2

3

4

1

2

3

4

User User

A CA B

AB

A CA B

AB

A CA B

AB

Availablechannel




channelAccess

channelLabel

1 00 02

1

2

1

1

0

1 00 0

2

1

1

CA

A CA B

AB

CA

B













1198770119900 = 119861 log2 (1 + 119892119904119875119900119904 ) 1198771119900 = 0


11987701015840119900 = 119861 log2 (1 + 119892119904119875119900119904119892119901119875119901 + 1) 11987711015840119900 = 0

(4)


1198770119906 = 119861 log2 (1 + 119892119904119875119906119904 ) 1198771119906 = 119861 log2 (1 + 119892119904119875119906119904119892119901119875119901 + 1)

(5)


119879 = 1 minus 119901out (6)




119901119900out = Pr [1198770119900 lt 119877119904] 1199011199001015840out = Pr [11987701015840119900 lt 119877119904]

(8)



(9)


119879119908 = 119879119905 minus 119879RTS (10)





119890119905+1119903 = min [119890119905119903 + 119890ℎ minus (119890119905 + 119890119904 + 119890119888) 119864max] (11)

5 Simulations



obab

ility

of f

alse

alar

m

001

002

003

004

005

006

007

008

009







OverlayUnderlay


Aver

age t

hrou

ghpu

t of S

Us

0

5

10

15

20

25





erag

e wai

ting

time o

f SU

s

15

20

25

30

35

40

45

50


OverlayUnderlay



Aver

age r

esid

ual e

nerg

y of

SU

s

0

05

1

15

2

25

3

2 4 6 8 10

Simulation time




6 Conclusion


Competing Interests


Acknowledgments


References


















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




SU T5

SU T6

SU T7

SU T8

SU T3

SU T1

SU T2

SU T4

SU R5

SU R2

SU R1

SU R3

SU R4

SU R7

SU R8

SU R6

Primary user





2 System Model








1198670 119909 (119905) = 119899 (119905) 1198671 119909 (119905) = 119904 (119905) + 119899 (119905) (1)




Channelsensing



CS DTEHTS




(3)




PU-T1 PU-R1

PU-TM PU-RM

Busy channel

Idle channel

Dataqueue

queueEnergy

SU-TX










Start

Data queueand

energy queuestate







Energyharvesting

model

OverlayModel

UnderlayModel

No

N

No

o

Yes

Yes

Yes

QDi = empty

QEi = Emax

QDi = empty

QEi = Emax

QDi = empty

QEi = empty

E(s) lt 120582O

120582O lt E (s) lt 120582U

E (s) gt 120582U





















0

1 00 0

1 1 12

2 23

3 34

4 4

1

2

3

4

1

2

3

4

User User

A CA B

AB

A CA B

AB

A CA B

AB

Availablechannel




channelAccess

channelLabel

1 00 02

1

2

1

1

0

1 00 0

2

1

1

CA

A CA B

AB

CA

B













1198770119900 = 119861 log2 (1 + 119892119904119875119900119904 ) 1198771119900 = 0


11987701015840119900 = 119861 log2 (1 + 119892119904119875119900119904119892119901119875119901 + 1) 11987711015840119900 = 0

(4)


1198770119906 = 119861 log2 (1 + 119892119904119875119906119904 ) 1198771119906 = 119861 log2 (1 + 119892119904119875119906119904119892119901119875119901 + 1)

(5)


119879 = 1 minus 119901out (6)




119901119900out = Pr [1198770119900 lt 119877119904] 1199011199001015840out = Pr [11987701015840119900 lt 119877119904]

(8)



(9)


119879119908 = 119879119905 minus 119879RTS (10)





119890119905+1119903 = min [119890119905119903 + 119890ℎ minus (119890119905 + 119890119904 + 119890119888) 119864max] (11)

5 Simulations



obab

ility

of f

alse

alar

m

001

002

003

004

005

006

007

008

009







OverlayUnderlay


Aver

age t

hrou

ghpu

t of S

Us

0

5

10

15

20

25





erag

e wai

ting

time o

f SU

s

15

20

25

30

35

40

45

50


OverlayUnderlay



Aver

age r

esid

ual e

nerg

y of

SU

s

0

05

1

15

2

25

3

2 4 6 8 10

Simulation time




6 Conclusion


Competing Interests


Acknowledgments


References


















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




Channelsensing



CS DTEHTS




(3)




PU-T1 PU-R1

PU-TM PU-RM

Busy channel

Idle channel

Dataqueue

queueEnergy

SU-TX










Start

Data queueand

energy queuestate







Energyharvesting

model

OverlayModel

UnderlayModel

No

N

No

o

Yes

Yes

Yes

QDi = empty

QEi = Emax

QDi = empty

QEi = Emax

QDi = empty

QEi = empty

E(s) lt 120582O

120582O lt E (s) lt 120582U

E (s) gt 120582U





















0

1 00 0

1 1 12

2 23

3 34

4 4

1

2

3

4

1

2

3

4

User User

A CA B

AB

A CA B

AB

A CA B

AB

Availablechannel




channelAccess

channelLabel

1 00 02

1

2

1

1

0

1 00 0

2

1

1

CA

A CA B

AB

CA

B













1198770119900 = 119861 log2 (1 + 119892119904119875119900119904 ) 1198771119900 = 0


11987701015840119900 = 119861 log2 (1 + 119892119904119875119900119904119892119901119875119901 + 1) 11987711015840119900 = 0

(4)


1198770119906 = 119861 log2 (1 + 119892119904119875119906119904 ) 1198771119906 = 119861 log2 (1 + 119892119904119875119906119904119892119901119875119901 + 1)

(5)


119879 = 1 minus 119901out (6)




119901119900out = Pr [1198770119900 lt 119877119904] 1199011199001015840out = Pr [11987701015840119900 lt 119877119904]

(8)



(9)


119879119908 = 119879119905 minus 119879RTS (10)





119890119905+1119903 = min [119890119905119903 + 119890ℎ minus (119890119905 + 119890119904 + 119890119888) 119864max] (11)

5 Simulations



obab

ility

of f

alse

alar

m

001

002

003

004

005

006

007

008

009







OverlayUnderlay


Aver

age t

hrou

ghpu

t of S

Us

0

5

10

15

20

25





erag

e wai

ting

time o

f SU

s

15

20

25

30

35

40

45

50


OverlayUnderlay



Aver

age r

esid

ual e

nerg

y of

SU

s

0

05

1

15

2

25

3

2 4 6 8 10

Simulation time




6 Conclusion


Competing Interests


Acknowledgments


References


















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




Start

Data queueand

energy queuestate







Energyharvesting

model

OverlayModel

UnderlayModel

No

N

No

o

Yes

Yes

Yes

QDi = empty

QEi = Emax

QDi = empty

QEi = Emax

QDi = empty

QEi = empty

E(s) lt 120582O

120582O lt E (s) lt 120582U

E (s) gt 120582U





















0

1 00 0

1 1 12

2 23

3 34

4 4

1

2

3

4

1

2

3

4

User User

A CA B

AB

A CA B

AB

A CA B

AB

Availablechannel




channelAccess

channelLabel

1 00 02

1

2

1

1

0

1 00 0

2

1

1

CA

A CA B

AB

CA

B













1198770119900 = 119861 log2 (1 + 119892119904119875119900119904 ) 1198771119900 = 0


11987701015840119900 = 119861 log2 (1 + 119892119904119875119900119904119892119901119875119901 + 1) 11987711015840119900 = 0

(4)


1198770119906 = 119861 log2 (1 + 119892119904119875119906119904 ) 1198771119906 = 119861 log2 (1 + 119892119904119875119906119904119892119901119875119901 + 1)

(5)


119879 = 1 minus 119901out (6)




119901119900out = Pr [1198770119900 lt 119877119904] 1199011199001015840out = Pr [11987701015840119900 lt 119877119904]

(8)



(9)


119879119908 = 119879119905 minus 119879RTS (10)





119890119905+1119903 = min [119890119905119903 + 119890ℎ minus (119890119905 + 119890119904 + 119890119888) 119864max] (11)

5 Simulations



obab

ility

of f

alse

alar

m

001

002

003

004

005

006

007

008

009







OverlayUnderlay


Aver

age t

hrou

ghpu

t of S

Us

0

5

10

15

20

25





erag

e wai

ting

time o

f SU

s

15

20

25

30

35

40

45

50


OverlayUnderlay



Aver

age r

esid

ual e

nerg

y of

SU

s

0

05

1

15

2

25

3

2 4 6 8 10

Simulation time




6 Conclusion


Competing Interests


Acknowledgments


References


















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in















0

1 00 0

1 1 12

2 23

3 34

4 4

1

2

3

4

1

2

3

4

User User

A CA B

AB

A CA B

AB

A CA B

AB

Availablechannel




channelAccess

channelLabel

1 00 02

1

2

1

1

0

1 00 0

2

1

1

CA

A CA B

AB

CA

B













1198770119900 = 119861 log2 (1 + 119892119904119875119900119904 ) 1198771119900 = 0


11987701015840119900 = 119861 log2 (1 + 119892119904119875119900119904119892119901119875119901 + 1) 11987711015840119900 = 0

(4)


1198770119906 = 119861 log2 (1 + 119892119904119875119906119904 ) 1198771119906 = 119861 log2 (1 + 119892119904119875119906119904119892119901119875119901 + 1)

(5)


119879 = 1 minus 119901out (6)




119901119900out = Pr [1198770119900 lt 119877119904] 1199011199001015840out = Pr [11987701015840119900 lt 119877119904]

(8)



(9)


119879119908 = 119879119905 minus 119879RTS (10)





119890119905+1119903 = min [119890119905119903 + 119890ℎ minus (119890119905 + 119890119904 + 119890119888) 119864max] (11)

5 Simulations



obab

ility

of f

alse

alar

m

001

002

003

004

005

006

007

008

009







OverlayUnderlay


Aver

age t

hrou

ghpu

t of S

Us

0

5

10

15

20

25





erag

e wai

ting

time o

f SU

s

15

20

25

30

35

40

45

50


OverlayUnderlay



Aver

age r

esid

ual e

nerg

y of

SU

s

0

05

1

15

2

25

3

2 4 6 8 10

Simulation time




6 Conclusion


Competing Interests


Acknowledgments


References


















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




0

1 00 0

1 1 12

2 23

3 34

4 4

1

2

3

4

1

2

3

4

User User

A CA B

AB

A CA B

AB

A CA B

AB

Availablechannel




channelAccess

channelLabel

1 00 02

1

2

1

1

0

1 00 0

2

1

1

CA

A CA B

AB

CA

B













1198770119900 = 119861 log2 (1 + 119892119904119875119900119904 ) 1198771119900 = 0


11987701015840119900 = 119861 log2 (1 + 119892119904119875119900119904119892119901119875119901 + 1) 11987711015840119900 = 0

(4)


1198770119906 = 119861 log2 (1 + 119892119904119875119906119904 ) 1198771119906 = 119861 log2 (1 + 119892119904119875119906119904119892119901119875119901 + 1)

(5)


119879 = 1 minus 119901out (6)




119901119900out = Pr [1198770119900 lt 119877119904] 1199011199001015840out = Pr [11987701015840119900 lt 119877119904]

(8)



(9)


119879119908 = 119879119905 minus 119879RTS (10)





119890119905+1119903 = min [119890119905119903 + 119890ℎ minus (119890119905 + 119890119904 + 119890119888) 119864max] (11)

5 Simulations



obab

ility

of f

alse

alar

m

001

002

003

004

005

006

007

008

009







OverlayUnderlay


Aver

age t

hrou

ghpu

t of S

Us

0

5

10

15

20

25





erag

e wai

ting

time o

f SU

s

15

20

25

30

35

40

45

50


OverlayUnderlay



Aver

age r

esid

ual e

nerg

y of

SU

s

0

05

1

15

2

25

3

2 4 6 8 10

Simulation time




6 Conclusion


Competing Interests


Acknowledgments


References


















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




11987701015840119900 = 119861 log2 (1 + 119892119904119875119900119904119892119901119875119901 + 1) 11987711015840119900 = 0

(4)


1198770119906 = 119861 log2 (1 + 119892119904119875119906119904 ) 1198771119906 = 119861 log2 (1 + 119892119904119875119906119904119892119901119875119901 + 1)

(5)


119879 = 1 minus 119901out (6)




119901119900out = Pr [1198770119900 lt 119877119904] 1199011199001015840out = Pr [11987701015840119900 lt 119877119904]

(8)



(9)


119879119908 = 119879119905 minus 119879RTS (10)





119890119905+1119903 = min [119890119905119903 + 119890ℎ minus (119890119905 + 119890119904 + 119890119888) 119864max] (11)

5 Simulations



obab

ility

of f

alse

alar

m

001

002

003

004

005

006

007

008

009







OverlayUnderlay


Aver

age t

hrou

ghpu

t of S

Us

0

5

10

15

20

25





erag

e wai

ting

time o

f SU

s

15

20

25

30

35

40

45

50


OverlayUnderlay



Aver

age r

esid

ual e

nerg

y of

SU

s

0

05

1

15

2

25

3

2 4 6 8 10

Simulation time




6 Conclusion


Competing Interests


Acknowledgments


References


















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




obab

ility

of f

alse

alar

m

001

002

003

004

005

006

007

008

009







OverlayUnderlay


Aver

age t

hrou

ghpu

t of S

Us

0

5

10

15

20

25





erag

e wai

ting

time o

f SU

s

15

20

25

30

35

40

45

50


OverlayUnderlay



Aver

age r

esid

ual e

nerg

y of

SU

s

0

05

1

15

2

25

3

2 4 6 8 10

Simulation time




6 Conclusion


Competing Interests


Acknowledgments


References


















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




erag

e wai

ting

time o

f SU

s

15

20

25

30

35

40

45

50


OverlayUnderlay



Aver

age r

esid

ual e

nerg

y of

SU

s

0

05

1

15

2

25

3

2 4 6 8 10

Simulation time




6 Conclusion


Competing Interests


Acknowledgments


References


















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in

















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in

















Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in










Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in



Documents

Research Article Channel Selection Policy in Multi-SU and