Multi-Channel Collaborative Spectrum Sensing in Cognitive

Radio Networks

Saud Althunibat, Tung Manh Vuong and Fabrizio Granelli

University of Trento, DISI, Trento, Italy

althunibat@disi.unitn.it, manhtung.vuong@studenti.unitn.it, granelli@disi.unitn.it

AbstractCollaborative spectrum sensing (CSS) consumes asignificant amount of energy during sensing and reporting thesensing results. Such an issue becomes a challenge in the multi-channel systems. In this paper we propose two different energy-efficient CSS schemes, namely, Reduced-Energy Sensing Scheme(RESS) and Reduced-Energy Reporting Scheme (RERS). InRESS, the channels that have been identified as occupied willnot be sensed for a number of next sensing rounds, while inRERS, only the sensing results of a subset of the sensed channelswill be reported channels. In both schemes, the energy andtime resources expended in CSS will be saved. Moreover, thesaved time will be exploited in data transmission, improving thethroughput. Simulation results show a significant improvementin the energy efficiency of multi-channel CSS.

I. INTRODUCTION

In the light of the increasing demand on limited spec-

trum resources, cognitive radio (CR) has been proposed as

a smart solution for spectrum shortage problem. CR enables

an efficient usage of the licensed spectrum bands, where it

gives unlicensed users, also called cognitive users (CUs), the

capability to exploit the temporally-unused portions of the

licensed spectrum [1].

The initial necessary process of a cognitive transmission is

called spectrum sensing [2]. Spectrum sensing aims at iden-

tifying the instantaneous spectrum status in order to use the

unoccupied portions. It is greatly important to perfectly per-

form spectrum sensing as it guarantees an efficient resources

utilization and avoids collisions with the licensed users. Thus,

aiming at accurate sensing results, spectrum sensing is usually

performed in a collaborative approach, called collaborative

spectrum sensing (CSS) [3], [4], [5].

CSS implies that individual spectrum sensing results should

be reported to a common entity, called fusion center (FC). The

FC is in charge of processing the received results and making a

global decision regarding spectrum availability. Although CSS

improves the reliability of the spectrum decision by mitigating

the shadowing and multi-path fading experienced by individu-

als [6], it creates other challenges including transmission delay,

energy consumption and security threats [7]. In this work, we

focus on the energy consumption challenge in multi-channel

CSS.

CSS in multi-channel systems expends a significant portion

of time and energy resources due to the large number of

sensed channels. Thus, many works in the literature have

investigated this problem, where some energy-efficient CSS

schemes have been proposed. For example, in [8], the CUs

are divided into non-disjoint subsets such that only one subset

senses the spectrum while the other subsets enter a low power

mode. The energy minimization problem is formulated as a

network lifetime maximization problem with constraints on

the detection accuracy. Another algorithm for user selection

This work is funded by the Research Project GREENET (PITN-GA-2010-264759).

is proposed in [9], where the user subset that has the lowest

cost function and guarantees the desired detection accuracy

is selected. The cost function is related to the system energy

consumption. A distributed approach for selecting the partic-

ipating CUs is proposed in [10], where the expected energy

consumption is calculated by each CU prior to the beginning

of the CSS process: if it is lower than a given threshold,

the corresponding CU will participate; otherwise, it will not

participate. The multi-channel spectrum sensing problem is

formulated as a coalition formation game in [11]. A utility

function of each coalition takes into account both the sensing

accuracy and energy efficiency, and a distributed algorithm

is proposed to find the optimal partition that maximizes the

aggregate utility of all the coalitions in the system.

In this work, we propose two different CSS schemes,

namely, Reduced-Energy Sensing Scheme (RESS) and

Reduced-Energy Reporting Scheme (RERS). In RESS, the

channels that have been identified as occupied will not be

sensed for a number of next sensing rounds. The idea is based

on exploiting the correlation in the licensed users activity. Re-

ducing the number of channels can save the energy consumed

in sensing the non-sensed channels and the energy consumed

in reporting their sensing results. Moreover, the time dedicated

for CSS will be shorter, and hence more time will be reserved

for data transmission, improving the achievable throughput.

However, the achievable throughput might be degraded if the

number of non-sensing rounds is large. Thus, the role of the

number of non-sensing rounds has been discussed. In the

other proposed scheme, RERS, only the sensing results of a

subset of the sensed channels will be reported. Therefore, the

amount of energy consumed in reporting the sensing results

will be reduced. Similar to the RESS, the reporting time will be

shortened, and hence, the data transmission will be extended.

As a result, less energy consumption and more data throughput

can be attained. However, the number of reported channels has

an important role that has been discussed and shown in the

simulation results.

The rest of the paper is organized as follows. Section II

describes the system model, while the conventional CSS for

multi-channel systems is discussed in Section III. The two

proposed schemes are presented in Section IV. Section V

presents the simulation results of both schemes compared to

the conventional scheme, and the conclusions are drawn in

Section VI

II. SYSTEM MODEL

Consider a primary network (PN) that includes a set of pri-

mary users (PUs). The licensed spectrum of the considered PN

is divided into L identical channels. The channel occupancy

by PUs is modeled as a two-state Markov process as shown

in Fig. 1. As indicated in Fig. 1, each channel can be in one

of the two states, either busy (1) or free (0). The probabilities

shown in Fig.1 represent the transition probabilities between

the two states. For example, P01 is the transition probability

from the free state to a busy state. On the other hand, we

consider another set of N CUs that try to exploit the licensed

channels once they are unused by the PUs. To do so, CUs

sense the licensed spectrum and forward their results to the

FC in order to identify the free channels and perform their

own data transmission.

Fig. 1. The considered channel state model.

The simplest and most efficient spectrum sensing technique

is energy detection. In energy detection, the channel occupancy

is checked by measuring the contained energy for a specific

time called sensing time (ts). The local sensing results (col-

lected energy samples) are then forwarded to the FC where

they are summed and compared to a predefined threshold ().

If the summed results is larger than , the global decision of

the corresponding channel will identify it as busy. Otherwise,

the channel is identified as free.

The reliability of the global decision is evaluated by two

probabilities, namely, detection probability (PD) and false-

alarm probability (PF ). The former represents the probability

that the channel is identified as busy given that it is actually

busy, while the latter represents the probability that the channel

is identified as busy given that it is actually free. Based on the

made decisions, only the channels that have been identified as

free will be exploited by CUs. The other channels will not be

targeted in order to avoid collisions with PUs.

III. MULTI-CHANNEL COLLABORATIVE SENSING

In the conventional scheme, all channels should be sensed

and the results should be reported to the FC [12]. The

expended time and consumed energy in local sensing by all

CUs according to the conventional scheme are given as follows

TCs = Lts (1)

ECs = NLes (2)

where es is the consumed energy in sensing one channel by

one CU.

Likewise, if we denote the amount of time required to report

one sensing result regarding one channel by tr, The expended

time and consumed energy in results reporting by all CUs

according to the conventional scheme are given as follows

TCr = NLtr (3)

ECr = NLer (4)

where er is the consumed energy in reporting one sensing

result of one channel by one CU. The difference in calcu-

lating TCs and TCr is due to the assumption that CUs sense

simultaneously and report the results based on TDMA scheme.

The total energy consumption can be given as follows

ECT = NL(es + er) + L(1 P0PF P1PD)et (5)

where the last term represents the average energy consumption

in the data transmission stage, et is the energy amount con-

sumed in data transmission, and the factor (1P0PFP1PD)represents the transmission probability.

Regarding the average achievable throughput by the cogni-

tive radio network (CRN), it can be expresses in terms of the

successfully delivered data in bits as follows

DCT = LP0(1 PF )RTt (6)

where R is the average data rate, Tt is the data transmission

time, and the factor P0(1 PF ) represents the probability ofthe correct identification of the free spectrum. Notice that we

consider that the data will be successfully delivered only if

the channel is unoccupied by PUs.

IV. THE PROPOSED ENERGY-EFFICIENT SCHEMES

Clearly, increasing the number of channels increases the

energy consumption in sensing and reporting stages, as shown

in (2) and (4). However, a large number of channels helps

to improve the achievable throughput of the CRN. Therefore,

energy efficient algorithms are highly demanded in order to

limit the energy consumption in multi-channel sensing while

achieving an acceptable amount of the throughput. In the

following we propose two reduced-energy schemes for sensing

and reporting stages.

A. Reduced-Energy Sensing Scheme

Although the CUs can sense a channel simultaneously, the

sensing process still consumes a significant amount of energy

and time in the multi-channel systems. Thus, a limit on the

number of the sensed channels should be kept in order to avoid

high energy/time expenditure. However, reducing the number

of channels may degrade the achievable throughput due to the

probable missing of free channels. Thus, the selection of the

sensed channels among the whole set of the channels plays an

important role in striking a balance between the saved energy

ad the achievable throughput.

In this subsection, we exploit the statistics of the PU

activity in order to limit the number of sensed channels.

As described earlier, the PU activity is modeled as two-state

Markov process, which implies that the next channel state can

be predicted based on the current state. In other words, if a

PU was detected in a specific channel in a sensing round,

there is a high probability that the PU will keep using the

channel during the next sensing round(s). Thus, it is better

not to sense the corresponding channel during these rounds as

the result is almost known. It is noticed that the same action

could be applied if a channel was identified as free and then

the FC realized that the PU is miss-detected due to the failure

in delivering the transmitted data.

Accordingly, the proposed reduced-energy sensing scheme

is performed as follows

At the initialization, all channels should be sensed and

classified as free or busy.

Any channel that has been classified as busy will not be

sensed for B next sensing rounds.

Any channel that has been classified as free and the

transmitted data were not successfully delivered will be

reclassified as busy, and will not be sensed for B next

sensing rounds.

Any channel that has been classified as free and the trans-

mitted data were successfully delivered will be sensed in

the next sensing round.

Fig. 2 shows a flow chart describing the proposed RESS

algorithm. Clearly, as the number of sensed channels in the

proposed RESS is less than or equal to the total number of

available channels (i.e. Ls L), a reduction in the consumed

time in sensing and reporting phases will be gained. This saved

time will be dedicated to data transmission, if any. Moreover,

reducing the sensed channels will save energy as well. On the

other hand, as non-sensed channels will be marked as used,

this will increase both detection and false-alarm probabilities.

Consequently, the amount of transmitted data will decrease

and the transmit energy as well.

Fig. 2. The flow chart of the proposed RESS for each channel.

The value of B has an important role in the performance

of the proposed RESS. High values of B lead to more energy

saving in the CSS process, whereas the achievable throughput

is degraded since the probability of missing unused channels

increases as B increases. Therefore, B should be carefully

adjusted as will be carried out in the Section V.

B. Reduced-Energy Reporting Scheme

Another significant amount of energy is spent in reporting

the local sensing results to the FC. In addition, a significant

portion of the time frame is consumed in the reporting phase

since CUs report their results consecutively to the FC. Thus,

energy efficient reporting schemes are highly demanded. In

this section we propose a reduced-energy reporting scheme

(RERS) aiming at limiting the amount of the reported data

from the CUs to the FC.

The proposed RERS implies that a CU does not need to

report the local sensing results of all the sensed channels.

Instead, each CU should report only the sensing results of a

selected subset of the channels. The selection of the reported

subset should be carefully chosen in order not to degrade

the overall detection accuracy, and consequently, reduce the

overall achievable throughput. In the proposed RERS, we

allow each CU to report only Lr channels that have the lowest

sensing results among its own sensing results. Therefore, each

CU will report results for a different set of channels based on

its local sensing results.

The proposed RERS reduces only the reporting time, while

the sensing time is identical to the conventional scheme.

However, it should be noticed that the index of each channel

whose sensing result will be reported should be attached and

sent to the FC. Similar to the RESS, the saved reporting time

will be exploited in data transmission, leading to improve

throughput. On the other hand, since the number of reporting

CUs is different among channels ( N ), both detection

and false-alarm probabilities should be affected. As a result,

the achievable throughput, energy consumption and energy

efficiency will experience contrasting influences as will be

shown in Section V.

The number of reported channels per user (Lr) has a dom-

inating role in the performance of the proposed RERS. Low

values of Lr yield in a lower energy consumption accompanied

with a lower throughput, while high values of Lr lead to a less

energy saving but the throughput will be slightly affected.

V. SIMULATION RESULTS

A cognitive radio network of 20 CUs is considered. Thenumber of channels is assumed 10 identical channels. For eachchannel, P00 and P11 are considered 0.7 and 0.9, respectively.The variance of the licensed signal and the noise are assumed

0.1 and 1, respectively. The total frame length is assumed100ms. The time dedicated for sensing one channel is 1mswith a sampling frequency of 0.1GHz, while the time spent inreporting one sensing result is 0.4ms. The consumed powersduring sensing, reporting and data transmission stages are

assumed 0.1W , 0.25W and 0.25W , respectively. As theenergy is the product of time and power, for a channel, the

sensing energy is 0.1mJoule and the reporting energy is0.1mJoule. At the FC, the global decision regarding theavailability of a channel is obtained by comparing the average

of the sensing results to a predefined fusion threshold that

is assumed = 1.05. The data rate over one channel isconsidered 100Kbps.

A. The performance of the proposed RESS

In this subsection, we evaluate the proposed RESS in terms

of energy consumption, throughput and energy efficiency.

Fig. 3, Fig. 4 and Fig. 5 show the effect of the variable B

on the total energy consumption, total achievable throughput

and energy efficiency, respectively.

As presented in Fig. 3, increasing the values of B results

in decreasing the total energy consumption. This is expected

since as we increase B the number of sensed channels will

be reduced, and thus the energy consumed in sensing and

reporting will be lower. However, reducing the number of

sensed channels degrades the total achievable throughput of

the considered CRN, as shown in Fig. 4. In Fig. 4, the initial

improvement in the throughput, i.e. at B = 2 and B = 3, isdue to not sensing the channels that are almost occupied. This

interesting property of the proposed RESS gives us the ability

to decrease energy consumption and simultaneously improve

the throughput.

The energy efficiency behavior as B increases, shown in

Fig. 5, is almost improving. Such a behavior is due to the

larger reduction in the energy consumption accompanied by a

less reduction in the throughput.

The shown results motivate the need of optimizing the

variable B. High values of B may save high amount of

0 1 2 3 40.02

0.025

0.03

0.035

0.04

0.045

0.05

B

To

tal

en

erg

y c

on

su

mp

tio

n [

Jo

ule

]

Propsed RESS

Conventional CSS

Fig. 3. The total energy consumption versus B for the proposed RESS andthe conventional CSS.

0 1 2 3 42000

3000

4000

5000

6000

7000

8000

9000

10000

B

To

tal

ach

ievab

le t

hro

gh

pu

t [b

its]

Proposed RESS

Conventional CSS

Fig. 4. The total achievable throughput versus B for the proposed RESSand the conventional CSS.

energy and achieve high energy efficiency. On the other hand,

the throughput is deeply degraded at high values of B. An

option is to optimize B for throughput maximization setup in

which the optimal B is the value that achieves the highest

throughput. Another option is to find the optimal B to the

value that maximizes energy efficiency. However, to avoid the

negative effect on the throughput, a threshold can be set on

the minimum achievable throughput.

B. The performance of the proposed RERS

We consider the same simulation parameters that have been

considered in the previous subsection. Additionally, the time

and the energy consumed to report one channel-index are

assumed 50s and 50Joule , respectively. In order to reducethe number of reported indexes, we report the minimum

of L Lr and Lr. Fig. 6 plots the energy consumption

only sensing and reporting process, i.e. without including the

transmit energy, versus the number of reported channels per

CU. Clearly, the energy expenditure increases as Lr increases

until it reaches the maximum energy consumption (when

all channels are sensed and reported) which corresponds the

conventional CSS. However, taking into account the total

0 1 2 3 4

0.5

1

1.5

2

2.5

3x 10

5

B

En

erg

y E

ffic

ien

cy [

bit

/Jo

ule

]

Proposed RESS

Conventional CSS

Fig. 5. The energy efficiency versus B for the proposed RESS and theconventional CSS.

energy consumption, the conventional CSS consumes less

energy than the proposed RERS whatever the value of Lr,

as shown in Fig. 7. This is due to the fact that the saved time

from reporting process will be exploited in data transmission,

and although both reporting and transmission spend the same

power, the transmit energy should be less as it conditioned

by the transmission probability, i.e. it only exists if a data

transmission occurs.

2 4 6 8 100.022

0.024

0.026

0.028

0.03

0.032

0.034

0.036

0.038

0.04

0.042

The number of reported channels (Lr)

En

erg

y c

on

su

mp

tio

n i

n C

SS

[Jo

ule

]

Proposed RERS

Conventional CSS

Fig. 6. The energy consumption in sensing and reporting versus the numberof reported channels per CU Lr for the proposed RERS and the conventional

CSS.

Fig. 8 shows the total achievable throughput versus the

number of reported channels Lr. As the number of reported

channels increases, the saved time from reporting process will

be decreased. Thus, the transmission time will be less, and

consequently, the total throughput will be degraded as shown

in Fig. 8.

Notice that the minimum energy consumption occurs at

Lr = L (when all channels are reported), see Fig. 7, whilethe maximum throughput can be achieved at Lr = 1 (whenonly one channel per CU is reported), see Fig. 8. Therefore,

it is worth considering another comparison metric to evaluate

the performance of the proposed RERS. In Fig. 9, the energy

2 4 6 8 10

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

The number of reported channels (Lr)

To

tal

en

erg

y c

on

su

mp

tio

n [

Jo

ule

]

Proposed RERS

Conventional CSS

Fig. 7. The total energy consumption versus the number of reported channelsper CU Lr for the proposed RERS and the conventional CSS.

efficiency versus the number of reported channels per CU is

shown for the proposed RERS and the conventional CSS.

Obviously, the proposed RERS can achieve higher energy

efficiency than the conventional CSS for the whole range of

Lr. Furthermore, it is noticed that based on the results shown

in Fig. 8 and Fig. 9 both the maximum throughput and the

maximum energy efficiency occur when only once channel

per CU is reported (Lr = 1).

2 4 6 8 100

0.5

1

1.5

2

2.5x 10

4

The number of reported channels (Lr)

To

tal

ach

ievab

le t

hro

ug

hp

ut

[bit

s]

Proposed RERS

Conventional CSS

Fig. 8. The total achievable throughput versus the number of reportedchannels per CU Lr for the proposed RERS and the conventional CSS.

VI. CONCLUSIONS

In this paper, the problem of high energy consumption in

multi-channel cooperative spectrum sensing is investigated.

Two energy-efficient schemes have been proposed, namely,

reduced-energy sensing scheme and reduced-energy reporting

scheme. The proposed sensing scheme aims at reducing the

consumed energy in sensing stage by reducing the number of

sensed channels, while the proposed reporting scheme reduces

the number of reported sensing results. The effects on energy

consumption, achievable throughput and energy efficiency

have been discussed in the paper. Simulation results have

2 4 6 8 105

6

7

8

9

10

11

12x 10

4

The number of reported channels (Lr)

En

erg

y e

ffic

ien

cy [

bit

/Jo

ule

]

Proposed RERS

Conventional CSS

Fig. 9. The energy efficiency versus the number of reported channels perCU Lr for the proposed RERS and the conventional CSS.

shown considerable amount of improvement on the overall

performance compared to the conventional schemes.

REFERENCES

[1] J.Mitola and G.Q.Maguire,Cognitive radio: Making software radios morepersonal, IEEE Personal Communications, vol. 6, no. 4, pp. 13-18, 1999.

[2] A. Ghasemi and E.S. Sousa, Spectrum sensing in cognitive radionetworks: requirements, challenges and design trade-offs, IEEE Com-munications Magazine, vol. 46, no. 4, pp. 32-39, 2008.

[3] S. M. Mishra, A. Sahai and R. Brodersen, Cooperative Sensing amongCognitive Radios, IEEE ICC, Istanbul-Turkey, June 2006.

[4] I. F. Akyildiz, B. F. Lo, R. Balakrishnan, Cooperative spectrum sensingin cognitive radio networks: A survey, Physical Communication (Else-vier), vol. 4, no. 1, pp. 40-62, March 2011.

[5] F. Adelantado, A. Juan and C. Verikoukis, Adaptive Sensing UserSelection Mechanism in Cognitive Wireless Networks, IEEE Comm.Letters, vol. 14, no. 9, pp. 800-802, Sept. 2010.

[6] M. Di Renzo, L. Imbriglio, F. Graziosi and F. Santucci, CooperativeSpectrum Sensing over Correlated Log-Normal Sensing and ReportingChannels, IEEE GlobeCom, pp.1-8, 2009.

[7] S. Althunibat, V. Sucasas, H. Marques, J. Rodriguez, R. Tafazolli and F.Granelli, On the Trade-off between Security and Energy Efficiency inCooperative Spectrum Sensing for Cognitive Radio, IEEE Communica-tion Letters, vol. 17, no. 8, pp. 1564-1567, 2013.

[8] Cheng, Peng, Ruilong Deng, and Jiming Chen. Energy-efficient coop-erative spectrum sensing in sensor-aided cognitive radio networks, IEEEWireless Communications, vol. 19, no. 6, pp. 100-105, 2012.

[9] Najimi, M., et al., A Novel Sensing Nodes and Decision Node SelectionMethod for Energy Efficiency of Cooperative Spectrum Sensing inCognitive Sensor Networks, IEEE Sensors Journal, vol.13, no.5, pp.1610-1621, May 2013.

[10] S. Althunibat et al., Energy-Efficient Partial-Cooperative SpectrumSensing in Cognitive Radio over Fading Channels, IEEE VTC-Spring,2013.

[11] H. Xiaolei, et al., A coalition formation game for energy-efficientcooperative spectrum sensing in cognitive radio networks with multiplechannels. Global Telecommunications Conference (GLOBECOM 2011),2011 IEEE. IEEE, 2011.

[12] A. Mesodiakaki, F. Adelantado, L. Alonso, and C. Verik-oukis,Performance Analysis of a Cognitive Radio Contention-AwareChannel Selection Algorithm, to appear in IEEE Trans. Veh. Technol.,2014.