Energy conserving relay assistance for reporting users in ...

11
Energy conserving relay assistance for reporting users in cognitive radio networks M S SUMI * and R S GANESH Department of Electronics and Communication Engineering, Noorul Islam Centre for Higher Education, Kumaracoil, Tamil Nadu, India e-mail: [email protected]; [email protected] MS received 28 August 2020; revised 18 April 2021; accepted 15 July 2021 Abstract. In Cognitive Radio (CR) networks, due to fading and shadowing the consistency of single user sensing is deeply affected thereby degrading its detection performance. Cooperative Spectrum Sensing (CSS) paves the way for improving the reliability of sensing in CR networks. In this work, an energy conserving relay assistance method is proposed for modifying the soft fusion with improved Equal Gain Combining (EGC) method to balance the energy consumption among the reporting users. The residual energies of the reporting users are conserved in such a way that reporting is assisted by nearby non reporting users. The overall network energy consumption is also not affected by the proposed method. Also the improved EGC soft fusion method is compared and analysed with the conventional AND hard fusion method and the detection performance is observed to be improved in the improved soft fusion method. In addition to energy conservation, complexity and missed detection analysis are carried out for the proposed algorithm. MATLAB based simulations are performed to compare and justify the algorithm proposed and to estimate the energy conservation analysis. Keywords. Spectrum sensing; soft fusion; distance; signal-to-noise ratio; equal gain combining; energy. 1. Introduction The deployment of 5G is expected to happen soon [1], which will lead to enhanced connectivity in various aspects of wireless sector. As a result, there will be a huge hike in the number of devices being used. It will also result in spectrum scarcity, since more spectrum resources are required for meeting the demands of the increasing number of devices. At the same time, according to Federal Com- munications Commission (FCC), spectrum is not com- pletely utilized and remains idle at certain locations and durations [2]. This condition is referred to as spectrum under-utilization. Therefore to address the issues of spec- trum scarcity and under-utilization, an intelligent solution like CR is being suggested [3]. It is defined as a smart technology which follows the principle of Dynamic Spec- trum Access (DSA). Its transceiver can sense the sur- rounding radio frequency environment and find out the unoccupied channels and instantly move into them. Once the incumbent users or the Primary Users (PUs) arrive back at their original licensed frequency band, then the cognitive or Secondary Users (SUs) have to leave another available free band without creating any interference to the PUs. This process of DSA has been assisted by certain functionalities such as spectrum sensing, decision, sharing and mobility [4]. The purpose of spectrum sensing is to identify the unused frequency bands that can be allotted to SUs during the absence of PUs [5]. Since the comeback of PUs can occur at any time, sensing has to be performed regularly. But due to certain unfavourable conditions like fading, shadowing, receiver uncertainty etc., the performance of single user spectrum sensing is highly affected. Therefore cooperative sensing is carried out, in which more than one SU perform sensing and their local decisions are combined together to make the final decision with or without the help of a centralized Fusion Center (FC) [2]. The final decision about channel availability can be obtained by using hard or soft fusion rules. While in the case of CSS, a trade-off arises between spectrum efficiency as well as energy efficiency. That is as the number of users increase, spectrum efficiency is improved, simultaneously increasing energy consumption of the individual cooperating users [6]. An improved EGC based soft fusion method is proposed in [7] to address the trade-off existing between both efficiencies in CR net- works. The above said is modified in this work in such a way that an energy conserving relay assistance method is proposed in order to combat energy conservation of the individual SUs in the improved EGC method. Here the limited SUs with larger Signal-to-Noise Ratio (SNR) values participating in results reporting can be assisted with non- reporting relay users to conserve energy if required. The *For correspondence Sådhanå (2021)46:169 Ó Indian Academy of Sciences https://doi.org/10.1007/s12046-021-01686-1

Transcript of Energy conserving relay assistance for reporting users in ...

Energy conserving relay assistance for reporting users in cognitiveradio networks

M S SUMI* and R S GANESH

Department of Electronics and Communication Engineering, Noorul Islam Centre for Higher Education,

Kumaracoil, Tamil Nadu, India

e-mail: [email protected]; [email protected]

MS received 28 August 2020; revised 18 April 2021; accepted 15 July 2021

Abstract. In Cognitive Radio (CR) networks, due to fading and shadowing the consistency of single user

sensing is deeply affected thereby degrading its detection performance. Cooperative Spectrum Sensing (CSS)

paves the way for improving the reliability of sensing in CR networks. In this work, an energy conserving relay

assistance method is proposed for modifying the soft fusion with improved Equal Gain Combining (EGC)

method to balance the energy consumption among the reporting users. The residual energies of the reporting

users are conserved in such a way that reporting is assisted by nearby non reporting users. The overall network

energy consumption is also not affected by the proposed method. Also the improved EGC soft fusion method is

compared and analysed with the conventional AND hard fusion method and the detection performance is

observed to be improved in the improved soft fusion method. In addition to energy conservation, complexity and

missed detection analysis are carried out for the proposed algorithm. MATLAB based simulations are performed

to compare and justify the algorithm proposed and to estimate the energy conservation analysis.

Keywords. Spectrum sensing; soft fusion; distance; signal-to-noise ratio; equal gain combining; energy.

1. Introduction

The deployment of 5G is expected to happen soon [1],

which will lead to enhanced connectivity in various aspects

of wireless sector. As a result, there will be a huge hike in

the number of devices being used. It will also result in

spectrum scarcity, since more spectrum resources are

required for meeting the demands of the increasing number

of devices. At the same time, according to Federal Com-

munications Commission (FCC), spectrum is not com-

pletely utilized and remains idle at certain locations and

durations [2]. This condition is referred to as spectrum

under-utilization. Therefore to address the issues of spec-

trum scarcity and under-utilization, an intelligent solution

like CR is being suggested [3]. It is defined as a smart

technology which follows the principle of Dynamic Spec-

trum Access (DSA). Its transceiver can sense the sur-

rounding radio frequency environment and find out the

unoccupied channels and instantly move into them. Once

the incumbent users or the Primary Users (PUs) arrive back

at their original licensed frequency band, then the cognitive

or Secondary Users (SUs) have to leave another available

free band without creating any interference to the PUs. This

process of DSA has been assisted by certain functionalities

such as spectrum sensing, decision, sharing and mobility

[4]. The purpose of spectrum sensing is to identify the

unused frequency bands that can be allotted to SUs during

the absence of PUs [5]. Since the comeback of PUs can

occur at any time, sensing has to be performed regularly.

But due to certain unfavourable conditions like fading,

shadowing, receiver uncertainty etc., the performance of

single user spectrum sensing is highly affected. Therefore

cooperative sensing is carried out, in which more than one

SU perform sensing and their local decisions are combined

together to make the final decision with or without the help

of a centralized Fusion Center (FC) [2]. The final decision

about channel availability can be obtained by using hard or

soft fusion rules.

While in the case of CSS, a trade-off arises between

spectrum efficiency as well as energy efficiency. That is as

the number of users increase, spectrum efficiency is

improved, simultaneously increasing energy consumption

of the individual cooperating users [6]. An improved EGC

based soft fusion method is proposed in [7] to address the

trade-off existing between both efficiencies in CR net-

works. The above said is modified in this work in such a

way that an energy conserving relay assistance method is

proposed in order to combat energy conservation of the

individual SUs in the improved EGC method. Here the

limited SUs with larger Signal-to-Noise Ratio (SNR) values

participating in results reporting can be assisted with non-

reporting relay users to conserve energy if required. The*For correspondence

Sådhanå (2021) 46:169 � Indian Academy of Sciences

https://doi.org/10.1007/s12046-021-01686-1Sadhana(0123456789().,-volV)FT3](0123456789().,-volV)

rest of the sections are described as follows. Related works

are explained in section 2, while section 3 demonstrates the

system model for CSS with the proposed algorithm. Sec-

tion 4 explains about the significance of fusion methods in

CSS. EGC soft fusion methods are explained along with the

proposed energy conserving relay assistance algorithm in

section 5. Simulations along with energy conservation

estimation and complexity analysis are demonstrated in

section 6. Detection performance of hard and soft fusion

methods are also compared in this section. Following which

conclusions and future perceptions are given in section 7.

2. Related works

To balance the spectrum and energy efficiencies in CR

networks, certain characteristics like energy consumption

minimization, sensing parameters optimization, fusion

rules’ analysis, etc., are given importance in literature [8].

An optimal soft fusion with Neyman - Pearson criterion is

being proposed by Ma et al [9] to improve detection. Also a

two bit softened hard fusion is further proposed to address

the trade-off between complexity and detection perfor-

mance. Shen and Kwak [10] proposed a linear soft fusion

method to balance interference and spectrum usage. The

sensing users are allocated with weights based on the

received PU energies and noise power levels. Authors have

proposed a selective weighted CSS scheme in [11], where

the users with higher SNR are involved in cooperation.

Various fusion schemes are studied in [12], among which

soft fusion is observed to have better performance with an

increase in complexity. While quantized fusion provides a

trade-off between performance and complexity. Particle

Swarm Optimization algorithm with Mini Max criterion is

proposed by El-Saleh et al [13], to minimize error proba-

bility and address convergence problem in soft fusion

methods. CSS with constant PU protection and constant SU

spectrum usability scenarios for EGC are analysed in [14].

Hossain and El-Saleh [15] proposed a soft fusion method

with binary genetic algorithm for improving bandwidth

utilization and detection results with faster convergence.

An adaptive sensing method based on user’s SNR value is

proposed in [16], where EGC fusion is employed for low

SNR users; else single user sensing is being used. A

reliability based fusion method presented in [17] is found

to minimize sensing error probability and improve per-

formance than AND and OR fusion methods. Also per-

formance is further enhanced with increase in the number

of SUs as different weights are assigned to SUs. The

sensed information is simultaneously transmitted to

FC in the sensing with EGC method, proposed by Hamza

et al [18].A soft linear CSS is studied in [19] and SNR based

algorithm is proposed and compared with the EGC method

to improve performance. Heterogeneous CR networks are

investigated in [20], where the SUs are employed with

different types of detectors and their final results are

combined with hard and soft fusion rules. Also an optimal

soft combiner is proposed, which has better performance

than EGC. The SU signals are weighted based on channel

quality and the detector type employed. Verma and Singh

[21] proposed a semi soft fusion method to overcome the

trade-off between bandwidth cost and sensing performance.

Here the test statistics is reconstructed at the FC from the

received bits. To overcome the noise uncertainty problem

and improve the performance, a soft fusion rule is proposed

by Farag and Mohamed [22], where the threshold levels at

the FC are dynamic. In [23], soft fusion is addressed with a

quantization based multi-bit sensing method, where the

user transmits multi-bit quantized data to FC. As a result,

performance with reduced overhead is achieved.

EGC is studied and applied for energy detection method

as well as Covariance Absolute Value method in [24] and

its performance is observed to be improved at low SNR.

Energy detection with cyclic redundancy check is per-

formed along with soft fusion in [25] and the sensed

information is retransmitted to FC if error is detected. The

detection is repeated and further erroneous SUs are dropped

from CSS. Improved performance and better throughput is

achieved even though sensing time required and overhead

are increased. Yuan et al [26] proposed a secure soft fusion

strategy, where maximum mean discrepancy is used to

distinguish malicious users from honest users. A combined

data and decision fusion is performed in [27], where the

least faded users having lower channel estimation error are

following hard fusion while the remaining SUs perform soft

combining. CSS is performed in the presence of multiple

PU in [28], where a quantization sensing method is pro-

posed to report the status of these multiple PUs to FCs by

SUs and final decisions of all PUs are obtained.

A multi-bit fusion is employed in [8] to maximize energy

efficiency, where the sensing parameters are optimized

along with the optimization of the number of cooperating

users, quantization bits to be received at FC and global

threshold for detection. Kumar et al [29] proposed a double

threshold based hard soft combining method, enabling log

likelihood detector. Here both sensing and reporting chan-

nel SNR are considered for weight estimation. Maximiza-

tion of energy efficiency by optimizing fusion rule and

reducing the number of SUs are investigated in [30] for non

fading AWGN and frequency flat fading channels. An

improved EGC method is proposed in [7] to enhance the

detection performance in CSS, where the number of local

decisions forwarded to FC is reduced, thereby saving

reporting energy.

3. System model

A centralized CSS based CR network representing our

proposed scheme is presented in figure 1. It has N SUs

sensing the availability of PU and the local decisions are

169 Page 2 of 11 Sådhanå (2021) 46:169

reported as per the proposed algorithm to a single FC. Each

SU is assumed to have V antennas, where V � 1. Each

SU’s position is not the same and they have different SNR

(� j) with the PU channel. Each SU participates in spectrum

sensing and the local decisions of different users are com-

bined in FC using fusion rules to make the global decision

regarding channel status. In order to achieve an interference

free spectrum access, the network is assumed to follow an

interweave access method [31]. Here, the channel is uti-

lized by the SU, only when it is observed to be free. This

can be obtained by performing sensing by the SUs in order

to identify the absence of the incumbent user in its

respective channel.

4. Significance of fusion rules in cooperativespectrum sensing

Sensing with energy detection is performed by each SU to

sense the PU signal, as stated by equation (1). Test statistics

is estimated from the received signal to be compared with

the threshold value [4]. Here the transmitted input signal

from PU is assumed to be a random input signal which is a

real valued Gaussian signal. It is combined with additive

white Gaussian noise signal.

The signal received by the SU is given below.

s nð Þ ¼ g nð Þy nð Þ þ r nð Þ � � � �H1

r nð Þ � � � �Ho

�ð1Þ

where n = 1,2,3,…LHere y(n) is the received PU signal with g(n) as the

amplitude gain of the channel. While r(n) is additive whiteGaussian noise with zero mean and variance rr

2. L is the

total number of samples at the received signal. The pres-

ence and absence of PU signal is indicated by hypotheses

H1 and Ho respectively [4]. The test statistics is calculated

for every SU as in equation (2).

X ¼ 1

L

XLn¼1

s nð Þj j2 ð2Þ

In the case of hard fusion based CSS, the test statistics of

individual SUs are compared with the determined value of

threshold and if found greater than that threshold, the

channel is stated as busy [32]. A single bit data indicating

the channel status is reported to FC, where all received

decisions are combined using any one of the methods of

hard fusion to determine the global decision. Whereas in

soft fusion, the estimated test statistics at each user are

forwarded to the FC, where they are combined as per the

soft fusion method and compared with their respective

threshold estimated in order to determine the global status

of the channel [12].

Some of the most popularly used hard fusion techniques

are AND, OR, Majority, Chair-Varshney methods etc.

While EGC, Square Law Selection (SLS), Maximum Ratio

Combining (MRC), Selection Combining (SC) etc., are

some of the popular soft fusion methods employed [12, 21].

Comparing hard and soft fusion methods, soft fusion

affords better detection performance. But at the same time,

it requires more overhead complexity and bandwidth as it

has to transmit the entire test statistics or energy signal

instead of forwarding a single bit of data to FC via the

control channel [32]. Equation (3) given below describes

the PU status of AND hard fusion method, where Bj is the

single bit decision of jth SU. If the SUs observe the PU to be

present, then the channel is stated as busy [12].

Figure 1. System model for centralized CSS with energy conserving relay assistance method.

Sådhanå (2021) 46:169 Page 3 of 11 169

ifPNj¼1

Bj ¼ N;H1

else;Ho

8<: ð3Þ

Also, the global detection and false alarm probabilities

of AND fusion [32] are listed by equations (4) and (5).

Qd ¼YNj¼1

Pdj ð4Þ

Qfa ¼YNj¼1

Pfj ð5Þ

where Pdj and Pfj are the probabilities of detection and

false alarm for the jth SU. Also the global missed detection

probability is expressed as in equation (6).

Qmd ¼ 1� Qd ð6Þ

5. Soft fusion to improve detection performanceand energy efficiency

Among the various soft fusion methods, EGC is one of the

simplest methods, where the test statistics from individual

sensing users are transmitted to the FC. At the FC, these

signals are combined with equal weights and the combined

value obtained is the global test statistics. This value will be

compared with the final threshold to determine the PU

status [23]. The global test statistics is calculated as men-

tioned in equation (7).

Eegc ¼XNj¼1

Ej ð7Þ

Two important parameters justifying the performance

of sensing in CR networks are detection and false alarm

probabilities [33]. Lower the value of false alarm prob-

ability, higher the possibility for spectrum utilization.

But, it can also rise the number of missed detections

during which the sensing user could not detect a PU

using the channel, leading to collision and thereby

interference to both users. The global detection and false

alarm probabilities for EGC are stated as in equations (8)

and (9)

Qd ¼Tegc �

PNj¼1 1þ � j

� �r2r

� �ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

2

LPN

j¼11þ� jð Þr4r

!vuut0@

1A

0BBBBBB@

1CCCCCCA

ð8Þ

Qfa ¼Tegc �

PNj¼1 r

2r

� �ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

2

LPN

j¼1r4r

r !0BBBB@

1CCCCA ð9Þ

where � j is the SNR of jth SU, rr2 is the noise variance and

Tegc given by equation (10) is the estimated global thresh-

old and Ej is the energy at the jth user [7, 23].

Tegc ¼ Q�1 Qfa

� � ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2

LPN

j¼1 r4r

s ! !þXNj¼1

r2r ð10Þ

Here Q(.) and Q-1(.) are defined as the standard Q and

inverse Q functions respectively, where Q xð Þ ¼1ffiffiffiffiffiffiffi2pð Þ

p R1x e

�t2

2

� �dt [34]. In MRC type of soft fusion, the

energy value at each user will be provided with an indi-

vidual weight corresponding to their SNR value. These

weighted signals are combined at the FC and compared

with the relative threshold value to determine the PU status

[12]. Although better detection is observed in MRC

method, it has certain drawbacks like higher complexity,

need of larger bandwidth for reporting as well as prior

knowledge of channel state information [18], which makes

EGC method more adaptable.

5.1 Improved EGC soft fusion rule

In order to overcome the trade-off between detection per-

formance and bandwidth conservation, an improved EGC

method is proposed in [7]. The test statistics of an indi-

vidual sensing user is forwarded to FC only if its SNR is

found to be higher than the average SNR of the network.

Better detection probability is obtained, as sensing results

of low SNR SUs are not reported to be combined to make

the global decision. As the number of reporting users is

reduced, reporting energy can also be saved. The average

value of SNR of all sensing users in the CR network is

calculated as mentioned in equation (11).

SNRavg ¼ 1

N

XNj¼1

� j ð11Þ

The global test statistics Eimpegc is estimated by com-

bining the received high SNR test statistics at FC. It is

compared with the threshold Timpegc to determine the

availability of PU in the channel. Equations (12) and (13)

state the expressions for estimating Eimpegc and Timpegc.

Eimpegc ¼XMj¼1

Ej ð12Þ

169 Page 4 of 11 Sådhanå (2021) 46:169

Timpegc ¼ Q�1 Qfa

� � ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2

LPM

j¼1 r4r

s ! !þXMj¼1

r2r ð13Þ

M is the number of high SNR test statistics reported to

FC, where M B N.

5.2 Proposed energy conserving relay assistancealgorithm for improved EGC method

In improved EGC method [7], the SUs with high SNR

value alone participate in reporting test statistics to FC.

Even though energy conservation is achieved with

respect to the entire CR network, individual users which

are involved in reporting to FC, face energy consumption

imbalance. Hence to overcome this issue, relay users can

be considered. The non-reporting SUs present in the CR

network, which are both nearer to the FC as well as to the

reporting users that are far away from the FC are allowed

to act as supporting relay users. An alternative reporting

path is established between the reporting SUs and FC

through the selected relay SUs. The improved EGC

algorithm as in [7], is further modified using Algorithm 1

with this energy conserving relay assistance method. As

in Algorithm 1, the long distance reporting users which

require relay support are determined based on the aver-

age distance value of reporting users from FC. The relay

users are predicted with respect to their distance from

FC. The long distance reporting users are matched with

the relay users to find out the relevant one, which can

report the test statistics to the FC with minimum energy

consumption. Hence energy consumption of the reporting

Figure 2. Flowchart representing energy conserving relay assistance algorithm.

Sådhanå (2021) 46:169 Page 5 of 11 169

process is shared among the reporting and the relay users,

thereby securing the reporting SU from running out of

energy. The flowchart in figure 2 explains the process in

Algorithm 1.

Algorithm 1. Energy conserving relay assistance

method

1. Determine the average SNR of sensing users SNRavg

2. Compare the single user SNR with SNRavg

3. Identify the reporting and non-reporting SUs

4. Determine distance of each reporting user from FC

RFCj

5. Find the mean distance between reporting users and FC

as RFCavg

6. Identify the long distance reporting users R by compar-

ing RFCj with RFCavg

7. Determine distance of each non-reporting user from FC

NFCj

8. Find the mean distance between non-reporting users and

FC as NFCavg

9. Identify the suitable relay users RL by comparing NFCj

with NFCavg

10. Match R and RL and identify the appropriate RL for each

R to minimize energy consumption

5.3 Energy consumption analysis

One of the important aspects paving the way for better

network characteristics in CR networks is obtaining

energy efficiency. It is the ratio of network throughput to

total energy consumption [30]. It is measured in bits/

joule. Hence reducing energy consumption is much

essential not only for improving the performance of the

entire network, but also for the survival of the individual

sensing users. Energy consumption includes the total

energy spend on sensing, results reporting and data

transmission. Also energy consumption increases with

the increase in the number of cooperating users. In the

method of improved EGC [7], since the number of

reporting users are reduced, corresponding reporting

energy can be saved. But there is an issue of the reporting

users to drain out of energy, if their respective residual

energies are low with respect to their reporting distances

from FC. Reporting energy of jth SU [35] is given as in

equation (14).

Erj ¼ Tu � Prfcj ð14ÞTF ¼ Ts þ Tr þ Tt ð15Þ

where Tu is the reporting time for a single user as per the

time frame (TF), stated by equation (15). Also, the total

reporting duration Tr is N times Tu. The total time frame is

comprised of the sensing and reporting energies along with

the transmission slot (Tt). Prfcj is the reporting power of jth

SU to FC. The reporting energy consumed by the alternate

path with relay for jth user is given by equation (16).

Errj ¼ Tu � Prlj

� �þ Tu � Plfcj

� � ð16Þwhere Prlj is the reporting power of jth reporting user

which requires relay assistance to its respective relay

user and Plfcj is the reporting power of the corresponding

relay user to the FC. The reporting power of a SU is a

function of distance to be transmitted [36] and is pro-

portional to the reporting distance. It is defined as in

equation (17).

Pr ¼ G � Rd ð17Þwhere Pr is the reporting power of any user and Rd is its

corresponding reporting distance.

6. Simulation results

In this section, the simulation of the proposed energy

conserving relay assistance method is performed and

energy consumption analysis is carried out along with

complexity analysis. Also, the improved EGC fusion

method is simulated along with AND fusion method to

compare the detection performance of hard and soft fusion

methods. Missed detections are analysed for different false

alarm values.

6.1 Simulation of energy conserving relayassistance method

Energy detection based CSS is employed with 1000 sam-

ples for each SU. Simulations are performed in MATLAB

with 1000 Monte Carlo runs. The analysis is carried out for

varying number of SUs and a single PU and FC. These

users are assumed to be employed with single antenna each.

Figure 3 analyses the Receiver Operating Characteristics

(ROC), plotting the detection performance against false

alarm probability comparing the conventional and

improved EGC methods as in [7] for 20 SUs at SNR

ranging between -30 and -11 dB.

Relay assistance is provided to the improved EGC

method [7] based on the simulation of energy conserving

relay assistance method (Algorithm 1). Table 1 gives an

overview of reporting and relaying users with respect to the

simulation of algorithm for 20 SUs at SNR range of -30 to -

11 dB.

From table 1, it is observed that out of 20 sensing users, 7

SUs will be capable of reporting to FC. Among which 4

SUs are found to be far away from the FC and require relay

assistance to conserve their respective reporting energies.

Among the 13 non reporting users, 8 SUs are observed to

be capable of assisting as relay users. The relay user

169 Page 6 of 11 Sådhanå (2021) 46:169

suitable for the corresponding reporting user is estimated as

in Algorithm 1 and is given in table 2.

As in table 2, an alternate reporting path can be estab-

lished for those 4 reporting users which are found to be far

away from the FC, to report their decisions to the FC.

Thereby with the help of relay users, these reporting users

are able to conserve their reporting energies. The reporting

duration for each user in this soft fusion is assumed as

0.0008 [35]. The reporting power of individual users and

relay users, which depends on their respective reporting

distances are determined as in equation (17), where G is

assumed to be 0.001. Energies consumed by reporting users

with and without relay assistance are estimated using

equations (16) and (14) and the obtained values are repre-

sented in table 4. It also analyses the reporting energy

consumed by the network by implementing the energy

conserving relay assistance method. The reporting power

estimated is displayed in table 3, where equation (17) is

employed.

From the above tables, it is observed that, the reporting

users are benefitted in such a way that the energies con-

sumed by them for reporting decisions are being reduced.

Their respective residual energies are being conserved with

the assistance of relay users, who spent a little amount of

their residual energies to forward the decisions to the FC.

Also it has been observed that, the increase in total energy

consumed for forwarding results by these users via the

alternate relay path are too small and negligible, thereby

maintaining overall network energy usage.

6.2 Complexity analysis

Even though the soft fusion methods are noted to have

better detection performance than hard fusion methods,

they have larger overhead and increased complexity. This is

Figure 3. ROC representing detection performance of improved EGC and EGC methods [7]

Table 1. Available number of reporting and relaying users.

No. of

sensing

users

No. of reporting

users as per

improved EGC

algorithm [7]

No. of reporting

users requiring

relay assistance

No. of non-

reporting users

capable of

relaying

20 7 4 8

Table 2. Reporting user - relay user pair.

Reporting user R Relaying user RL

1 6

2 3

3 1

4 2

Table 3. Reporting power estimation (in milliwatts).

Reporting

user id

Reporting power

of user without

relay assistance

Reporting

power of user

with relay

assistance

Reporting

power of

corresponding

relay user

1 208.325 81.7125 128.3125

2 198.9875 91.3375 107.6625

3 196.6125 93.0375 103.5875

4 186.0875 80 106.1

Sådhanå (2021) 46:169 Page 7 of 11 169

because the SUs are forwarding their entire observed

energies to FC. In improved EGC methods [7], the number

of reporting users is reduced, thereby resulting in increased

energy conservation and bandwidth efficiency. In this

extended work, as the reporting distances of individual

users are minimized, their reporting overhead complexity is

addressed. The reporting distances of users for two condi-

tions, that is without and with relay assistance are estimated

and displayed in table 5.

From table 5, it is observed that the reporting distances

of users are reduced once relay support is being

employed. As a result their individual reporting com-

plexity is minimized.

6.3 Comparison of improved EGC methodwith AND fusion method

Improved EGC and conventional EGC methods are com-

pared with AND fusion method and their simulation results

are illustrated in figure 4, for 15 users at SNR range of -25

to -11 dB. From the results obtained it is observed that, the

soft fusion methods have higher detection performance than

AND fusion method. Table 6 demonstrates the comparison

of detection performance of hard and soft fusion methods

for different number of SUs at different SNR ranges.

From the results obtained, it is observed that both hard

and soft fusion methods have improved detection perfor-

mance at higher SNR ranges. Also, soft fusion has better

performance for different number of users and under dif-

ferent SNR conditions.

Table 4. Energy consumption analysis (in millijoules).

Reporting

user id

Energy consumed for user

without relay assistance

Energy consumed for user

with relay assistance

Energy consumed by

corresponding relay user

Total energy consumed by

alternate relay path

1 0.16666 0.06537 0.10265 0.168024

2 0.15919 0.07307 0.08613 0.159205

3 0.15729 0.07443 0.08287 0.15731

4 0.14887 0.06400 0.08488 0.14889

Table 5. Reporting distance estimation (in metres).

Reporting

user id

Reporting

distance of user

without relay

assistance

Reporting

distance of

user with relay

assistance

Reporting

distance of

corresponding

relay user

1 208325 81712.5 128312.5

2 198987.5 91337.5 107662.5

3 196612.5 93037.5 103587.5

4 186087.5 80000 106100

Figure 4. Comparison of detection performance of improved EGC and conventional EGC with AND fusion methods.

169 Page 8 of 11 Sådhanå (2021) 46:169

6.4 Missed detections and interferencemanagement in soft fusion methods.

From the above figures and tables, it is understood that at

lower values of false alarm probability, the detection per-

formance is observed to be lower. This also increases the

number of missed detections. When the PU channel is

falsely detected to be free and is utilized by a SU for data

transmission, then collisions take place, creating interfer-

ence between both secondary and incumbent users. When

compared with the hard fusion and conventional EGC

methods, improved EGC method has lower missed detec-

tions even at lower values of false alarm probability. Also

these missed detections are lower at higher SNR conditions.

Figure 5 illustrates the comparison of the missed detections

for the hard and soft fusion methods. Here the number of

missed detections is observed to increase with the decrease

in the values of false alarm. Also, the improved EGC

method has reduced missed detections than the conven-

tional methods. As the users satisfying the SNR criteria

alone forward their sensed information, inappropriate

detections are avoided, reducing collisions and interference

between secondary and incumbent users.

7. Conclusion

This work proposes an energy conserving relay assistance

method for modifying the existing improved EGC method,

in such a way that the energy consumed for reporting

decisions to FC is shared by both the reporting and non-

reporting users. As a result, the residual energy of the

reporting SUs are saved from getting drained out without

affecting the overall energy consumption of the network.

Reduction in energy consumption as well as reporting

distances paves the way for reduced complexity for the

individual reporting users. Also, improved EGC soft fusion

method is observed to have increased detection perfor-

mance than hard fusion methods for varying SNR condi-

tions. When compared with the hard fusion and

conventional EGC methods, improved EGC method has

lower missed detections even at lower values of false alarm

Table 6. Comparison of detection performance of hard and soft fusion methods.

No.of SUs SNR range (dB) AND fusion method (Qd) Improved EGC method (Qd)

10 (- 30 to - 21) 0.0000 0.0000 0.0000 0.0000 0.0170 0.0690 0.1330 0.2170

0.0000 0.0002 0.0015 0.0066 0.3360 0.4560 0.5960 0.7380

0.0595 1.0000 0.8670 1.0000

10 (- 20 to - 11) 0.0000 0.0000 0.0001 0.0008 0.5910 0.7720 0.8990 0.9440

0.0034 0.0138 0.0431 0.1217 0.9670 0.9770 0.9940 0.9970

0.3058 1.0000 1.0000 1.0000

20 (- 30 to - 11) 0.0000 0.0000 0.0000 0.0000 0.6070 0.7900 0.9010 0.9560

0.0000 0.0000 0.0001 0.0009 0.9700 0.9890 0.9940 0.9990

0.0171 1.0000 1.0000 1.0000

Figure 5. Comparisons of missed detection probabilities of improved EGC and conventional EGC with AND fusion methods.

Sådhanå (2021) 46:169 Page 9 of 11 169

probability. Also, these missed detections are lower at

higher SNR conditions. As a part of future enhancement,

this relay assistance work can be further modified in such a

way that the entire network energy consumption as well as

power required are further reduced, bringing energy effi-

ciency to the complete CR network. Also, the proposed

energy conserving soft fusion method can be analysed for

channels with adverse fading conditions and employing

multiple receiver antennas at the SUs. The reporting

channel condition of the relay users can also be explored to

reduce further missed detections and interference to users.

List of symbolsBj Single bit decision of jth SUEegc Global test statistics of EGC fusion

Eimpegc Global test statistics for improved EGC

Ej Local test statistics of jth SUErj Reporting energy of jth SU to FC

Errj Reporting energy of R-RL path to FC

G Constant

Ho Binary hypothesis denoting PU is absent

H1 Binary hypothesis denoting PU is present

L Number of samples in the PU signal received

M Number of results forwarded to FC in improved

EGC method

N Number of cooperating SUs

NFCavg Mean distance between non-reporting users and

FC

NFCj Distance of jth non-reporting user from FC

Pdj Probability of detection of jth SUPfj Probability of false alarm of jth SUPlfcj Reporting power of jth relay to FC

Pr Reporting power for any user

Prfcj Reporting power of jth SU to FC

Prlj Reporting power of jth reporting user to relay

Q(.) Standard Q function

Q-1(.) Inverse Q function

Qd Global detection probability

Qfa Global false alarm probability

Qmd Global missed detection probability

R Long distance reporting users requiring relay

assistance

Rd Reporting distance for any user

RFCavg Mean distance between reporting users and FC

RFCj Distance of jth reporting user from FC

RL Relay users

SNRavg Entire CR network’s average SNR

SUj jth SUTegc Global threshold for EGC fusion

TF Total time frame

Timpegc Final threshold for improved EGC

Tr Reporting slot

Ts Sensing duration

Tt Data transmission slot

Tu Reporting duration of a single user

V No. of antennas in each SU transceiver

X Test statistics to be compared with threshold

� j SNR of jth SUrr

2 Variance of noise

References

[1] Bhandari S and Joshi S 2018 Cognitive radio technology in

5G wireless communications. In: Proceedings of the 2ndIEEE International Conference on Power Electronics, Intel-ligent Control and Energy Systems (ICPEICES),pp. 1115–1120

[2] Muchandi N and Khanai R 2016 Cognitive radio spectrum

sensing: a survey. In: Proceedings of the InternationalConference on Electrical, Electronics, and OptimizationTechniques (ICEEOT), pp. 3233–3237

[3] Song M, Xin C, Zhao Y and Cheng X 2012 Dynamic

spectrum access: from cognitive radio to network radio.

IEEE Wireless Commun. 19: 23–29.[4] Alom M Z, Godder T K and Morshed M N 2015 A survey of

spectrum sensing techniques in cognitive radio network. In:

Proceedings of the International Conference on Advances inElectrical Engineering (ICAEE), pp. 161–164

[5] Ali A and Hamouda W 2017 Advances on spectrum sensing

for cognitive radio networks: theory and applications. IEEECommunications Surveys & Tutorials 19: 1277–1304

[6] Awasthi M, Kumar V and Nigam M J 2017 Energy—

efficiency techniques in cooperative spectrum sensing: a

survey. In: Proceedings of the 3rd International Conferenceon Computational Intelligence & Communication Technol-ogy (CICT), pp. 1–6

[7] Sumi M S and Ganesh R S 2019 Improved EGC method for

increasing detection in cognitive radio networks. Comput.Commun. 147: 127–137

[8] Wu H, Zhang T, Chen Y and Liu Y 2019 Multi-bit fusion

based energy-efficient collaborative spectrum sensing for

cognitive radio network. In: Proceedings of the 19th IEEEInternational Conference on Communication Technology(ICCT), pp. 776–780

[9] Ma J, Zhao G and Li Y 2008 Soft combination and detection

for cooperative spectrum sensing in cognitive radio net-

works. IEEE Trans. Wireless Commun. 7: 4502–4507[10] Shen B and Kwak K S 2009 Soft combination schemes for

cooperative spectrum sensing in cognitive radio networks.

ETRI journal 31: 263–270[11] Wu S W, Zhu J K, Qiu L and Zhao M 2010 SNR-based

weighted cooperative spectrum sensing in cognitive radio

networks. J. China Univ. Posts Telecommun. 17: 1–7

[12] Teguig D, Scheers B and Le Nir V 2012 Data fusion schemes

for cooperative spectrum sensing in cognitive radio net-

works. In: Proceedings of the Military Communications andInformation Systems Conference (MCC), pp. 1–7

[13] El-SalehAA, IsmailM,AkbariM,ManeshMRandZavareh S

A R T 2012 Minimizing the detection error of cognitive radio

networks using particle swarm optimization. In: Proceedings

169 Page 10 of 11 Sådhanå (2021) 46:169

of the International Conference on Computer and Communi-cation Engineering (ICCCE), pp. 877–881

[14] Teguig D, Scheers B and Le Nir V 2013 Throughput

optimization for cooperative spectrum sensing in cognitive

radio networks. In: Proceedings of the 7th InternationalConference on Next Generation Mobile Apps, Services andTechnologies, pp. 237–243

[15] Hossain M and El-Saleh A A 2013 Cognitive radio engine

model utilizing soft fusion based genetic algorithm for

cooperative spectrum optimization. Int. J. Comput. Netw.Commun. IJCNC 5: 23–36

[16] Kaviarasu A and Devapriya S 2014 SNR based adaptive

spectrum sensing in cognitive radio networks. Int. J. Eng.Res. Technol. IJERT 3: 1438–1442

[17] Khalid L and Anpalagan A 2014 Reliability-based decision

fusion scheme for cooperative spectrum sensing. IET Com-mun. 8: 2423–2432

[18] Hamza D, Aı̈ssa S and Aniba G 2014 Equal gain combining

for cooperative spectrum sensing in cognitive radio net-

works. IEEE Trans. Wireless Commun. 13: 4334–4345[19] Guo J, Gu Y and Jing D H 2014 An improved SNR-based

cooperative spectrum sensing in cognitive radio networks.

Appl. Mech. Mater. Trans Tech Publ. Ltd 631–632: 874–877

[20] Ejaz W, Hattab G, Cherif N, Ibnkahla M, Abdelkefi F and

Siala M 2018 Cooperative spectrum sensing with heteroge-

neous devices: hard combining versus soft combining. IEEESyst. J. 12: 981–992

[21] Verma P and Singh B 2017 On the decision fusion for

cooperative spectrum sensing in cognitive radio networks.

Wireless Netw. 23: 2253–2262[22] Farag H M and Mohamed E M 2017 Soft decision

cooperative spectrum sensing with noise uncertainty reduc-

tion. Pervasive and Mobile Computing 35: 146–164

[23] Fu Y, He Z and Yang F 2017 A simple quantization-based

multibit cooperative spectrum sensing for cognitive radio

networks. In: Proceedings of the 14th International Com-puter Conference on Wavelet Active Media Technology andInformation Processing (ICCWAMTIP), pp. 220–223

[24] Apurva K and Lakshmi P D 2018 To improve the probability

of detection in spectrum sensing by using equal gain

combining technique. Int. J. Comput. IJC 29: 99–106

[25] Bhatti D M S, Ahmed S, Saeed N and Shaikh B 2018

Efficient error detection in soft data fusion for cooperative

spectrum sensing. AEU-Int. J. Electr. Commun. 88: 141–147

[26] Yuan S, Li L and Chigan C 2018 Maximum mean

discrepancy based secure fusion strategy for robust

cooperative spectrum sensing. In: Proceedings ofthe IEEE International Conference on Communications(ICC), pp. 1–6

[27] Verma G, Dhage V and Chauhan S S 2018 Analysis of

combined data-decision fusion scheme for cognitive

radio networks. In: Proceedings of the 2nd InternationalConference on Inventive Systems and Control (ICISC),pp. 1324–1327

[28] Shah H A, Kwak K S, Sengoku M and Shinoda S 2019

Reliable cooperative spectrum sensing through multi-bit

quantization with presence of multiple primary users in

cognitive radio networks. In: Proceedings of the 34thInternational Technical Conference on Circuits/Systems,Computers and Communications (ITC-CSCC), pp. 1–2

[29] Kumar A, Saha S and Tiwari K 2019 A double threshold-

based cooperative spectrum sensing with novel hard-soft

combining over fading channels. IEEE Wireless Commun.Lett. 8: 1154–1158

[30] Awasthi M, Nigam M J and Kumar V 2020 Energy

efficiency maximization by optimal fusion rule in fre-

quency-flat-fading environment. AEU-International Journalof Electronics and Communications 113: p.152965

[31] Kusaladharma S and Tellambura C 2017 An overview of

cognitive radio networks. Wiley Encyclopedia of Electricaland Electronics Engineering, pp.1–17

[32] Fu Y, Yang F and He Z 2018 A quantization-based multibit

data fusion scheme for cooperative spectrum sensing in

cognitive radio networks. Sensors 18: 473.[33] Sumi M S and Ganesh R S 2017 Performance enhancing

techniques in cognitive radio networks. In: Proceedings ofthe IEEE International Conference on Circuits and Systems(ICCS), pp. 172–178

[34] Liang Y C, Zeng Y, Peh E C Y and Hoang A T 2008

Sensing-throughput tradeoff for cognitive radio networks.

IEEE Trans. Wireless Commun. 7: 1326–1337[35] Althunibat S and Granelli F 2014 Energy efficiency

analysis of soft and hard cooperative spectrum sensing

schemes in cognitive radio networks. In: Proceedings ofthe IEEE 79th Vehicular Technology Conference (VTCSpring), pp. 1–5

[36] Ansari N and Han T 2017 Green Mobile Networks: ANetworking Perspective. Wiley, New York

Sådhanå (2021) 46:169 Page 11 of 11 169