Mobile Networks and Applications

https://doi.org/10.1007/s11036-018-1045-8

Group Based Multi-Channel Synchronized Spectrum Sensing

in Cognitive Radio Network with 5G
R. Giri Prasad 1 & P. Venkatesan 1

# Springer Science+Business Media, LLC, part of Springer Nature 2018

Abstract
This paper investigates the problem of Cognitive Radio Network (CRN) with Cooperative Spectrum Sensing (CSS), when
multiple idle channels are available. In this work CRN-CSS is modeled to resolve the problems of sensing, grouping and decision
making. To enlarge network connectivity and provide larger coverage for users, we integrate CRN-CSS with 5G. Trinary
partitioning is performed to group user and perform sensing in cooperative manner. Sensing of multiple channels leads to
interference that is overwhelmed by the novel Dynamic Multi-Channel Slot Allocation (DMCSA) algorithm which allocates
channel effectively. To address the challenges of spectrum decision, we have presented a special entity (i.e.) Spectrum Agent
which is deployed to perform only spectrum sensing and report to fusion center. Fusion center is responsible for decision making
and spectrum allocation, for global decision fusion center constructs a graph based on the reports obtained from secondary users
and spectrum agent. These reports are compared for making final decision about spectrum. On the whole we describe with a
detailed architecture of the proposed CRN-CSS model with seamless integration of 5G cellular networks that achieves higher
throughput efficiencies. The obtained simulation results demonstrate the proposed CRN-CSS model with 5G is a remarkable
cellular network design to improve throughput, detection probability, delay and sensing overhead.

Keywords Cognitive radio . Cooperative spectrum sensing . Throughput . Multi-channel sensing

1 Introduction sensing scheme reduces fading effects and improves detection

[4]. This scheme uses a predefined threshold for decision mak-
Growth of advanced technologies in wireless communication ing about the presence of licensed signals. Energy detection in
has tremendously increased the utilization of spectrum re- OR rule based CSS, the frame structure is splitted with multi-
sources, this leads to spectrum scarcity problem [1]. This ple transmission time slots and a single sensing time slot [3].
was overwhelmed by CRN, in which cognitive Radio (CR) In this work the CRs selects the channels that have higher
is used. In CRN, Cooperative Spectrum Sensing involves with Signal to Noise Ratio (SNR). In CSS, multiple CR (i.e.)
key functions of sensing, reporting and information fusion [2]. Secondary Users (SUs) participate in sensing, hence interfer-
Spectrum sensing approaches include energy detection, fea- ence is a significant metric to be noted [5, 9]. In [5], a weight-
ture detection, pilot based detection and covariance based de- ed cooperative bandwidth spectrum sensing method is pro-
tection [3–6]. CSS radios also uses different fusion rules as posed to minimize interference by optimizing the sensing
AND logic, OR logic and K-out-N logic [3, 6–8]. Spectrum bandwidth. In rule based multichannel CSS, threshold optimi-
sensing deals with two significant parameters as probability of zation methods is proposed as Non-Restrained Multichannel
detection and probability of false alarm. Energy detection Threshold Optimization (NRMTO) and Restrained MTO
(RMTO) [7]. Threshold optimization is focused in CSS to
minimize total error detection probability. This optimization
* R. Giri Prasad is formulated by comparing obtained energy statistics with the
[email protected]
licensed sub-channel thresholds. For effective performance of
P. Venkatesan CSS in CRN, many authors have designed with clustering
[email protected] algorithms. Clustering in CRN for CSS is one of the intelligent
procedure held to minimize sensing overhead, interference
1
Department of ECE, SCSVMV University, Kanchipuram, Tamil avoidance and also to enrich sensing performances [9–12].
Nadu, India
 Mobile Netw Appl

Cluster based CSS framework in [12], is built with two-phase SUs and SA for evaluating the final decision and allocate
reporting protocol in which the sensing reports are sent to spectrum to SUs.
cluster heads and then to fusion center. Then authors Yan
Jiao, Peitong Yin, Inwhee Joe et al. proposed a clustering Then, rest of this paper is organized as follows: Section 2
scheme (i.e.) comprised with pruning, selection and clustering illustrates with previous research works held on CRN-CSS,
[10]. The Secondary User (SU) providing with reliable sens- Section 3 formulates the problems defined in prior research.
ing data is assigned as cluster head. Here cluster is re- Section 4 elaborates our entire proposed CRN-CSS archi-
constructed with respect to the correlation of sensing results tecture with novel algorithms, then Section 5 supports the
and not with changes in network topology. To optimize the evaluation of performances obtained for our proposed work
cluster head selection, an optimized cluster head rotation and and finally Section 6 concludes our overall proposed work and
selection scheme is proposed [13]. In this scheme a cluster extends it with future work.
head and a super node is selected. Super node is considered
to be a fusion center. To avoid interference, two clusters are
constructed and SUs in each cluster perform sensing and esti- 2 Literature review
mate Signal to Noise Ratio (SNR) values [9]. Sensing is per-
formed individually by two clusters and the results of SNR In this section a detailed discussion is undergone on several
value are compared by heads and maximum SNR value is works focused in CRN-CSS. Sensing in CRN follow certain
chosen. The works on clustering based CSS, consume higher predefined algorithms as mentioned in previous section. To
time for cluster formation and selection of cluster head. enhance conventional energy detector sensing algorithm, an
Grouping of CRs is recently followed for CSS in CRN; Improved Energy Detector (IED) was designed which
grouping does not require any headers to monitor members. equipped the cognitive users with mini-slots for performing
Grouping was focused by many researchers in [11, 14, 15]. spectrum sensing [6]. Spectrum sensing by fusion rules with
SUs are grouped based on their reliability [15], reporting based Two Rayleigh faded channels (S-channel and R-channel) was
grouping [11] and more. Historical sensing information of SU’s involved. Majority / Maximal Ratio Combining fusion rules
is utilized for grouping which groups up reliable nodes [15]. The were used for local decisions. Here FC is responsible for the
reliable nodes in group are assured to provide accurate results in final decision by considering two fusion rules (i.e.) whether
sensing. Since the unreliable nodes are rejected, it is assumed primary users are present or absent, whereas the local deci-
that the sensing results are accurate. In [15], for reducing sions of sensing are predicted by using OR logic. This paper
reporting overhead, the grouping is multi-bit CSS scheme. investigated two fusion rules as mentioned earlier, on compar-
The reporting slot is modeled with mini-slots, after reporting ison Maximal Ratio combining performs better than Majority
the final decision in Fusion Center (FC) is based on number of fusion rule. Mini-slots are not only defined for sensing, it was
counts of obtained sensing results. This paper work focuses on also defined for reporting phase in [11]. A group based multi-
CRN-CSS by grouping of SUs for sensing multiple channels bit CSS scheme was proposed to minimize reporting over-
and evaluating its performance against previous works. head. As per this scheme the SUs send report in one-bit in-
The major contributions of this work are summarized as stead of multi-bit information. Reporting phase was com-
follows: prised with mini-slots, which was split based on the number
of SUs. Here in case if the numbers of SUs are increased then,
& Integration of CRN-CSS with 5G for achieving higher the FC needs to select new mini-slots for processing [11].
throughput efficiencies and also supports to avoid inter- In the reporting channel the best CR is selected by using
ference, by using Multiple-Input-Multiple-Output rank-based censoring and threshold-based censoring [16]. In
(MIMO) and Single-Input-Single-Output (SISO). rank-based censoring, the SNR of reporting channel decreases
& CSS in CRN involves with grouping of SU nodes with which leads to increase in overhead whereas in threshold-
trinary partitioning that is executed by Fusion Center (FC) based censoring the value of thresholds are significant, from
which splits the groups with respect to its coverage. which CR are selected. Threshold values are determined with
& Then we propose novel Dynamic Multi-Channel Slot channel and network parameters for improvising the detection
Allocation (DMCSA) algorithm for allocating a set of probability.
channels to a group of SUs, which minimizes interference. Reporting scheme was focused for minimizing the energy
& To detect accurate information about the availability of PU consumption metric [8]. Decision making with the received
signals, we have designed our network with a special en- results was based on fusion rule. In this scheme, the reporting
tity (i.e.) Spectrum Agent (SA) which is responsible for scheme was terminated when final decision is made. Two
sensing the spectrum and works only for FC. techniques Soft-CSS (S-CSS) and Hard-CSS (H-CSS) were
& Global decision of Primary User signals is analyzed by FC proposed, in which the average number of reporting slots are
with graphical representation, which plots the reports of formulated. Global decision in S-CSS is given based on pre-
Mobile Netw Appl

defined threshold whereas in H-CSS the number of counts CRN-CSS have been focused by many researchers,
based decision is given. Rule based sensing used are OR rule, especially over the topics of sensing, grouping / cluster-
AND rule and K-out-of-N rule. Faster and earlier termination ing and methods to maximize its throughput efficiency.
of reporting phase is considered to reduce energy consump- The problems and limitations of these previous works
tion but minimizing energy consumption will eventually re- are solved by designing a novel CRN-CSS design with
duce the detection efficiency. In [17], authors have proposed 5G to support a huge number of users and also enhance
semi-soft fusion scheme for decision making. This scheme user experience of users participating in our network.
involves with two threshold values and the global decision is Following sections will discuss more about our novel
obtained by assigning weights based on test statistics. The CRN-CSS design with it solutions.
computed test statistic is compared with thresholds and then
final decision is taken. Involvement of complex formulations
leads to delay in decision. 3 Problem formulation
In CSS, allocation of sensing slots plays a significant role
in CRN. Contention free approach was built for managing CRN-CSS main focus is to solve spectrum scarcity
spectrums by intelligent-MAC (i-MAC) [18], with two trans- problem by allowing SUs to utilize licensed spectrums
ceivers based approach. Two transceivers based hybrid ap- by performing sensing, reporting and then transmission.
proach which is a combination of cooperative decision and In [21], Listen-And-Talk (LAT) protocol was proposed
contention-free approach. The intelligent-MAC (i-MAC) is which was supported by full-duplex technique. In this
ease to access spectrums and also it does not affect QoS. A protocol the cognitive users are enabled to simulta-
cycle time is comprised with three frames as sensing and shar- neously perform sensing and data transmission. This
ing, contention (for channel allocation) and transmission. process is executed, by using two antennas in a cogni-
Sensing is assumed to be perfect without any false alarms tive user. The spectrum sensing is based on half-duplex
and mis-detections. This frame is divided according to the sensing that is similar to conventional method. Sensing
number of primary channels and each slot is sub-divided with decisions are based on the thresholds which are dynam-
three sub-slots, in which the radios perform sensing in 1st sub- ically determined by user activity. The secondary user
slot then sharing data in 2nd and 3rd sub-slot on control chan- activity is monitored with state transition model.
nel. State information is represented in 2 bits as 00, 01, and 11 Discrete-Time Markov Chain (DTMC) is the system
for idle, busy and not sensed. Next the Contention frame is used for analyzing state transition model. This protocol
also divided into slots equal to the number of cognitive users, leads to self-interference since each SU user is equipped
this slot division reduces collision. SUs are allocated with with two antennas and then system tends to be complex
single sensing slot that tends to give inaccurate sensing report due to repeated computation of threshold values.
and if number of SUs increases then the spectrum allocation Clustering was performed for minimizing mutual in-
time needs to be increased. terference among cognitive users [22]. CRN was de-
Interference-aware CSS was proposed by authors in [19], signed by considering multiuser OFDM. Fusion center
which uses Logic-OR fusion rule for sensing. The interference is responsible for gathering channel information from
probability is determined by satisfying the condition with secondary users and then clustering is performed. For
levels of received SNR to that of desired SNR values. construction of a cluster, a weighted interference graph
Probability is formulated by power received from primary is built with the corresponding topology of the network.
transmitter and power obtained at primary receiver, this inter- Weights for each secondary user’s are given based on
ference probability was estimated with respect to the distance the edge that represents the interference degree between
between Primary user and secondary user. Here sensing time two secondary users. Clusters are formed with the
is optimized based on the decision fusion rule process. This weighted interference graph based on heuristic algo-
process was focused to maximize throughput by optimizing rithm. In this paper, the proposed algorithm is enabled
sensing time. CSS was also involved in utility based schedul- to manage the number of clusters and size of the clus-
ing scheme [20]. This scheme was supported by evolutionary ters. For cluster size maintenance, the clusters are either
game which considered utility history and improves utility. A merged or splitted (i.e.) if distance between two cluster
coalition formation algorithm was also developed with respect centers is smaller than the threshold, then they are
to the channel status (i.e.) measured according to the uncer- merged. In this work it is complex to maintain clusters
tainty. Scheduling scheme executes on two stages, one is to due to the requirement of updating the values of graph.
decide the contributor based on utility history, then next is to Then adaptive assignment of heterogeneous
decide which channel to sense. Identifying each sensing chan- cooperating cognitive users was concentrate for increas-
nel for each secondary user becomes complex when many ing throughput efficiency. Group based sensing scheme
secondary users involves in sensing. was established over Additive White Gaussian Noise
 Mobile Netw Appl

(AWGN) channels. Greedy Heuristic algorithm is in- Center (FC), Secondary Users (SUs) and Spectrum
volved for assigning with maximum number of channels Agent (SA). Here PBS, SA, FC are statically deployed
for each sensing rounds. Further proposed with a in a particular position, whereas the SUs are dynamic
Channel Selection then Best User Assignment (CSBUA) and hence they move from one place to another within
algorithm, this algorithm sorts the sensed channel in the network coverage. Figure 1 illustrates our architec-
descending order with respect to achievable rates [14]. ture with 5G and novel solutions in solving previous
So the best channel is selected and the user with best problems. Since 5G is integrated we fed Multiple-
detection probability is allocated with that channel. Input-Multiple-Output in FC and Single-Input-Single-
After selecting with a best channel, Best User Output over SUs. MIMO is used in FC to aggregate
Assignment then Channel Selection (BUACS) algorithm all the reports as Multiple-Inputs and response the SUs
is involved. Both CSBUA and BUACS algorithm adap- with Multiple-Outputs, whereas SISO in SUs is used to
tively executes with user-grouping and group-sensing avoid self-interference, which can sense and then report
for channel discovery. The main aim of these algorithms its sensing result. For CSS, the SUs are grouped by
is to maximize the spectrum efficiency in cognitive ra- performing trinary partitioning in FC. Grouped SUs per-
dio networks. Each user is enabled to sense only one form sensing according to their time slots (i.e.) defined
channel in each round, so each sensed user needs to from Dynamic Multi-Channel Slot Allocation (DMCSA)
wait until its next sensing round. All these problems algorithm.
existing in CRN-CSS is defined with perfect solutions DMCSA algorithm supports to allocate sensing slots
in our work which improves the throughput efficiencies for group of SUs and also allocate with specified set of
that denotes maximization of detection accuracy. channels to sense and report. Then global decision in
FC is predicted by comparing local sensing report and
report received from SA. This SA is responsible for
4 System model sensing the signals of PU and provides accurate results
to FC. SA is a novel entity that was introduced only for
4.1 System overview sensing and it does not makes use of the sensing spec-
trum, it is guaranteed with its sensing reports so this
Our work on CRN-CSS is performed effectively by in- used to assure with accurate detection results. FC makes
tegrating it with 5G. Design of our novel CRN-CSS a graphical plot with the received local sensing reports
architecture is comprised with deployment of Primary and it compares the report with that of the SA’s report
Base Station (PBS), Primary Users (PUs), Fusion and then global decision is made about PU. Local

Communication
Range of FC

Macro Cell Stability based node

grouping

SA
Femto Cell

Pico Cell PBS

R FC

PU

Trinary
Fig1. System Architecture Partitioning

Fig. 1 CRN-CSS system model

Mobile Netw Appl

sensing of SUs is based on energy detection method. changes in user’s speed. A network with dynamically
Each significant process is detailed in the following moving users, affects the detection performances and
sub-sections. so this stability co-efficient in CSS is required to be
taken in account.
4.2 Trinary partitioning Pseudo Code 1 – Trinary Partitioning.

Trinary partitioning is the initial process which is performed

1. Begin
once, when the network topology is deployed. Trinary means
2. Initialize
‘3’ (i.e.) it specified the coverage is partitioned into three parts.
={ , ,……., } // = 1,2, ….
As per our architecture FC is static in their positions, which is ={ , ,……., } // = 1,2, ….
in-charge of aggregating all the reports from SUs and make a 3. Determine CR FC // Begin Trinary Partitioning
global decision about the presence / absence of PU signals. By 4. FC = =
this partitioning three rings are formed in which each ring is For ( = )
comprised with set of SUs. Obtain // From equ (2)
Trinary partitioning formation is initiated we the innermost For ( = )
radius ‘r1’, we define the first partitioning area as ‘A1’. The Obtain // From equ (4)
area of first partition is given based on the distance from the 5. FC (CR) = ( , , ) // End Trinary Partition
position of FC to the ‘kth’ ring in the partitioning. Here, in 6. SUs estimate ( ) from equ (5) // Grouping
trinary partitioning all the three partitions have equal area 7. { ( ( )), … … . , ( ( ))} FC
(i.e.) ‘A1 = A2 = A3’. The areas are equal as ‘A1 = A2’ and 8. ( , , ) { , , … … … , } // = 1,2, … ….
‘A2 = A3’ that is given as, 9. { , . . , } = {( ,… ), … , ( ,.. )}
10. End
(i) ‘A1 = A2’,
Pseudo code 1, illustrates with the procedure of
πr21 ¼ πr22 −πr21 ð1Þ trinary partitioning and formation of groups in each
partitioning. This partitioning is static, since it is per-
r22 ¼ 2r21 ð2Þ formed only once based on the Communication Range
(CR) of FC. In the above pseudo code the terms ‘(P1,
P2, P3)’ represents partitions, and‘{G1, G2, …, Gz}’ rep-
resents the number of groups present in three partitions.
(ii) ‘A2 = A3’, To minimize the grouping time we segregate the CR
into three parts and form groups based on the estima-
tion of stability co-efficient of SUs.

πr22 −πr21 ¼ πr23 −πr22 ð3Þ

4.3 Dynamic multi-channel slot allocation algorithm
r23 ¼ 3r21 ð4Þ
Grouped SUs are allocated with a set of channels for sensing
Therefore, we have estimated area for trinary partitioning and provide sensed reports to FC. Local sensing in SUs is
based on the communication range of FC. Then, SUs are performed by using conventional Energy detection sensing
grouped based on their stability co-efficient with respect to in CSS. In this energy detection model, we define two hypoth-
user’s mobility speed. Neighboring users compute their stabil- eses as follows,
ity co-efficient and intimate it to FC which forms groups of

SUs in each partition. Stability co-efficient is defined with the v k ð nÞ H 01
x k ð nÞ ¼ ð6Þ
following formulation, hk sðnÞ þ vk ðnÞ; H 00
vi −vmin
ρðiÞ ¼ e v j −vmax ð5Þ From the above Eq. (6), ‘xk(n)’ denotes the received
signal from ‘kth’ SU during ‘nth’ time, then ‘hk’ represents
The terms ‘vi’ and ‘vj’ represents user’s speed, then ‘vmin’ the ‘kth’ SU’s channel gain. This channel gain is consid-
and ‘vmax’ represents the minimum and maximum speed of ered to be constant at the time of detection. ‘s(n)’ is the
users. User’s minimum and maximum speed are pre-defined PU’s transmitted signal (i.e.) independent identically dis-
during network construction. Neighboring users with relative tributed (i.i.d.) random process with zero mean and vari-
stability co-efficient are grouped into one in a partition. The ance ‘σ2s ’. The hypotheses ‘H01’ and ‘H00’ represents the
value of stability co-efficient keeps varying with respect to the absence or presence of PU signals. Further the
 Mobile Netw Appl

probabilities of detection ‘Pd’ and probability of false Table 1 Status of Sensing

alarm ‘Pf ’ is given as, State Information (SI) Status
0   1
2 2 H00 Presence of PU signal
B γ k −N σv þ jhk j σs
2
C
Pd ¼ PðY k > γ k jH 1 Þ ¼ B
@ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  A ð7Þ
C H01 Absence of PU signal
2
2N σ2v þ 2jhk j σ2s σ2v H10 SU is already performing transmission
H11 Not sensed
!
γ k −N σ2v
P f ¼ PðY k > γ k jH 0 Þ ¼ Q pffiffiffiffiffiffiffiffiffiffiffi ð8Þ
2N σ4v one after the other in sensing slot and a reporting slot to send
the sensed information about the channels allocated. For in-
Equation (7) and (8), involves with ‘γk’ that denotes deci- stance, let us assume a group ‘G1’ with four SUs as ‘SU1, SU2,
sion threshold, ‘σ2v ’ is the noise variance which is estimated SU3, SU4’ and these SUs in G1 is allocated with a set of
from power levels of channel, then ‘Yk’ is defined as the test channels ‘CH1, CH2, CH3, CH4, CH5’. To sense five channels
statistic that is given as, we allocate each SU with five sensing slots ‘s1, s2, s3, s4, s5’.
N In the allocated first slot, SU1, SU2, SU3, SU4 perform
Y k ¼ ∑ j x k ð nÞ j 2 ð9Þ sensing of four different channels allocated to their
n¼1
group G1. In s1, four channels are sensed by four SUs present
The CSS model is involved with two fields; one is for in the group (i.e.) SU1 senses CH1, SU2 senses CH2, SU3
sensing, reporting and another one for transmission. The senses CH3 and SU4 senses CH4. This is a cyclic process, in
frame structure is comprised with these two fields as men- which the channels are sequentially changed for each SU and
tioned above. Fig. 2 shows the CSS model by its frame struc- so all the five channels are sensed within five time slots by G1
ture for a cycle with sensing and reporting time ‘Ts, r’ and SUs.
transmission time ‘Ttx’. According to the groups of SUs their Figure 3 depicts with the slot allocation for sensing multi-
sensing time for a channel is allocated with respect to the ple channels dynamically without causing interference to oth-
number of SUs present in a group. er SU. Interference is avoided since each secondary user sense
In proposed DMCSA, each SU is enabled to sense more independent channel (i.e.) the allocated channel is sensed by a
than one channel sequentially. Reporting information from SU SU at a time. Assigned time slots sequentially changes with
includes ‘H00’, ‘H01’, ‘H10’ and ‘H11’, which represents idle respect to the number of channels allocated for a group. With
(i.e.) PU signals are available, busy (i.e.) PU signals are ab- this allocated slots, SUs perform local sensing by using
sent, occupied (i.e.) SU is already performing data transmis- Energy detection method.
sion and no sensing is performed (i.e.) the SU need not require Procedure for sensing slot allocation.
any spectrum, so it does not perform any sensing else it can
sense particular channels that is allocated for sensing. The
Step 1 : Begin
sensing states are listed in Table 1. Step 2 : SUs are grouped, Let
Grouped SUs are allocated with a set of channels, which ={ , , , }
performs sensing multi-channel sequentially and dynamically Step 3 : Grouped users are organized randomly
Step 4 : Set of channels allocated by FC to , Let
{ , , , , }
Step 5 : Equal number of channels are allocated with
respect to group members
Step 6 : is sequentially allocated with all five
channels in five sensing slots, Let { , , , , }
Step 7 : Next shifts the sensing slot from the prior
sequence of slot allocation in
Step 8 : Repeat until all SUs are allocated with sensing
slots
Step 9 : After allocation user begins to sense each time
slot with different channel
Step 10 : End

The above procedure illustrates slots allocation for SUs

using DMCSA algorithm with an instance of five SUs, chan-
Fig. 2 Proposed CSS model nels and sensing slots.
Mobile Netw Appl

Fig. 3 Sensing slot allocation

According to our process SUs are grouped and allocated than one channel. The assigned sets of channel are sensed in
with a set of channel for sensing as shown in Fig. 4. Here set of SUs sequentially without causing any interference to each
channels are represented as ‘S 1 (CHs), . . S w − 1 (CHs), other. Hence the SUs sense more than one channel and so they
Sw(CHs)’ in which ‘w’ denotes the total number of channel are allocated with channels in short period of time. The sensed
sets involved. Based on the groups of SUs, set of channels are information is stored by SU, and then in reporting time slot it
determined and each set of channel is comprised with more sends all the sensed channel’s information to FC for global

Fig. 4 Group based CRN-CSS

 Mobile Netw Appl

Table 2 Sensing report information Table 2 illustrates with a sample sensing report aggregated
SUs CH1 CH2 CH3 CH4 CH5 by FC from SUs for a group of users with five different chan-
nels. From Table 2, SU1 detects CH2 as idle, SU2 detects CH1,
SU1 H01 H00 H01 H11 H01 CH2, CH4 as idle, SU3 detects CH2, CH4, CH5 as idle and SU4
SU2 H00 H00 H01 H00 H01 detects CH4 as idle. Hereby the CH3 is always detected as
SU3 H10 H00 H01 H00 H00 busy (i.e.) it means absence of PU signals. Also some chan-
SU4 H01 H11 H01 H00 H01 nels are left un-sensed by SUs and in other cases SUs are
already performing transmission. These reports are represent-
ed graphically along with the sensing report obtained from
decision making. DMCSA achieves efficient slot allocation SA.
for sensing multi-channels without causing any interference Figure 5a shows the representation of aggregated channel
among SUs; this algorithm can be guaranteed for large scale information from SUs, each shape represents each channel.
networks. Upon the obtained local sensing, the hypotheses based evalu-
ation is represented. Next in Fig. 5b, the SA’s sensing report is
4.4 Global decision making also represented in black colored shapes for differentiating
from SUs report. From this reports representation, final deci-
In this sub-section, we discuss about the retrieved sion of presence / absence of PU are taken. The appropriately
sensed information for groups of secondary users. sensed SU will be allocated with channels for their communi-
Each secondary user reports with the number of sensed cation. The global decision by FC is accurate due to the utili-
c h a n n e l ’s i n f o r m a t i o n w i t h f o u r p o s s i b i l i t i e s zation of new entity called spectrum agent. SA plays a major
as H00, H01, H10 and H11. Global decision in FC aggre- role in this work which enhances CRN-CSS environment.
gates sensing reports and spectrum agent’s sensing re- CRN is completely supported by PU signals, here we allo-
port is obtained. SA is a special entity in CRN that was cate channel for SUs to sense the availability of idle PU
introduced to minimize energy consumption and utiliza- signals.
tion of resources [23]. But here our goal is to achieve PUs utilizes the signals from PBS, at idle instances it is
accurate result about channel’s availability, so as per our utilized by SUs.
work SA is used for providing sensing results about the Therefore, our work gives a remarkable change in CRN-
channels whether PU signals is present or not. Therefore CSS for sensing multiple channels and final global decision is
in our CRN-CSS design SA works only for FC and not more accurate, which cannot be achieved by using any other
for SUs. complex algorithms.
For global decision we design with a graphical representa-
tion which includes sensing reports from SUs and SA. The
report from SA is always assured with accurate results of PU’s 5 Performance estimation
signals. The use of SA is to support accurate spectrum detec-
tion as well as, this leads to efficient allocation of channel to In this section, simulation results are provided to validate our
SUs for communication. novel CRN-CSS network with DMCSA algorithm. This

Fig. 5 Global decision making –

representation
0 1
CH1
CH2
CH3
CH4
0
CH5
1

(a)
(b)
Mobile Netw Appl

Table 3 Simulation Parameters

Parameters Values

Simulation Area 400 × 400

Network Environment entities Number of PUs 20
Number of SUs 20
Number of SA 3
Number of PBS 1
Fusion Center (With MIMO) Number of FC 3
Communication range 200 m
Number of Antennas 4×4
Number of PU channels 20
Data rate (5G) 10 Gbps
Mobility Model Random Way point
Transmission Power 20 Amps
Mobility Speed Minimum 0 mbps
Maximum 100 mbps
Simulation time 5 Seconds

section is categorized into three significant subsections as [18] that is caused due to SUs interfere PU signals is given as
mathematical evaluations, simulation setup and comparative follows,
analyses. Hereby we evaluate our novel algorithm in this sec-
tion and list out the merits of this work, which can be used PðiÞ ¼ P½T ≤ T i  ¼ 1−eλT i ð10Þ
real-time.
Here ‘T’ denotes inter-arrival time occur between PUs that
has exponential distribution rate of ‘λ’. Then the probability of
channel being idle is followed by binomial distribution that is
5.1 Mathematical evaluation
expressed as,
 
In this section the significant metrics are formulated which PCH
includes probability of interference, probability of channels PðN idle Þ ¼ ð1−αÞNidle  αPCH −N idle ð11Þ
N idle
being idle, probability of the sensed channel being idle and
more. As per our work, interference is completely minimized The above Eq. (11), includes ‘PCH’ as the total number of
even though we formulate the probability of interference ‘P(i)’ primary user channels, ‘Nidle’ denotes the total number of

Fig. 6 CRN-CSS Network model

 Mobile Netw Appl

Table 4 Demerits of Previous

CRN-CSS algorithms Previous Demerits
algorithms

i-MAC [18] • Sensing is performed in single time slot leads to inaccurate sensing
• Increase in number of nodes, increases the spectrum allocation time.
CSBUA, BUACS • Each user is enabled to sense only one channel in each round, so a user needs to wait until
[14] its next sensing round.
Fusion Schemes • More complex computations leads to higher time consumption
[6]

channels to be idle, here utilization of PU is denoted as By using these mathematical formulations the probabilities
‘α’ (i.e.) this term can be taken in account of PU chan- of channels and SUs are determined. Main aim of our work is
nels being busy. Hence the probability of idle channel to minimize interference and completely improve the through-
followed by binomial distributions is formulated as put efficiency of the designed CRN-CSS by utilizing DMCSA
shown in Eq. (11). Then the probability of idle sensed algorithm and SA’s support for global decision making.
channel ‘Sidle’ is given as,
 
S MAX 5.2 Simulation setup
PðS idle Þ ¼ ð1−αÞS idle  αS MAX −S idle ð12Þ
S idle
We perform simulation of our proposed work in Network
The condition for Eq. (12) is stated as ‘0 ≤ Sidle ≤ SMAX’, Simulator – 3 (NS3) which is installed in Ubuntu Operating
further the average number of sensed idle channels are given System. Our design of CRN-CSS is integrated with 5G that
as follows, supports with a huge number of users. NS3 is one of the best
S MAX simulators which are sustained for different type of networks.
AvgðS idle Þ ¼ ∑ S idle  PðS idle Þ ð13Þ We have preferred this simulation environment which sup-
S idle ¼0 ports all the characteristics and models of CRN network and
so the result gives better accuracy for our novel algorithm. All
From the probabilities ‘P(Nidle)’ and ‘P(Sidle)’ estimated in
the significant simulation parameters taken in account for our
Eqs. (11) and (12), is further involves in determination of
CRN-CSS are discussed.
probability ‘η’ that is nothing but an idle channel is being
Table 3 illustrates with the significant specifications that
sensed among all the idle channels that is assigned for that
are taken in account for our simulation of CRN-CSS network
group.
environment with 5G. Specifications are not limited upto this
Sensed idle Channels Table 3, it extends with all characteristics of 5G technology.
η¼
With these specifications our simulation environment is de-
Number ofidle channel
ð14Þ ployed by the topology as shown in Fig. 6.
Avg ðS idle Þ
ði:e:Þ ¼
Avg ðN idle Þ 4.5

4
5
Throughput Efficiency (Mbps)

3.5
4.5
Throughput Efficiency (Mbps)

3
4
3.5 2.5 CSBUA
3 CSBUA BUACS
2
2.5 BUACS DMCSA
2 1.5
DMCSA
1.5 1
1
0.5
0.5
0 0
4 6 8 10 2 3 4 5 6
Number of Cooperating Users Number of Groups
Fig. 7 Variation of throughput with respect to cooperative users Fig. 8 Variation of throughput with respect to groups formed
Mobile Netw Appl

100000 4
90000
3.5
Average Throughput (bytes/s)

Average Sensing Overhead(Mb/s)

80000
CR-iMAC 3
70000 DSR
DMCSA 2.5
60000
CR-iMAC CSBUA
50000 AODV 2
BUACS
40000 1.5 DMCSA
30000
1
20000
0.5
10000

0 0
1 2 3 4 5 6 7 8 9 4 6 8 10
Number of Cooperative Users
Simulation Time (s)
Fig. 11 Comparison on sensing overhead
Fig. 9 Variation of average throughput with respect to simulation time

Throughput efficiency is plotted against varying simulation

5.3 Comparative analyses time, cooperative users, number of groups and SU’s sensing
time.
In this section, we perform comparative analyses between our In Fig. 7 we compare the throughput efficiency of CSBUS,
proposed result and results achieved in previous algorithms / BUACS and DMCSA algorithm. In both CSBUA and
schemes in CRN-CSS. Previous works as intelligent-MAC. BUACS algorithms SUs in a Group ‘G’ are designed to sense
[18], fusion schemes [6] and Channel Selection then Best only a single channel. Multiple SUs in a group, sensing a
User Assignment (CSBUA) Algorithm and Best User single channel cause’s interference, due to this energy in each
Assignment then Channel Selection (BUACS) Algorithm SU is consumed higher and also their sensing results are not
[14], are taken into considerations for showing improvements accurate. From the figure, we observe that the throughput
of our work. efficiency of CSBUA and BUACS are too low when com-
Table 4 lists out the major demerits involved in previous pared with our proposed DMCSA algorithm. From this an
works which are significantly considered for comparison. optimal increase of about 2.5% to 3% of throughput efficiency
Comparative results are mainly focused on improvising is achieved. As per the gradual increase in number of cooper-
throughput efficiencies, sensing overhead and detection ative users, throughput also increases due to the reason of SUs
probability. sensing more than one channel and they are allocated with
available channels for communication.
Figure 8 shows the comparison plot of throughput efficien-
5.3.1 Throughput Efficiency cy for CSBUA, BUACS and DMCSA algorithms. Among the
three algorithms, DMCSA shows tremendous improvement
We investigate the impact of our proposed CRN-CSS design achieved on sensing multiple channels by different SUs in a
with DMCSA algorithm over other works held on CRN-CSS. group. This comparison indicates that algorithms CSBUA,
25
Fusion Scheme (M=1) Fusion Scheme (M=2)
Fusion Scheme (M=3) DMCSA (M=1)
20 Fusion Scheme 0.8
(N=1)
Throughput (bits/s)

Fusion Scheme 0.7

Detection Probability

(N=5)
15 DMCSA (N=1) 0.6
0.5
DMCSA (N=5)
10 0.4
0.3
5 0.2
0.1
0 0
1 2 3 4 5 1 2 3 4 5 6 7 8 9 10
Sensing time (s) Sensing Time (s)
Fig. 10 Variation of throughput with respect to sensing time Fig. 12 Comparison on detection probability
 Mobile Netw Appl

BUACS are able to achieve consistent throughput but each group. Sensing overhead for a group-based CSS is for-
not upto the level of DMCSA. According to increases mulated as follows [14],
in the number of groups, the throughput increases, from
Q q Ts
this we notice that the increase in number of groups SOH ¼ ∑ Pq ðg i Þ ∑ ∑ ∑ Ri;k ð15Þ
also increases the participation of SUs in sensing. So q¼1 h¼1 gi ∈GðhÞ k∈gi Tt
even at increase in number of SUs, our algorithm is
capable to withstand with a tremendous improvement From the above formulation sensing overhead ‘SOH’ is
in throughput. estimated with ‘G(h) = {gi| 1 ≤ i≤| G| }’ represents set of groups
When the number of groups ‘G = 2’ the throughput is in which SUs performing sensing at ‘hth’ sensing round, then
increased upto 0.5%, as per the group increases as ‘G = ‘Ri, k’ denotes the rate at which the cooperative secondary user
6’ the throughput efficiency is increase approximately of ‘k’ performs transmission in channel, ‘Q’ represents the total
3.2%. The increase in throughput in DMCSA is also number of sensing rounds that is required for sensing the
strongly supported by 5G which is capable to tolerate licensed channels. ‘Ts’ is sensing period and ‘Tt’ is transmis-
the participation of huge users and provide best data sion time.
rate to their communication. Utilization of 5G in This metric sensing overhead is evaluated with respect to
CRN-CSS have results with higher throughput and the number of cooperative users. If the number of channels
hence this can create a brighter network environment lesser when compared with the number of users, then the
in future to maintain many number of users in groups average sensing overhead increases with respect to the number
for sensing. of cooperative users. Figure 11 illustrates the comparison of
Next, Fig. 9 depicts the simulated average throughput sensing overhead with respect to the number of cooperative
in bytes per second over the time our network is simu- users. In previous algorithms of CSBUA and BUACS, sens-
lated. In out simulation, we have considered one chan- ing overhead increases when the number of cooperative users
nel for each PU so totally in our CRN-CSS ‘20’ PU increases, this is due to the higher number of rounds taken to
channels is applicable in our simulation. The state-of- determine a single channel. Hence these two algorithms re-
the-art, CR-iMAC was designed and its performance is quire more number of rounds for detecting idle channels.
evaluated with two different conventional routing proto- Therefore our proposed DMCSA algorithm shows lesser sens-
cols as DSR and AODV. This plot shows that DMCSA ing overhead even when the number of cooperative users
algorithm is better than CR-iMAC which results an av- increases.
erage throughput of 38,208 bytes/s for AODV and
38,368 bytes/s for DSR, whereas DMSCA reaches an 5.3.3 Detection Probability
approximate throughput of 68,000 bytes/s.
Throughput is also evaluated with respect to sensing Detection probability increases when the users perfectly iden-
time of cognitive secondary users. The value of ‘N’ tify the available idle channels. Probability of detection varies
represent the number of PU channels present. Based according to the Signal-to-Noise Ratio (SNR) present in the
on the number of channels the throughput value varies sensing channels.
with respect to sensing time. In Fig. 10, a comparative From Fig. 12 it is observed that that the detection probabil-
study is illustrated that our DMSCA shows better ity of our proposed CRN-CSS increases with respect to the
throughput than the previous fusion schemes. This sensing time as per the increase in sensing time, the detection
throughput increase in network, intimates that the num- probability also increases gradually. The value of ‘M’ is varied
ber of cooperative SUs participation is also increased. which represents the number of slots for sensing channels.
On the whole throughput is plotted with respect to sig- Table 5 depicts the comparison on proposed DMCSA with
nificant metrics involved in our work. All the compari- prior research works held in CRN-CSS. From this comparison
sons of throughput give higher results than the other it is clear that proposed DMCSA better than previous work
research works. These results have proven that our pro- due to the efficient design of CRN-CSS model with 5G.
posed CRN-CSS is a remarkable algorithm.
Table 5 Comparison on performance metrics
5.3.2 Sensing overhead
Parameter CSBUA BUACS DMCSA

Sensing overhead is a significant metric in CRN, where the Throughput (Mbps) 0.77 0.69 1.7
SU perform sensing and identify the idle channels and make Sensing Overhead (Mbps) 3.2 2.7 0.62
use for their own communication. This metric depends on the Detection Probability Fusion Scheme DMCSA
total number of sensing rounds that is required for a SU to 0.08 0.46
predict an idle channel and number of users cooperating in
Mobile Netw Appl

Therefore, all the performance evaluations results with im- 4. Jaglan RR, Mustafa R, Sarowa S, Agrawal S (2016) Performance
evaluation of energy detection based cooperative sensing in cogni-
provements of our proposed novel CRN-CSS with DMCSA
tive radio network. Proceedings of First International Conference
algorithm which is enabled to sense multi-channel without on Information and Communication Technology for Intelligent
causing interference among SUs. Our main aim is to solve Systems, Springer 2:585–593
the problem of interference and cooperative sensing of SUs 5. Xin L, Chen KQ, Yan JH (2016) A novel weighed cooperative
bandwidth spectrum sensing for spectrum occupancy of cognitive
participating in the network. The reduction of interference is
radio network. J Cent South Univ, Springer 23(7):1709–1718
established by the improvements in throughput which is ana- 6. Bhowmick A, Yadav K, Roy SD, Kundu S (2017) Multi slot-
lyzed with respect to certain significant metrics. throughput tradeoff in an improved energy detector based faded
cognitive radio network. Wireless Networks, Springer, pp. 1–14
7. Liu X, Jia M, Tan X (2013) Threshold optimization of cooperative
spectrum sensing in cognitive radio networks. Radio Sci 48(1):23–
6 Conclusion 32
8. Althunibat S, Granelli F (2016) On Results’ Reporting of
Cooperative Spectrum Sensing in Cognitive Radio Networks.
In this paper, the key objective of CRN-CSS model is Telecommun Syst, Springer 62(3):569–580
achieved (i.e.) to sense multi-channel in minimized interfer- 9. Gaurav V, Sahu OP (2017) Interference Aware Sensing Scheme in
ence and without degrading the sensing accuracy. This work Cognitive Radio System. Wireless Personal Communications,
integrates CRN-CSS model with 5G to support participation Springer 94(3):1405–1425
10. Jiao Y, Yin P, Joe I (2016) Clustering scheme for cooperative spec-
of multiple users with higher data rates. This work is initiated trum sensing in cognitive radio networks. IET Commun, IEEE
with trinary partitioning which groups SUs for performing 10(13):1590–1595
cooperative sensing. FC allocates channels and sensing slots 11. So J, Sung W (2016) Group-based Multi-bit Cooperative Spectrum
for the grouped SUs using the novel DMCSA algorithm. Then Sensing for Cognitive Radio Networks. IEEE Trans Veh Technol
65(12):10193–10198
according to the slots SUs perform sensing and send their 12. Wang Y, Lin W, Huang Y, Ni W (2014) Optimization of Cluster-
reports to FC for global decision making. Further for accurate Based Cooperative Spectrum Sensing Scheme in Cognitive Radio
global decision making, we have included Spectrum agent Networks with Soft Data Fusion. Wirel Pers Commun, Springer
which give guaranteed report about the presence of PU spec- 77(4):2871–2888
13. Sharma P, Abrol V (2013) Optimized cluster head selection & ro-
trum. FC compares the reports of SUs and spectrum agent by tation for cooperative spectrum sensing in cognitive radio networks.
forming a graphical representation for obtained reports. Final In: Tenth International Conference on Wireless and Optical
decision is given by FC and then it allocates channels to SUs Communications Networks, IEEE, pp. 1–5
for their communication. Due to efficiency of the proposed 14. Khalid L, Anpalagan A (2016) Adaptive Assignment of
Heterogeneous Users for Group-Based Cooperative Spectrum
novel algorithm, throughput efficiency is highly increased Sensing. IEEE Trans Wirel Commun 15(1):232–246
when compared with previous research works. Hereby it in- 15. Song LL, Liu Y, Wang Z, Zhang SB (2014) Cooperative spectrum
crease detection probability and minimizes sensing overhead sensing based on node recognition in cognitive radio networks.
in proposed CRN-CSS model. On the whole our major re- International Conference on Information and Communications
Technologies, IEEE, pp. 1–6
search findings of proposed novel CRN-CSS model are the
16. Bhowmick A, Nallagonda S, Roy SD, Kundu S (2015) Cooperative
effective grouping of users and utilization of spectrums by Spectrum Sensing with Double Threshold and Censoring in
channel allocation with reduced interference that achieves ap- Rayleigh Faded Cognitive Radio Network. Wirel Pers Commun,
proximately 90% of throughput efficiency. In future, we have Springer 84(1):251–271
17. Verma P, Singh B (2016) On the decision fusion for cooperative
planned to extend our work with this novel CRN-CSS archi-
spectrum sensing in cognitive radio networks. Wireless Networks,
tecture design with improvising the use of spectral efficiency Springer, pp. 1–10
by spectrum allocations methods. 18. Singh JSP, Rai MK (2017) Cognitive radio intelligent-MAC (CR-i-
MAC): channel-diverse contention free approach for spectrum
management. Telecommun Syst, Springer 64(3):495–508
19. Song F, Kan C, Wu Q, Ding G (2015) Optimal Cooperative
Spectrum Sensing in Interference-Aware Cognitive Radio
References Networks. Wirel Pers Commun, Springer 82(4):2171–2184
20. Li H, Xing X, Zhu J, Cheng X, Li K, Bie R, Jing T (2017) Utility-
Based Cooperative Spectrum Sensing Scheduling in Cognitive
1. Ali A, Hamouda W (2017) Advances on spectrum sensing for cog- Radio Networks. IEEE Trans Veh Technol 66(1):645–655
nitive radio networks: theory and applications. IEEE 21. Liao Y, Wang T, Song L, Zhu H (2017) Listen-and-Talk: Protocol
Communications Surveys & Tutorials PP(99):1–29 Design and Analysis for Full-duplex Cognitive Radio Networks.
2. Umara R, Sheikh AUH (2012) A comparative study of spectrum IEEE Trans Veh Technol 66(1):656–667
awareness techniques for cognitive radio oriented wireless net- 22. Dai J, Wang S (2017) Clustering-Based Spectrum Sharing Strategy
works. Physical communication, vol. 9. Elsevier, pp. 1–23 for Cognitive Radio Networks. IEEE Journal on Selected Areas in
3. Liu X, Zhong WZ, Chen KQ (2015) Optimization of sensing time Communications 35(1):228–237
and cooperative user allocation for OR-rule cooperative spectrum 23. Zhang Z, Zhang W, Zeadally S, Wang Y, Liu Y (2015) Cognitive
sensing in cognitive radio network. J Cent South Univ, Springer Radio Spectrum Sensing Framework Based on Multi-Agent
22(7):2646–2654 Architecture for 5G Networks. IEEE Wirel Commun 22(6):34–39