You are on page 1of 5

This fUll-text paper was peer-reviewed and accepted to be presented at the IEEE WiSPNET 2016 conference.

Performance Analysis of Cooperative Spectrum


Sensing in Cognitive Radio using Game Theory
Fenila Janet M
Lavanya S Bhagyaveni M.A
Electronics and Communication Electronics and Communication Electronics and Communication
Engineering Engineering Engineering
College of Engineering,Guindy College of Engineering,Guindy College of Engineering,Guindy
Chennai, India Chennai,India Chennai,India
fenilajnt@gmail.com lavanyashanmugavel@gmail.com bhagya@annauniv.edu

Abstract- Cognitive Radio (CR) aims to provide efficient the unoccupied channel. Cooperative Spectrum Sensing (CSS)
utilization of the scarcely available radio spectrum. Spectrum exploits spatial diversity thereby improving the sensing
sensing is indispensable to identify spectrum holes. Inaccurate performance. A comprehensive survey on the various CSS
sensing results in interference to PU. Therefore, sensing accuracy
techniques is elaborated [4]. Game theory is a mathematical
is an important parameter. Cooperative Spectrum Sensing (CSS)
tool that analyzes the strategic relations among numerous
is an efficient technique to improve the sensing accuracy. It
makes use of cooperation between Secondary Users (SUs) to
decision makers. In recent years, game theory has been found
detect the presence of Primary User (PU). Game theoretical to be very useful in various applications including cognitive
models help in understanding and modeling interaction among radio [5]. In [6], Saad el al has given a detailed overview on
users. In this paper, a CR network with multiple PUs and SUs is coalitional game theory, which also deals with the major
considered. CSS is modeled as hedonic coalition formation game challenges in applying coalition games to modem
where SUs sensing the same channel form coalitions. The communication systems. A more focused study on modeling
coalitions are formed on the basis of probability of detection. For CSS as coalition formation game is presented in [7]. In [8],
each PU, the SUs which have maximum probability of detection
Hao et at formulated multichannel spectrum sensing and
form coalition and cooperatively make the final sensing decision
channel access as a hedonic coalition formation game which
by making use of OR rule. Channel estimation for ST -SR link is
done using WARP testbed. The sensing accuracy and energy
yielded improvement in performance than non-cooperative
consumption for the coalitions when link is modeled using free sensing in terms of average utility of su.
space pathloss model and using WARP testbed are found out. In this paper, an algorithm for coalition formation is
Simulation results show that CSS modeled using hedonic presented. After coalitions are formed, Decision Node (DN) is
coalition formation has better sensing accuracy than non­ elected which gives the final sensing decision. The link
cooperative sensing. between Secondary Transmitter (ST) and Secondary Receiver
(SR) is modeled using free space path loss model and in real
Keywords-Cognitive Radio, Cooperative Spectrum Sensing,
time using Wireless Open-Access Research Platform (WARP).
hedonic, utility, WARP.
The novelty of this paper lies in the fact that, for channel
INTRODUCTION
access, two constraints are defined. If inactive PU is detected,
SUs in the coalition access the sensed channel. If PU is active,
The ever-increasing pace of development in wireless SUs in a distant coalition access the channel so that the
services has led to a substantial rise in the demand for radio interference provided by SU to the active PU is minimum as in
spectrum. However, the radio spectrum is limited and a large the case of underlay network.
amount of it has already been licensed to existing operators. The remainder of the paper is organized as follows: Section
According to studies conducted by Federal Communications II describes the algorithm for coalition formation. Section III
Commission (FCC), fixed spectrum allocation is not always deals with channel estimation using WARP. Section IV
efficient as a large amount of the licensed spectrum remains explains the calculation of energy consumption and sensing
unoccupied (spectrum holes) for significant periods of time [ 1]. accuracy. Section V presents the results and future work.
Cognitive Radio (CR) systems have been proposed [2] to
efficiently exploit this under-utilized spectrum whereby the ALGORITHM FOR COALITION FORMATION
unlicensed users (secondary users) dynamically and
The procedure involved in coalition formation :
opportunistically access the bands allocated to the licensed
i. Energy detection based sensing is performed by all ST
users (primary users) when the spectrum remains unused.
and the probability of detection of all PUs with respect
Spectrum sensing is vital [3] because misdetection leads to
to all STs is found out.
interference to the Primary Users (PU) whereas false alarm
causes the Secondary Users (SU) to lose an opportunity to use

978-1-4673-9338-6/16/$31.00 ©2016 IEEE 2061


This fUll-text paper was peer-reviewed and accepted to be presented at the IEEE WiSPNET 2016 conference.

ii. SUs which have maximum probability of detection for In order to account for 20 SU, the nodes (RF C, RF D) are
each PU join to form coalition and perform CSS to placed at random locations and channel estimation is done. The
find out whether the PU is active or inactive. average channel gain is found out.
iii. Number of coalitions formed will be equal to the
CALCULATION OF SENSING ACCURACY AND ENERGY
number of PUs in the network. Only non-overlapping
CONSUMPTION
coalitions are formed. Each ST can be a part of one
coalition only. In game theory, payoff is used to capture the tradeoff
IV. After coalitions are formed for all PU, Decision Node between reward and penalty. In this work, reward refers to the
(DN) has to be selected. In each coalition, ST with amount of data transmitted by ST (in case of correct detection
greatest probability of detection is chosen as DN. of inactive PU). Penalty is the loss due to interference to PU (in
v. DN uses OR rule as decision fusion rule to combine case of misdetection). Sensing accuracy is given by A.
the sensing results of the members in the coalition.
The primary reason for the choice of OR decision A =
P . L,jESiRj.; C T -
0) -
PCW,ODo (T -
0)
fusion rule is that it helps to minimize the problem of (0 1 O" )--I is -I-
misdetection which would otherwise lead to adverse
effects due to interference to PU. where P01o,i is the probability that correct decision occurs,
VI. The probability of detection of each coalition is found R j,i is the transmission rate of STj, (T-8) is the data transmission
out. duration, ISil is the number of members in the coalition, Do is
vii. Game theory is used to model the sensing accuracy the penalty for misdetection, 8 is the sensing duration, POII,i is
and energy consumption of the coalitions. The link the probability that misdetection occurs.
between ST and SR is modeled using Energy consumption is given by
1. Free Space model
2. WARP testbed
Vlli. After step (vii), channel access has to be performed.
• If the sensing result concludes PU is inactive, SUs Here, the time period T is divided into two time slots:
which have sensed the PU can access the channel. Sensing duration (8), data transmission duration (T-8). Data
• If the sensing result concludes PU is active, SUs transmission takes place during correct detection (inactive PU
which have sensed the PU cannot access the sensed is detected correctly) denoted by POIO,i and during misdetection
channel but can access the channel of other PUs for (Active PU is not detected) denoted by polI,i. For misdetection
which it would provide lower interference. POII,i penalty Do is given as it has led to interference to PU.
Sensing power is given by Ps1Si18. The power required for data
CHANNEL ESTIMATION USING WARP transmission is given by (POIO,i + POII,i)Pt(T-8) .
The WARP platform shown in Fig 1, was designed at Rice In the calculation of transmission rate,
University and is used by number of academic and industrial
research lab for clean state protocol implementation of the
R··"
1 z Pt
B·logz(l + h· ·I -
MAC and PHY. =

],
'
,]
(J�,i
the average channel gain of ST-SR link Ihi is modeled
using free space pathloss model as Ihi=l/d and also using
2
WARP testbed .

RESULTS AND DISCUSSIONS

Cogniti\e radio network


100
,------�--�.�6--�

80r-�����_+· "����"�-r- -----i


.1
+ 17
Fig 1. Experimental Setup ofWARP 60 r-----+_----____
,7
A WARPV3 node usually contains two RF node (RF A, RF 40 �----+_----_+--��F�----�----__4
B) which is suitable for SISO transmission, each RF node can
be configured as either transmitter or receiver. Two more RF 20r----=+-- -�-+- -----+-.�-�. 9 -r- ---�
8
node (RF C, RF D) can be added via FMC (FPGA Mezzanine
0 �____�____�__ · 1_�
� 4 ______ ����
Connector) for MIMO transmission. Multiple WARP nodes o 20 40 60 80 100

can be connected to same computer via Ethernet switch. One


pair of nodes (RF A, RF B) can be considered as pu. The other
pair of nodes (RF C, RF D) can be considered as ST-SR pair. Fig 2. CR Network Scenario

2062
This fUll-text paper was peer-reviewed and accepted to be presented at the IEEE WiSPNET 2016 conference.

The CR network scenario under consideration consists of 5 In Fig 4, the variation of sensing accuracy with respect to
PU, 20 Su. SUs are identified as ST-SR pair. Bandwidth of sensing time is plotted. From the plot, it is observed that as
each primary channel is assumed to be 10 MHz. Fig. 1 shows sensing time increases, sensing accuracy increases till sensing
the CR network scenario considered. The dark circles represent duration of 5 ms, after which it decreases. If the sensing
pu. The dark colored diamonds represent ST. The light colored duration is 1ms, the nodes do not have enough time to sense,
diamonds represent SR. Transmit power is taken as 50mW. hence sensing accuracy is less. As the sensing duration
Sensing power is taken as 10mW. The detection threshold E is increases, time for data transmission is less. Therefore sensing
chosen as 0.2mW. Noise power is assumed to be O.lmW. accuracy decreases for larger sensing time. The optimal sensing
Energy detection is used as the sensing mechanism. The ST time of 5 ms is considered for the simulations performed in this
perform local sensing initially and sense all pu. For each PU, work.
STs satisfying the threshold criterion combine to form
coalitions. The nodes (STs) in each coalition is given in Table
I.

NODES IN COALITION

Coalition for PU SUs in the coalition


I 14

2 5,17

3 2,8, 10,12, 13, 15, 18,20

4 1,4,9,19
Fig 4. Plot ofSensing Accuracy vs Sensing Time (WARP)
5 3,6,7, 1 1, 16
. . . .
In each coalItIOn, node WIth hIghest probabIlIty of detectIOn
(DN) combines the sensing results using OR rule. The
probability of detection and probability of false alarm for the
coalitions are tabulated in Table II.

TAB LE II. PF AND PD FOR THE COALITIONS

Coalition Probability of false Probability of


alarm (Pc) detection (Pd)
I 0.0 127 0. 1288

2 0.0253 0.87 19
Fig 5. Plot ofSensing accuracy vs Sensing time (Free Space model)
3 0. 1 1.0000

4 0.0507 1.0000

5 0.0634 1.0000

Fig 6.Plot ofSensing accuracy Vs Probability ofPU being inactive (WARP)

Fig 3. Comparison of Cooperative and Non-cooperative sensing

Fig 3 depicts a comparison of cooperative and non­


cooperative sensing in terms of probability of detection. From
the plot, it is observed that cooperative sensing always
outperforms non-cooperative sensing.
Fig 7.Plot ofSensing accuracy Vs Probability ofPU being inactive (Free Space
pathloss model)

2063
This fUll-text paper was peer-reviewed and accepted to be presented at the IEEE WiSPNET 2016 conference.

Fig S.Plot ofSensing accuracy Vs Number ofPU (WARP) Fig 12. Plot ofEnergy Consumption Vs Number ofPU

The plot of sensing accuracy vs probability of PU being


inactive is shown in Fig.6. For a range of probabilities from O.l
to 1, sensing accuracy increases from a lower value to
maximum. This is mainly due to the fact that, greater
probability of PU being inactive increases the opportunity for
SU to utilize the spectrum and hence greater sensing accuracy.
Fig 8 shows plot of sensing accuracy with respect to the
number of PU. As the number of PU increases, SUs have more
channels to sense and sensing accuracy improves . From Fig
10, it is evident that as sensing time increases, energy
consumption increases. As energy consumption is based on
both sensing and data transmission, energy consumed
Fig 9.Plot ofSensing accuracy Vs Number ofPU (Free increases. Fig 11 shows plot of energy consumption with
Space pathloss model) respect to probability of PU being inactive. When PU is
inactive, data transmission takes place and hence energy
consumption is high. As the number of PU increases, sensing
becomes uncomplicated and hence energy consumption
decreases as shown in Fig 12.

CONCLUSION AND FUTURE WORK


Coalitional game theory has been used to model CSS. It is
evident from results that CSS modeled using game theory
produces better sensing performance than non-cooperative
sensing techniques. Moreover, ST-SR link has been modeled
using free space model and it has also been implemented in real
time using WARP. This work could be further extended to
include spectrum management.
Fig 10. Plot ofEnergy consumption vs Sensing time
REFERENCES

[I]. Federal Communications Commission, "Spectrum Policy


Task Force Report," Report ET Docket no. 02-135, Nov.
2002.

[2]. S. Haykin, "Cognitive radio: Brain-empowered wireless


communications," IEEE J. Select. Areas Commun. , vol. 23,
pp. 201-220, Feb. 2005.

[3]. Yucek T, Arslan H. "A survey of spectrum sensing


algorithms for cognitive radio applications," IEEE Commun
Surv Tut, pp.116-130, 2009.

[4]. I. F. Akyildiz, B. F. Lo, and R. Balakrishnan, "Cooperative


spectrum sensing in cognitive radio networks: a survey,"
Fig II. Plot ofEnergy consumption Vs Probability ofPU being inactive
Elsevier Physical Commun.,vol. 4, no. I, pp. 40-62, Mar.
201!.

2064
This fUll-text paper was peer-reviewed and accepted to be presented at the IEEE WiSPNET 2016 conference.

[5]. Bebei Wang, Yongle Wu, K. J. Ray Liu,"Game theory for sensing," IEEE Trans. Veh.Technol., vol. 60, no. 1, pp.
cognitive radio networks: An overview," Elsevier journal 276-297, Jan. 2011.
on Computer networks, 2010. [8]. Xiaolei Hao, Man Hon Cheung, Vincent W. S. , "Hedonic
[6]. W. Saad, Z. Han, M. Debbah, A. Hjorungnes, and T. Basar, Coalition Formation Game for Cooperative Spectrum
"Coalitional game theory for communication networks," Sensing and Channel Access in Cognitive Radio
IEEE Signal Processing Magazine, vol. 26, no. 5, pp. 77- Networks", IEEE Transactions On Wireless
97, Sep. 2009. Communications, Vol. II, No. 11, 2012.

[7]. W. Saad, Z. Han, T. Basar, M. Debbah, and A. Hjorungnes,


"Coalition formation games for collaborative spectrum

2065

You might also like