Coordination of Energy Efficient Sensing and
Transmission in Cognitive Radio Networks
Wuchen Tang, Muhammad Ali Imran, Rahim Tafazolli
Centre for Communication Systems Research, University of Surrey
Guildford GU1 7XH, Surrey, United Kingdom
Email: [email protected], [email protected], [email protected]
Abstract—To improve spectrum efficiency, cognitive radio (CR)
is designed to allow unlicensed (secondary) users to utilize the
licensed spectrum when it is not used by the licensed (primary)
users. Because of the dynamic nature of cognitive radio network,
the resource allocation targeting at best system performance
becomes more and more important. In this paper, a proposal
aiming at improving the energy efficiency for cognitive radio
networks and at the same time keeping an optimal system
performance, is provided. On one hand, the secondary users aim
to accurately sense the spectrum availability which costs them
in terms of energy. On the other hand higher sensing accuracy
is required to minimize the interference with primary users. An
optimal balance between these contradicting goals is required.
In this work, we simulate the eigenvalue ratio based detector
to evaluate the relation between sensing accuracy and energy
consumption. A sensing-transmission structure of secondary user
in CR networks is introduced and the channel access method is
simulated to analyze the normalized throughput.
Index Terms—Cognitive radio network; spectrum sensing;
network throughput; energy efficiency; eigenvalue ratio based
detection.
I. I NTRODUCTION
As one of the important finite resource in wireless communication, spectrum bands should be utilized efficiently. However
it is reported by Federal Communications Commission(FCC)
that 70% of the allocated spectrum in US is not fully utilized
[1]. Under this motivation, the concept of cognitive radio
was firstly provided by [2] in 1999. With cognitive radio
technology secondary user is allowed to use free spectrum
bands licensed to primary users [3], [4], [5]. At the same
time, the primary user should be protected from interference
brought by secondary users. Besides spectrum efficiency, energy consumption is another significant parameter for current
wireless communication system because of slowly developing
battery technology and the environment issue. So how to use
cognitive radio technology in an more green way should be
carefully considered.
This paper provides a proposal aiming at improving the
energy efficiency for cognitive radio networks and at the same
time keeping an optimal system performance. Main objective
is to design an optimal scenario on how to build a balance
between energy consumptions of different functions in CR
networks and the designed optimal scenario should
ISBN: 978-1-902560-25-0 © 2011 PGNet
achieve the highest system throughput with energy constraint
• achieve the shortest transmission delay with energy constraint
• be able to implement coexistence of numbers of CR
networks with best performance
The main functions in cognitive radio networks are nothing
but sensing and data transmission. To build such a balance,
the relations between energy consumption and these two
functions should be analyzed, respectively. The rest of this
paper is organized as follows. Section II consists of two
subsections. Subsection A presents the state of art on spectrum sensing technology. Subsection B gives an overview on
energy consumption and throughput analysis in CR networks;
Methodology and simulation results of these two aspects are
given in section III. Finally we conclude the paper in Section
IV.
•
II. S TATE OF ART AND P ROGRESS BEYOND
In this part, the state of art on spectrum sensing algorithm,
and on energy consumption and throughput analysis in CR
networks are presented. Some current progress are introduced
and explained.
A. Spectrum sensing algorithms in CR networks
In cognitive radio networks, depending on sensing results
the secondary user will choose to access licensed channels
or to keep silent. So the spectrum sensing accuracy is a very
important parameter when an optimal scenario is designed.
The present literature for spectrum is still in its early stages
of development [6]. A number of different methods are proposed for identifying the presence of signal transmissions,
which could be divided into energy detection based sensing,
waveform based sensing, cyclostationarity based sensing, covariance based sensing and matched filter sensing. As one of
covariance based sensing methods, eigenvalues based detection
of received signal covariance matrix is currently one of the
most effective solutions. Although the exact eigenvalues ratio
distribution is difficult to compute, this method is still an
ideal solution because it does not require prior information
of primary signal [7], [8], [9].
For the current progress on this issue, the close-formed expression of exact eigenvalues ratio distribution is exceptionally
Sensing
Fig. 1.
Data Transmission
Frame structure of secondary user.
complex to compute in practice. So an non-asymptotic spectrum sensing approach to approximate the extreme eigenvalues
was introduced [10] and an upperbound of extreme eigenvalues joint distribution was cited to derive the eigenvalues
ratio distribution [11]. All the corresponding analytical results
were compared with simulation results. Furthermore, another
eigenvalue based detector named geometric mean detector was
introduced [12] and the performance of the proposed detector
outperforms the traditional energy detector.
B. Energy consumption and throughput analysis in CR networks
To achieve the objective of this scenario, a tradeoff between
energy consumption and other system performance parameters
should be targeted. For the sensing function of CR networks,
[13] gave an energy efficient distributed spectrum sensing technology based on the combination of censoring and sleeping
policies. The results showed that an optimal sleeping and
censoring parameters could be found to achieve significant
energy savings. For data transmission in CR networks, an
sensing-throughput trade off problem was formulated in [14]
to choose an optimal sensing time for maximizing the achievable throughput. And [15] proposed energy efficient based
transmission duration design and power allocation methods to
improve energy efficiency greatly while the target throughput
is also achieved. However, all these previous work considered
and optimized several parameters or functions separately, an
overall optimal strategy need to be organized.
In addition to the present work above, in this paper a CR
network with one secondary user and one primary user is
simulated to analysis the throughput of secondary user. The
sensing and transmission of secondary user are based on a
frame structure which is shown by Figure 1. A frame consists
of fixed sensing length τ and various transmission length T −τ .
The methodology and simulation results along with eigenvalue
ratio based detection in last subsection are given in the next
section.
III. M ETHODOLOGY AND S IMULATION R ESULTS
A. Eigenvalue ratio based detection in CR networks
To design the energy efficient scenario for CR networks,
we need to find how the consumed energy impacts the sensing
accuracy. Sensing accuracy always depends on the the number
of samples N and the number of sensors K. The bigger of N
and K are, the more energy consumed for each sensing. Here
we take the eigenvalue ratio(ER) based detection for example.
Consider there are K collaborating secondary sensors and
each sensor collects N samples during the sensing time. The
collected samples from the K sensors will be forwarded to a
fusion center for combined processing. The combined K × N
data matrix X is


x1,1 x1,2 · · · x1,N
 x2,1 x2,2 · · · x2,N 


(1)
X= .
..
.. 
..
 ..
.
.
. 
xK,1 xK,2 · · · xK,N
Detection problem is based on two hypotheses: H0 and H1 ,
where H0 denotes the target spectrum band is not occupied
by a primary signal (available):
H0 : xk,n = nk,n
(2)
and H1 denotes the target spectrum band is occupied by a
primary signal (not available):
H1 : xk,n = hk,n sn + nk,n
(3)
where k = 1, ..., K and n = 1, ..., N . nk,n here is complex
Gaussian noise with zero mean and unit variance. In (3),
the vector hk,n typically represents the propagation channel
between primary user and secondary users and s(n) stands
for the source signal to be detected. The received covariance
matrix is R = XX H , where H is the Hermitian conjugate
operator, and the largest and smallest eigenvalues of this matrix
are λ1 and λK respectively. The test statistic proposed for
eigenvalue based detection is
Z=
λ1
λK
(4)
Then Z is compared with the decision threshold γ to decide
if the target spectrum resource is occupied or not. If Z < γ
the detector outputs H0 , otherwise H1 . The decision threshold
should be precalculated, which is determined by the distribution of the test statistic Z.
Based on the explanations and assumptions above, we simulate this maximum-minimum eigenvalue detector with various
number of N and K to verify this issue. Two main parameters,
probabilities of missed detection and false alarm are calculated
and shown by the receiver operating characteristics (ROC)
plot. Figure 2 illustrates the performance of eigenvalue ratio
detector with various numbers of collected samples and fixed
number of sensors while Figure 3 shows the result with fixed
number of collected samples and various numbers of sensors.
Both the figures verify the fact that the bigger the values of N
and K are, the better sensing performance could be achieved,
which implies that sensing with better performance consumes
more energy.
B. Throughput maximization in CR networks
Compared with spectrum sensing, data transmission consumes much more energy. When the secondary user should
transmit and how long the transmission should last need
to be carefully considered. The former problem depends on
Throughput of Secondary User in CR Networks
Performance of Eigenvalue Ratio Based Detection
0
10
0.8
VOIP traffic
Heavy traffic
10
Normalized throughput
Miss Detection Probability
0.7
−1
−2
10
−3
10
K= 10 N=100
K= 10 N=150
K= 10 N=200
−4
10
10
−4
Fig. 2.
−3
10
0.6
0.5
0.4
0.3
−2
−1
10
10
False Alarm Probability
0.2
0
0
10
Fig. 4.
Receiver operating characteristics for ER detector.
0.05
0.15
The normalized throughput for VOIP traffic and heavy traffic.
Performance of Eigenvalue Ratio Based Detection
0
0.1
Frame duration T (sec)
Interference between Primary User and Secondary User
10
Primary user
Secondary user
Probability of collision Pc
Miss Detection Probability
0.25
−1
10
−2
10
−3
10
K=10 N=80
K=20 N=80
K=30 N=80
−4
10
10
−4
Fig. 3.
−3
10
0.2
0.15
0.1
0.05
−2
10
False Alarm Probability
−1
10
0
10
Receiver operating characteristics for ER detector.
the detection result of each sensing so here we consider
a frame structure which consists of fix sensing length and
various transmission lengths to maximize the throughput. We
assume that one secondary user and one primary user in the
CR network and the traffic of primary user is exponential
distribution, with mean of idle and occupied duration denoted
as α0 = 650 msec, α1 = 352 msec (VOIP traffic). The same
work was done in [16] so here we show the throughput not
only under VOIP traffic but also under heavy traffic with α0
= 350 msec, α1 = 600 msec. The sensing length is fixed and
set to 1 ms. We assume primary user is always transmitting
data at the rate of 20 msec/packet when it is active while the
transmitting rate of secondary user is 1 msec/packet when it
is on transmitting.
The simulation result is shown by Figure 4. It should be
note that the normalized achievable throughput is calculated
0
0
0.05
Fig. 5.
0.1
Frame duration T (sec)
0.15
Probability of collision for VOIP traffic.
by the ratio of absolute throughput to channel capacity. Figure
4 shows the trend of normalized achievable throughput of
secondary user with various transmission durations. Two traffic
models are compared and it is observed that the throughput under VOIP is higher than that under heavy traffic because under
heavy traffic the channel is more often occupied by the primary
user. Under each traffic model there is an optimal transmission
length which maximizes the normalized throughput.
Figure 5 shows the interference level between primary and
secondary user under VOIP traffic model. We assess the
interference using the probability of collision which is calculated by the ratio of collided packets to the total transmitted
packet for primary user and secondary user, respectively. Any
collided packed will be considered lost. The collisions happen
only when primary user turns on during secondary user is
transmitting data. It is observed that both of users have higher
packet lost as the frame duration increases. This is because
the longer the frame is, the higher probability that the primary
user turns on when the secondary user is transmitting is. From
the curve of primary user, we can also see the extent that how
much the primary user is protected from the interference.
IV. C ONCLUSION AND FUTURE WORK
In this paper, a proposal targeting at coordination of energy
efficient sensing and transmission in CR networks is provided.
An overall review of different functions in cognitive radio
networks and some related energy efficiency considerations
is given. Aiming at higher energy efficiency, current spectrum sensing algorithms are presented and eigenvalue based
detection is simulated to verify that the sensing accuracy
increases as sensing energy consumption increases. A basic
sensing-transmission structure in cognitive radio networks is
introduced and simulated to analyze the network throughput.
To implement this proposal, there are still other issues
to be resolved in the future. For example, how to allocate
power between sensing and transmission in CR networks for
maximizing throughput. In future most of the cognitive radio
networks will coexist with each other using CR technology
and optimal balance between sensing and transmission will
be a key to achieve high efficiency.
R EFERENCES
[1] F. C. Commission., “Spectrum policy task force report,fcc 02-155,” Tech.
Rep., 2002.
[2] I. Mitola, J. and J. Maguire, G.Q., “Cognitive radio: making software
radios more personal,” Personal Communications, IEEE, vol. 6, no. 4,
pp. 13 –18, Aug. 1999.
[3] S. Haykin, “Cognitive radio: brain-empowered wireless communications,” Selected Areas in Communications, IEEE Journal on, vol. 23,
no. 2, pp. 201 – 220, 2005.
[4] Q. Zhao and B. Sadler, “A survey of dynamic spectrum access,” Signal
Processing Magazine, IEEE, vol. 24, no. 3, pp. 79 –89, May 2007.
[5] I. Akyildiz, W.-Y. Lee, M. Vuran, and S. Mohanty, “A survey on
spectrum management in cognitive radio networks,” Communications
Magazine, IEEE, vol. 46, no. 4, pp. 40 –48, 2008.
[6] T. Yucek and H. Arslan, “A survey of spectrum sensing algorithms for
cognitive radio applications,” Communications Surveys Tutorials, IEEE,
vol. 11, no. 1, pp. 116 –130, 2009.
[7] Y. Zeng and Y. chang Liang, “Eigenvalue-based spectrum sensing
algorithms for cognitive radio,” Communications, IEEE Transactions on,
vol. 57, no. 6, pp. 1784 –1793, 2009.
[8] F. Penna, R. Garello and M. A. Spirito, “Cooperative spectrum sensing
based on the limiting eigenvalue ratio distribution in Wishart matrices,”
in IEEE Communs. Letters, vol. 13, no. 7, pp. 507–509, Jul. 2009.
[9] L. Wei and O. Tirkkonen, “Spectrum sensing with Gaussian approximated eigenvalue ratio based detection,” in Proc. IEEE Intl. Symp.
Wireless Commun. Systems, ISWCS’2010, pp. 961–965, Sep. 2010.
[10] M. Z. Shakir, W. Tang, A. Rao, M. A. Imran, M. Alouini, “Eigenvalue
Ratio Detection Based On Exact Analytical Moments Of Smallest And
Largest Eigenvalues,” to appear in ICST Conf. Cognitive Radio Oriented Wireless Networks and Communications, CrownCom2011, Osaka,
Japan, May 2011.
[11] M. Z. Shakir, A. Rao, W. Tang, M. A. Imran, M. Alouini , “Geometric
Mean Detector For Spectrum Sensing,” submitted to IEEE Communication Theory Workshop, CTW2011, Sitges, Spain, Jun. 2011.
[12] M. Z. Shakir, W. Tang, M. A. Imran, M. Alouini, “Collaborative
spectrum sensing based on the upper bound on joint PDF of extreme eigenvalues,” submitted to European Conf. Signal Processing,
EUSIPCO2011, Barcelona, Spain, Aug. 2011.
[13] S. Maleki, A. Pandharipande, and G. Leus, “Energy-efficient distributed
spectrum sensing for cognitive sensor networks,” Sensors Journal, IEEE,
vol. 11, no. 3, pp. 565 –573, 2011.
[14] Y.-C. Liang, Y. Zeng, E. Peh, and A. T. Hoang, “Sensing-throughput
tradeoff for cognitive radio networks,” Wireless Communications, IEEE
Transactions on, vol. 7, no. 4, pp. 1326 –1337, 2008.
[15] L. Li, X. Zhou, H. Xu, G. Li, D. Wang, and A. Soong, “Energy-efficient
transmission in cognitive radio networks,” in Consumer Communications
and Networking Conference (CCNC), 2010 7th IEEE, 2010, pp. 1 –5.
[16] Y. Pei, A. T. Hoang, and Y.-C. Liang, “Sensing-throughput tradeoff in
cognitive radio networks: How frequently should spectrum sensing be
carried out?” in Personal, Indoor and Mobile Radio Communications,
2007. PIMRC 2007. IEEE 18th International Symposium on, 2007, pp.
1 –5.