scispace - formally typeset
Search or ask a question
Journal ArticleDOI

Performance Analysis of Selective Opportunistic Spectrum Access With Traffic Prediction

TL;DR: The proposed SOSA scheme can decrease the probability of packet losses in the discontinuous spectrum environment and improve the spectrum efficiency, and the practical issues encountered by an SU in a wireless environment are considered.
Abstract: In cognitive radio (CR) networks, the ability to capture a frequency slot for transmission in an idle channel has a significant impact on the spectrum efficiency and quality of service (QoS) of a secondary user (SU). The radio frequency (RF) front-ends of an SU have limited bandwidth for spectrum sensing with the target frequency bands dispersed in a discontinuous manner. This results in the SU having to sense multiple target frequency bands in a short period of time before selecting an appropriate idle channel for transmission. This paper addresses this technical challenge by proposing a selective opportunistic spectrum access (SOSA) scheme. With the aid of statistical data and traffic prediction techniques, our SOSA scheme can estimate the probability of a channel appearing idle based on the statistics and choose the best spectrum-sensing order to maximize spectrum efficiency and maintain an SU's connection. By means of doing so, this SOSA scheme can preserve the QoS of an SU while improving the system efficiency. In contrast to previous work, we consider the practical issues encountered by an SU in a wireless environment, such as discontinuous target frequency bands and limited spectrum-sensing ability. We examine the spectrum-sensing scheme in terms of packet loss ratio (PLR) and throughput. The simulation results show that the proposed SOSA scheme can decrease the probability of packet losses in the discontinuous spectrum environment and improve the spectrum efficiency.

Summary (3 min read)

Introduction

  • As far as the SU’s performance is concerned, SUs within the CR system should intelligently search and exploit vacant channel resources via dynamic spectrum access.
  • The authors derive the theoretical upper bound for an SU’s system performance by employing traffic prediction, which in turn could serve as a performance evaluation criterion for different traffic prediction methods in CR networks.
  • The simulation results are analyzed in Section VI.

A. Generic Selective Spectrum-Sensing Cycle

  • In [4], the SU is defined as a machine that can learn from the surrounding environment in an intelligent manner and adjust its transmission parameters to meet certain objectives, such as link reliability and transmission rate, in the light of the learning outcome.
  • 2) Learning stage—This phase involves the analysis of statistical data, the modeling of PU traffic, and the estimation of parameters serving as inputs to the spectrum sensing and access strategy.
  • 4) Adaption stage—This final phase involves adjusting the transmitter and receiver parameters according to the outputs acquired from the spectrum-sensing and access strategy.
  • First, using a sniffing process, the SU collects, stores, and updates information regarding the usage of target frequency bands.
  • In their work, the authors assume that spectrum overlay is the only access method available to an SU, unless stated otherwise.

B. System Architecture and Notation

  • The authors consider a distributed slotted wireless access environment where the multiple frequency bands span a wide spectrum range.
  • In their work, these frequency bands are defined as primary channels and are independent of each other.
  • All primary channels serve as target-sensing channels for an SU, with some of them acting as potential access channels for an SU in a specific time slot.
  • The following two sections present the design assumptions and the implementation issues associated with the PU and SU systems, respectively.

C. PU Traffic Model

  • PU traffic can generally be modeled in one of two distinct ways: using either a deterministic model or a stochastic model, depending on the traffic pattern of the primary channels under consideration.
  • In the current CR research, the traffic activity for PUs operating on a licensed frequency band is modeled as an alternating renewal process consisting of busy and idle periods [3], [11], which correspond to the stochastic model.
  • The authors use binary digits 0 and 1 to accordingly denote the idle state (OFF) and the busy state (ON) for each primary channel in every time slot.
  • In the aforementioned literature, the ON and OFF periods of the primary channels are independent identically distributed (i.i.d.), where the alternating renewal process is modeled as a two-state birth–death process with death rate αn and birth rate βn.
  • The lengths of the OFF and ON periods follow an exponential distribution with mean value equal to αn and βn, respectively [11].

D. SU Spectrum Sensing and Access

  • The first function validates the real state of the primary channels to be accessed to avoid interference.
  • The second function involves collecting long-term usage data with regard to the activity taking place on the primary channels to aid traffic prediction.
  • Existing spectrum sensing technologies cannot achieve faultless spectrum detection, and undetected errors degrade system performance [12].
  • Physical layer issues are beyond the scope of this paper and, hence, are not dealt with in their work.
  • Following transmission, the SU continues to sense the remaining primary channels to collect their channel state, which will subsequently be used to predict the channel states in the next time slot.

IV. SELECTIVE OPPORTUNISTIC SPECTRUM ACCESS

  • The authors formulate their SOSA framework by presenting their theoretical models evaluating SU system performance under constant and variable bit rate (CBR and VBR) traffic.
  • At the start of each time slot, the SU needs to determine the sensing order by jointly considering the transmission capacity of each channel and its probability of appearing idle in the next time slot.
  • The authors also assume that the SU knows its transmission rate on the primary channels; in other words, the SU knows the channel capacity vector Ci in time slot i.
  • 7) If deemed necessary, then the SU adjusts the predictor parameters according to the outcome obtained from the sniffing stage in the last time slot.
  • The vector ŵi+1, which represents the probability of a channel being idle in the next time slot, is determined by the channel’s long-term statistical characteristics.

V. PERFORMANCE ANALYSIS

  • The authors will evaluate system performance in terms of two key performance metrics, namely, packet loss ratio (PLR) and throughput.
  • The transmission quality of an SU is evaluated under CBR and VBR traffic in terms of PLR and throughput, respectively.
  • In the following two sections, the authors derive the theoretical models to evaluate the performance of CR systems with and without traffic prediction.
  • The authors determine the system performance upper bound under the assumption that the SU has limited spectrum-sensing and access capability and is operating in a wireless environment, which consists of discontinuous target frequency bands.
  • Since all the primary channels are i.i.d., it follows that each channel has the same probability of being idle, which is denoted by P .

A. PLR

  • For CBR traffic, the target SU needs to transmit R information bits in one time slot.
  • The authors define the packet loss as the event during which an SU cannot find an idle channel for transmission that has already sensed S primary channels, where S is the sensing threshold.
  • The authors can identify three different scenarios in relation to packet loss.
  • The authors define the joint pdf of these three random variables as VK,M,L(k,m, l).

VI. SIMULATION RESULTS AND ANALYSIS

  • In the previous sections, the authors introduced and developed their proposed selective opportunistic spectrum-sensing and access cycle.
  • The authors present the corresponding simulation results, which were generated using MATLAB, as well as the numerical results obtained from theory.
  • All primary channels have the same traffic model but with different traffic characteristics.
  • The lengths of the idle and busy channel state periods follow an exponential distribution.
  • The magnitudes of these probabilities are then used to determine the optimum spectrum-sensing order.

A. PLR Analysis

  • The authors compare the PLRs for the three scenarios identified earlier, namely, the spectrum-sensing cases with and without traffic prediction, as well as the system performance upper bound.
  • Packet loss occurs when the SU fails to find an idle channel during the sensing time threshold S. In Fig. 3, the authors compare the results under varying threshold values for a fixed number of primary channels.
  • This curve corresponding to the case with traffic prediction should always lie between the curve representing the lowest PLR and the curve corresponding to the case where the sensing order remains unaltered.
  • Once again, the two curves representing the simulation and theoretical results for the case where traffic prediction is not employed almost overlap with each other.
  • As can be seen in Fig. 4, the PLR for the case without traffic prediction does not improve as the authors increase the number of available primary channels.

VII. CONCLUSION

  • The authors have proposed an intelligent selective opportunistic spectrum-sensing and access cycle to enable CR users to make efficient use of the wireless spectrum, which consists of discontinuous frequency bands.
  • 2F1 in (25) is the regularized hypergeometric function.
  • Yum, “Analysis of cognitive radio spectrum access with optimum channel reservation,” IEEE Commun.
  • He is currently working toward the D.Eng. degree with the Telecommunications Research Group, University College London, London, U.K.

Did you find this useful? Give us your feedback

Content maybe subject to copyright    Report

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 59, NO. 4, MAY 2010 1949
Performance Analysis of Selective Opportunistic
Spectrum Access With Traffic Prediction
Guangxiang Yuan, Ryan C. Grammenos, Yang Yang, Member, IEEE, and Wenbo Wang, Member, IEEE
Abstract—In cognitive radio (CR) networks, the ability to cap-
ture a frequency slot for transmission in an idle channel has a
significant impact on the spectrum efficiency and quality of service
(QoS) of a secondary user (SU). The radio frequency (RF) front-
ends of an SU have limited bandwidth for spectrum sensing with
the target frequency bands dispersed in a discontinuous manner.
This results in the SU having to sense multiple target frequency
bands in a short period of time before selecting an appropriate
idle channel for transmission. This paper addresses this technical
challenge by proposing a selective opportunistic spectrum access
(SOSA) scheme. With the aid of statistical data and traffic pre-
diction techniques, our SOSA scheme can estimate the probability
of a channel appearing idle based on the statistics and choose the
best spectrum-sensing order to maximize spectrum efficiency and
maintain an SU’s connection. By means of doing so, this SOSA
scheme can preserve the QoS of an SU while improving the system
efficiency. In contrast to previous work, we consider the practical
issues encountered by an SU in a wireless environment, such
as discontinuous target frequency bands and limited spectrum-
sensing ability. We examine the spectrum-sensing scheme in terms
of packet loss ratio (PLR) and throughput. The simulation results
show that the proposed SOSA scheme can decrease the probability
of packet losses in the discontinuous spectrum environment and
improve the spectrum efficiency.
Index Terms—Cognitive radio (CR), spectrum sensing, traffic
prediction.
I. INTRODUCTION
C
URRENT radio systems employ inflexible spectrum al-
location strategies, resulting in inefficient spectrum uti-
lization [1]. Cognitive radio (CR), which has attracted the
attention of many researchers across the globe, is a novel
technology aimed at more efficiently utilizing the spectrum [2].
Manuscript received May 25, 2009; revised October 2, 2009. First published
December 28, 2009; current version published May 14, 2010. This work was
supported in part by the China Scholarship Council under the postgraduate
student exchange program by the Ministry of Science and Technology of China
under Grant 2009DFB13080, by the National Basic Research Program of China
(973) under Grant 2007CB310602, and by the Research Councils U.K. under
the U.K.–China Science Bridges Project: R&D on (B)4G Wireless Mobile
Communications (EP/G042713/1). The review of this paper was coordinated
by Prof. H. Aghvami.
G. Yuan and W. Wang are with the Wireless Signal Processing and Networks
Laboratory, Beijing University of Posts and Telecommunications, Beijing
100876, China (e-mail: gxyuan@gmail.com; wbwang@bupt.edu.cn).
R. C. Grammenos is with the Department of Electronic and Electrical
Engineering, University College London, WC1E 7JE London, U.K. (e-mail:
r.grammenos@ee.ucl.ac.uk).
Y. Yang is with the Department of Electronic and Electrical Engineering,
University College London, WC1E 7JE London, U.K., and also with the
Shanghai Research Center for Wireless Communications, Chinese Academy
of Sciences, Shanghai 200335, China (e-mail: y.yang@ee.ucl.ac.uk).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TVT.2009.2039155
CR technology enables a secondary user (SU) with cognitive
access ability to use the idle channel resources of the primary
radio systems. This idle channel resource, which is temporarily
unoccupied by the primary user (PU), can be used by the SU
to exchange information, provided it does not interfere with the
smooth communication of the primary system [3], [4].
Current research in CR focuses on solving two main chal-
lenges, namely, 1) the avoidance of interference to the normal
communication of primary systems and 2) the performance
optimization of SU’s transmission. For the former challenge,
several methods for accurate signal detection have been pro-
posed, including matched filter detection, energy detection, and
cyclostationary feature detection [3]. To combat the negative
effect caused by fading to the transmitter detection accuracy,
a novel detection method based on CR user cooperation is
proposed in [5] and [6]. As far as the SU’s performance is con-
cerned, SUs within the CR system should intelligently search
and exploit vacant channel resources via dynamic spectrum
access. Nevertheless, SUs do not have the ability to gather data
from the primary systems; hence, they can only obtain informa-
tion regarding the channel status of different frequency bands
using periodic spectrum detection, which in turn consumes a
nonnegligible length of time [11].
With regard to the deployment of wireless access networks
in practice, SUs will be present in an environment comprising
numerous wireless access networks spanning a wide spectrum
range in a discontinuous manner [1]. Consequently, the SU is
likely to lose some transmission opportunities made available
from vacant channels, unless it can support spectrum sensing
over a large frequency range spanning several gigahertz. State-
of-the-art radio frequency hardware capable of sensing weak
signals over a large dynamic range, such as multispeed analog-
to-digital converters with a high resolution, wideband antennas,
and highly linear power amplifiers, are still subject to many
application constraints [3]. Thereby, an SU will have limited
ability to carry out spectrum sensing and access, leading to an
overall decrease in throughput.
In this paper, we propose a generic selective opportunis-
tic spectrum access (SOSA) framework to carry out intelli-
gent and selective spectrum sensing and access. This SOSA
framework enables an SU to sense and select target spectrum
bands in an optimum order while achieving quaity-of-service
(QoS) requirements and maximizing spectrum efficiency. This
is achieved with the aid of traffic prediction, which allows an
SU to determine the optimum channel-sensing order by taking
into account the probability of a channel appearing idle in the
next time slot and its QoS requirements. In this paper, we also
define the criteria and devise a theoretical model for evaluating
0018-9545/$26.00 © 2010 IEEE

1950 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 59, NO. 4, MAY 2010
system performance under the application of prediction-based
spectrum sensing and access. We derive the theoretical upper
bound for an SU’s system performance by employing traffic
prediction, which in turn could serve as a performance eval-
uation criterion for different traffic prediction methods in CR
networks.
The rest of this paper is organized as follows: In Section II,
the related work on traffic prediction and its application in CR
are summarized. In Section III, the proposed generic intelligent
spectrum sensing cycle is described, and the system parameters
for our research scenario, the PU traffic model, and the SOSA
scheme are defined. Section IV examines the proposed selective
spectrum sensing and access strategy in detail. In Section V, the
performance evaluation models are presented, and the SOSA
scheme performance is investigated. The simulation results are
analyzed in Section VI. Finally, the conclusions are given in
Section VII.
II. R
ELATED WORK
The ability to predict the variation of parameters reflect-
ing network conditions, such as bandwidth, interference level,
bit rate, and bit error rate, in future time intervals is a key
challenge in network management and control. As a result,
traffic prediction i s becoming increasingly important. Previous
work has focused on improving network efficiency and QoS
performance through traffic control and avoidance of traffic
congestion. Fuzzy logic and linear autoregressive prediction
algorithms have been implemented for this case [14], [15].
CR technology enables users to share the available spectrum
in an opportunistic manner. Current research in CR networks
focuses on optimizing specific parameters, such as link outage
probability and delay, rather than capturing transmission op-
portunities at the physical layer to maximize the overall system
throughput. As far as spectrum handoff is concerned, [8] and [9]
analyze the probability of link maintenance during a spectrum
handoff by employing different spectrum handoff schemes or a
different number of redundant channels, respectively. In [10], a
protocol is proposed to support link maintenance during a spec-
trum handoff for orthogonal frequency-division-multiplexing-
based CR systems. In recent years, several traffic-prediction
methods have been introduced into CR technology. Reference
[13] presents different prediction rules for different types of
ON/OFF traffic models with the aim of minimizing the number
of handovers between channels, as well as the associated delay.
In [14], an autoregressive spectrum hole prediction model is
presented, which adopts an AR-2 model with a Kalman filter to
reduce the collisions between PUs and SUs. In [15], a proactive
spectrum-access approach is introduced, which differs from the
existing reactive sense-and-avoid approaches in that channel
histories are utilized to make predictions regarding future spec-
trum conditions. Furthermore, the idea of predictive dynamic
spectrum access is also explored and developed in [15], which
is then implemented in [16] using cyclostationary detection and
hidden Markov models.
Although several models that employ traffic prediction have
been proposed, none of them consider traffic prediction and
spectrum sensing in conjunction in a practical heterogeneous
Fig. 1. Generic selective spectrum-sensing and access cycle.
wireless network environment where SUs have limited sensing
and access ability. Furthermore, there is still no unified theoret-
ical model for evaluating system performance when proactive
spectrum selection and sensing are employed.
III. S
YSTEM MODEL AND PROBLEM FORMULATION
A. Generic Selective Spectrum-Sensing Cycle
In [4], the SU is defined as a machine that can learn from the
surrounding environment in an intelligent manner and adjust
its transmission parameters to meet certain objectives, such as
link reliability and transmission rate, in the light of the learning
outcome. We can divide the CR entire sensing and access
process into four parts: 1) sniffing;2)learning;3)decision; and
4) adaption.
1) Sniffing stage—During this stage, relevant information
concerning the surrounding wireless access networks is
collected and stored, such as bandwidth, traffic operation,
interval duration, and channel quality.
2) Learning stage—This phase involves the analysis of sta-
tistical data, the modeling of PU traffic, and the esti-
mation of parameters serving as inputs to the spectrum
sensing and access strategy.
3) Decision stage—In this stage, the spectrum sensing order
is determined, as well as the access scheme for the next
time unit, depending on the values of the input parameters
obtained from the learning stage.
4) Adaption stage—This final phase involves adjusting the
transmitter and receiver parameters according to the
outputs acquired from the spectrum-sensing and access
strategy.
The generic spectrum-sensing and access framework, which
is based on the foregoing four operations, is illustrated in Fig. 1.
First, using a sniffing process, the SU collects, stores, and
updates information regarding the usage of target frequency
bands. During the learning stage, the SU will estimate the PU
traffic pattern and the relevant traffic parameters with the aim of
predicting the PU traffic trend for a future time interval. The de-
cision stage is where the SU will determine the optimum sens-
ing order based on QoS requirements, the probability of each
channel appearing idle in the next time interval, and the trans-
mission capability of each frequency band under consideration.
The spectrum sensing and access operation will take place

YUAN et al.: PERFORMANCE ANALYSIS OF SOSA WITH TRAFFIC PREDICTION 1951
during the adaptation phase based on the outcome obtained
from the decision phase. This adaptation may involve tuning
transceiver parameters to adapt to the instantaneous channel
variations. In our work, we assume that spectrum overlay is the
only access method available to an SU, unless stated otherwise.
Moreover, taking into account that, with spectrum overlay, the
transceiver can be configured in a more straightforward fashion
compared with the underlay method, we will also assume that
the SU’s transceiver transmits data with a fixed power, which
will not cause interference to the PUs operation in the licensed
spectrum.
B. System Architecture and Notation
We consider a distributed slotted wireless access environ-
ment where the multiple frequency bands span a wide spectrum
range. In our work, these frequency bands are defined as pri-
mary channels and are independent of each other. All primary
channels serve as target-sensing channels for an SU, with some
of them acting as potential access channels for an SU in a
specific time slot. The following notations are used throughout
the remaining sections of this paper:
N total number of target frequency bands or primary chan-
nels that can be accessed by the SU;
K random variable representing the number of channels
that are idle in a time slot, where 0 K N;
i time slot index in the primary s ystem;
w
i
n
channel state of the nth channel at time slot i. W
n
(i)=1
means that the channel is occupied by a PU, namely ON
(Busy) state; W
n
(i)=0means that the channel is not
being used by a PU, namely OFF (Idle) state;
T
s
time slot length in the primary system;
C
i
n
transmission capacity of the target SU on the nth pri-
mary channel in the ith time slot;
t
s
average sensing time per primary channel;
t
h
average handoff time between two primary channels;
M random variable representing the number of channels
that are assumed to be busy in a time slot but are in fact
idle, where 0 M K;
L random variable representing the number of channels
that are assumed to be idle in a time slot but are in fact
busy, where 0 L N K;
R target transmission rate of an SU;
α
n
mean value of the OFF-period for the nth primary
channel;
β
n
mean value of the ON-period f or the nth primary
channel;
S spectrum sensing time threshold;
P
e
average probability of error associated with the predictor
of an SU;
P
n,i
off
predicted probability of the nth primary channel appear-
ingidleintheith time slot;
P
n,i
on
predicted probability of the nth primary channel appear-
ingbusyintheith time slot.
The following two sections present the design assumptions
and the implementation issues associated with the PU and SU
systems, respectively.
C. PU Traffic Model
PU traffic can generally be modeled in one of two distinct
ways: using either a deterministic model or a stochastic model,
depending on the traffic pattern of the primary channels under
consideration. In the current CR research, the traffic activity
for PUs operating on a licensed frequency band is modeled
as an alternating renewal process consisting of busy and idle
periods [3], [11], which correspond to the stochastic model.
An example of this alternating renewal process is illustrated in
Fig. 2(a). We use binary digits 0 and 1 to accordingly denote
the idle state (OFF) and the busy state (ON) for each primary
channel in every time slot. In the aforementioned literature, the
ON and OFF periods of the primary channels are independent
identically distributed (i.i.d.), where the alternating renewal
process is modeled as a two-state birth–death process with
death rate α
n
and birth rate β
n
. The lengths of the OFF and
ON periods follow an exponential distribution with mean value
equal to α
n
and β
n
, respectively [11].
D. SU Spectrum Sensing and Access
The spectrum-sensing procedure is indispensable to an SU
since it entails two important functions that allow an SU to ob-
tain access to primary systems. The first function validates the
real state of t he primary channels to be accessed to avoid inter-
ference. The second function involves collecting long-term us-
age data with regard to the activity taking place on the primary
channels to aid traffic prediction. Existing s pectrum sensing
technologies cannot achieve faultless spectrum detection, and
undetected errors degrade system performance [12]. However,
physical layer issues are beyond the s cope of this paper and,
hence, are not dealt with in our work. Instead, we assume that
the sensing results generated by the spectrum sensor of an SU
are accurate enough to establish the state of a primary channel.
In each time slot, the SU senses all the primary channels
one by one and transmits its data on the first idle channel that
becomes available. Following transmission, the SU continues to
sense the remaining primary channels to collect their channel
state, which will subsequently be used to predict the channel
states in the next time slot. The detailed channel-sensing and
access process is illustrated in Fig. 2(b). It is evident that the
overall time required to sense N primary channels s hould not
exceed the length of a time slot T
s
; otherwise, the SU will not
be able to learn the states of all primary channels, making it
impossible to predict the channel state for the next time interval.
IV. S
ELECTIVE OPPORTUNISTIC SPECTRUM A CCESS
In this section, we formulate our SOSA framework by pre-
senting our theoretical models evaluating SU system perfor-
mance under constant and variable bit rate (CBR and VBR)
traffic.
The wireless access environment under consideration con-
sists of N primary channels, which are indexed as 1, 2,...,N.
These N channels form a primary channel set Φ, which con-
forms to no specific sequence criteria. The transmitter can
access any of these channels without having to inform the

1952 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 59, NO. 4, MAY 2010
Fig. 2. Example of (a) the renewal process for a primary system and (b) the SU sensing process in a time slot.
primary system beforehand. Operations are carried out in the
discrete time domain, which is indexed with i.Atthestartof
each time slot, the SU needs to determine the sensing order by
jointly considering the transmission capacity of each channel
and its probability of appearing idle in the next time slot. These
operations will take place during the learning and decision
stages.
The target SU has a transmission capacity C
i
n
on the nth
primary channel in time slot i. We have defined the variable
w
i
n
to indicate the state of the nth primary channel in time
slot i taking binary values 0 or 1. In this case, 0 represents
an idle primary channel, whereas 1 represents a busy primary
channel. If the SU finds an idle channel in time slot i, then
the number of information bits in this time s lot is I(i).This
is the theoretical product of an SU’s channel capacity with its
effective transmission time in one time slot having subtracted
the time consumed for spectrum sensing and handoff. To protect
a PU’s signal, the SU is not permitted to transmit on a primary
channel before sensing and detecting its real state.
We assume that the SU obtains the accurate channel state
information vector w
i
of N primary channels, where w
i
=
[w
i
1
,w
i
2
,...,w
i
N
], and w
i
∈W, where W is the set of all
possible primary channel state combinations. We also assume
that the SU knows its transmission rate on the primary channels;
in other words, the SU knows the channel capacity vector C
i
in
time slot i. This information can be collected via the sniffing
process that takes place in each time slot. During the learning
stage, the SU will predict the primary channel states for time
slot i +1 based on the most recent channel state vector w
i
and will update the channel history record accordingly. The SU
can adjust the relevant parameters of its predictor to improve
the accuracy of the results obtained; however, this depends on
the initial design requirements outlined for the predictor. The
prediction result can be represented by the vector
ˆ
w
i+1
, which
takes into account the probability of each primary channel being
idle P
n,i+1
off
or occupied P
n,i+1
on
in time slot i +1. Without
loss of generality, we can arrange the index of the N primary
channels according to their probability of being idle P
n,i+1
off
in
time slot i +1 such that P
1,i+1
off
P
2,i+1
off
···P
N,i+1
off
0. In contrast to previous work, the SU in our case does not
predict the length of time for which a channel will remain
idle. Taking such a criterion into account for the transmission
of an SU’s data would lead to a higher probability of outage
compared with a slotted system scenario where the time unit is
considered to be fixed. We formulate this process as
L : F(w
i
, α, β)
R
N
−→
ˆ
w
i+1
,n Φ, w
i
∈W (1)
where L indicates the learning process, F represents the pre-
diction method used in the learning process, and R
N
is the
traffic record of all N primary channels. The data format for
this record will be designed according to the prediction method
employed. α and β are the vectors for the death and birth rates,
respectively, of the periods for all primary channels. The SU
has to arrange the spectrum-sensing sequence of decreasing
probability of a channel appearing idle while considering t rans-
mission requirements and sensing efficiency at the same time.
This procedure can be modeled as follows:
D : G(
ˆ
w
i+1
, C
i+1
, ψ) −→ U
,n Φ (2)
where ψ is the set of transmission requirements for the SU, G
is the ordering strategy used by the SU to determine the sensing
sequence for the primary channels, which jointly considers the
transmission requirements of the SU and the sensing efficiency,
and U
is the optimum spectrum-sensing order acquired during
the decision phase. The following pane summarizes the steps of
our proposed spectrum sensing cycle:
SOSA Scheme
1) The SU retrieves the channel statistics regarding the state
of each obtained through sniffing and based on the predictor
requirements defined at the s tart of time slot i +1.
2) The SU predicts the probability of each channel appearing
idle or occupied in time slot i +1using the predefined predic-
tion model.
3) The SU uses the predetermined ordering strategy to
arrange the sensing sequence for time slot i +1.
4) The SU begins to sense primary channels according to the
sensing order generated in step 3).

YUAN et al.: PERFORMANCE ANALYSIS OF SOSA WITH TRAFFIC PREDICTION 1953
5) The SU locks on to the first idle channel that be-
comes available, provided that there are no special transmission
requirements.
6) The SU continues to sense the remaining primary chan-
nels, finally recording the states of all N primary channels in
time slot i +1.
7) If deemed necessary, then the SU adjusts the predictor
parameters according to the outcome obtained from the sniffing
stage in the last time slot.
8) The SU repeats these steps until it reaches the last time
slot.
For the purpose of performance evaluation and comparison,
we adopt a relatively simple but effective channel-ranking
method. This method takes the long-term statistical traffic
characteristics as the critical input parameters to the learning
process formulated in (1). The two input parameters α and β
are the vectors corresponding to the mean values of the OFF and
ON periods of all primary channels, respectively. These mean
values are obtained using the maximum-likelihood estimation
method, which in turn uses statistical data relevant to the
channel state of all primary channels. For simplicity, during the
learning process, the channel state vector w
i
is not considered
in the channel-state-prediction process. The vector
ˆ
w
i+1
, which
represents the probability of a channel being idle in the next
time slot, is determined by the channel’s long-term statistical
characteristics. If we denote p
n
as the probability of the nth
primary channel appearing idle in the next time slot, then p
n
is
given by
p
n
=
α
n
α
n
+ β
n
(3)
where 1 n N . As channel-state statistical data are gradu-
ally collected over the period of time, the mean lengths of the
ON and OFF periods, as well as the probability of each primary
channel appearing idle in the next time slot, will be updated
accordingly.
Equation (2) indicates that the channel capacity and the
transmission requirements of the SU will be taken into account
during the decision procedure. In our work, we assume that all
primary channels have the same transmission capacities C
0
and
that their channel-state variation is slow enough so that it does
not distort the results obtained from our theoretical evaluation
model. Hence, C
i+1
in (2) is simplified into a constant vector.
The i nput parameter ψ in (2) reflects the SU’s transmission
requirements and should be considered in conjunction with the
transmission capacity. The optimum channel-sensing sequence
indicated by U
should be arranged in order of decreasing
probability according to the corresponding magnitudes defined
in vector
ˆ
w
i+1
.
V. P
ERFORMANCE ANALYSIS
In the previous sections, we looked at the challenges encoun-
tered by an SU when wishing to capture a transmission oppor-
tunity. In this section, we will evaluate system performance in
terms of two key performance metrics, namely, packet loss ratio
(PLR) and throughput. In this paper, the transmission quality of
an SU is evaluated under CBR and VBR traffic in terms of PLR
and throughput, respectively. In the following two sections, we
derive the theoretical models to evaluate the performance of CR
systems with and without traffic prediction. We determine the
system performance upper bound under the assumption that the
SU has limited spectrum-sensing and access capability and is
operating in a wireless environment, which consists of discon-
tinuous target frequency bands. This performance upper bound
could in turn be used to compare different traffic prediction and
spectrum-sensing methods.
In Section III, we defined the random variable K as the
number of idle primary channels in one time slot i. Hence,
the number of occupied primary channels in one time slot is
N K. Since all the primary channels are i.i.d., it follows
that each channel has the same probability of being idle, which
is denoted by P . Therefore, the random variable k
i
follows a
binomial distribution with a mean value equal to NP.
A. PLR
For CBR traffic, the target SU needs to transmit R informa-
tion bits in one time slot. We define the packet loss as the event
during which an SU cannot find an idle channel for transmission
that has already sensed S primary channels, where S is the
sensing threshold. In other words, S is defined as the maximum
number of channels the SU is permitted to sense to find a va-
cant channel. This threshold is determined by the transmission
capacity of the target primary channel to be accessed by the
SU, as well as t he number of information bits that need to
be transmitted in one time slot [19]. S is given by
S =
T
s
R/C
0
t
s
+ t
h
(4)
where 1 S<N.
1) Spectrum Sensing Without Traffic Prediction: Without
the aid of traffic prediction, capturing a transmission opportu-
nity is totally dependent on the traffic activity taking place on
each primary channel. Since the SU needs to transmit its data
at a CBR, packet loss will occur when the SU cannot locate an
idle primary channel for transmission within the sensing time
threshold S. We assume that there are K idle primary channels
in time slot i. The probability that the SU cannot find an idle
channel from K idle primary channels is given by
P
i,K
wo
=
S
j=1
1
K
N j +1
, 0 K N. (5)
This is the instantaneous packet loss probability for time
slot i. As earlier defined, K is a random variable following a
binomial distribution with mean value NP and represents the
number of idle primary channels in time slot i. Hence, the aver-
age probability of packet loss for an SU in time slot i is given by
P
i
wo
=
N
k=0
P
r
{K
i
= k}
S
j=1
1
K
i
N j +1
=
N
k=0
N
k
P
k
(1P )
Nk
S
j=1
1
k
N j +1
. (6)

Citations
More filters
Journal ArticleDOI
TL;DR: In this survey, various spectrum occupancy models from measurement campaigns taken around the world are investigated and spectrum occupancy prediction is also discussed, where autoregressive and/or moving-average models are used to predict the channel status at future time instants.
Abstract: Spectrum occupancy models are very useful in cognitive radio designs. They can be used to increase spectrum sensing accuracy for more reliable operation, to remove spectrum sensing for higher resource usage efficiency, or to select channels for better opportunistic access, among other applications. In this survey, various spectrum occupancy models from measurement campaigns taken around the world are investigated. These models extract different statistical properties of the spectrum occupancy from the measured data. In addition to these models, spectrum occupancy prediction is also discussed, where autoregressive and/or moving-average models are used to predict the channel status at future time instants. After comparing these different methods and models, several challenges are also summarized based on this survey.

221 citations


Cites methods from "Performance Analysis of Selective O..."

  • ...In [8], the authors used the spectrum occupancy information to predict the channel status in the following time slots so that optimal spectrum sensing order can be achieved by comparing them....

    [...]

Journal ArticleDOI
TL;DR: This paper proposes an intelligent and distributed channel selection strategy, SURF, that classifies the available channels and uses them efficiently to increase data dissemination reliability in multi-hop cognitive radio networks.

114 citations

Journal ArticleDOI
TL;DR: A novel MAC design for DCRN is proposed which provides an efficient approach to address quality of service requirements of delay sensitive applications by defining higher priority to such applications during channel reservation and outperforms the existing protocols.
Abstract: Dynamic resource availability and lack of central control unit offer many challenges while designing medium access control (MAC) protocol for a distributed cognitive radio network (DCRN). In this paper, we propose a novel MAC design for DCRN which provides an efficient approach to address quality of service (QoS) requirements of delay sensitive applications by defining higher priority to such applications during channel reservation. It also combats other major challenges such as efficient spectrum utilization, multi-channel hidden terminal problem (MHTP) and collision with primary user (PU) due to sensing error at SU. We develop an analytical framework to study the performance of the proposed protocol. We then compare the performance of proposed protocol with those of two existing protocols. Comparison results show that proposed MAC outperforms the existing protocols by providing better throughput and reducing DCRN users' collision probability with PUs in presence of sensing error. The results achieved from the analytical model and validated by simulations show that our simple yet efficient design identifies and fulfils the QoS requirements of delay sensitive applications, achieves excellent spectrum utilization, shows superb robustness in presence of sensing errors and handles MHTP effectively.

86 citations

Journal ArticleDOI
TL;DR: This paper develops a priority-based channel allocation scheme to assign channels to the mobile stations based on their QoE requirements, and proposes a handoff management technique to overcome the interruptions caused by the handoff.
Abstract: Cognitive radio (CR) is among the promising solutions for overcoming the spectrum scarcity problem in the forthcoming fifth-generation (5G) cellular networks, whereas mobile stations are expected to support multimode operations to maintain connectivity to various radio access points. However, particularly for multimedia services, because of the time-varying channel capacity, the random arrivals of legacy users, and the on-negligible delay caused by spectrum handoff, it is challenging to achieve seamless streaming leading to minimum quality of experience (QoE) degradation. The objective of this paper is to manage spectrum handoff delays by allocating channels based on the user QoE expectations, minimizing the latency, providing seamless multimedia service, and improving QoE. First, to minimize the handoff delays, we use channel usage statistical information to compute the channel quality. Based on this, the cognitive base station maintains a ranking index of the available channels to facilitate the cognitive mobile stations. Second, to enhance channel utilization, we develop a priority-based channel allocation scheme to assign channels to the mobile stations based on their QoE requirements. Third, to minimize handoff delays, we employ the hidden markov model (HMM) to predict the state of the future time slot. However, due to sensing errors, the scheme proactively performs spectrum sensing and reactively acts handoffs. Fourth, we propose a handoff management technique to overcome the interruptions caused by the handoff. In such a way that, when a handoff is predicted, we use scalable video coding to extract the base layer and transmit it during a certain interval time before handoff occurrence to be shown during handoff delays, hence providing seamless service. Our simulation results highlight the performance gain of the proposed framework in terms of channel utilization and received video quality.

62 citations


Cites background from "Performance Analysis of Selective O..."

  • ...where ν and υ are the rate parameter of the exponential distribution [25] for busy and idle periods with the average 1 ν and 1 υ , respectively....

    [...]

Journal ArticleDOI
TL;DR: This paper develops a Robust Online Spectrum Prediction (ROSP) framework, with incomplete and corrupted observations, and forms the ROSP as a joint optimization problem of matrix completion and recovery by effectively integrating the time series forecasting techniques and develop an alternating direction optimization method to efficiently solve it.
Abstract: A range of emerging applications, from adaptive spectrum sensing to proactive spectrum mobility, depend on the ability to foresee spectrum state evolution. Despite a number of studies appearing about spectrum prediction, fundamental issues still remain unresolved: 1) The existing studies do not explicitly account for anomalies, which may incur serious performance degradation; 2) they focus on the design of batch spectrum prediction algorithms, which limit the scalability to analyze massive spectrum data in real time; 3) they assume the historical data are complete, which may not hold in reality. To address these issues, we develop a Robust Online Spectrum Prediction (ROSP) framework, with incomplete and corrupted observations, in this paper. We first present data analytics of real-world spectrum measurements to reveal the correlation structures of spectrum evolution and to analyze the impact of anomalies on the rank distribution of spectrum matrices. Then, from a spectral–temporal 2-D perspective, we formulate the ROSP as a joint optimization problem of matrix completion and recovery by effectively integrating the time series forecasting techniques and develop an alternating direction optimization method to efficiently solve it. We apply ROSP to a wide range of real-world spectrum matrices of popular wireless services. Experiment results show that ROSP outperforms state-of-the-art spectrum prediction schemes.

49 citations


Cites background from "Performance Analysis of Selective O..."

  • ...tative applications include adaptive spectrum sensing [4]–[6], proactive spectrum mobility [7], [8], dynamic spectrum access [9]–[11], smart topology control [12], and competitive...

    [...]

References
More filters
Book
01 Jan 1943
TL;DR: Combinations involving trigonometric and hyperbolic functions and power 5 Indefinite Integrals of Special Functions 6 Definite Integral Integral Functions 7.Associated Legendre Functions 8 Special Functions 9 Hypergeometric Functions 10 Vector Field Theory 11 Algebraic Inequalities 12 Integral Inequality 13 Matrices and related results 14 Determinants 15 Norms 16 Ordinary differential equations 17 Fourier, Laplace, and Mellin Transforms 18 The z-transform
Abstract: 0 Introduction 1 Elementary Functions 2 Indefinite Integrals of Elementary Functions 3 Definite Integrals of Elementary Functions 4.Combinations involving trigonometric and hyperbolic functions and power 5 Indefinite Integrals of Special Functions 6 Definite Integrals of Special Functions 7.Associated Legendre Functions 8 Special Functions 9 Hypergeometric Functions 10 Vector Field Theory 11 Algebraic Inequalities 12 Integral Inequalities 13 Matrices and related results 14 Determinants 15 Norms 16 Ordinary differential equations 17 Fourier, Laplace, and Mellin Transforms 18 The z-transform

27,354 citations

Journal ArticleDOI
Simon Haykin1
TL;DR: Following the discussion of interference temperature as a new metric for the quantification and management of interference, the paper addresses three fundamental cognitive tasks: radio-scene analysis, channel-state estimation and predictive modeling, and the emergent behavior of cognitive radio.
Abstract: Cognitive radio is viewed as a novel approach for improving the utilization of a precious natural resource: the radio electromagnetic spectrum. The cognitive radio, built on a software-defined radio, is defined as an intelligent wireless communication system that is aware of its environment and uses the methodology of understanding-by-building to learn from the environment and adapt to statistical variations in the input stimuli, with two primary objectives in mind: /spl middot/ highly reliable communication whenever and wherever needed; /spl middot/ efficient utilization of the radio spectrum. Following the discussion of interference temperature as a new metric for the quantification and management of interference, the paper addresses three fundamental cognitive tasks. 1) Radio-scene analysis. 2) Channel-state estimation and predictive modeling. 3) Transmit-power control and dynamic spectrum management. This work also discusses the emergent behavior of cognitive radio.

12,172 citations


"Performance Analysis of Selective O..." refers background in this paper

  • ...In [4], the SU is defined as a machine that can learn from the surrounding environment in an intelligent manner and adjust its transmission parameters to meet certain objectives, such as link reliability and transmission rate, in the light of the learning outcome....

    [...]

  • ...to exchange information, provided it does not interfere with the smooth communication of the primary system [3], [4]....

    [...]

Journal ArticleDOI
TL;DR: With RKRL, cognitive radio agents may actively manipulate the protocol stack to adapt known etiquettes to better satisfy the user's needs and transforms radio nodes from blind executors of predefined protocols to radio-domain-aware intelligent agents that search out ways to deliver the services the user wants even if that user does not know how to obtain them.
Abstract: Software radios are emerging as platforms for multiband multimode personal communications systems. Radio etiquette is the set of RF bands, air interfaces, protocols, and spatial and temporal patterns that moderate the use of the radio spectrum. Cognitive radio extends the software radio with radio-domain model-based reasoning about such etiquettes. Cognitive radio enhances the flexibility of personal services through a radio knowledge representation language. This language represents knowledge of radio etiquette, devices, software modules, propagation, networks, user needs, and application scenarios in a way that supports automated reasoning about the needs of the user. This empowers software radios to conduct expressive negotiations among peers about the use of radio spectrum across fluents of space, time, and user context. With RKRL, cognitive radio agents may actively manipulate the protocol stack to adapt known etiquettes to better satisfy the user's needs. This transforms radio nodes from blind executors of predefined protocols to radio-domain-aware intelligent agents that search out ways to deliver the services the user wants even if that user does not know how to obtain them. Software radio provides an ideal platform for the realization of cognitive radio.

9,238 citations


"Performance Analysis of Selective O..." refers background in this paper

  • ...Cognitive radio (CR), which has attracted the attention of many researchers across the globe, is a novel technology aimed at more efficiently utilizing the spectrum [2]....

    [...]

Journal ArticleDOI
TL;DR: The novel functionalities and current research challenges of the xG networks are explained in detail, and a brief overview of the cognitive radio technology is provided and the xg network architecture is introduced.

6,608 citations


"Performance Analysis of Selective O..." refers background or methods in this paper

  • ...as an alternating renewal process consisting of busy and idle periods [3], [11], which correspond to the stochastic model....

    [...]

  • ...For the former challenge, several methods for accurate signal detection have been proposed, including matched filter detection, energy detection, and cyclostationary feature detection [3]....

    [...]

  • ...Stateof-the-art radio frequency hardware capable of sensing weak signals over a large dynamic range, such as multispeed analogto-digital converters with a high resolution, wideband antennas, and highly linear power amplifiers, are still subject to many application constraints [3]....

    [...]

  • ...to exchange information, provided it does not interfere with the smooth communication of the primary system [3], [4]....

    [...]

Journal ArticleDOI
24 Apr 2009
TL;DR: This information-theoretic survey provides guidelines for the spectral efficiency gains possible through cognitive radios, as well as practical design ideas to mitigate the coexistence challenges in today's crowded spectrum.
Abstract: Cognitive radios hold tremendous promise for increasing spectral efficiency in wireless systems. This paper surveys the fundamental capacity limits and associated transmission techniques for different wireless network design paradigms based on this promising technology. These paradigms are unified by the definition of a cognitive radio as an intelligent wireless communication device that exploits side information about its environment to improve spectrum utilization. This side information typically comprises knowledge about the activity, channels, codebooks, and/or messages of other nodes with which the cognitive node shares the spectrum. Based on the nature of the available side information as well as a priori rules about spectrum usage, cognitive radio systems seek to underlay, overlay, or interweave the cognitive radios' signals with the transmissions of noncognitive nodes. We provide a comprehensive summary of the known capacity characterizations in terms of upper and lower bounds for each of these three approaches. The increase in system degrees of freedom obtained through cognitive radios is also illuminated. This information-theoretic survey provides guidelines for the spectral efficiency gains possible through cognitive radios, as well as practical design ideas to mitigate the coexistence challenges in today's crowded spectrum.

2,516 citations

Frequently Asked Questions (1)
Q1. What have the authors contributed in "Performance analysis of selective opportunistic spectrum access with traffic prediction" ?

This paper addresses this technical challenge by proposing a selective opportunistic spectrum access ( SOSA ) scheme. In contrast to previous work, the authors consider the practical issues encountered by an SU in a wireless environment, such as discontinuous target frequency bands and limited spectrumsensing ability. The authors examine the spectrum-sensing scheme in terms of packet loss ratio ( PLR ) and throughput.