Al

madarJ
our
nal
I
SSN2313-
156X

f
orCommuni
cat
ions
,Inf
ormat
ionT
echnol
ogy
,andAppl
icat
ions
J
uly2
01,Vol
4 .0,No.0
1 1

Cogni
ti
veRadi
o
Pr
omi
singT
echnol
ogyf
ort
heConges
tedSpec
trum

Pu
bl
is
he
dbyR&DO ce
Al
mad
arAl
j
adi
dCo
.Li
by
a
www.
al
mada
r.l
y
 Almadar Journal for Communications, Information
Technology, and Applications
Vol. 01, No. 01, July 2014
AJCITA
Almadar Journal for Communications, Information Tech-
nology, and Applications (AJCITA) (ISSN 2313−156X),
(Local registration number at the Libyan National Li-
brary 369/2014) is published semi-annually by the Research
and Development Office, Almadar Aljadid Co. Address:
Gurji−Tripoli−Libya, P.O.Box 83792, +218 91 919 0500, Ext.
3520, ajcita@almadar.ly, www.almadar.ly.
AJCITA aims to provide high-level scientific papers with
value in academic research and industry. The published papers
can be:
• Original research articles with clear contributions,
• Review articles,
• Application articles with industrial values,
• Invited articles.
D IRECTOR OF J OURNAL
Abdulla A. Abouda, Manger of Research and De-
velopment Office, Almadar Aljadid Co. Libya, Email:
a.abouda@almadar.ly
E DITOR IN C HIEF
Mohammed S. Elmusrati, Professor and Head of Communi-
cations and Systems Engineering Group, University of Vaasa,
Finland. He is also Professor at Electrical and Electronic En-
gineering Department, University of Benghazi, Libya, Email:
mohammed.elmusrati@uwasa.fi
E DITORIAL B OARD
Mohamed G. El−Tarhuni, Professor and Head of Electrical
Engineering Department, American University of Sharjah,
United Arab Emirates, Email: mtarhuni@aus.edu
Ali A. Ganoun, Assistant Professor at Electrical and Elec-
tronic Engineering Department, University of Tripoli, Libya,
Email: ali.ganoun@ee.edu.ly
Majdi A. Ashibani, Associate Professor at Electrical and
Electronic Engineering Department, Collage of Industrial
Technology, Libya, Email: mashibani@lptic.ly
Mohamed M. Elammari, Associate Professor and Head of
Software Engineering Department, University of Benghazi,
Libya, Email: mohamed.elammari@uob.edu.ly
Abdulkader A. Akki, Professor at Electrical and Electronic
Engineering Department, University of Tripoli, Libya, Email:
asakki@lttnet.net
P UBLICATION S TAFF
Asmaiel A. Ahteebah
Ahmed A. Aljarray
R EVIEWERS OF THIS I SSUE
Rajab Legnain
Naser El−Tarhuni
Abdulla Abouda
Mohammed Elfituri
Duan Ruifeng
Omar Abu-Ella
Tarek Sheltami
Mohamed El−Tarhuni
Mohammed Elmusrati
C OVER PAGE D ESIGN
Abdulnaser E. Elsousi
Abdulla I. Aborwies
S UBMISSIONS
The journal welcomes submissions of original research work
in the area of telecommunications, information technology and
applications. Articles both in English and Arabic language
are welcomed. A 500 LYD is granted for accepted papers.
Best paper award every year will be granted 5000 LYD. All
communications between editorial board and correspondent
authors will be through ajcita@almadar.ly. The submitted
papers will be checked first by the editors to confirm that
it follows the main journal requirements in terms of topics
and style. If the paper passed the first check it will be sent to
at least three independent peer reviewers to give acceptance
statement about the paper. The reviewers’ identities will be
masked from the authors.
A DVERTISING
Advertising in AJCITA is accepted at the discretion of the
publisher. In the next couple of issues accepted advertisement
will be published free of charge. All communications related
to advertising should be directed to ajcita@almadar.ly.
C OPYRIGHT
All rights reserved for AJCITA, however, abstracting is per-
mitted with credit to the source and libraries are permitted to
photocopy. AJCITA is not responsible for opinions presented
in its publication, they represent the views of the individuals.
 Almadar Journal for Communications, Information
Technology, and Applications
Vol. 01, No. 01, July 2014
Contents

Editorial Page: Mobile Networks: The Platform for Smart Connected World
Mohammed S. Elmusrati, Page 1.

The Future of WCDMA/HSPA: Delivering Exceptional MBB User Experience Everywhere

White Paper from Ericsson, Page 2.

Cognitive Radio−An Overview

Monirosharieh Vameghestahbanati and Mohamed El−Tarhuni, Invited Paper, Page 7.

Power Allocation for Cognitive Radios: A Survey

Ruifeng Duan, Mohammed S. Elmusrati, and Reino Virrankoski, Page 13.

On Effective Capacity of Cognitive Radios with TAS and MRC

Ruifeng Duan and Mohammed S. Elmusrati, Page 25.

Interference Mitigation Using Optimal Successive Group Decoding for Interference Channels
Omar Abu−Ella and Mohammed S. Elmusrati, Page 37.

Advertisments in Arabic

Libyan Innovation Prize 2014 by National Authority for Research, Science, and Technology.

Introduction to Research and Development Office at Almadar Aljadid Co. Libya.

Introduction to Almadar Journal for Communications, Information Technology, and Applications, the
Arabic version.

EDITORIAL PAGE
Mobile Networks: The Platform for Smart Connected World
Mohmmed S. Elmusrati1
M OBILE networks have expanded dramatically in all
dimensions during the last two decades. Based on pre-
diction studies carried out by CISCO, the growth of required
wireless data rate will increase by more than 11 fold (1100%)
between 2013 and 2018. In this year 2014, the number of
Internet accounts over cellular networks has passed 7.2 billion
connections, meaning that there are more device connections
than people on Earth!. This refers to the fact that average user
has more than one device connected to Internet networks. This
indicates the starting of the era of Internet of Things (IoT).
IoT includes countless applications beside the conventional
VoIP and Internet applications. Some applications are in fields
of security, smart cities, machine-to-machine, automations, e-
government, smart homes, remote monitoring and control, and
safety applications. According to ABI Research, more than
30 billion devices will be wirelessly connected to the Internet
backbone by 2020.
The recent surprised improvements in the performance and
reliability of wireless communication technologies motivate
the majority of Internet and IP traffics to use wireless as the
first hop. Hence, fast and reliable wireless networks become
very critical requirement for the growing Internet and IP
application demands.
At abstract level we have wireless communication platform
to transfer information signal (voice, video, sensor signal, data
from certain machine, etc.) from one location to another. Infor-
mation Technology (IT) may create several new applications
to utilize this communication platform for the sake of higher
quality of life as well as higher safety and security.
Beside the opportunities of wireless networks and appli-
cations, there are many challenges as well. The challenges
are everywhere in the wireless communication, IT, and the
applications. Therefore, we have this journal to discuss ideas,
analyze problems, propose solutions, suggest applications, and
1 Mohammed S. Elmusrati is a full Professor and Head of Communication
and Systems Engineering Group, University of Vaasa - Finland (On Leave).
He is also Professor with Electrical and Electronic Engineering Department -
University of Benghazi - Libya (email: mohammed.elmusrati@uva.fi).
1
enrich the scientific society of wireless communication and IT.
One of the major challenges of wireless communication is
the acute shortage (scarcity) in the spectrum. The whole useful
spectrum (except small ISM bands) has been already licensed
for specific applications. Therefore, it becomes a problem to
define new services or to increase the communication capacity
(which is linearly related to the bandwidth). Several novel
technologies have been developed during the last two decades
to overcome such problems. One method is based on the
recycling of licensed bands at the condition that no harmful
interference could be allowed for the primary user. This is
known as cognitive radio. One invited paper in this regard is
presented in this issue. The paper is titled “Cognitive Radio:
An Overview” by M. Vameghestahbanati and M. El-Tarhuni.
This invited paper presents a quite general background about
the topic. Moreover, two more specific technical papers in
the area of cognitive radios are presented in this issue as
well. The first paper is titled “Power Allocation for Cognitive
Radios: A Survey”. This paper gives an important survey
for the power allocation techniques in cognitive radios. It
is clear that transmit power is the critical parameter for
successful cognitive radio system. The second paper in the
area of cognitive radio is titled “On Effective Capacity of
Cognitive Radio with TAS and MRC”. This paper presents
novel analysis of the effective capacity of cognitive radio
with multiple transmit antennas and using transmit antennas
selection scheme with MRC.
Another method to overcome the problem of limited spec-
trum is by increasing the spectrum efficiency. This can be
achieved with different techniques such as MIMO antennas,
interference management, global radio resources scheduling,
and multiuser detection methods. One important contribution
in the area of joint multiuser detection is presented in this is-
sue. The paper is titled “Interference Mitigation Using Optimal
Successive Group Decoding for Interference Channels”.
Beside one invited paper and three peer reviewed papers,
we have also one white paper submitted by Ericsson Co.
Ericsson’s white paper discusses the future of WCDMA/HSPA
as powerful technology for providing mobile broadband during
the current decade.
Finally we would like to thank all authors for submitting
their great contributions to the first issue of AJCITA. Further-
more we are thankful for the reviewers for their great efforts to
review papers and give important comments and corrections.
2 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

The Future of WCDMA/HSPA: Delivering

Exceptional MBB User Experience Everywhere
(White Paper from Ericsson1 )

Abstract—WCDMA/HSPA enables hundreds of millions of

people to access mobile broadband (MBB) through their smart-
phones every day. Today, new, low-priced WCDMA/HSPA smart-
phones are entering the market, and they will enable MBB
for new hundreds-of-millions-sized markets. WCDMA/HSPA has
evolved into a highly effective MBB technology that will continue
to serve both new and traditional markets for years to come, ei-
ther as the main MBB technology or as an important complement
to LTE.

I. E SSENTIAL T ODAY AND T OMORROW

T He number of mobile broadband (MBB) subscribers and

the level of traffic continue to grow at an unprecedented
pace. The key driver behind this growth is an accelerating shift
from voice-centric phones and feature phones to increasingly
affordable MBB-enabled smartphones. In some markets, as
many as 90 percent of all new handset sales are accounted for
by MBB enabled smartphones. While this figure varies across
markets with different levels of maturity and consumer buying
power, the global trend is clear as illustrated in Figure 1. And
it is being driven first and foremost by the falling price of
highly capable WCDMA/HSPA smartphones.
This transition to smartphones presents a significant new
revenue opportunity for operators, as experience shows there
is substantially higher average revenue per user (ARPU) from
smartphone users compared with non-smartphone users.
WCDMA/HSPA networks will have an essential role to
play in enabling operators to take advantages of this market
opportunity, both today and for the foreseeable future. This is
driven by mobile operators’ need to satisfy four fundamental
market requirements, wherever they are in the world and
whatever the stage of market development. These market
requirements are:
• Serving the growing volumes of affordable MBB-capable
WCDMA/HSPA smartphones that are arriving on the
market.
• Ensuring a “megabit experience”, better than 1Mbps
download speed, for all MBB subscribers, and a
“superior-”MBB experience, better than 2Mbps download
speed, typically 10Mbps, for high-end-device users wher-
ever they go.
• Delivering high-quality voice services everywhere.
• Meeting the roaming needs of all MBB subscribers
around the world.
1 This article is a white paper from Ericsson, provide to Almadar Journal
for Communications, Information Technology, and Applications by Paolo
Lamberti. (emails: paolo.lamberti@ericsson.com).
Fig. 1. The rise of smartphones, mobile PCs, mobile routers and tablet
subscriptions with a cellular connection, 2009-2018 [1].
As with most consumer electronics technology, smartphones
are subject to Moore’s Law. In simple terms, smartphones are
roughly doubling their performance. For example, processing
power and data speeds, while halving in cost, every two years.
We have reached a stage where the cost of WCDMA/HSPA
chipsets is bringing highly capable smartphones into the price
bracket previously occupied only by voice-centric and feature
phones. At such sub-USD 100 prices, smartphones appear to
have reached a tipping point and are entering a true mass-
market phase.
While the rollout of 4G LTE radio networks is proceed-
ing rapidly, especially in developed markets, coverage is
still only a fraction of that provided by GSM/EDGE and
WCDMA/HSPA and, from a global perspective, will be for
many years to come. Operators need to ensure that 4G sub-
scribers MBB experience does not “fall off a cliff” whenever
they leave LTE coverage. MBB subscribers will always want
good-quality voice services. WCDMA/HSPA offers a well-
proven and efficient voice solution that meets very good KPIs,
including voice retainability and voice accessibility. It is also
able to provide HD voice quality, which is already being
deployed by several operators.
LTE is being deployed in more than 20 radio bands around
the world, unlike WCDMA/ HSPA, which has only four main
bands worldwide. It will take time before one single LTE
smartphone will be able to operate on every LTE network
around the world. In addition, some markets are still years
away from awarding LTE licenses. Nonetheless, operators
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014
Fig. 2. Fixed and mobile subscription growth 2009-2018 [1].
need to ensure that high-value roaming subscribers can enjoy
a consistent MBB experience.
While LTE is often viewed as the key solution for handling
high-capacity situations, WCDMA/HSPA is equally capable of
handling large numbers of smartphone users. What is more,
this capability is being dramatically extended over the coming
years, mainly through the implementation of software features.
Fundamentally, all operators whether they are voice-centric
GSM operators or MBB-centric operators who are deploying
LTE will benefit from having a strong WCDMA/HSPA net-
work, as HSPA is the only technology that will be used in all
smartphones for the foreseeable future.
Operators deploying LTE in the same frequency band
in which they already offer HSPA will benefit from
WCDMA/HSPA functionality that allows effective spectrum
refarming.
II. T HE H IGH - GROWTH M ARKET
In the past couple of years, MBB has become firmly
established as people have grown accustomed to having high-
speed-internet access wherever they go. According to market
analyst firm Wireless Intelligence [2], MBB technologies now
account for about one-quarter of total global connections.
WCDMA/HSPA makes up the vast majority of MBB con-
nections and is the fastest-growing wireless technology so far.
Now MBB is entering its next phase of expansion as
the availability and affordability of devices grows, especially
in developing markets. Ericsson’s estimates show that there
were 6.6 billion mobile subscriptions (excluding machine-to-
machine subscriptions) at the end of 2012 [1]. Of these, about
1.5 billion were MBB subscriptions (including feature phones,
smartphones, mobile PCs, tablets, mobile routers and dongles).
According to Wireless Intelligence, mobile subscriptions in
the developing world passed the 5 billion mark in the third
quarter of 2012, and now comprise almost 80 percent of the
world’s total. By the end of 2018, total mobile subscriptions
are expected to grow to 9.3 billion, and about 6.5 billion of
these will be for MBB, as shown in Figure 2.
3
Fig. 3. Global smartphone sales forecast by wholesale price tier. Smart-
phones below USD 190 are the biggest contributor to volume increases
(source:Strategy Analytics) [5].
A. Enter the affordable smartphone
At the centre of this more than fourfold predicted increase in
MBB subscriptions is the rising tide of affordable smartphones
and, to a lesser extent, tablets. According to the market analyst
firm Strategy Analytics [3], [4], 217 million smartphones were
sold in the fourth quarter of 2012 some 40 percent more than
in the same quarter of 2011 overall. Within this figure, sales of
Android smartphones grew close to 90 percent year-on-year.
And there is a similar story for tablets: fourth-quarter 2012
sales were 45 million; 67 percent higher year-on-year overall
and 85 percent higher year-on-year for Android devices. The
overall growth trend is set to continue. Total smartphone
subscriptions are set to rise from 1.1 billion at the end of
2012 to about 3.3 billion in 2018. One of the most significant
factors behind this rapid growth in smartphone adoption will
be their significantly lower average selling price, driven by
the availability of lower-cost chipsets, especially from Asian
manufacturers.
We are already starting to see a number of sub-USD
100 WCDMA/HSPA smartphones with 14.4Mbps, dual-band,
dual-core processor capabilities. There is also strong growth
in the midrange smartphone market (USD 100-200), thanks
to the arrival of much lower-cost chipsets for 42Mbps, quad-
core, HD (1280x720) devices. According to market research
firm Strategy Analytics [5], unit sales of entry and mid-range
smartphones are set to grow at a compound annual rate of 45
percent between 2011 and 2016, while unit sales of premium
and high-end smartphones will grow by a compound rate of
just 1.8 percent, as illustrated in Figure 3.
This increase in entry and mid-range smartphones will
power demand for MBB coverage, capacity and throughput,
especially in developing markets, where mobile devices will
provide many people’s first taste of high-speed internet con-
nectivity. So, which networks will this growing band of MBB
subscribers be using?.
4 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014
Fig. 4. Mobile subscription growth by technology [1].
B. WCDMA/HSPA scale beats LTE pace
In the second half of 2012, there was extremely rapid growth
in LTE connections, mainly driven by smartphone uptake
in Japan, South Korea and the US. LTE subscriptions are
expected to grow from just under 100 million at the end of
2012 to about 1.6 billion by the end of 2018. However, over
the same period, WCDMA/HSPA subscriptions are predicted
to grow from just over 1 billion to 4.4 billion, as shown in
Figure 4 [1]. In other words, there are likely to be almost
three times as many WCDMA/HSPA subscriptions as LTE
subscriptions in 2018. It is also worth noting that GSM/EDGE
subscriptions have continued to grow in number, and are
only expected to start declining during 2013. In addition,
Ericsson’s predictions forecast that global population coverage
for WCDMA/HSPA will increase from 50 percent in 2012
to more than 85 percent in 2017. Furthermore, people are
expected to be using their MBB devices a lot more than today.
They will consume roughly four times as much data across all
device types by 2018, as shown in Figure 5. For example, the
average monthly data usage for a smartphone is expected to
rise from 450MB in 2012 to nearly 2GB in 2018 [1].
Whichever way we look at the numbers relating to MBB,
it is clear that it is a high-growth market that offers great
potential for revenue growth in developed and developing
markets. What is also clear is that WCDMA/HSPA is going
to be the MBB workhorse for some years to come. Whether
or not mobile operators have access to LTE spectrum, they
will need to ensure the availability and good performance
of WCDMA/HSPA networks in order to serve the rapidly
growing numbers of non-LTE smartphones and other devices.
Even if they are rolling out LTE, operators need to ensure
their WCDMA/HSPA networks provide a comparable quality
experience when users move outside LTE coverage. As de-
mand grows, MBB services are putting immense pressure on
limited radio spectrum, and operators will need to find new
ways of using this spectrum ever more efficiently across 2G,
3G and 4G technologies. So, how can operators ensure they
have the WCDMA/HSPA coverage, capacity, performance and
business models they need to meet the rapid increases in MBB
uptake, usage and expectations?.
Fig. 5. How average monthly mobile-data usage will grow for mobile PCs,
tablets and smartphones [1]
III. D ELIVERING MORE
Whether operators are expanding their WCDMA/HSPA
coverage into new areas, or upgrading WCDMA/HSPA cov-
erage and capacity as a complement to LTE rollout, the key
challenges are the same:
• Delivering cost-efficient WCDMA/HSPA coverage.
• Efficiently handling very high numbers of smartphones,
characterized by a mix of voice and very bursty data
traffic.
• Ensuring a consistent user experience with the focus on
KPIs, such as call retainability and accessibility, as well
as high uplink and downlink throughput and short latency.
Such KPIs will, to a large extent, define the perception
of the operator among smartphone users.
• Constantly monitoring the smartphone population and
taking action to shape it, for example, by using subsidies
to encourage users to replace their old network-inefficient
smartphones with newer models that enhance both net-
work efficiency and user experience.
Unlike any other radio technology, WCDMA/HSPA is both
a proven voice solution and also a very capable MBB solution
that can deliver very high peak rates in the uplink and
downlink, as well as high cell-edge throughput. Cell-edge
throughput has a dominant influence over the WCDMA/ HSPA
system’s smartphone capacity.
WCDMA has already gone through a lengthy process of
evolution, and has come a long way from its first Release 99
incarnation more than a decade ago. WCDMA/HSPA already
exceeds the requirement to deliver a megabit experience for
users everywhere, thanks to:
• Support for up to 42Mbps in the downlink and 5.8Mbps
in the uplink.
• Superior radio performance with a comprehensive base
station portfolio for optimized coverage and capacity.
• Excellent in-service performance built on scalable and
future-proof 3G platforms.
• A clear evolution path to HSPA Evolution, which will
provide speeds of more than 84Mbps in the downlink
and more than 12Mbps in the uplink.
Behind these headline features, there are numerous improve-
ments being developed and applied to WCDMA/HSPA that
will further boost its ability to deliver a better user experience
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014
through improved smartphone capacity and higher uplink and
downlink bit rates.
Many newer smartphones on the market support features
that enable a substantial rise in overall smartphone capacity
(number of users per cell). For instance, multicarrier tech-
nology enables the WCDMA/HSPA system to use multiple
5MHz carriers for one user in both the uplink and downlink.
From 3GPP Release 10 onwards, WCDMA/HSPA supports
multicarrier operation on up to four carriers in the downlink
(which can be spread across one or two frequency bands) and
up to two carriers in the uplink. Multicarrier technology pro-
vides both substantial capacity gains, as well as throughput and
peak rate gains in the cell. The first step in WCDMA/HSPA
multicarrier development (2x5MHz downlink) is available in
smartphones today. Multicarrier in the uplink is expected to
arrive in smartphones toward the end of 2013.
Another function that boosts capacity is the 3GPP-specified
Fast Dormancy Release 8. This enables a smartphone to
move to an energy-efficient state (Universal Terrestrial Ra-
dio Access Network (UTRAN) Registration Area Forward
Access Channel (URA_FACH)) as soon as it has no data
to send, dramatically decreasing the time the smartphone is
in the most resource-intensive state (Cell Dedicated Channel
(Cell_DCH)).
The WCDMA/HSPA radio uplink is non-orthogonal by
nature, meaning that all users in a cell interfere with each
other on the radio interface. The most efficient way to counter
this interference is to eliminate unnecessary, or excessive,
network chatter, such as control signalling. This is the purpose
of a feature called Continuous Packet Connectivity (CPC),
which has the effect of dramatically improving uplink capacity
by limiting interference. Many commercially available smart-
phones already use fast dormancy and growing numbers are
appearing with CPC.
In addition, advances in base station signal processing
provide a clearer received uplink signal, which reduces the
cell’s total interference while sustaining uplink quality. Four-
way (instead of two-way) receive diversity further amplifies
the positive effects of advanced uplink receivers.
WCDMA/HSPA enables users that are sending and receiv-
ing only small bursts of data (as is typical with smartphones
most of the time) to handle that data in a semi-active state
called CELL_Forward Access Channel (CELL_FACH). To
enable even more efficient CELL_FACH operation, 3GPP
has specified High-Speed (HS)_FACH for the downlink and
Enhanced Uplink (EUL)_FACH for the uplink. HS_FACH-
enabled smartphones are already available and EUL_FACH
capable smartphones are expected during 2013.
A. Enhancing coverage
The ability to extend WCDMA/HSPA coverage efficiently
is vital to turning the proliferation of low-cost HSPA smart-
phones into increased revenue from MBB services. One key
way of doing this is to re-farm the 900MHz spectrum from
GSM to WCDMA/HSPA. This spectrum typically gives a 6dB
link budget advantage over the 2100MHz spectrum, which
translates into substantial coverage advantages. According to
5
the Global mobile Suppliers Association (GSA), 57 commer-
cial WCDMA 900MHz networks have been deployed in 39
countries (as of December 2012) [6]. Other WCDMA/HSPA
coverage-enhancing measures include four-way receiver diver-
sity (rather than two-way), lower speech rate for better voice
coverage, and the capability for improved scaling of control
and traffic channels.
B. Adapting data plans to market needs
One important aspect of driving the uptake of MBB, and
increasing ARPU overall, is to ensure that data plans meet
subscriber needs, especially in markets where the ARPU has
traditionally been low. The MBB pricing models used in
more established markets may not be appropriate for many
developing markets where ARPU can be one-tenth that of a
developed market, for example. With the trend to bring your
own device (BYO D) in many mature markets, there are sev-
eral new plans that are designed to attract new user categories.
Often, operations are run by a mobile virtual network operator
(MVNO) to differentiate from the major (owner) brand. These
offerings tend to be characterized by a distinct internet flavour,
no operator subsidies for devices and less customer support.
Such approaches have proved successful in mature markets
and have attracted new smartphone users.
Another trend in mature markets is “prepaid as postpaid”.
Traditionally, before data buckets were introduced, postpaid
customers often represented a higher ARPU group than pre-
paid customers. Data buckets are often defined by a fixed
traffic amount for a fixed price, which has made the distinction
between prepaid and postpaid less important for operators.
Bucket allowances are typically not exceeded, which provides
increased revenue opportunities for operators. With the in-
creasing popularity of tablets in developed markets, opera-
tors have an opportunity to monetize this trend by offering
attractive tethering add-ons and sharing plans. Several oper-
ators have already successfully transformed their subscriber
plans from traditional voice minutes and SMS volume-based
charging to charging based on actual data use.
The increasing availability of low-priced smartphones will
enable 2G/feature phone users to move to a highly capable
WCDMA/HSPA smartphone next time they invest in a new
device, for about the same price. As it is quite likely that
the WCDMA/HSPA smartphone will be the main broadband-
access device for many consumers, operators need to offer
appropriate, affordable plans. One way to enable MBB on
these devices is to permit small payments such as pay-per-
hour or pay-per-day. Further, some operators have developed
plans that permit pay-per- (small) data volume, and here it is
important that the consumer is given control of the actual data-
traffic consumption. For instance, using the operator’s own
portal would not consume any of the data allowance, while
accessing Facebook would. With consumers in control, over
time they may be more willing to pay for a data bucket for
internet access, resulting in increased operator revenue.
Yet another way to boost revenue is to allow consumers
time-limited free MBB access to any predefined internet ser-
vices in return for being shown advertisements before access
6 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

is granted. An example could be 30 minutes of free internet

use after a one-minute commercial.

IV. C ONCLUSION
Whether or not operators are rolling out LTE 4G networks,
they will need to focus their attention on the performance
of their WCDMA/HSPA networks if they are to deliver a
consistent, high-quality user experience throughout their cov-
erage areas. Fundamentally, it is vital that network technology
matches the capabilities and cost of the devices subscribers are
choosing to use to access MBB services. For the foreseeable
future, WCDMA/HSPA will be by far the biggest technol-
ogy by subscription numbers and by population coverage.
CDMA/HSPA already provides the backbone for most MBB
services, and is being continuously developed to efficiently
deliver a true broadband experience that is on a par with 4G
to any device, in any location. No other technology can make
that claim.

V. GLOSSARY
ARPU Average Revenue per User
BYOD Bring Your Own Device
CELL_DCH Cell Dedicated Channel
CELL_FACH Cell Forward Access Channel
CPC Continuous Packet Connectivity
EUL_FACH Enhanced Uplink Forward Access Channel
GSA Global mobile Suppliers Association
HS_FACH High-Speed Forward Access Channel
MBB mobile broadband
MVNO mobile virtual network operator
TD-SCDMA time division synchronous code division multiple access
URA_FACH UTRAN Registration Area Forward Access Channel
UTRAN Universal Terrestrial Radio Access Network

R EFERENCES
[1] Ericsson Mobility Report, November 2012, available at: http://www.
ericsson.com/ericsson-mobility-report
[2] Global cellular market trends and insights, Wireless Intelligence,
Q4, 2012, available at: https://wirelessintelligence.com/analysis/2012/01/
global-cellular-market-trends-andinsight-q4-2011/
[3] Android and Apple iOS Capture a Record 92 Percent Share of
Global Smartphone Shipments in Q4 2012, Strategy Analytics, January
2013, available at: http://www.strategyanalytics.com/default.aspx?mod=
reportabstractviewer&a0=8155
[4] Global Tablet OS Market Share: Q4 2012, Strategy Analytics, January
2013, available at: http://www.strategyanalytics.com/default.aspx?mod=
reportabstractviewer&a0=8147
[5] Global LTE Smartphone Shipments Will Reach 275 Million
Units in 2013, Strategy Analytics press release, December 2012,
available at: http://www.strategyanalytics.com/default.aspx?mod=
pressreleaseviewer&a0=5310
[6] UMTS900 Global Status Report, Global mobile Suppliers Associa-
tion, http://www.gsacom.com, November 2012, available at:http://www.
gsacom.com/php/access.php4
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014
Cognitive Radio−An Overview
(Invited Paper)
Monirosharieh Vameghestahbanati1 and Mohamed El-Tarhuni2
Abstract—There has been an increasing demand for high
data rate services, which necessitates the development of more
efficient schemes for using the scarce radio spectrum. Traditional
frequency allocation schemes are static and, thus, not capable
of accommodating the growing number of wireless users and
services. Hence, it is required to have some form of spectrum
sharing between existing and new users of the radio spectrum.
Cognitive Radio (CR) is a system that allows for sharing the
spectrum among users, which offers a highly flexible alternative
to the traditional fixed frequency band assignment. In this paper,
we survey the recent progress along with some issues related to
cognitive radio technology.
I. I NTRODUCTION TO C OGNITIVE R ADIO
T He emerging multimedia type applications have resulted
in increasing the need for higher data rate services. On
the contrary, the precious natural frequency is limited and
cannot fulfill all the requirements. Traditionally, radio channels
are assigned to specific users who have a license for the
exclusive use of that channel to avoid interference and provide
a certain quality of service. If the license holder is not using
the radio channel then that capacity is wasted. To avoid such
loss, new techniques should be developed to efficiently utilize
the radio spectrum.
Spectrum occupancy measurements in the US indicate that
the licensed spectrum is only utilized for 15% to 85% of
the time [1], indicating that conventional static frequency
allocation schemes utilize a significant portion of the spectrum
in an irregular manner. Therefore, dynamic spectrum sharing,
in which a radio unit is able to autonomously alter its features
according to the channel conditions in real time, has been
proposed in [2] in order to improve the spectral efficiency.
Dynamic spectrum sharing requires the use of cognitive radios
(CR) [3].
CR is an intelligent, low-cost, and highly flexible radio
that is able to adapt itself to its surrounding environment by
dynamically changing its radio parameters. It is an innovative
alternate to the classic static frequency devices and improves
the usage of the scarce frequency spectrum due to its cognitive
capability and reconfigurability. By its cognitive capability, it
is able to identify the unused portion of the spectrum, while
reconfigurability helps it adjusting its operational parameters
in order to perform in the best possible manner [1], [4].
Cognitive radio network contains primary owner of the
spectrum that are the legitimate users and have a higher
priority on the usage of the spectrum, and secondary users
1 Vameghestahbanati is with Department of Systems and Computer Engi-
neering, Carleton University, Ottawa, ON, Canada.
2 El-Tarhuni is with Department of Electrical Engineering, American Uni-
versity of Sharjah, Sharjah, United Arab Emirates. (email: mtarhuni@aus.edu).
7
that have a lower priority on the spectrum and try to use it
opportunistically when it is not used by the licensees [5].
II. C OGNITIVE R ADIO F UNCTIONS AND N ETWORK
A RCHITECUTE
CR system functions involve the following steps: spectrum
sensing, analysis, reasoning, and adaptation [4]. Spectrum
sensing is needed in order to detect the unutilized portion of
the spectrum as will be discussed in Section V. It then performs
the spectrum management and handoff so as to select the best
frequency band with a smooth transition and low latency.
Cognitive radio architecture involves a primary network,
which contains a set of primary users or license holders
who communicate either directly or through one or more
base stations. Similarly, the secondary network has a set
of secondary users that can communicate either directly or
through secondary base stations. The secondary users and the
secondary base stations have cognitive radio capabilities. In
the case when several secondary networks share the same
frequency band, a spectrum broker [6] can manage the usage
of the spectrum by assembling information from each network.
Cognitive radios are envisioned to be used in diverse areas
such as wireless networks for biomedical applications, smart
grid, commercial markets for wireless technologies such as
cellular systems, military communications, and public safety
and enhancement of the security of homelands [4].
III. T YPES OF C OGNITIVE R ADIO
Cognitive radio systems fall into three different categories,
interweave, underlay and overlay cognitive radio [7].
In an interweave cognitive radio, both the secondary and the
primary users occupy the same frequency band without inter-
fering to each other as if their signals are orthogonal to one
another. This can be accomplished by using multiple access
techniques such as time-division-multiple-access (TDMA) or
frequency-division-multiple-access (FDMA).
In an underlay cognitive radio, the secondary user can
utilize the spectrum simultaneously with the primary user but
under the primary users interference constraint. The name
underlay cognitive radio results from the fact that the CR
signal looks like a noise under the primary signal [4].
The interference level to the primary user is kept below an
acceptable level by limiting the power of the secondary user.
This approach becomes more challenging due to the appear-
ance of unpredictable new sources of interference. As such,
Federal Communications Commission (FCC) in the United
States has proposed a metric named interference temperature
to judge the interference levels. In this way, an interference
8 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

restriction is imposed on the receivers. The Interference tem- 1) Hardware requirements: In traditional systems, the re-
perature is defined as “the temperature equivalent to the RF ceivers are tuned to receive signals over a limited frequency
power available at a receiving antenna per unit bandwidth” band and the problem of estimating the noise and interference
[4], that is is easier. However, cognitive radio terminals should be capable
PI (fc , B) of sensing signals over a wide bandwidth, which required wide
TI (fc , B) = (1)
KB band amplifiers and antennas. Furthermore, for the cognitive
radios to perform computationally complex functions with a
PI (fc ,B) in (1) is the average interference power that is
low delay, high speed Digital Signal Processing (DSP) units
measured in Watt and centered at frequency of fc with a
are required.
bandwidth of B measured in Hertz. K is the Boltzmann’s
Spectrum sensing can be performed based on either single-
constant (K = 1.38×10−23 ) with the unit of Joules per degree
radio or dual-radio architecture. The former is simple and
Kelvin.
cheap but has a specific time slot dedicated for the spectrum
The interference temperature limit is defined as a maximum
sensing with a bounded sensing duration and poor spectrum
amount of interference that can be tolerated for a particular
accuracy. Also, since a fraction of the time slot is used for
frequency band and location. By this constraint, the transmis-
sensing instead of transmitting data, the spectrum efficiency is
sion plus noise and interference of any unlicensed transmitter
low. The dual radio architecture allows sensing and data trans-
must not go beyond the interference temperature limit at the
mission at the same time and, hence, provides higher spectrum
receiver of the licensee. For example, if TL is the interference
efficiency and better sensing accuracy. However, it has a higher
temperature limit of the legitimate user for a given frequency
cost, more complexity and more power consumption.
band having a bandwidth of B, the average interference of the
2) Hidden primary user problem: In the case of severe
secondary transmitter should fall below KBTL .
multipath fading or shadowing, there is a possibility for the
In an overlay cognitive radio, the secondary user can coexist
primary user to be hidden to the secondary system. Thus, the
with the primary user over the same spectrum by knowing
primary user might not be detected and thus unwanted interfer-
its channel with the licensee along with its operation. For
ence is introduced to the legitimate user by the secondary user
instance, it is required to be aware of the primary user’s
transmission. The hidden primary user problem is illustrated
codebook that can help in decoding the legitimate user’s
in Figure 1. Cooperative spectrum sensing is used to combat
transmission [8].
this problem [12].

IV. C OGNITIVE R ADIO C HALLENGES

Despite the many advantages that cognitive radio systems
offer such as improving the spectrum utilization and availing
high bandwidth to users, they impose some challenges that are
substantially in behalf of the fluctuating nature of the available
frequency band, the coordination and coexistence requirements
with the primary users, and the need for various quality of
service (QoS) in diverse applications. The main challenges
with the CR systems are minimizing interference, as pointed
above in Section III, and guaranteeing the QoS since the CR
system utilizes the frequency band opportunistically, which
would change the quantity of the used spectrum [9].
V. S PECTRUM S ENSING
IEEE defines Spectrum sensing as “the act of measur-
ing information indicative of spectrum occupancy” [10]. As
such, spectrum sensing deals with the implementation of self-
governing process of the secondary users that investigate the
spectrum based on the received signals [11], which should be
performed in a delicate manner in order not to interfere with
the primary owner of the spectrum. It is thus a crucial com-
ponent for the establishment of the cognitive radio. Spectrum
sensing has different aspects that are discussed below:
A. Challenges
Some of the challenges associated with spectrum sensing
include:
Fig. 1. Hidden primary user problem in cognitive radio systems.
3) Detecting spread spectrum primary users: There are
two main types of primary users assigned to use the radio
spectrum. The first type has in general narrow band chan-
nels with the transmitted energy confined to the assigned
channel. The second type uses wide-band spread spectrum
systems that transmit data over a wider bandwidth than the
minimum required bandwidth. Spread spectrum systems might
use a frequency-hopping (FHSS) scheme by changing the
carrier frequency of the transmitted signal dynamically over
a large number of possible frequencies (channels) according
to a known sequence. Another type of spread spectrum uses
direct sequence (DSSS) to transmit a wideband signal over a
single frequency. If a primary user employs spread spectrum
devices, the transmitted power is spread over a wide range
of frequencies, which makes it difficult to be detected by the
secondary user. In order to solve this issue, the secondary user
needs to know the spreading code or hopping sequence used
by the primary user.
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014 9

4) Sensing duration and frequency: As pointed above, the

licensee has the priority to utilize the spectrum over the
secondary user and can demand it whenever needed. Hence,
cognitive radio should be able to identify the presence of the
legitimate user within specific time interval in order not to
cause interference. As such, the sensing frequency that deals
with the regularity of performing the spectrum sensing task,
is required to be considered in the design of the CR system.
5) Security: A selfish user can alter its air interface to mock
the primary owner of the spectrum. In order to overcome this
issue, legal primary users can transmit an encrypted value for
their validation. Yet, the primary and the secondary users are
required to coordinate and synchronize.

B. Multi-dimensional spectrum sensing

In order to improve the spectral efficiency, spectrum sensing
can be performed in multi-dimensions that are summarized
below [12].
1) Frequency dimension: The spectrum is divided into a
large number of non-overlapping sub-channels and sensing is
done to decide on which sub-channels are available at any
time.
2) Time dimension: A new opportunity can be provided to
the secondary user by considering the fact that the band is not
used all the times by the primary user, and some portions of
the spectrum might be always available. The frequency and
time dimensions are shown in Figure 2.
Fig. 3. Illustration of code dimension.
5) Angle dimension: By employing the recent progresses
in multi-antenna technology that allows multiple users to be
multiplexed into the same channel at the same time in the same
area, the direction of the primary user beam i.e. the azimuth
and the elevation angles and the location of the licensee, can
be sensed. In that case, the secondary system can transmit
in another direction without introducing interference to the
primary one. The angle dimension is illustrated in Figure 5.

Fig. 2. Frequency and time dimensions.

3) Code dimension: Simultaneous transmission without in-

terfering with the licensee can be feasible by transmitting a
code orthogonal to the one used by the primary user. However,
secondary user needs to synchronize itself with the primary
user employing time hopping or frequency hopping, and thus
timing information is required. Figure 3 illustrates the code
dimension.
4) Geographical space dimension: By noting that the fre-
quency band is used in some parts of a geographical area
while it is available in other parts, geographical space can be
viewed as one dimension. This requires sensing the location
of the primary user or in particular sensing the latitude,
longitude, elevation, and the distance of the legitimate user.
This dimension is clarified in Figure 4.
Fig. 4. Geographical space dimension.
C. Spectrum sensing methods
In this section, we review some of the spectrum sensing
techniques used in CR systems.
1) Energy detector-based sensing: A commonly used ap-
proach in order to detect an unknown signal in the presence of
noise in CR systems is the energy detection based sensing. The
block diagram of an energy detector in depicted in Figure 6.
The input band pass filter selects the desired channel at
frequency fs and with a bandwidth W . A squaring device and
10 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014
Fig. 5. Angle space dimension.
integrator measure the energy of the received signal over an
observation interval, T . The integrator output Y is compared
to a threshold λ to decide if a signal is present in the selected
channel or not [13].
Fig. 6. Block diagram of an energy detector.
This sensing technique is easy to implement and does
not need prior knowledge about the primary signal due to
the dependency of the threshold on the noise floor. On the
contrary, it is not simple to select the threshold. This approach
also fails in low Signal-to-Noise ratio (SNR) scenarios since
it is not able to discriminate between the primary user and
noise [4], [12].
2) Waveform-Based Sensing: Waveform-based sensing is
more reliable than energy detector based sensing. In this case,
known patterns or pilots are used for synchronization purposes.
The known sequence can be transmitted either before each
burst or slot and is called a preamble or in the middle of
each burst and is called a midamble. At the receiver side, the
received signal is correlated with a reference to sense if the
received signal contains the known pattern. Let us assume that
the received signal has the following form
y(k) = s(k) + w(k) (2)
where s(k) is the primary signal to be detected, w(k) is
the additive white Gaussian noise (AWGN) sample, and k is
sample index.
The decision metric for the waveform-based sensing can be
written as "K # secondary user has a perfect knowledge of the primary signal.
X
∗
M = <e y(k)s (k) (3)
k=1
Where is the complex conjugation operation and K is the
∗
number of samples.
When the primary user is not present, the metric becomes
"K #
X
∗
M = <e w(k)s (k) (4)
k=1
while when the primary user is present, the metric is
K
"K #
X X
2 ∗
M= |s(k)| + <e w(k)s (k) (5)
k=1 k=1
The decision can then be made by comparing the decision
metric M with a fixed threshold λw .
3) Cyclostationarity- Based Sensing: The cyclostationarity
features of the received signals can be used to detect the pres-
ence of the primary user. This feature deals with the periodicity
in the signal or its statistics like the mean or the autocorrela-
tion. Cyclostationary-based sensing has the advantage of being
capable of differentiating noise from primary users’ signals,
since noise is a wide sense stationary (WSS) process and has
no correlation with delayed versions, while a modulated signal
is cyclostationary having spectral correlation [12].
Cyclostationarity-Based Sensing differs from the energy-
based detector in conducting a hypothesis test in the frequency
domain rather than time domain. The hypothesis model of the
received signal is [4]
H0 : y(t) = n(t),
H1 : y(t) = hx(t) + n(t) (6)
where x(t) is the primary user’s signal to be detected, n(t)
is AWGN, and h is the channel gain from the primary user
transmitter to the secondary user receiver. H0 is the null
hypothesis that signifies the absence of the primary user in
the frequency band of interest, while H1 shows the presence
of the primary user.
The cyclic autocorrelation function (CAF) of the received
signal is
 
Ryα (τ ) = E y(t + τ )y ∗ (t − τ )ej2παt (7)
E[.] is the expectation and α is the cyclic frequency. For a
digitally modulated signal, the CAF of the received signal is
periodic. By applying the Fourier series expansion to (7), the
cyclic spectrum density (CSD) function, can expressed as
X∞
S(f, α) = Ryα (τ )e−j2πf τ (8)
τ =−∞
The CSD function have peaks whenever the cyclic fre-
quency, α, equals to the fundamental frequency of the trans-
mitted signal, x(t). On the other hand, since the noise is not
cyclostationary signal, the CSD has no peaks under the H0
hypothesis. As a result, a peak detector can differentiate the
two hypotheses.
4) Matched Filtering: A matched-filter with a threshold test
is an optimal detector in stationary Gaussian noise when the
However, it is hard to implement such a detector [13] as the
secondary user needs to know about the primary user’s signal
like its bandwidth, operating frequency, modulation type, etc.
In addition, this technique requires diverse receiver algorithms
for detection and thus consumes large amounts of power [12].
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014
D. Cooperative Spectrum Sensing
The three main factors that limit the spectrum sensing
performance include noise, shadowing, and multi-path fading.
When the SNR of the received primary signal is below a
predetermined threshold, reliable detection is not possible
even with long sensing time. Therefore, the secondary user
cannot detect the presence of the primary user and would
interfere with it. As an example, the primary owner of the
spectrum can only be detected by one secondary user in a
scenario depicted in Figure 7. due to a deep shadowing effect.
Cooperative sensing can be used to overcome this problem.
Some advantages of cooperation include:
Fig. 7. Cooperative spectrum sensing.
• Increasing the probability of detection that leads to better
protecting the primary signal.
• Decreasing the false alarm that yields in utilizing the
vacant spectrum efficiently.
• Solving the hidden primary user problem (introduced
above).
• Decreasing the sensing time.
However, cooperative sensing has some challenges; for
instance, in the case of having a wideband cooperative sensing,
multiple unlicensed users needs to investigate a wide range of
channels that causes a high consumption of energy, a lower
data throughput, and a large amount of data transfer [4], [12].
Cooperative sensing is classified into three main categories
that are discussed below:
1) Centralized Sensing: In this technique, a central entity
collects sensing information from cognitive devices in order to
identify the available vacant spectrum. It then broadcasts this
data to the other CR devices or it is able to directly organize
11
the cognitive radio traffic. In the case of large number of users,
a huge bandwidth is required to report the information. In
order to decrease this bandwidth, CR devices with the same
data can report their decisions to the central body, which
requires censoring some of the data.
2) Distributed Sensing: In this approach, cognitive nodes
interchange information with each other, however, they make
their own decisions about the portion of the spectrum that
they use. Distributed Sensing does not require a backbone
infrastructure, which implies less cost, and hence differs
from the centralized sensing. However, the network overhead
needed for sharing the information among collaborating radios
will increase resulting in reduced spectrum efficiency.
3) External Sensing: In this technique, an external agent or
base station performs the spectrum sensing of primary users.
The results about spectrum availability are then broadcast to
the cognitive radios. External sensing reduces the impact of
the hidden primary user problem and improves the power and
spectrum efficiencies for the CRs since the processing is done
by an external agent and CRs do not waste time or energy in
sensing the spectrum [3].
4) Recent Spectrum Sensing Schemes: There has been
some recent techniques for spectrum sensing using machine
learning schemes initially introduced in [14] for single carrier
modulation schemes and then extended to multicarrier Or-
thogonal Frequency Division Multiplexing (OFDM) schemes
in [15]. Those schemes train a polynomial classifier based
on different primary user pattern and then used the trained
classifier to decide upon the availability of the channel. These
schemes provide significant improvement in performance in
terms of improved detection probability without increasing the
false alarm probability, especially when used in a cooperative
architecture. Since the classifier training is done offline then
there is no major increase in complexity. Finally, a new CR
system using an overlay structure has been proposed in [16]
that alleviates the problem of spectrum sensing and allows for
simultaneous use of the radio channel.
VI. W ORLDWIDE S TANDARDIZATION ACTIVITIES
Major organizations making various standards for cognitive
radio systems (CRS) include the Institute of Electrical and
Electronic Engineers (IEEE), International Telecommunication
Union Radio-communication (ITU-R), European Telecommu-
nications Standards Institute (ETSI), and European Computer
Manufacturers Association (ECMA).
A. Standardization in IEEE
In the IEEE, cognitive radio systems and their components
are standardized in several working groups (WG) in standards
coordination committee (SCC) and 802 LAN/MAN Standards
Committee. IEEE SCC 41 is dealing with the standards asso-
ciated to dynamic spectrum access networks with the purpose
of improving the utilization of spectrum. On the other hand,
IEEE 802 WGs define CRS and its components. The task
of defining CRS is accomplished in the 802.22 and 802.11
working groups, while 802.21, 802.22, and 802.19 working
groups identify components of a CRS.
12 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014
B. Standardization in the ITU
In the ITU, ITU-R WPs 1B and 5A are preparing reports
that describe the cognitive radio concepts and the regulatory
measures associated with that. WP 1B has developed defi-
nitions for the software defined radio (SDR) and cognitive
radio system (CRS) [17]. The technical and operational studies
and relevant ITU-R Recommendations that are related to the
SDR and CRS are summarized in WP 1B together with the
relationship between SDR and CRS [18].
C. Standardization in ETSI
A series of ETSI technical reports (TR 102 838), which
examine the opportunities associated with reconfigurable radio
systems like SDR and CR and their related standardization
needs, was released in October 2009 by ETSI’s Technical
Committee on reconfigurable radio systems (RRS). The fea-
sibility studies performed by the committee was summarized
in the reports. The RRS Standardization is now proceeding
with four different working groups, system aspect studies,
reconfigurable radio equipment studies, cognitive network
management studies, and public safety studies, examining
different characteristics of reconfigurable radio systems [19].
D. Standardization in ECMA
Task Group 1 of Technical Committee 48 carries out the
standardization of the cognitive radio systems. In December
2009, standard ECMA-392, “MAC and PHY for Operation in
TV White Space,” was issued. This standard identifies MAC
and physical layers for personal/portable cognitive wireless
systems that function in TV bands. Moreover, a number
of protection mechanisms that might be useful in meeting
regulatory requirements are stated in this standard [18].
VII. F UTURE T RENDS IN C OGNITIVE R ADIO R ESEARCH
The evolving cognitive radio technology is anticipated to
become a vital component in the future generation of “smart
radio” technology. As a result, it will optimize the utilization
of the limited radio resources dynamically, and tries to reduce
the issue of the radio traffic jam along with improving the effi-
ciency of the future wireless devices. The CR is envisioned to
emerge as a general purpose radio implemented through SDR
and works as an ubiquitous platform for the developments
in the wireless devices. There are still substantial challenges
in the domain of the spectrum policy and valuing dynamic
spectrum access, even though significant amount of research
has been done in various aspects of cognitive radio systems.
In addition, the implementation and testing of the cognitive
radio technology should be performed in the real world [19].
VIII. C ONCLUSIONS
Cognitive radio is a new paradigm for improved radio
spectrum utilization and is expected to be a main driver
towards the goals of future wireless connectivity of billions
of radios. An overview of cognitive radio technologies and
issues along with standardization activities is presented in this
paper.
R EFERENCES
[1] F. Akyildiz, W. Lee, M. Vuran, and S. Mohanty, “Next genera-
tion/dynamic spectrum access/cognitive radio wireless networks: A
survey,” Comput. Netw., Vol. 50, pp. 2127-2159, May 2006.
[2] N. Devroye, P. Mitran, and V. Tarokh, “Limits on communications in a
cognitive radio channel,” IEEE Communication Magazine, Jun. 2006.
[3] J. Mitola, “Cognitive Radio: making software radio more personal,”
IEEE Personal Communications, Vol. 06, No. 04, pp. 48-52, Aug. 1999.
[4] B. Wang and K. Ray Liu, “Advances in cognitive radio networks: A
survey,” IEEE Journal of Selected Topics in Signal Processing, Vol. 5,
No. 1, Feb. 2011.
[5] E. Axell, G. Leus, E. Larsson, H. Poor, “Spectrum sensing for cognitive
radio : State-of-the-Art and recent advances,” IEEE Signal Processing
Magazine, Vol.29, No.3, pp.101-116, May 2012.
[6] C. Raman, R. Yates, and N. Mandayam, “Scheduling variable rate links
via a spectrum server,”in Proc. of IEEE Int. Symposium on New Frontiers
in Dynamic Spectrum Access Networks, Baltimore, Maryland, USA,
pp.110-118, Nov. 2005.
[7] M. Pischella, D. Ruyet, “Adaptive resource allocation and decoding
strategy for underlay multi-carrier cooperative cognitive radio systems,”
Journal Transactions on Emerging Telecommunications Technologies,
special issue on Cognitive Radio, Vol. 24, No. (7-8), pp. 748-761, 2013.
[8] N. Devroye, “Information theoretical limits on cognitive radio networks,”
Cognitive Radio Communications and Networks; Principles and Prac-
tice, A. Wyglinski, M. Nekovee and Y.T. Hou Ed., Elsevier, 2009.
[9] S. Kumar, J. Sahay, G. Mishra, and S. Kumar, “Cognitive radio concept
and challenges in dynamic spectrum access for the future generation
wireless communication systems,” Wireless Personal Communications,
Vol. 59, pp. 525-535, 2011.
[10] IEEE Standard Definitions and Concepts for Dynamic Spectrum Ac-
cess: Terminology Relating to Emerging Wireless Networks, System
Functionality, and Spectrum Management. Amendment Addition of New
Terms and Associated Definitions (2013).
[11] A. Mariani, A. Giorgetti, and M. Chiani, “Recent Advances on Wide-
band Spectrum Sensing for Cognitive Radio,” in Cognitive Commu-
nication and Cooperative HetNet Coexistence, Springer International
Publishing, 2014, pp. 1-31.
[12] T. Yucek and H. Arslan, “A Survey of spectrum sensing algorithms for
cognitive radio application,” IEEE Communications Surveys Tutorials,
Vol. 11, No. 1, 2009.
[13] Ghasemi and E. Sousa, “Collaborative spectrum sensing for opportunis-
tic access in fading environments,” in Proc. of IEEE Int. Symposium
on New Frontiers in Dynamic Spectrum Access Networks, Baltimore,
Maryland, USA, pp. 131-136, Nov. 2005.
[14] Y. Hassan, M. El-Tarhuni, and K. Assaleh, “Learning-based spectrum
sensing for cognitive radio systems,” Journal of Computer Networks and
Communications, Vol. 2012, pp. 1-13, Nov. 2012.
[15] M. Muzaffar, M. El-Tarhuni, and K. Assaleh, “Learning-based Spec-
trum Sensing in OFDM Cognitive Radios,” The Second International
Conference on Advances in Cognitive Radio, May, 2012.
[16] M. Vameghestahbanati, H. Mir, and M. El-Tarhuni, “An MMSE Overlay
Cognitive Wireless System,” International Journal of Interdisciplinary
Telecommunications and Networking, Vol. 4, No. 4, pp. 64-76, 2012.
[17] ITU-R SM.2152, “Definitions of Software Defined Radio (SDR) and
Cognitive Radio System (CRS),” Sept. 2009.
[18] S. Filin, H. Harada, H. Murakami, K. Ishizu, “International standard-
ization of cognitive radio systems,” IEEE Communications Magazine,
Vol. 49, No.3, pp.82-89, Mar. 2011.
[19] E. Hossain, D. Niyato and D. Kim,“ Evolution and future trends of
research in cognitive radio: a contemporary survey,” Wireless Commu-
nicating and Mobile Computing, 2013.
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014
Power Allocation for Cognitive Radios: A Survey
Ruifeng Duan, Mohammed S. Elmusrati, and Reino Virrankoski
Abstract—From the literature we know that power allocation
is of great importance in managing the interference in spectrum
sharing networks, maximizing the spectrum reuse, increasing
communication capacity, and making our living environment
greener. In this paper we reviewed the most important and up-
to-date results of the power allocation approaches proposed in
literature from an information-theoretic perspective. Therefore,
we will take a look at the optimal power allocation strategies of
the secondary users in order to maximize their ergodic capacity
and effective capacity over fading channels. This survey improves
the understanding of ultimate performance limits of the cognitive
radios and the cognitive radio systems design.
Index Terms—Cognitive radio, ergodic capacity, effective ca-
pacity, optimum power allocation.
I. I NTRODUCTION
Patial considerations for frequency reuse have been exten-
S sively studied in cellular systems. However, these systems
largely differ from the cognitive radio (CR) systems [1]. As
the command-and-control structure of frequency allocation
for traditional wireless communications, the within-system
interference is the dominant interference to the users operating
with the same operator. This kind of interference can be well
controlled through planning. For these systems, power control
has been studied in SIR-based, e.g. [2], and information-
theoretic contexts for fading and non-fading channels, for
instance [3]–[6]. However, in cognitive radio networks, the
interference is caused not only by the secondary users (SUs),
or cognitive users, sharing the same spectrum, but also by
the primary users (PUs), or licensed users, who share the
spectrum. Additionally, the secondary users should not cause
unacceptable interference to the primary users [7], [8].
In this paper we focus on the information-theoretic ap-
proaches, i.e., reviewing the optimal power allocation ap-
proaches for the SUs to maximize the achievable rate under
certain constraints. The framework employed to evaluate the
power allocation schemes and other performance matrices is
mainly based on information theory [9]. There is a growing
body of literature on power control/allocation in CR systems.
In [10] and [11], power control for one pair of secondary
users coexisting with one pair of primary users is considered.
In [10], the secondary transmitter adjusts its transmission
power to maximize its data rate without increasing the out-
age probability at the primary receiver. The authors in [11]
proposed the optimal power control schemes based on the
The authors are with Communications and Systems Engineering Group,
University of Vaasa, Finland. (emails: ruifeng.duan@ieee.org; {moel,
rvir}@uva.fi).
The content of this paper is part of Duan’s thesis, and is reprinted with the
permission of the University of Vaasa. This work was supported in part by
the SMACIW (Statistical Modeling and Control of Aggregate Interference in
Wireless Systems, Decision no. 265077) project funded by the Academy of
Finland.
13
soft sensing information, and the capacity of the secondary
user was maximized under a peak power constraint at the
primary receiver. Power control for opportunistic spectrum
access (OSA) in TV bands is investigated in [12] and [13],
where the primary users transmit all the time and spatial
(rather than temporal) spectrum opportunities are exploited by
secondary users. For the interference control of the secondary
users over television white spaces, Koufos et al. in [14]
proposed the power density and deployment based transmit
power control of the secondary users such that the quality of
the TV services is not violated by the aggregated interference.
Gastpar investigated the ergodic capacity of different non-
fading additive-white-Gaussian noise (AWGN) channels [15],
[16]. The transmit power of the SU is regulated by the average
interference power received at a third-party receiver. The
author illustrated that the received-signal constraints can lead
to substantially different results as compared to transmitted-
signal constraints. There are some important findings which
are different from a conventional point-to-point communica-
tion. Without fading the author showed that in the point-to-
point case, the transmitted- and received-power constraints are
largely equivalent. While in network cases, they can lead to
quite different conclusions, for example, multiple access chan-
nels with dependent sources and feedback, and collaborative
communication scenarios. Ghasemi and Sousa, in [17], showed
that in many cases significant capacity gains may be achieved
if the channels are varying due to fading and shadowing under
either the average or the peak interference power constraint. In
[18], the authors extended the work in [17] by investigating the
achievable capacity gains in asymmetric fading environments.
Then, Musavian and Aı̈ssa in [19] studied the capacity gains
offered by the spectrum-sharing approach in a Rayleigh fading
environment subject to both average and peak received-power
constraints at the primary receiver. In [20], Kang et al. studied
the optimal power allocation strategies to achieve the ergodic,
delay-limited, and outage capacities of a secondary fading
channel subject to a diverse combinations of peak/average
transmit and/or peak/average interference power constraints.
The authors observed that fading of the channel from sec-
ondary transmitter to primary receiver can be a good phe-
nomenon for maximizing the capacity of SU fading channel.
Zhang concluded in [21] that the average-interference-power
(AIP) constraint can be more advantageous over the peak-
interference-power for minimizing the resultant capacity loss
of the primary fading channel, and AIP should be used for
the purposes of both protecting the PR communications as
well as maximizing the CR capacity. Therefore, in this paper
we review the channel model and the concepts of capacity,
and then survey the main results of optimal power allocation
approaches for cognitive radios.
14 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014
II. C HANNEL M ODEL AND C ONCEPTS OF C APACITY
In this section we introduce the channel model, and review
two important concepts, i.e., ergodic capacity and effective
capacity. We consider independent and identically distributed
additive white Gaussian noise (AWGN) block-fading channels.
Generally, we do not specify the fading distributions, but we
mention them together with the associated results. The block-
fading, or quasi-static, channel model was introduced in [22]
and has been commonly used in the literature for studying
wireless communications systems over slowly-varying fading
channels [22], [23], through which a codeword spans only a
certain number of fading blocks. During each fading block,
the channel gain remains constant while varying from block
to block. We assumed that the primary user(s) are located
far away from the secondary receiver so that there is no
significant interference to the secondary user [17], [19], [24]–
[26]. In addition, the interference from the primary user to the
secondary user could be considered being absorbed into the
noise if the random Gaussian codebooks are applied at the
primary transmitters [27]–[29].
For imperfect channel information scenarios, we adopt
the following channel estimation methods for measuring the
channel gain of ST-PR link, which has been widely used in
literature, e.g. [19]. For Rayleigh fading channels, the complex
channel gain from the secondary transmitter to the primary
receiver, cps , is zero mean circularly symmetric complex
Gaussian distributed variable with the imaginary and real parts
having variances of 0.5. However, the CR transmitter is only
provided with partial channel information of cps , namely c̃ps ,
where cps and c̃ps are jointly ergodic and stationary Gaus-
sian processes. The secondary user performs minimum mean
square error estimation (MMSE) of cps given c̃ps , such that
ĉps [n] = E {cps [n] | c̃ps [n], c̃ps [n − 1], ...}, where [n] denotes
the time index. The MMSE estimation error can be presented
as c̆ps [n] = cps [n]−ĉps [n], and c̆ps [n] and ĉps [n] are zero mean
circularly symmetric complex Gaussian distributed variables
2 2
with variances 1−σ 2 and σ2 respectively. So the associated
2 2
channel power gain can be presented as g = |cps | , ĝ = |ĉps | ,
2
and the channel power gain estimation error by ğ = |c̆ps | . The
probability density function of estimated channel power gain,
ĝ, is characterized by [30]:
 
1 ĝ
fĝ (ĝ) = exp − , ĝ ≥ 0 (1)
1 − σ2 1 − σ2
A. Ergodic Capacity
This subsection reviews the ergodic capacity formulation
of the secondary user. With perfect channel state information
(CSI) of the secondary link (ST-SR) and the secondary trans-
mitter to the primary receiver (ST-PR), the ergodic capacity
of the secondary user is given in [17] by
  
ps (gss , gps )gss
maximize Egss ,gps log 1 + (2)
ps (gss ,gps )≥0 N1 B
where ps (gss , gps ) is the transmit power of the secondary
transmitter, gss and gps denote the channel power gains of
ST-SR and ST-PR, respectively. N1 represents the additive
white noise density at the secondary receiver. log(·) denotes
the natural logarithm operator, and Ex denotes the expectation
operator over x in this paper. The secondary user chooses the
optimal transmit power to maximize the achievable rate ac-
cording to the instantaneous CSI of the two channels instead of
only its own CSI as in the traditional wireless communications
systems. The maximization is over power allocation functions
that are being discussed later in certain problems.
B. Effective Capacity
From literature, we know that the ergodic capacity has
no transmission delay limitation, while the outage capacity
does not allow any delay [31]. In order to study the delay
performance, the concept of effective capacity (EC) was
developed in [32], [33] to define the maximum arrival data rate
that can be supported by the channel subject to the required
communication delay. It is a link-layer channel model and
can be interpreted as the dual of effective bandwidth [34].
The quality of service (QoS) is represented by a term, named
QoS exponent θ ∈ R++ defined in Eqn.(3). The EC bridges
the ergodic capacity and the outage capacity. When the QoS
exponent θ → 0, it means that there is no delay limitation,
and the EC equals the ergodic capacity. On the other hand, the
link cannot tolerate any delay as θ → ∞. This concept has
received much attention in the point-to-point communication
scenarios, e.g., [35], [36], as well as in cognitive radios, e.g.,
[37], [38] and references therein. The effective capacity along
with energy efficiency was also investigated in [39].
Let q(x) be the queue length of a stationary ergodic arrival
and service process. The probability that q(x) exceeds a certain
threshold Tq decays exponentially as a function of Tq , and the
delay QoS exponent is defined in [32] as
log(Pr {q(∞) > Tq })
θ = − lim . (3)
Tq →∞ Tq
It is worth noting that θ → 0 indicates that the system has
no delay constraint, while θ → ∞ implies a stringent delay
constraint. The effective capacity is defined in [32, eqn. (12)]
by
1 h  Pt i
EC(θ) = − lim log E e−θ i=0 R[i] , t ≥ 0 (4)
t→∞ θt
where {R[i], i = 1, 2, ...} denotes a discrete-time service pro-
cess of the maximum achievable instantaneous service rate of
time [i], which is assumed to be ergodic and stationary. For a
block fading channel, the EC can be reduced to [35],
h  i
1
EC(θ) = − log E e−θR[i] . (5)
θ
The maximum achievable instantaneous service rate R[i] of
block i can be expressed as R[i] = T B log (1 + γ[i]), where T
denotes the block length duration, B is the channel bandwidth,
and γ[i] is the instantaneous SINR of block i.
III. E RGODIC C APACITY
This section reviews the optimal power allocation policies
of the secondary user in order to maximize its ergodic ca-
pacity (maximum achievable rate) under various constraints
categorized as short-term and long-term constraints. In the
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014 15

literature, many results have been proposed for cognitive

1.2
radios. In [17], Ghasemi and Sousa studied the optimal power gss: Rayleigh; gps: Rayleigh
gss: AWGN; gps: AWGN
allocation strategies for the secondary user through showing

Ergodic Capacity of the SU (nats/sec/Hz)

1 gss: AWGN; gps: Rayleigh
that with the same limit on the received power level, the gss: Rayleigh; gps: AWGN
channel ergodic capacity for a range of fading models (e.g., 0.8
Rayleigh, Nakagami-m and log-normal fading) exceeds that
of the non-fading AWGN channel. 0.6 Qpk= 0dB

The remainder of this section review some main results

in terms of the ergodic capacity of the SU under the short- 0.4 Q = −5dB
pk

term constraints, long-term constraints, or the combination

0.2
of short-term and long-term constraints. The constraints are
categorized as follows. 1) short-term constraints: peak transmit
0
power, peak interference power, and outage probability at cer- −10 −5 0 5
pmax (in dB)
10 15 20

tain channel state; 2) long-term constraints: average transmit

power, and average interference power. Intuitively, the short-
term constraints are more stringent than the long-term ones.
The following results hold in [17]–[19], [21], [30], [40]. We
omit the proofs which can be found in the associated papers.
A. Short-Term Constraints
In this subsection, we review the optimal power allocation
strategies for cognitive radio in order to maximize its ergodic
capacity constrained on various combinations of short-term
constraints, i.e., peak transmit power (PTP) denoted as Pmax ,
and peak interference power (PIP) denoted as Qpk . One
formulation of the optimization problem is given by
  
ps (gss , gps )gss
O1 : maximize E log 1 + (6)
ps (gss ,gps )≥0 N1 B
1
C1 : ps (gss , gps )gps ≤ Qpk (7)
C21 : ps (gss , gps ) ≤ Pmax (8)
where C11 and C12 denote the PIP and PTP constraints, respec-
tively, associated to the objective function O1 . Intuitively, the
Qpk
SU transmits using the power of min(Pmax , gps ), which is
also given in [20]. In this scenario, the SU transmitter exploits
only the interference channel state information (CSI).
The fading of the ST-PR channel determines the ergodic
capacity of the SU. In consequence, given Pmax the SU
achieves higher ergodic capacity when ST-PR channel experi-
ences severe fading, e.g. Rayleigh, than the case that ST-PR
is an AWGN without fading or Rician channel. Such that the
fade state of ST-PR is a good phenomenon for maximizing
the capacity of the SU. The challenging issue for this scheme
is to provide the accurate CSI of ST-PR at the secondary
Fig. 1. Ergodic capacity of the secondary user with different values of peak
interference constraints over AWGN and/or Rayleigh fading channels.
some transmission opportunities due to its peak transmit power
constraint.
1) Mean Value-based Power Allocation: The optimal power
strategy discussed in previous subsection requires the in-
stantaneous CSI of the ST-PR link. In this subsection, we
review the mean-value based power allocation (MVPA) and
the ergodic capacity of a cognitive-share radio under the
outage probability constraint of the interference power to the
primary user [42]. This means that the secondary user has
only the statistical information of the channel ST-PR. The
ergodic capacity optimization problem based on MVPA can
be formulated as [42]
Z ∞  
gss ps (g ps , gss )
maximize log 1 + fgss (gss ) dgss .
g ps ,gss 0 N1 B
(9)
 th
s.t. Pr gps ps (g ps , gss ) ≥ Qpk ≤ PO
(10)
where B denotes the bandwidth, gps represents the mean
value of gps that is assumed to be known at the secondary
transmitter, ps (gps , gss ) denotes the transmit power of the ST,
fgss (gss ) is the probability density function of gss which is
the channel power gain of the ST-SR link, and POth denotes the
predefined outage probability threshold that the instantaneous
interference is allowed to exceed the predefined peak inter-
ference power constraint Qpk . For Rayleigh fading,
n which o is
1 gss
assumed in this subsection, fgss (gss ) = g exp − g , and
ss ss

transmitter. The instantaneous CSI can be fed back to the
secondary transmitter [20], [41].
Figure 1 illustrates the achieved ergodic capacity of the
secondary user versus various peak transmit powers along
with different values of the peak interference power, where
we assumed that all the mean values of the channel power
gains are 1. We can observe that when the PIP constraint
is dominant, i.e. Pmax ≪ Qpk , the secondary user may
simply transmit at the maximum power to achieve its ergodic
capacity. Additionally, under Pmax ≪ Qpk , the fades of gps
are not beneficial to the ergodic capacity of the SU. This is
because over Rayleigh fading the SU is not able to exploit
N1 is the additive white Gaussian noise density at the SR. log
denotes natural logarithm operation.
The ergodic capacity of the SU with MVPA can be achieved
through employing a frame-work presented by Zouheir Rezki
and Mohamed-Slim Alouini in [43]. According to [43], the
interference outage probability constraint in the above opti-
mization problem is equivalent to
Qpk
ps (gss , gps ) ≤ (11)
Fg−1
ps (1 − POth )
where Fg−1
ps
(1 − POth ) denotes the inverse c.d.f. of gps . For
Rayleigh fading scenarios, the probability density function of
16 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

0.8
Qpk = 5dB
0
10 Qpk = 1dB
0.7

Ergodic capacity of the SU (nats/s/Hz)

Qpk = 0dB
Ergodic capacity of the SU (nats/s/Hz)

0.6 Qpk = −5dB

0.5

−1
10 0.4

Pth=0.01, ρ = 1 0.3
O
Pth
O
=0.2, ρ = 1
0.2
Pth
O
=0.01, ρ = 2
th
PO =0.2, ρ = 2 0.1
−2
10

0
−10 −8 −6 −4 −2 0 2 4 6 8 10 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Qpk (dB) Pth
O

th
Fig. 2. Ergodic capacity of the SU versus Qpk for different values of PO th for different values of Q
Fig. 3. Ergodic capacity of the SU versus PO pk
and ρ = g ps /g ss . and ρ = g ps /g ss = 1.

the channel power gain is continuous and not null so that of the optimization problem is given by
Fg−1 (·) exists. This new transformed constraint is called a   
ps ps (gss , gps )gss
variable peak transmit power constraint in [43]. In MVPA O2 : maximize E log 1 + (14)
ps (gss ,gps )≥0 N1 B
the secondary transmitter has the statistical information in
stead of the instantaneous ST-PR channel state information. In C12 : E {ps (gss , gps )gps } ≤ Qav (15)
addition, Fg−1ps
(1 − POth ) takes a fixed value [43]. This means C22 : E {ps (gss , gps )} ≤ Pav (16)
that the secondary user uses fixed transmit power which is
not variant with respect to gss . Based on the setting that where and C12
denote the AIP constraint and ATP con- C22
gps is exponentially distributed straint, respectively, associated to objective function O2 . Qav
 with a mean of gps , i.e., represents the predefined average interference power caused by
Fg−1
ps
(1 − POth ) = g ps log P1th . Consequently, the fixed the SU at the primary receiver, and Pav denotes the average
O
transmit power for the secondary transmitter is transmit power.
Qpk Here we have to point out that besides the mentioned
ps (gss , g ps ) ≤   (12) long-term constraints above there is another constraint called
g ps log P1th primary capacity loss constraint (PCLC) proposed in [45].
O
This method was shown to be better than the common ones,
where we may use the notation ps (g ps ) rather than
e.g. the average and/or peak interference power constraints,
ps (gss , gps ). We can obtain the ergodic capacity of the SU
in terms of achievable ergodic capacities of both the primary
exploiting MVPA as following
and the secondary links. It protects the primary transmission
Z ∞  
gss ps (g ps ) by ensuring that the maximum ergodic capacity loss of the
C= log 1 + fgss (gss ) dgss . primary link, due to the secondary transmission, is no greater
0 N1 B
Z ∞   than some predefined value. However, to enable the scheme,
gss ps (g ps ) 1 − ggss
= log 1 + e ss dgss . not only the CSI of the secondary fading channel and the
0 N1 B g ss
  fading channel from the secondary transmitter to the primary
N1 B
N1 B receiver, but also the CSI of the primary direct link. For details
= −e gss ps (gps ) Ei − (13)
gss ps (g ps ) please refer to [45].
where in the last two 1) AIP constraint only with perfect CSI: In this scenario,
R x steps
t
we have the help of [44, 4.337-
the secondary user aims to maximize its ergodic capacity under
2], and Ei (x) = −∞ et dt, x < 0 denotes the exponential
integral function [44, 8.211-1]. This result also was shown in the average interference power (AIP) constraint predefined by
[42] using a different method of proof. The ergodic capacity the primary user. This problem is denoted as (O2 , C12 ). The
versus Qpk and POth are plotted in Figure 2 and in Figure optimal power allocation scheme is waterfilling, which is given
3, respectively. We have to point out that in the discussed in [17] by
environment if POth → 0, the secondary user needs to stop  +
∗ 1 N1 B
transmission. ps (gss , gps ) = − , (17)
λgps gss
+
where [x] = max(x, 0), and gss and gps denote the channel
B. Long-Term Constraints power gains from the secondary transmitter to the secondary
We consider the long-term constraints are as follows: aver- receiver (ST-SR) and primary receiver (ST-PR), respectively.
age transmit power constraint (ATP) and average interference N1 is the noise density at the SR, B denotes the bandwidth,
power constraint (AIP) at the primary user. One formulation and λ ≥ 0 is the Lagrangian multiplier satisfying the AIP
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014 17

Ergodic capacity of the SU (nats/sec/Hz)

Ergodic capacity of the SU (nats/s/Hz)

0
0 10
10

AWGN
Rayleigh
2
AWGN Rayleigh, σe=0.01
−1
10 Lognormal,σ=4dB Rayleigh, σ2e=0.1
Lognormal,σ=8dB
2
Rayleigh Rayleigh, σe=0.4
Nakagami, m=2 −1
10
−15 −10 −5 0 5 10 15 20 −10 −8 −6 −4 −2 0 2 4 6 8 10
Qav/N1B (dB) Qav/(N0B) in dB

Fig. 4. Ergodic capacity of the SU with perfect CSI under AIP constraint in Fig. 5. The ergodic capacity of the SU under AIP constraint for AWGN, and
different fading scenarios. Rayleigh fading with/without estimation errors of ST-PR.

link has been presented in [19] as follows.

constraint given by (15). From the power allocation strategy, it
is obvious that when the secondary link is in a good condition,
the secondary user may not transmit if the interference link
is also in a good condition, which is dislike the conventional
waterfilling strategy in [9]. The authors in [17] named this
strategy as a 2-dimensional waterfilling. The ergodic capacity
is given by
" power gain estimation error. In addition, Egss ,ĝps [·] defines
C = E log 1 +
λN1 Bgps
ZZ
= log
gss
λN Bg ≥1
1 ps
Table I, which holds in [17], illustrates expressions of the
ergodic capacity of the SU given the channel distributions.
Figure 4 depicts the ergodic capacity. Fading is beneficial to
the cognitive radios. In the low Qav (normalized by N1 B)
regime, with fading the CR achieves mush better ergodic
capacity than the AWGN case. On the other hand, in high
Qav (normalized by N1 B) regime, all the ergodic capacity
approaches to the AWGN ergodic capacity. For some ranges
of Qav (normalized by N1 B), the fading degrades the ergodic
capacity. This can be explained as that the CR can not
utilize all the transmission opportunities because of the average
interference power constraint.
The above considers the perfect CSI. However, as we know
that the channel gains may be obtained through measurements
which suffers estimate errors. Therefore, the following illus-
trates the effects of imperfect CSI on the power allocation and
ergodic capacity of the secondary user.
2) AIP constraint with imperfect CSI: The previous strategy
considers the perfect channel state information. However, there
may have estimate errors on the channel gains. According
to the channel estimate mode reviewed in Section II, the
optimization problem and the OPA for the SU in fading envi-
ronments with imperfect channel information of interference
  
gss ps (gss , ĝps )
O3 : maximize E log 1 + (19)
ps (gss ,ĝps )≥0 N1 B
C13 : Egss ,ĝps [ps (gss , ĝps )ĝps ]
+ σe2 Egss ,ĝps [ps (gss , ĝps )] ≤ Qav (20)
where ĝps denotes the estimated channel power gain of the
ST-PR link, and σe2 represents the variance of the channel
 + !# the expectation over joint probability density function of gss
gss
−1 and ĝps . We can see that there is a penalty on the transmission
  (18) power of the secondary user because of the imperfect channel
gss estimation. Then Using the Lagrangian method, the optimal
dgss dps
λN1 Bgps power allocation can be directly obtained as
 
∗ 1 N1 B
ps (gss , ĝps ) = max 0, − (21)
λ (ĝps + σe2 ) gss
where the Lagrangian multiplier λ ≥ 0 satisfies the average
interference power constraint, and max {0, ·} operator guaran-
tees nonnegative transmit power.
The analytic results with/without perfect CSI are illustrated
in Figure 5. The SU loses its capacity because of the channel
estimate error that the SU has to lower its transmit power to
satisfy the AIP constraint. In addition, it is worth noting that
at higher values of AIP constraint, the ergodic capacity of the
SU with estimate error under Rayleigh fading (green curve)
is less than the one (blue curve) under AWGN, since the SU
has to use less transmit power in order to satisfy the AIP
constraint so that loses some opportunities for transmission.
Therefore, it is important to study the ergodic capacity with
other techniques for mitigating the influence of the estimation
error, e.g., diversity technique. The diversity technique is an
efficient means to increase the channel capacity [46]–[48].
We studied the OPA strategy of the secondary user under the
imperfect CSI and receiving MRC diversity, and the resultant
ergodic capacity in [49], [50].
3) ATP and AIP Constraints: Under average transmit
power constraint and average interference power constraint
predefined by the primary user, the optimization problem is
given by (O2 , C12 , C22 ) in (14), (15), and (16), respectively.
18 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

TABLE I
E RGODIC C APACITY OF THE SU WITH PERFECT SCI UNDER AIP CONSTRAINT.

gss gps Ergodic Capacity (nats/s/Hz)

 
Qav
AWGN AWGN log 1 + N B
1
h  i  2

log γ0
Lognormal (σ2 ) Lognormal (σ2 ) 2
1 + erf log
2σ
γ0
+ √σ
π
exp − log
4σ 2
γ0

Exponential (1) Exponential (1) log (1 + γ0 )

γ0
Nakagami (m = 2) Nakagami (m = 2) log (1 + γ0 ) − (1+γ0 )2

* where γ0 = 1/λN1 B.

4.5
Pav= −10dB
4 Pav= −7.5dB

Ergodic capacity of the SU (bits/s/Hz)

P = −5dB 5
Ergodic capacity of the SU (bits/s/Hz)

av
3.5
Pav= −2.5dB
4
3 Pav= 0dB
Pav= 2.5dB 3
2.5 Pav= 5dB
Pav= 7.5dB 2
2
Pav= 10dB
1
1.5 Pav= 12dB
Pav= 15dB 0
1 20
15
10 10
0.5 5
0 0
0 −5
−10 −5 0 5 10 15 Qav (in dB) −10 −10
Qav (in dB) Pav (in dB)

Fig. 6. The ergodic capacity of the SU under AIP and ATP constraints. Fig. 7. The ergodic capacity of the SU under AIP and ATP constraints.

The associated optimal power allocation scheme for the sec- ST-PR channel state information under the AIP constraint are
ondary user to maximize the ergodic capacity is given in [20] presented in detail in [50].
by From previous analysis, we can see that under the long-
 +
1 N1 B term constraints the optimal power allocation approaches are
p∗s (gss , gps ) = − , (22) (modified) waterfilling, and the water-level is jointly decided
µ + λgps gss
by the long-term constraints. In the following, we will take
where µ and λ are the nonnegative Lagrangian variables a look at how the combined constraints, long-term and short-
associated with the average transmit power constraint in (16) term, influence the power allocation and the ergodic capacity.
and average interference power constraint in (15), respectively.
We can see that this scheme is also waterfilling. However,
the water lever is related to not only the interference channel C. Combined Long-term and Short-term Constraints
condition, but also the average transmit power. Intuitively, even This section reviews the optimal power allocation schemes
having enough power budget for transmission and the CR link and the ergodic capacity of the secondary user under the
has a very good condition, the secondary user may not able to combined long-term and short-term constraints, which is pretty
transmit if the interference channel in a very good condition. different from the long-term constraints cases [53]. The opti-
To solve this problem (O2 , C12 , C22 ) we used ellipsoid method mization problem may be formulated as
[51], [52], shown in Table II.   
gss ps (gss , gps )
We show the simulation results in the Figure 6 and in Figure O4 : maximize E log 1 + (23)
ps (gss ,gps )≥0 N1 B
7, where we illustrate the ergodic capacity of the SU using
bits/s/Hz instead of nats/s/Hz only for a purpose of comparison C14 : ps (gss , gps )gps ≤ Qpk (24)
with the original results shown in [20]. We can see from the C24 : ps (gss , gps ) ≤ Ppk (25)
figures that at low Pav case the ergodic capacity is mainly C34 : E {ps (gss , gps )gps } ≤ Qav (26)
affected by the average transmit power constraint, in other
C44 : E {ps (gss , gps )} ≤ Pav (27)
words, ATP dominates AIP. On the other hand, at high Pav
regime, AIP dominates ATP. where (24) represents the peak interference power (PIP)
4) AIP constraint with imperfect CSI and receive MRC: constraint, (25) is the peak transmit power (PTP) constraint
The optimal power allocation schemes and the associated indicating the maximal transmit power of the SU, (26) and
ergodic capacity of the SU with receive MRC and imperfect (27) are the average interference power (AIP) constraint and
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014 19

TABLE II
E LLIPSOID M ETHOD : P SEUDOCODE .

1) Initialization: (subscript or superscript, k, denotes the kth loop)

λ1 : Lagrangian multiplier associated to AIP.
µ1 : Lagrangian multiplier associated to ATP.
A1 : a 2 × 2 positive-definite
(  matrix. An Ellipsoid, Ek , can be defined as )
  T   
λk λk λk
Ek (xk , Ak ) = : − xk A−1 − x k ≤ 1 , where
µk µk k µk
xk is the center of Ek .
2) repeat{
(k)
a) calculate ps using (22).  
λk
b) calculate the subgradients at using
 n µk o
(k)
Qpk − E ps (gss , gps )gps
sg =  n o 
(k)
Pav − E ps (gss , gps )
p
and normalized subgradients sg ˜ = sg/ sg T Ak sg
c) update
 the multipliers
  and the ellipsoid by
λk+1 λ
= k − A2+1 k sg
˜
.
µk+1 µk  
2
2 2
Ak+1 = (2+1) 2 −1 Ak − 2+1 Ak sg ˜ T Ak .
˜ sg
n o n o
(k) (k)
} until E ps (gss , gps )gps − Qpk ≤ 0, E ps (gss , gps ) − Pav ≤ 0,
p
and sg T Ak sg < ǫ, where ǫ is the desired accuracy.

the average transmit power (ATP) constraint, respectively.

The optimization problem can be solved by using Lagrangian
method. The numerical results can be obtained through using
Ergodic Capacity of the SU (nats/sec/Hz)

bisection method.
1) PIP and AIP Constraints: Under the PIP and AIP 0
10

constraints predefined by the primary user, the optimization

problem is given by (O4 , C14 , C34 ) in (23), (24), and (26).
The resultant optimal power allocation for the secondary user
holds in [30] by
AWGN, no PIP
 Qpk gps λ0 Rayleigh, ρ=0.5
 gps ,
 gss < N1 B Rayleigh, ρ=1
Rayleigh, ρ=1.5
p∗ (gss , gps ) = λ11gps − Ng1ssB , Nλ10B ≤ ggps ≤ Nλ11B (28) −1 Rayleigh, no PIP

 ss 10
−10 −8 −6 −4 −2 0 2 4 6 8 10
0, otherwise Qav/(N1B) in dB

where the Lagrangian multipliers λ0 ≥ 0 and λ1 ≥ 0 are asso-

Fig. 8. Ergodic capacity of the SU under average and peak interference
ciated to the PIP constraint given by (24) and AIP constraint by Q
constraints for AWGN and Rayleigh fading with different values of ρ = Qpk .
(26), respectively. We can see that the optimal power control av

to achieve the secondary maximum ergodic capacity under

joint peak and average interference power constraints at the
as
primary receiver is a function of the channel state information  1
of the secondary user and of the link ST-PR. Compared to 
 Ppk , gps ≤ N B
 λ(Ppk + g1ss )
the case that there is only AIP constraint, this strategy is a
combination of channel inversion and water-filling. The ratio p∗s (gss , gps ) = λgps
1
− N1 B
gss ,
1
N B < gps < gss
λN1 B
 λ(Ppk + g1ss )
g 

of the channel gains, gps , plays a key role in this case. In gss
ss 0, λN1 B ≤ gps
Figure 8, the ergodic capacity of the SU is plotted for AWGN (29)
Qpk
and Rayleigh fading with different ratios of ρ = Qav . In The Lagrangian multiplier λ satisfies the following KTT
addition, for ρ > 1 the figure states that at higher values condition
of Qav /(N1 B), the PIP constraint can be ignored. Moreover, Egss ,gps [gps p∗s (gss , gps )] = Qav (30)
when ρ < 1 the secondary user loses some opportunities
for transmission resulting in lower ergodic capacity than the The optimal power allocation scheme is a combination of
AWGN case at higher regime of Qav /(N1 B). the fixed power transmission and the water filling approach.
2) PTP and AIP Constraints: The optimization problem Figure 9 depicts the simulation results. If there is no interfer-
under peak transmit power constraint (PTP) and average ence power constraint, the secondary user transmits by using
interference power constraint is given by (O4 , C24 , C34 ) in its maximal transmit power. With the combination of PTP and
(23), (25), and (26). The optimal power allocation for the AIP constraints, if the value of the AIP constraint is smaller
secondary user to maximize the ergodic capacity holds in [20] compared to the value of the PTP, the secondary user looses
20 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

3 1.8
Ppk=0dB
Ppk=5dB 1.6
Q = 5 dB
Ergodic Capacity of the SU (nats/sec/Hz)

2.5 P =5dB, no interf. constr. pk

Ergodic capacity of the SU (nats/s/Hz)

pk
1.4
Ppk=10dB, no interf. constr.

2 Ppk=10dB 1.2
P =0dB, no interf. constr.
pk
no transmit power constr. 1
1.5
Q = 0 dB
0.8 pk

1 0.6

0.4 Q = −5 dB
pk
0.5
0.2 Without ATP
With ATP
0 0
−10 −8 −6 −4 −2 0 2 4 6 8 10 −10 −5 0 5 10 15
Qav (dB) Pav (dB)

Fig. 9. Ergodic capacity of the SU under different values of PTP and AIP Fig. 10. Ergodic capacity of the SU under different values of ATP and PIP
constraints over Rayleigh fading. Constraints over Rayleigh fading.

some opportunities to transmit. This corresponds to the AIP- schemes are combined channel inversion and water filling or
dominant regime. On the other hand, if the PTP is dominant, two-dimensional water filling. As we know that in the analysis
the ergodic capacity is unbounded by the ergodic capacity with of ergodic capacity, the delay limit is not considered which
the ATP constraint. This is because that the SU has a lot of means that it can be approaching to infinity. In the successive
chances to transmit, but the PTP limits the ergodic capacity. section, we review the effective capacity which takes the delay
3) ATP and PIP Constraints: Under the average transmit into consideration.
power constraint and peak interference power constraint, the
optimization problem is formulated by (O4 , C14 , C44 ) in (23),
IV. E FFECTIVE C APACITY
(24), and (27). The resultant optimal power allocation for the
secondary user holds in [20] as The concept of effective capacity (EC) has been reviewed
 Qpk Q in subsection II-B. This section reviews some results of

 , gps ≥ 1 pk N1 B , gss > λN1 B
optimal power allocation strategies and effective capacity of
 gps λ − gss

p∗s (gss , gps ) = λ1 − Ng1ssB , gps < 1 QpkN1 B , gss > λN1 B
the secondary user over block fading channels. For simplicity,

 λ − gss we in this section omit the parameters for ps (θ, gss , gps ), i.e.

0, gss ≤ λN1 B the transmit power of the SU is denoted as ps . The objective
(31) function and possible constraints are listed in the following,
It is intuitive that the power allocation scheme is a combination     
of channel inverse and waterfilling. The waterfilling reflects 1 −θT B log 1+ pNs gss
O5 : maximize − log E e 1B (32)
the average transmit power constraint and the channel inverse ps (θ,gss ,gps )≥0 θ
reflects the peak interference power constraint. The simulation Constraints:
results are shownin Figure 10. The  ergodic capacity of the SU C15 : ps gps ≤ Qpk (33)
Qpk
is capped by log 1 + gps N1 B gss . In the low-ATP regime, the
C25 : ps ≤ Ppk (34)
ergodic capacity is dominated by the ATP constraint, while in
the high-ATP regime the ergodic capacity is limited by the PIP. C35 : E {ps gps } ≤ Qav (35)
These can be explained as follows: In the low-ATP regime, the C45 : E {ps } ≤ Pav (36)
power allocation scheme is mainly the water-filling, and in in
the high-ATP regime the power allocation scheme is performed where T denotes the block length duration, B is the channel
as the channel inversion. bandwidth, θ is the delay exponent, Ppk denotes the maximum
allowed peak transmit power, Pav denotes the average transmit
power constraint, and Qpk and Qav represent the peak and
D. Summary average interference power threshold, respectively. We can
We can see that the strategies how the secondary user use Lagrangian method to solve the optimization problems
allocates the optimal transmit power to maximize the ergodic with different combinations of constraints. Without loss of
capacity are decided by the types of constraints. When the generality we assume that T B = 1 and N1 B = 1 in following
constraint is the peak interference constraint, the OPA is the simulations.
channel inversion with respect to the interference channel
from the secondary transmitter to the primary receiver (ST-
PR). If the constraint is the average interference constraint, A. Short-term Constraint
the OPA is the water-filling scheme, where the water level is The same as in the previous sections that the short-term
decided by the interference channel, ST-PR, power gain. For constraints include the peak transmit power constraint and the
the combined long-term and/or short-term constraints, the OPA peak interference power constraint.
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014 21

1) PTP and PIP Constraints: The optimization problem is power constraint becomes to dominate. In the two figures, we
given by (O5 , C15 , C35 ) in (32), (33), and (34). The power also show the upper bounds given by Eq. (39), i.e. no peak
allocation strategy is straightforward
 obtained  that the SU transmit power constraint, for Qpk = −5dB and Qpk = 5dB.
Qpk
transmits using the power of min Ppk , gps . Then the ef-
fective capacity can be obtained as
    −θT B  0
10
ρ=0.001
 Qpk  ρ=0.01
1  min P ,
pk gps g ss  ρ=0.1
 

Effective capacity of the SU (nats/s/Hz)

EC = − log E  1 +   ρ=1
ρ=100
θ 
 N1 B 
 ECup, Q = 5dB
−1 pk
10
ECup, Qpk = −5dB
(37)
Given the distributions of the fading channels, the expression
of the EC can be obtained numerically, since, to the best −2
10
of our knowledge, there are no closed-form expressions for
the common fading scenarios, e.g., Rayleigh and Nakagami-
m. Thus we derive the upper bound expression, i.e. without −3
10
considering the peak transmit power constraint, under inde-
pendent Rayleigh fading for the effective capacity of this case −2
10
−1
10
0
10
θ
1
10 10
2

as a verification. Let h denote the ratio of two independent

exponential variables ggps
ss
, and h̄ be the mean ratio of g ss /gps . Fig. 11. Effective capacity of the SU under different values of θ over Rayleigh
Then we have the p.d.f. of h by using [54, 5-15] as fading with Ppk = −5dB, where ρ = Qpk /Ppk .

h
f (h) =
2 (38)
h+h
ρ=0.001
( " −θT B #)
0
10 ρ=0.01
ρ=0.1
1 Qpk gss
Effective capacity of the SU (nats/s/Hz)

EC ub = − log E 1 + ρ=1
ρ=10
θ gps N1 B ρ=100
  ECup, Q = 5dB

 
 −1
pk


 " −θT B #

10 ECup, Qpk = −5dB

1  Qpk h 
= − log E 1 +
θ 
 N1 B 


 }

| {z  −2
C1 10

where
Z ∞ −θT B
Qpk h h
C1 =
−2 −1 0 1 2
1+
2 dh 10 10 10
θ
10 10

0 N 1 B h+h
 
Qh Fig. 12. Effective capacity of the SU under different values of θ over Rayleigh
= B(1, 1 + θT B) 2 F1 θT B, 1; θT B + 2, 1 − fading with Ppk = 5dB, where ρ = Qpk /Ppk .
N1 B
where, in the last step, we have used [44, 3.197-1],
2 F1 (a, b; c, d) is hypergeometric function [44, 9.14], and
B(a, b) denotes the beta function [44, 8.38]. B. Long-term Constraints
Then we have In the following, we review the optimal power allocation
1 strategies and the simulation results of the effective capacity
EC ub = − log {B(1, 1 + θT B) of the SU under long-term constraints.
θ   (39) 1) AIP Constraint: Under average interference power con-
Qh
× 2 F1 θT B, 1; θT B + 2, 1 − straint and secondary QoS constraint, the optimization problem
N1 B
is given by (O5 , C35 ) in (32) and (34). The resultant optimal
Figures 11 and 12 illustrate the effective capacity of the power allocation for the secondary user to maximize effective
SU versus different values of the delay component along with capacity holds in [38] by
the different ratios of PIP and PTP over Rayleigh fading. In  " #+
the simulation we assume that the mean value of the channel 

1
β 1+α
N1 B − g1ss , gps ≤ βgss
power gains are 1, and the AWGN power at the receiver is p∗s = 1
1+α 1+α
gps gss
α
(40)


1. First, it is intuitive that when the value of Qpk decreases, 0, otherwise
i.e. ρ decreases, the effective capacity of the SU decreases.
Second, when the value of ρ is bigger than 1 , for instance, where [x]+ = max(0, x), α = θT B, β = λNα1 B , and λ is the
ρ = 1 and ρ = 100, the effective capacity of the SU increases non-negative Lagrangian variable associated with the average
slowly and will converge. This is because the peak transmit interference power constraint. Based on the above optimal
22 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

1.8 1.4
Qav=−5dB, Pav=−5dB
θ=0.001
1.6 θ=0.01 Qav=5dB, Pav=−5dB
1.2
θ=0.1 Q =−5dB, P =5dB
Effective capacity of the SU (nats/s/Hz)

Effective capacity of the SU (nats/s/Hz)

av av
1.4 θ=1 Q =5dB, P =5dB
θ=10 1
av av

1.2 θ=20
θ=50
1 θ=100 0.8

0.8 0.6

0.6
0.4
0.4

0.2
0.2

0 0 −2 −1 0 1
−5 −4 −3 −2 −1 0 1 2 3 4 5 10 10 10 10
Qav (in dB) θ

Fig. 13. Effective capacity of the SU under different values of θ over Rayleigh Fig. 14. Effective capacity of the SU over Rayleigh fading.
fading.

suggests us that when ATP≪AIP, the secondary user can

power allocation scheme, the closed-form expression of the ignore the average interference constraint.
effective capacity of the SU over Nakagami-m fading channels
was derived in [38]. Here we show the simulations over i.i.d.
Rayleigh fading channels. C. Combined Long- and Short-Term Constraints
From the simulation shown in Figure 13 we can see that To the best of our knowledge, the results of the effective
when the delay is stringent (θ has large values), the average capacity under the combined long-term and short-term con-
interference power constraint sightly influences the effective straints have not been proposed in literature. In following we
capacity of the SU. This is because the SU needs to transmit show our results.
at very low rate in order to fulfill the delay requirement. This 1) ATP and PIP Constraints: Under average transmit
is, the delay component dominates the effective capacity of power and peak interference power constraints, and delay
the SU. On the other hand, θ is small, the AIP has dramatic constraint, the optimization problem is given by (O5 , C15 , C46 )
influence on the effective capacity. The reason is that in lower in (32), (33) and (36). Using Lagrangian method, the resultant
AIP regime the SU has very limited amount of opportunities to optimal power allocation for the secondary user to maximize
transmit; however, in higher AIP regime, the SU could utilize effective capacity is obtained as
almost all the opportunities for its transmission. Now the AIP  (  1 + )
dominates the effective capacity of the SU. 
min Qpk , N B β 1+α 1
1 α − gss , 1 ≤ βgss
2) ATP and AIP Constraints: The results of the effective p∗s = gps 1+α
gss


capacity under the average interference power and average 0, otherwise
transmit power constraints along with the QoS constraint, (42)
to the best of our knowledge, have not been proposed in where α = θT B, β = λNα1 B , and λ is the non-negative La-
literature. In following we show our results. grangian variable associated with the average transmit power
The optimization problem is given by (O5 , C35 , C45 ) in (32), constraint. This power allocation is a water-filling scheme but
(35), and (36). The resultant optimal power allocation for the capped by the peak interference power constraint. Therefore,
secondary user to maximize effective capacity can be obtained the water level is defined by these two constraints. To the best
by using Lagrangian method as of our knowledge, there is no closed-form expression for the
" #+ effective capacity for this case. Here we show the simulations
1
∗ α 1+α 1 for i.i.d. Rayleigh fading channels in Figure 15.
ps = N 1 B α − (41)
1
1+α gss In Figure 15, we can discover two differences from the
(N1 B (λgps + µ)) 1+α gss
previous case which is under AIP and ATP constraints. First,
where α = θT B, and λ and µ are the non-negative Lagrangian in low ATP, Pav = −5dB, cases, the values of the effective
variables associated with the AIP constraint in (35) and ATP capacity for Qpk = −5dB and Qpk = 5dB are slightly
constraint in (36), respectively. To the best of our knowledge, different. When the value of the delay component θ is small,
with the above optimal power allocation scheme there is no i.e. the delay is not stringent, the higher Qpk value, the
closed-expression of the effective capacity. Here we show the larger effective capacity. However, when θ is big, the EC
simulations in Figure 14 for i.i.d. Rayleigh fading channels. has lower values when Qpk = 5dB than Qpk = −5dB.
The optimal power can be obtained applying the pseudocode Second, in the case that Pav = 5dB, the gap of the effective
in Table II by using proper transmit power and interference capacity of the two cases, Qpk = 5dB and Qpk = −5dB,
constraints. From the Figure, one thing we need to point out increases. The effect of the Qpk is similar in the lower and
is that in the low ATP, Pav = −5dB, the ATP constraint higher value regimes of θ. The phenomenon can be explained
dominates the effective capacity of the secondary user. This as follows: when θ is small, the secondary user is able to
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014 23

1.4 1.4
Qpk= −5dB, Pav= −5dB Qav= −5dB, Ppk= −5dB
Qpk= −5dB, Pav= 5dB Qav= −5dB, Ppk= 5dB
1.2 1.2
Q = 5dB, P = −5dB Q = 5dB, P = −5dB
Effective capacity of the SU (nats/s/Hz)

Effective capacity of the SU (nats/s/Hz)

pk av av pk
Q = 5dB, P = 5dB Q = 5dB, P = 5dB
pk av av pk
1 1

0.8 0.8

Pav= 5dB
0.6 0.6

0.4 0.4

0.2 P = −5dB 0.2

av
Ppk= −5dB, Qav = +/− 5dB
0 −2 −1 0 1
0 −2 −1 0 1
10 10 10 10 10 10 10 10
θ θ

Fig. 15. Effective capacity of the SU over Rayleigh fading under PIP and Fig. 16. Effective capacity of the SU over Rayleigh fading under AIP and
ATP constraints. PTP constraints.

utilize higher power to transmit when the opportunities appear V. C ONCLUSION

with the average transmit power budget; however, when delay
In this paper, we reviewed the main results of the optimal
requirement is stringent, the secondary user needs to maintain
power allocation schemes for cognitive radios under different
a constant rate as possible not to use a higher power to obtain
constraints and objectives.
a higher instantaneous transmission rate. In later case, the
The optimal power allocation schemes mainly can be cate-
secondary user has more opportunities than the former case
gorized as following:
for transmission.
2) PTP and AIP Constraints: Under peak transmit power • Channel inversion : when only the peak interference

and average interference power constraints, and delay con- power constraint is applied, i.e. short-term constraint.
straint, the optimization problem is given by (O5 , C25 , C35 ) in • Two-dimensional waterfilling: when the average transmit

(32), (34) and (35). Using Lagrangian method, the resultant and/or interference power constraints are applied, i.e.
optimal power allocation for the secondary user to maximize long-term constraints.
effective capacity is obtained as • Capped two-dimensional waterfilling: when the aver-
  " #+  age/peak transmit power constraint and peak/average in-

  1  terference power constraint are considered, i.e. combined
min P , N B β 1+α
− 1 g
, gps ≤β
pk 1
p∗s =
1 α
 1+α 1+α
gps gss
gss  ss long- and short-constraints.


 The ergodic capacity is mainly influenced by the interfer-
0, otherwise
ence channel from the secondary transmitter to the primary
(43)
receiver. In addition, the effective capacity is affected by the
where α = θT B, β = λNα1 B , and λ is the non-negative
delay component besides the interference channel. Especially,
Lagrangian variable associated with the average interference
the short-term constraints, i.e. peak transmit power and peak
power constraint. This power allocation scheme is capped by
interference power constraints, have different influences on
the peak transmit power. Thus the water level of the water
the effective capacity over lower and higher delay component
filling algorithm is different from the AIP-only case and the
regimes.
ATP-PIP scenario that the water level is changing from block
to block. To the best of our knowledge, there is no closed-form
expression for the effective capacity for this case. In Figure 16 R EFERENCES
we show the simulations for i.i.d. Rayleigh fading channels. [1] N. Hoven and A. Sahai, “Power scaling for cognitive radio,” in Proc.
From Figure 16 we can see that when the average inter- Int’l Conf. Wireless Netw., Commun. and Mobile Computing, Vol. 1,
ference power constraint is low, Qav = −5dB, higher peak pp. 250–255, Jun. 2005.
[2] M. Chiang, P. Hande, T. Lan, and C. W. Tan, “Power control in wireless
transmit power threshold is not an advantage at the higher cellular networks,” Foundations and Trends
inR Networking, Vol. 2,
Ppk regime. In addition, the effective capacity of the case of No. 4, pp. 381–533, 2008.
Qav = 5dB and Pav = −5dB is supreme over the case of [3] O. Kaya and S. Ulukus, “Optimum power control for cdma with
deterministic sequences in fading channels,” IEEE Trans. Inf. Theory,
Qav = −5dB and Pav = 5dB at the stringent delay regime. Vol. 50, No. 10, pp. 2449–2462, Oct. 2004.
This can be explained as that when θ is large, the secondary [4] R. Knopp and P. Humblet, “Information capacity and power control
user needs to maintain the rate as constant as possible, and in single-cell multiuser communications,” in Proc. IEEE ICC, Vol. 1,
pp. 331–335, Jun. 1995.
in lower Qav and higher Ppk case the SU has much less [5] R. Yates, “A framework for uplink power control in cellular radio
opportunities to transmit than the case of higher Qav and systems,” IEEE J. Sel. Areas Commun., Vol. 13, No. 7, pp. 1341–1347,
lower Ppk . Moreover, when the PTP constraint is lower and Sep. 1995.
[6] A. Goldsmith and P. Varaiya, “Capacity of fading channels with channel
not bigger than Qav , Ppk = −5dB, the AIP constraint can be side information,” IEEE Trans. Inf. Theory, Vol. 43, No. 6, pp. 1986–
ignored. 1992, Nov. 1997.
24 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014
[7] A. Goldsmith, S. A. Jafar, I. Maric, and S. Srinivasa, “Breaking spectrum
gridlock with cognitive radios: An information theoretic perspective,”
Proc. IEEE, Vol. 97, No. 5, pp. 894–914, May 2009.
[8] S. Haykin, “Cognitive radio: brain-empowered wireless communica-
tions,” IEEE J. Sel. Areas Commun., Vol. 23, No. 2, pp. 201–220,
Feb. 2005.
[9] T. M. Cover and J. A. Thomas, Elements of Information Theory, 2nd ed.
John Wiley & Sons, Inc., 2006.
[10] Y. Chen, G. Yu, Z. Zhang, H. hwa Chen, and P. Qiu, “On cognitive radio
networks with opportunistic power control strategies in fading channels,”
IEEE Trans. Wireless Commun., Vol. 7, No. 7, pp. 2752–2761, Jul. 2008.
[11] S. Srinivasa and S. A. Jafar, “Soft sensing and optimal power control
for cognitive radio,” IEEE Trans. Wireless Commun., Vol. 9, No. 12,
pp. 3638–3649, Dec. 2010.
[12] H. Islam, Y. chang Liang, and A. Hoang, “Joint power control and beam-
forming for cognitive radio networks,” IEEE Trans. Wireless Commun.,
Vol. 7, No. 7, pp. 2415–2419, Jul. 2008.
[13] L. Qian, X. Li, J. Attia, and Z. Gajic, “Power control for cognitive
radio ad hoc networks,” in Proc. IEEE LANMAN Workshop, pp. 7–12,
Jun. 2007.
[14] K. Koufos, K. Ruttik, and R. Jäntti, “Controlling the interference from
multiple secondary systems at the tv cell border,” in PIMRC, pp. 645–
649, 2011.
[15] M. Gastpar, “On capacity under received-signal constraints,” in
Proc.42nd Annual Allerton Conf. on Commun., Control and Comp.,
(Monticello, IL, USA), Sept. 29 - Oct. 1 2004.
[16] ——, “On capacity under receive and spatial spectrum-sharing con-
straints,” IEEE Trans. Inf. Theory, Vol. 53, No. 2, pp. 471–487,
Feb. 2007.
[17] A. Ghasemi and E. S. Sousa, “Fundamental limits of spectrum-sharing
in fading environments,” IEEE Trans. Wireless Commun., Vol. 6, No. 2,
pp. 649–658, Feb. 2007.
[18] H. Suraweera, P. Smith, and M. Shafi, “Capacity limits and performance
analysis of cognitive radio with imperfect channel knowledge,” IEEE
Trans. Veh. Technol., Vol. 59, No. 4, pp. 1811–1822, May 2010.
[19] L. Musavian and S. Aissa, “Fundamental capacity limits of cognitive
radio in fading environments with imperfect channel information,” IEEE
Trans. Commun., Vol. 57, No. 11, pp. 3472–3480, Nov. 2009.
[20] X. Kang, Y.-C. Liang, A. Nallanathan, H. Garg, and R. Zhang, “Opti-
mal power allocation for fading channels in cognitive radio networks:
Ergodic capacity and outage capacity,” IEEE Trans. Wireless Commun.,
Vol. 8, No. 2, pp. 940–950, Feb. 2009.
[21] R. Zhang, “On peak versus average interference power constraints for
protecting primary users in cognitive radio networks,” IEEE Trans.
Wireless Commun., Vol. 8, No. 4, pp. 2112–2120, Apr. 2009.
[22] L. Ozarow, S. Shamai, and A. Wyner, “Information theoretic consider-
ations for cellular mobile radio,” IEEE Trans. Veh. Technol., Vol. 43,
No. 2, pp. 359–378, May 1994.
[23] E. Biglieri, J. Proakis, and S. Shamai, “Fading channels: information-
theoretic and communications aspects,” IEEE Trans. Inf. Theory, Vol. 44,
No. 6, pp. 2619–2692, Oct. 1998.
[24] H. Ding, J. Ge, D. Benevides da Costa, and Z. Jiang, “Asymptotic
analysis of cooperative diversity systems with relay selection in a
spectrum-sharing scenario,” IEEE Trans. Veh. Technol., Vol. 60, No. 2,
pp. 457–472, Feb. 2011.
[25] J. Lee, H. Wang, J. G. Andrews, and D. Hong, “Outage probability
of cognitive relay networks with interference constraints,” IEEE Trans.
Wireless Commun., Vol. 10, No. 2, pp. 390–395, Feb. 2011.
[26] C. Zhong, T. Ratnarajah, and K.-K. Wong, “Outage analysis of decode-
and-forward cognitive dual-hop systems with the interference constraint
in nakagami- fading channels,” IEEE Transactions on Vehicular Tech-
nology, Vol. 60, No. 6, pp. 2875 –2879, Jul. 2011.
[27] T. Duong, P. L. Yeoh, V. N. Q. Bao, M. Elkashlan, and N. Yang,
“Cognitive relay networks with multiple primary transceivers under
spectrum-sharing,” IEEE Signal Process. Lett., Vol. 19, No. 11, pp. 741–
744, 2012.
[28] K. J. Kim, T. Duong, H. Poor, and L. Shu, “Performance analysis of
cyclic prefixed single-carrier spectrum sharing relay systems in primary
user interference,” IEEE Trans. Signal Processing, Vol. 60, No. 12,
pp. 6729–6734, Dec. 2012.
[29] R. Etkin, A. Parekh, and D. Tse, “Spectrum sharing for unlicensed
bands,” IEEE J. Sel. Areas Commun., Vol. 25, No. 3, pp. 517–528,
Apr. 2007.
[30] L. Musavian and S. Aissa, “Capacity and power allocation for spectrum-
sharing communications in fading channels,” IEEE Trans. Wireless
Commun., Vol. 8, No. 1, pp. 148–156, Jan. 2009.
[31] D. Tse and S. Hanly, “Multiaccess fading channels. I. polymatroid
structure, optimal resource allocation and throughput capacities,” IEEE
Trans. Inf. Theory, Vol. 44, No. 7, pp. 2796–2815, Nov. 1998.
[32] D. Wu and R. Negi, “Effective capacity: a wireless link model for
support of quality of service,” IEEE Trans. Wireless Commun., Vol. 2,
No. 4, pp. 630–643, Jul. 2003.
[33] ——, “Downlink scheduling in a cellular network for quality-of-service
assurance,” IEEE Trans. Veh. Technol., Vol. 53, No. 5, pp. 1547–1557,
Sep. 2004.
[34] C.-S. Chang and J. Thomas, “Effective bandwidth in high-speed digital
networks,” IEEE J. Sel. Areas Commun., Vol. 13, No. 6, pp. 1091–1100,
Aug. 1995.
[35] J. Tang and X. Zhang, “Quality-of-service driven power and rate
adaptation over wireless links,” IEEE Trans. Wireless Commun., Vol. 6,
No. 8, pp. 3058–3068, Aug. 2007.
[36] ——, “Quality-of-service driven power and rate adaptation for mul-
tichannel communications over wireless links,” IEEE Trans. Wireless
Commun., Vol. 6, No. 12, pp. 4349–4360, Dec. 2007.
[37] S. Akin and M. Gursoy, “Effective capacity analysis of cognitive radio
channels for quality of service provisioning,” IEEE Trans. Wireless
Commun., Vol. 9, No. 11, pp. 3354–3364, Nov. 2010.
[38] L. Musavian and S. Aissa, “Effective capacity of delay-constrained
cognitive radio in Nakagami fading channels,” IEEE Trans. Wireless
Commun., Vol. 9, No. 3, pp. 1054–1062, Mar. 2010.
[39] M. Gursoy, D. Qiao, and S. Velipasalar, “Analysis of energy efficiency in
fading channels under qos constraints,” IEEE Trans. Wireless Commun.,
Vol. 8, No. 8, pp. 4252–4263, Aug. 2009.
[40] C.-X. Wang, X. Hong, H.-H. Chen, and J. Thompson, “On capacity
of cognitive radio networks with average interference power con-
straints,” IEEE Trans. Wireless Commun., Vol. 8, No. 4, pp. 1620–1625,
Apr. 2009.
[41] J. Peha, “Sharing spectrum through spectrum policy reform and cogni-
tive radio,” Proc. IEEE, Vol. 97, No. 4, pp. 708–719, Apr. 2009.
[42] S. Lim, H. Wang, H. Kim, and D. Hong, “Mean value-based power
allocation without instantaneous CSI feedback in spectrum sharing
systems,” IEEE Trans. Wireless Commun., Vol. 11, No. 3, pp. 874–879,
Mar. 2012.
[43] Z. Rezki and M. Alouini, “Ergodic capacity of cognitive radio under
imperfect channel-state information,” IEEE Trans. Veh. Technol., Vol. 61,
No. 5, pp. 2108–2119, Jun. 2012.
[44] I. Gradshteyn and I. Ryzhik, Table of Integrals, Series, and Products,
Seventh Edition, 7th ed. Academic Press, 2007.
[45] R. Zhang, “Optimal power control over fading cognitive radio channel by
exploiting primary user CSI,” in Proc. IEEE GLOBECOM, Nov. 2008.
[46] D. Brennan, “Linear diversity combining techniques,” Proc. IEEE,
Vol. 91, No. 2, pp. 331–356, Feb. 2003.
[47] M.-S. Alouini and A. Goldsmith, “Capacity of rayleigh fading channels
under different adaptive transmission and diversity-combining tech-
niques,” IEEE Trans. Veh. Technol., Vol. 48, No. 4, pp. 1165–1181,
Jul. 1999.
[48] I. E. Telatar, “Capacity of multi-antenna gaussian channels,” Eur. Trans.
Telecomm., Vol. 10, No. 6, pp. 585–595, 1999.
[49] R. Duan, M. Elmusrati, R. Jäntti, and R. Virrankoski, “Capacity for
spectrum sharing cognitive radios with MRC diversity at the secondary
receiver under asymmetric fading,” in Proc. IEEE Global Telecommu-
nications Conference, pp. 1–5, Dec. 2010.
[50] R. Duan, R. Jäntti, M. Elmusrati, and R. Virrankoski, “Capacity for
spectrum sharing cognitive radios with MRC diversity and imperfect
channel information from primary user,” in Proc. IEEE Global Telecom-
munications Conference, pp. 1–5, Dec. 2010.
[51] R. G. Bland, D. Goldfarb, and M. J. Todd, “The ellipsoid method: A
survey,” Operations Research, Vol. 29, No. 6, pp. 1039–1091, Dec. 1981,
http://www.jstor.org/stable/170362.
[52] S. Boyd and L. Vandenberghe, Convex Optimization. New York, NY,
USA: Cambridge University Press, 2004.
[53] M. Khojastepour and B. Aazhang, “The capacity of average and peak
power constrained fading channels with channel side information,” in
Proc. IEEE WCNC, Vol. 1, pp. 77–82, Mar. 2004.
[54] A. Papoulis and S. U. Pillai, Probability, Random Variables and Stochas-
tic Processes, 4th ed. McGraw-Hill, 2002.
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014
On Effective Capacity of Cognitive Radios with
TAS and MRC
Ruifeng Duan and Mohammed S. Elmusrati
Abstract—We investigate the effective capacity of a cognitive-
shared channel with implementing transmit antenna selection at
the secondary transmitter and maximal ratio combining at the
secondary receiver under different transmit antenna selection
schemes, minimum interference selection, maximum secondary
composite channel gain selection, and the maximum channel ratio
selection. Closed-form expressions for the effective capacity are
presented and validated through simulations.
Index Terms—Cognitive radio, generalized selection combin-
ing, ergodic capacity, symbol error probability.
I. I NTRODUCTION
F Rom the literature we know that the ergodic capacity
has no transmission delay limitation, while the outage
capacity does not allow any delay. To this point, the concept
of effective capacity (EC) was developed in [1] to define the
maximum arrival data rate that can be supported by the channel
subject to the required communication delay. It is a link-layer
channel model and can be interpreted as the dual of effective
bandwidth [2]. The quality of service (QoS) is represented by a
term, named QoS exponent. When the QoS exponent θ → 0,
it means that there is no delay limitation, and the effective
capacity equals the ergodic capacity. On the other hand, the
link cannot tolerate any delay as θ → ∞. This concept has
received much attention in the point-to-point case, e.g., [3],
[4], as well as in cognitive radios, e.g., [5], [6] and references
therein.
A combined transmitter antenna selection and receiver
maximal ratio combing scheme was proposed in [7]. The
authors investigated the system performance in terms of the
average bit-error rate, and proved that the diversity order of
the product of the number of transmit and receive antennas
could be achieved, so that this scheme significantly improves
the diversity. On one hand, this combined scheme provides
not only the diversity gain as the selection combining at the
transmitter and the selection combining at the receiver (SC/SC)
but also provides the combining gain which the SC/SC does
not have. On the other hand, using the SC at the transmitter
reduces the complexity of the transmitter and transmit power
so that it is suitable in the uplink communication [7]. There-
after, the transmit antenna selection (TAS) and the maximal
ratio combining (MRC) techniques have been extensively
The authors are with Communications and Systems Engineering Group,
University of Vaasa, Finland. (emails: ruifeng.duan@ieee.org; moel@uva.fi).
The content of this paper is part of Duan’s thesis, and is reprinted with the
permission of the University of Vaasa. This work was supported in part by
the SMACIW (Statistical Modeling and Control of Aggregate Interference in
Wireless Systems, Decision no. 265077) project funded by the Academy of
Finland.
25
studied in the literature for traditional wireless communication
systems, e.g. [8]–[10]. In [8], [9], the authors investigated the
outage probability and bit error rate of TAS/MRC in Rayleigh
and Nakagami-m fading channels, respectively. In [10], the
authors proposed the expressions of the outage probability of
multiuser diversity for a TAS/MRC system in independent
and identically distributed Nakagami-m channels. The full
diversity is achieved at high SNR regime.
Recently, the authors in [11] studied the ergodic capacity
of a spectrum sharing secondary user link with TAS/MRC, in
the Rayleigh fading environment. The closed-form expression
of the ergodic capacity of the secondary user with peak
interference power constraint was derived, while the case with
additional peak transmit power for the SU was simulated.
The ergodic capacity depends only on the product of the
numbers of transmit and receive antennas and the product of
the peak signal-to-noise ratio (SNR) and the mean channel
power gain ratio. To the best of our knowledge, the EC of a
cognitive-shared channel with TAS/MRC in a Rayleigh fading
environment and the comparison of the EC with different
transmit antenna selection schemes have not been studied in
the literature. We investigated this problem under different
selection schemes in this paper.
The remainder of this paper is organized as follows. In
Section II, we present the system and channel model, and the
transmit selection schemes investigated in this paper, as well
as the probability density function (p.d.f) and cumulative dis-
tribution function (c.d.f) of the SNR at the secondary receiver.
In Section III, the EC of the SU under peak interference power
(PIP) constraint is studied with different antenna selection
techniques. In Section IV, the EC of the secondary user
under average interference power (AIP) constraint is analyzed.
The results are then depicted in Section V. The last Section
concludes this paper.
II. S YSTEM AND C HANNEL M ODEL
The spatial diversity scheme for a cognitive-shared channel
investigated in this paper is depicted in Figure 1. This model
forms a combination of transmit antenna selection (TAS) at
the secondary transmitter (ST) and maximal ratio combining
(MRC) at the secondary receiver (SR). The ST is equipped
with M antennas and the SR employs L receiving branches
that are assumed to be independent from each other. The
primary receiver (PR) has one antenna. We assume that the
secondary system is far away from the primary transmitter
(PT), so that the interference from the PT to the SR is
ignored. This pattern has been adopted widely in the research
26 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

(i)
corresponding p.d.f. and c.d.f. of gss are given as follows [21]:
hL−1 exp (−h/gss )
fg(i) (h) = , h ≥ 0, 1 ≤ i ≤ M (1)
ss gL
ss (L − 1)!
 
1 h
Fg(i) (h)= γ L,
ss (L − 1)! gss
  L−1  l
h X1 h
= 1 − exp − (2)
gss l! gss
l=0

where the noise at each branch is assumed to be uncorrelated

Fig. 1. System Model: the ith transmit antenna is selected.
of performance analysis of cognitive radio [6], [12]–[18]. In
this scenario, the secondary transmitter aims to maximize its
effective capacity, where the interference caused at the primary
receiver must be regulated by the predefined interference
constraint. Therefore, the ST will select one of its transmit
antennas by using proper transmit power to maximize its
effective capacity. in order to see how the different transmit an-
tenna selection schemes influence the capacity, we study three
scenario: minimum interference selection (Sel (1)), maximum
secondary composite channel gain selection (Sel (2)), and the
maximum channel ratio selection (Sel (3)).
In addition, we consider underlay paradigm which is one
of the cognitive radio paradigms. In underlay paradigm, the
secondary and primary users could transmit simultaneously,
if the interference caused by the secondary users to the
licensed users is below a predefined threshold [19]. This
paradigm assumes that the secondary user has the channel
state information (CSI) of the interference channel from the
secondary transmitter to the primary receiver, which can be
gathered by the spectrum manager, primary receiver or a third-
party device and then fed back to the secondary transmitter
[13]. Of course, this CSI can be assumed to be perfect for
simplicity. However, in practice it is always imperfect due
to, such as, fading, Doppler, limited feedback channel, and
measurement error. The interference can be regulated by the
interference temperature.
(il)
Let gss , which is assumed to be i.i.d ∀l = 1, · · · , L, be
the channel power gain from the ith antenna element of the
(i)
ST to the lth antenna element of the SR with mean g ss , gss
denote the composite channel power gain from the ith antenna
(i)
element of the ST to the SR, and gps represents the channel
power gain from the ith antenna element of the ST to the
PR. In our analysis, we assume that the wireless channels
experience block Rayleigh fading assumed to be ergodic and
stationary. Thus, the channel state does not change during
each block, and the channel states are uncorrelated between
blocks [20]. Let T denote the time duration of a block. In
consequence, in a Rayleigh fading the channel power gain
(i)
gps follows exponential distribution with mean value g ps .
(i)
The compound channel power gain gss is characterized by
a Chi-square distribution with 2L degrees of freedom, and the
(i)
additive white Gaussian noise (AWGN). The p.d.f. of gps is
represented by
1 −g/gps
fg(i) (g) = e , g ≥ 0, 1 ≤ i ≤ M. (3)
ps gps
The MRC technique requires perfect knowledge of the
branch amplitudes and phases, and provides the optimal di-
versity. Hence, it offers the maximal capacity improvement
relative to other linear combining techniques [22]. The trans-
mit power of the secondary transmitter is regulated by peak
interference power (Qpk ) constraint or the average interference
power (Qav ) constraint at the primary receiver. We investigate
the following three schemes of transmit antenna selection of
the secondary transmitter:
• Minimum interference selection (Sel (1)): the antenna
element at the secondary transmitter is selected which
causes minimum interference to the primary user. In
other words, the interference channel (from the secondary
transmitter to the primary receiver, ST-PR) power gain
is the minimum. In this case the ST uses only the
interference channel information for making selection.
• Maximum secondary composite channel gain selection
(Sel (2)): the antenna element is selected to maximize the
data rate of the secondary user. That is, the antenna ele-
ment which has maximum channel gain to the secondary
receiver is selected (from the secondary transmitter to the
secondary receiver, ST-SR).
• Maximum channel ratio selection (Sel (3)): the antenna
element is selected at the secondary transmitter which has
(i) (i)
the maximal channel gain ratio of gss /gps .
For deriving the expressions of the effective capacity of the
secondary user, we first investigate the related p.d.f. and c.d.f.
expressions in the following.
A. Minimum interference selection
The transmit antenna which minimizes the interference to
the primary is selected, i.e., the channel from the selected
antenna to the primary user has the minimal channel gain
among all antennas.
n o
(i)
k = arg∀i min gps , 1 ≤ i ≤ M. (4)
(k)
Using order theorem [23], the c.d.f. and the p.d.f. of gps are
given by
  M
g
Fg(k) (g) = 1 − exp − (5)
ps gps
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014 27

and Using order theorem [23], the c.d.f. of the maximum channel
(k)
gain, gss , can be represented as
  M−1

g exp −g/gps Fg(k) (x)
fg(k) (g)= M exp − ss
ps gps g ps M
Y M 
Y  
  M 1 x
M g = Fhl (x) = γ L, (12)
= exp − (6) (L − 1)! gss
l=1 l=1
g ps gps " #
YM
− gx
X 1  x kl
L−1
= 1 − e ss
kl ! g ss
Now we derive the probability distribution of the ratio, l=1 kl =0
" #nl
(k) (k) (k) (k) (k)
gss /gps . Let z = gss /gps and w = gps . The Jacobian X Y M
− gx
X 1  x kl
L−1
nl
is given by = (−1) e ss
kl ! g ss
n∈θM l=1 kl =0
  M
" L−1  kl #nl
w z  X Y x X 1 x
Jz,w = det  =w (7) = (−1)nl e− gss nl
0 1 kl ! g ss
n∈θM l=1 kl =0
X − x PM n Y M X Lnl −1  x kl nl
L−1
l=1 l nl
The joint p.d.f. of z and w. = e gss
(−1)
(kl !)nl g ss
n∈θM l=1 kl =0
X L−1
X L−1
X − gx
PM
nl
PM
kl nl
fz,w = fg(k) ,g(k) (h = zw, g = w) w (8) = ··· e ss l=1
x l=1
ss ps
  M n∈θM k1 =0 kM =0
M w (zw)L−1 e−zw/gss M  nl
= exp − w Y −1 Lnl −1
g ps gps gL
ss (L − 1)! ×
l=1
kl ! gkssl nl

The p.d.f. of z. X L−1

X L−1
X
= ··· Kn,k e−xBn xAn,k
n∈θM k1 =0 kM =0
Z ∞
fz (z) = fz,w (z, w) dw (9) where θM is defined as the binary number set with M elements
0
Z   YM  n
M z L−1 ∞
Mw zw −1 l Lnl −1
= wL exp − − dw Kn,k = (13)
gps g L (L − 1)! g g l=1
kl ! g kssl nl
ss 0 ps ss
M
M Lρz L−1 1 X
= Bn = nl (14)
(z + M ρ)L+1 g ss
l=1
M
X
where ρ = g ss /gps . The last step is with the help of [24, eqn. An,k = kl nl (15)
3.351-1] l=1

The c.d.f. of z. The above proof follows the Lemma in [25] and [26]. The
p.d.f. can be obtained as following
Z z Z z d
M LρxL−1 fg(k) (x) =
Fh (x)
Fz (z)= fz (x) dx = dx ss dx max
0 0 (x + M ρ)L+1
 L X L−1
X L−1
X
z = ··· Kn,k e−xBn xAn,k −1 [An,k − xBn ]
= (10)
z + Mρ n∈θM k1 =0 kM =0
(16)
(k) (k) (k)
Let z = gss /gps and w = gps . The Jacobian is given by
 
w z 
B. Maximum secondary composite channel gain selection Jz,w = det  =w (17)
0 1
The transmit antenna which has maximal channel gain of The joint p.d.f. of z and w can be obtained by
ST-SR is selected, i.e., the channel from the selected antenna
fz,w = fg(k) ,g(k) (h = zw, g = w) w
to the secondary receiver has the maximal composite channel ss ps
L−1 L−1
gain among all antennas. 1 −w/gps X X X
= e ··· Kn,k e−zwBn
g ps
n o n∈θM k1 =0 kM =0
(i)
k = arg∀i max gss , 1 ≤ i ≤ M. (11) ×(zw)An,k −1 [An,k − zwBn ] w (18)
28 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

The p.d.f. of z can be expressed as The c.d.f. can be obtained as

Z ∞ Z z Z z  L
LρxL−1 z
fz (z)= fz,w (z, w) dw Fzi (z) = fzi (x)dx = L+1
dv =
0 0 0 (x + ρ) z+ρ
L−1 L−1 (23)
1 X X X
= ··· Kn,k z An,k −1 Then according to the order theorem the p.d.f. of zk can be
g ps obtained as
n∈θ k =0 kM =0
Z ∞ M 1
z ML−1
× e−w/gps −zwBn wAn,k [An,k − zwBn ] dw fzk (z) = M (Fzi (z))M−1 fzi (z) = M Lρ (24)
ML+1
0 (z + ρ)
X L−1
X L−1
X
= ··· Kn,k g An,k III. E FFECTIVE C APACITY U NDER PIP C ONSTRAINT
n∈θM k1 =0 kM =0
In this section we study the effective capacity of the
An,k Γ(An,k + 1)z An,k −1 secondary user under peak interference power (PIP) constraint
×
A +1
1 + Bn g ps z n,k for different transmit antenna selection schemes.
! First, we again briefly review the concept of the effective
Bn gps Γ(An,k + 2)z An,k capacity and derive a closed-form expression for the EC. The
−
A +2 (19)
1 + Bn gps z n,k probability of the queue length q(x) of a stationary ergodic
arrival and service process exceeding a certain threshold Tq
where in the last step we have used [24, eqn. 3.381-4]. decays exponentially as a function of Tq [3]. The delay QoS
It is obvious that if Bn = 0, i.e. n ∈ 0M we also have exponent is defined as
An,k = 0, then fz (z) = 0. In the following, without loss of
generality we assume that n ∈ θM represents n ∈ θM and log(Pr {q(∞) > Tq })
θ = − lim . (25)
n 6= 0M , where 0M is binary set with M zero elements. Tq →∞ Tq
It is worth noting that θ → 0 indicates that the system has
C. Maximum channel ratio selection no delay constraint, while θ → ∞ implies a stringent delay
g(i)
constraint. The effective capacity is defined in [1, Eqn. (12)]
The transmit antenna which maximizes zi , ss(i) is selected, as follows,
gps
i.e., ( ) 1 h  Pt i
(i)
gss EC(θ) = − lim log E e−θ i=0 R[i] , t ≥ 0 (26)
t→∞ θt
k = arg∀i max (i)
, 1 ≤ i ≤ M. (20)
gps where {R[i], i = 1, 2, ...} denotes a discrete-time service pro-
The p.d.f. of zi can be obtained as follows. cess, which is assumed to be ergodic and stationary. For a
block fading channel, the EC can be reduced to [3],
LρxL−1 1 h  i
fzi (x) = L+1
(21) EC(θ) = − log E e−θR[i] . (27)
(x + ρ) θ
where ρ = g ss /gps . The maximum achievable instantaneous service rate R[i] of
(i)
gss (i) block i can be expressed as
Proof: Let zi = (i) and w = gps . Then the Jacobian, !
gps (k) (k) (k)
J(zi , w), is given by ps (θ, gss , gss )gss
R[i] = T B log 1 +
  N1 B
w z 
Jzi ,w (z, w) = det  =w (22)
0 1 where the superscript k denotes the kth transmit antenna
(k) (k)
branch is selected, ps (θ, gss , gps ) is the transmit power of the
fzi ,w (z, w)= fg(i) (h = zw) fg(i) (g = w) w secondary transmitter, T denotes the block length duration, N1
ss ps
L−1 −zw/gss
denotes the AWGN power density at the secondary receiver,
(zw) e 1 −w/gps and B is the channel bandwidth. In the following we derive
= e w
gL
ss (L − 1)! gps the expressions of the effective capacity of the secondary user
Then the marginal distribution of zi can be obtained by in different scenarios.
integrating fz,w (z, w) over w. Notations:
R1 Let B(x, y) denote the Beta function given
Z ∞ by 0 tx−1 (1 − t)y−1 dt, and 2 F1 (a, b; c; z) be the Gauss’s
fzi (z)= fzi ,w (z, w)dw hypergeometric function [27]. For convenience, let α = θT B.
0
Z ∞
(zw)L−1 e−zw/h 1 −w/g A. Minimum interference selection
= L
e wdw
0 h (L − 1)! g In this case, the p.d.f. of the channel power gains, zk , is
Lρz L−1 given by (9). The resultant effective capacity of the secondary
= L+1 user is given by the following theorem.
(z + ρ) Theorem 1: The effective capacity of the cognitive-shared
channel of Figure 1 under peak interference power constraint
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014 29

B. Maximum secondary composite channel gain selection

1.4
M=1, L=1
M=2, L=1 In this case, the p.d.f. of zk is given by (19). The resultant
M=4, L=1
1.2
M=1, L=2 effective capacity of the secondary user is given by the
Effective capacity of the SU (nats/s/Hz)

M=1, L=4
M=2, L=2
following theorem.
1
Theorem 2: The effective capacity of the cognitive-shared
0.8 channel of Figure 1 under peak interference power constraint
over Rayleigh fading with maximum secondary composite
0.6 channel gain selection is given by
( )
0.4
1 X L−1 X L−1
X Kn,k
EC(θ) = − log ··· A
[C1 + C2 ]
0.2
θ Bn n,k n∈θM k1 =0 kM =0
(29)
0 −2 −1 0 1 2
10 10 10 10 10
θ where C1 for Bn 6= 0 and C2 for Bn = 0 are given by

Fig. 2. Effective capacity of the SU versus QoS exponent under PIP constraint C1 = An,k Γ (An,k + 1) B(An,k , α + 1)
and minimum interference selection, where the mean channel power gain ratio  
ρ = 1, N1 B = 1, and Qpk = −5dB. Qpk
× 2 F1 α, An,k ; An,k + α + 1; 1 −
N1 Bg ps Bn
(30)
over Rayleigh fading with minimum interference selection
scheme is given by and

1
EC(θ)= − log [L B(L, 1 + α) C2 = −Γ (An,k + 2) B(An,k + 1, α + 1)
θ  
Qpk M ρ  
× 2 F1 α, L; L + 1 + α; 1 − (28) Qpk
N1 B × 2 F1 α, An,k + 1; An,k + α + 2; 1 −
N1 Bg ps Bn
Proof: (31)
    
1 zQ
−α log 1+ N pk Proof: see Appendix A for details.
EC(θ)= − log Ezk e 1 B

θ
"Z  −α #
∞
1 zQpk M Lρz L−1 1.4
= − log 1+ dz M=1, L=1
θ 0 N1 B (z + M ρ)L+1 M=2, L=1
M=4, L=1
1.2
M=1, L=2
1
Effective capacity of the SU (nats/s/Hz)

M=1, L=4
= − log [L B(L, 1 + α) M=2, L=2
θ  
1

Qpk M ρ
× 2 F1 a, L; L + 1 + α; 1 − 0.8
N1 B
0.6
where in the last step, we have used Eqn. (3.197-1) in [24].
0.4
In Fig. 2 we show the effective capacity of the secondary
user when the SU chooses the transmit antenna element having 0.2

the minimum channel gain from the secondary transmitter

0 −2
to the primary receiver over different transmit and receive 10
−1
10 10
0 1
10 10
2

degrees of diversity. Without loss of generality, N1 B = 1.
First, it is obvious that the TAS/MRC improves the achievable
effective capacity over a large range of the QoS exponent.
Second, the performance of using multiple receive antennas
surpasses the one using multiple transmit antennas, e.g. the
performance of M = 1, L = 2 is better than the one of
M = 2, L = 1, and M = 1, L = 4 is better than the one
of M = 4, L = 1. Third, in this case using M = 2 and
L = 2 can not achieve the full diversity, because we only
utilize partial channel information, i.e. only the interference
channel information. Forth, at high θ regime, M = 1, L = 2
is slightly better than M = 4, L = 1. This is because the
MRC is the optimal linear combing technique [22], and the
benefit of using TAS is not able to compensate the capacity
loss caused by the channel deep fading.
θ
Fig. 3. Effective capacity of the SU versus QoS exponent under PIP constraint
and maximum MRC channel gain, where the mean channel power gain ratio
ρ = 1, N1 B = 1, and Qpk = −5dB.
In Fig. 3 we show the effective capacity of the secondary
user when the SU chooses the transmit antenna element having
the maximum ST-SR channel gain over different transmit
and receive degrees of diversity. First, it is obvious that the
TAS/MRC improves the achievable effective capacity over a
large range of the QoS exponent. Second, in this case using
more receiving antennas always (over the simulated range of θ)
is superior using more transmit antennas given the same value
of M × L. This is different from the case of Sel (1) scheme at
the high θ regime. Third, the larger value of M × L, the better
30 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

is limited by the average interference power, Qav . We may

1.4
M=1, L=1 formulate the following optimization problem.
M=2, L=1
1.2 M=4, L=1  (k) (k) (k)
!
M=1, L=2 gss ps (θ,gps ,gss )
1 −α log 1+
Effective capacity of the SU (nats/s/Hz)

M=1, L=4 N1 B

1
M=2, L=2 maximize − E e  (33)
θ
0.8
subject to
h i
(k) (k) (k)
0.6 E gps ps (θ, gps , gss ) ≤ Qav (34)
(k) (k)
0.4 ps (θ, gps , gss ) ≥0 (35)

0.2 This optimization problem can be equivalently written as

 !−α 
(k) (k) (k)
0 −2 gss ps (θ, gps , gss )
10
−1
10 10
θ
0
10
1
10
2
minimize E  1 +  (36)
N1 B

Fig. 4. Effective capacity of the SU versus QoS exponent under PIP constraint subject to (34), (35)
and maximum channel radio selection, where the mean channel power gain
ratio ρ = 1, N1 B = 1, and Qpk = −5dB. The power allocation can be obtained through using La-
grangian method as
 +
1
performance the SU has. However, the performance of the β 1+α 1
, gss ) = N1 B 
(k) (k)
ps (θ, gps − (k)  (37)
scenario of M = 1, L = 4 surpasses the one of M = 2, L = 2. α 1
(k) 1+α (k) 1+α gss
This is because the former case can use the full diversity while gss gps
the later does not according to the antenna selection method. where α = θT B, β = λNα1 B , and λ is the Lagrangian
multiplier associated to (34).

C. Maximum Channel Ratio Selection

A. Minimum Interference Selection
In this case the channel information of ST-PR and ST-SR are In this case, the p.d.f. of the channel power gains, zk =
used to make the selection. Thus the secondary user utilizes (k) (k)
gss /gps , is given by (9). Although this case is not practical
the full diversity of multiple antennas which can be seen from that the SU does not utilize all the channel state information
the following theorem and the simulation. (CSI) for making transmitter antenna selection, we here just
Theorem 3: The effective capacity of the cognitive-shared study the impact of using CSI of ST-PR only on the effective
channel of Figure 1 with exploiting maximum channel ratio capacity and the power adaptation uses all the CSI. The
selection scheme under peak interference power constraint resultant effective capacity of the secondary user is given by
over Rayleigh fading is given by the following theorem.
 Theorem 4: The effective capacity of the cognitive-shared
1
EC(θ) = − log M L B(M L, α + 1) channel of Figure 1 with exploiting minimum interference
θ
  channel ratio selection scheme under average interference
Qpk ρ
× 2 F1 α, M L; α + M L + 1; 1 − power constraint over Rayleigh fading is given by
N1 B
(32) EC(θ)
  
1 1 1
Proof: see Appendix B for details. = − log 2 F1 L + 1, L; L + 1; −
θ (M ρβ)L M ρβ
It is proved that the EC depends on the diversity order M ×
L, the QoS exponent θ, the product of the channel bandwidth M Lρβ(1 + α)
+ (38)
and the channel coherence time (duration of fading block), the 1 + 2α 
interference power constraint, the noise power and the mean  
α α
channel power gain ratio. The simulated results are plotted in × 2 F1 L + 1, 1 + ;2+ ; −M ρβ 
1+α 1+α
Fig. 4.
We have investigated the performance of the secondary user Proof: see Appendix C for details.
in terms of the effective capacity under the peak interference The Lagrangian multiplier satisfies the following condition
power constraint. Next, we study the performance under 
average interference power constraint for the different antenna M LρN1 Bβ 2 2 + 2α
Qav =
selection schemes. 2 1 + 2α
 
α α
× 2 F1 L + 1, 1 + ;2 + ; −M ρβ
1+α 1+α
IV. E FFECTIVE C APACITY UNDER AIP C ONSTRAINT 
In this section we study the effective capacity of the − 2 F1 (L + 1, 2; 3; −M ρβ)
secondary user where the interference to the primary user (39)
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014 31

Proof: AIP-based power allocations are more flexible to utilize the

h i opportunities (channel fading states) for transmission, where
(k) (k) (k)
Qav =E gps ps (θ, gps , gss ) AIP-based power control depends on the both channels, ST-PR
Z ∞ " 1 #+ and ST-SR, while PIP-based one depends only ST-PR.
β 1+α 1
= N1 B α − fzk (z) dz
0 z 1+α z
Z ∞ " 1 # B. Maximum Secondary Composite Channel Gain Selection
β 1+α 1 M Lρz L−1 In this case, the p.d.f. of the channel power gains, zk , is
= N1 B α − dz
1/β z 1+α z (z + M ρ)L+1 given by (19). The same as in previous subsection that the
" Z ∞
1
α
z L− 1+α −1 SU uses the CSI of ST-SR for making transmitter antenna
=N1 BM Lρ β 1+α L+1
dz selection, however, the power adaptation uses all the CSI. The
1/β (z + M ρ) resultant effective capacity of the secondary user is given by
Z ∞ #
z L−2 the following theorem.
− L+1
dz Theorem 5: The effective capacity of the cognitive-shared
1/β (z + M ρ)
channel of Figure 1 with exploiting maximum secondary
we then have the result by employing [24, Eqn. (3.194-2)] to composite channel gain selection scheme under average in-
the last step. terference power constraint over Rayleigh fading is given by
( )
1 X L−1 X L−1
X An,k
1.4 EC(θ) = − log ··· Kn,k g ps [I1 + I2 ]
M=1, L=1 θ
M=2, L=1 n∈θM k1 =0 kM =0
Effective capacity of the SU (nats/s/Hz)

1.2 M=4, L=1

M=1, L=2 Proof: see Appendix D for details.
M=1, L=4
1
M=2, L=2 The Lagrangian multiplier satisfies the following condition
0.8 X L−1
X L−1
X A
Qav = N1 B ··· Kn,k g psn,k [Q1 − Q2 ] (40)
0.6 n∈θM k1 =0 kM =0

0.4 where Q1 is given by (43) and (45), and Q2 is given by (44)

and (46).
0.2
Proof: see Appendix E for details.
0 −2 −1 0 1 2
10 10 10 10 10
θ
1.4
M=1, L=1
M=2, L=1
Effective capacity of the SU (nats/s/Hz)

Fig. 5. Effective capacity of the SU versus QoS exponent under AIP constraint 1.2 M=4, L=1
and minimum interference selection, where the mean channel power gain ratio M=1, L=2
ρ = 1, N1 B = 1, and Qav = −5dB. 1
M=1, L=4
M=2, L=2

In Fig. 5 we show the effective capacity of the secondary 0.8

user when the SU chooses the transmit antenna element having

0.6
the minimum channel gain from the secondary transmitter to
the primary receiver over different transmit and receive degrees 0.4
of diversity. First, it is obvious that the TAS/MRC improves
0.2
the achievable effective capacity over a large range of the QoS
exponent. Second, the performance of using multiple receive 0 −2 −1 0 1 2
antennas slightly surpasses the one using multiple transmit 10 10 10
θ
10 10

antennas at the range of small values of θ, e.g. the performance

of M = 1, L = 2 is better than the one of M = 2, L = 1,
and M = 1, L = 4 is better than the one of M = 4, L = 1.
However, the gap rises when θ increases. This phenomenon is
different from the one under PIP constraint. Third, in this case
using M = 2 and L = 2 can not achieve the full diversity,
because we only utilize partial channel information, i.e. only
the interference channel information. Forth, at high θ regime,
M = 1, L = 2 is slightly better than M = 4, L = 1. This
can be explained as that the benefit of using TAS is not able
to compensate the capacity loss caused by the channel deep
fading.
Comparing the results in Fig. 3 and Fig. 5, it is clear that
AIP constraint is more favorable than PIP constraint. This
result also confirms the conclusions drawn in [28] that the
Fig. 6. Effective capacity of the SU versus QoS exponent under AIP constraint
and maximum MRC channel selection, where the mean channel power gain
ratio ρ = 1, N1 B = 1, and Qav = −5dB.
We depict the results in Fig. 6. Besides the improve-
ment through using multiple antennas, the performance of
the scenario of M1 , L = 2 is not superior to the one of
M = 4, L = 1 as in the cases using Sel (1) scheme. Moreover,
the performance of the cases with larger value of M × L
surpasses the ones with lower value of M × L.
C. Maximum Channel Ratio Selection
This scheme uses the all the channel state information for
transmit antenna selection and for transmit power adaptation.
32 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

Theorem 6: The effective capacity, EC(θ), of the cognitive-

1.4
shared channel of Figure 1 with exploiting maximum channel M=1, L=1
M=2, L=1
ratio selection scheme under peak interference power con-

Effective capacity of the SU (nats/s/Hz)

1.2 M=4, L=1
straint over Rayleigh fading is given by M=1, L=2
M=1, L=4
"   
1
M=2, L=2
ML
1 1 1
− log F
2 1 M L + 1, M L; 1 + M L; − 0.8
θ ρβ ρβ
 # 0.6
M Lρβ α α
+ α 2 F1 M L + 1, 1 + ;2 + ; −ρβ
1 + 1+α 1+α 1+α 0.4

(41) 0.2

Proof: see Appendix F for details. 0 −2 −1 0 1 2

It is obviously that the EC depends on the diversity order 10 10 10
θ
10 10

M × L, the QoS exponent θ, the product of the channel

bandwidth and the channel coherence time (duration of fading Fig. 7. Effective capacity of the SU versus QoS exponent under AIP constraint
block), the interference power constraint, the noise power and and maximum channel radio selection, where the mean channel power gain
ratio ρ = 1, and Qav = −5dB.
the mean channel power gain ratio. Fig. 7 depicts the results.
The Lagrangian multiplier satisfies the following condition

M LρN1 Bβ 2 2 + 2α 0.9
Qav = M=1, L=1
2 1 + 2α 0.8 Sel (1), M=2, L=1
 

Effective capacity of the SU (nats/s/Hz)

Sel (2), M=2, L=1
α α 0.7 Sel (3), M=2, L=1
×2 F1 M L + 1, 1 + ;2 + ; −ρβ Sel (1), M=1, L=2
1+α 1+α Sel (2), M=1, L=2
 0.6
Sel (3), M=1, L=2

− 2 F1 (M L + 1, 2; 3; −ρβ) (42) 0.5

0.4
Proof:
0.3
h i
(k) (k) (k)
Qav = E gps ps (θ, gps , gss ) 0.2

Z ∞ " 1 #+ 0.1
β 1+α 1
= N1 B α − fzk (z) dz 0 −2
0 z 1+α z 10 10
−1 0
10 10
1
10
2

" 1 # θ
Z ∞
β 1+α 1 M Lρz ML−1
= N1 B α − dz Fig. 8. Effective capacity comparison under PIP constraint and different TAS
1/β z 1+α z (z + ρ)ML+1 schemes, where, M × L = 2, the mean channel power gain ratio ρ = 1, and
" Z ∞ ML− 1+α α Qpk = −5dB.
−1
1 z
= N1 BM Lρ β 1+α
ML+1
dz
1/β (z + ρ)
Z ∞ #
z ML−2 1.4
− ML+1
dz
1/β (z + ρ)
Effective capacity of the SU (nats/s/Hz)

1.2

we then have the result by employing [24, Eqn. (3.194-2)] to 1

the last step.
0.8 M=1, L=1
Sel (1), M=4, L=1
V. S IMULATION R ESULTS 0.6
Sel (2), M=4, L=1
Sel (3), M=4, L=1
This section presents simulation results for the EC by Sel (1), M=1, L=4
0.4 Sel (2), M=1, L=4
comparison. The same as in previous simulations, we assume Sel (3), M=1, L=4
Rayleigh block fading channels, T B = 1, and additive white 0.2 Sel (1), M=2, L=2
Sel (2), M=2, L=2
Gaussian noise power N1 B = 1. Sel (3), M=2, L=2
0 −2
Fig. 8 and Fig. 9 compare the EC of three selection schemes 10 10
−1 0
10 10
1
10
2

versus θ under peak interference power (PIP) constraint with
diversity order M × L = 2 and M × L = 4, respectively.
From Fig. 8 we can see that Sel (1) scheme is superior to Sel
(2) at the low range of delay component θ which represents
delay-insensitive regime. However, along with increasing the
value of θ, Sel (2) becomes better than Sel (1). Moreover,
The effective capacity is capped by using full diversity order.
Fig. 9 shows the similar phenomenon as in Fig. 8. We may
θ
Fig. 9. Effective capacity comparison under PIP constraint and different TAS
schemes, where, M × L = 4, the mean channel power gain ratio ρ = 1, and
Qpk = −5dB.
conclude that at the stringent case, i.e. lager values of θ, with
the same diversity order the receiving diversity is superior to
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014 33

1 1.4
M=1, L=1
0.9 Sel (1), M=2, L=1
Effective capacity of the SU (nats/s/Hz)

Effective capacity of the SU (nats/s/Hz)

Sel (2), M=2, L=1 1.2
0.8 Sel (3), M=2, L=1
Sel (1), M=1, L=2
0.7 Sel (2), M=1, L=2 1
Sel (3), M=1, L=2
0.6
0.8 M=1, L=1
0.5 Sel (1), M=4, L=1
Sel (2), M=4, L=1
0.6
0.4 Sel (3), M=4, L=1
Sel (1), M=1, L=4
0.3 0.4 Sel (2), M=1, L=4
0.2 Sel (3), M=1, L=4
0.2 Sel (1), M=2, L=2
0.1 Sel (2), M=2, L=2
Sel (3), M=2, L=2
0 −2 −1 0 1 2
0 −2 −1 0 1 2
10 10 10 10 10 10 10 10 10 10
θ θ

Fig. 10. Effective capacity comparison under AIP constraint and different Fig. 11. Effective capacity comparison under AIP constraint and different
TAS schemes, where, M × L = 2, the mean channel power gain ratio ρ = 1, TAS schemes, where, M × L = 4, the mean channel power gain ratio ρ = 1,
and Qav = −5dB. and Qav = −5dB.

the transmit diversity.

Fig. 10 and Fig. 11 show the results under average inter-
transmit antenna selection and maximal ratio combining under
ference power constraint with diversity order M × L = 2
peak or average interference power constraint. The results
and M × L = 4, respectively. The significant difference
are compared for using different transmit antenna selection
in terms of effective capacity from the cases under peak
strategies, minimum interference selection, maximum sec-
interference power constraint is the improvement by using
ondary channel gain selection, and maximum channel gain
multiple antennas at the high θ regime.
ratio selection. The multiple antenna techniques improve the
Taking look at these results, we can see that in high QoS
communication quality significantly.
exponent θ and under AIP constraint the gap in capacity
achievement between different schemes remains slight differ- However, we have more work to do in the future. In the
ence in opposite to the case of PIP constraint. This is because previous study, we considered only the perfect channel state
of the behavior of hypergeometric function, 2 F1 (a, b; c; x). information cases. Thus the influence on the effective capacity
Also, when the QoS exponent goes to infinity, the EC is ap- of being provided imperfect channel state information, e.g.
proaching to the outage capacity. However, we leave the work delayed channel information, channel information with mea-
of deriving the scaling laws as our future work. Additionally, surement errors, or both, are also important for understanding
in high QoS exponent and under PIP constraint, the EC of cognitive radio systems. Moreover, we need to investigate the
MIMO, MISO, SIMO systems is reduced but still different multiple users scenarios, for instance, multiple primary users,
from the one of SISO system. This is another phenomenon multiple secondary users, or both. Furthermore, some new
that AIP constraint is more favorable than PIP constraint. selection schemes are needed to be developed and studied if
The secondary user could use some opportunities that the ST- we only have the channel statistics information without the
PR channel experiences deep fading. Though, the multiples instantaneous channel state information, for instance, mean
antennas can benefit the secondary transmission, there is a values. We will leave these as our future work.
PIP constraint such that the secondary user only can transmit
with certain amount power. Moreover, if at the same time, its
own channel is experiencing severe fading, the instantaneous
rate will be low. In contrast, AIP-based power control scheme
provides the secondary user transmit opportunities by taking
into consideration both the ST-PR and ST-SR channel states.
In sum, the EC increases as the number of TAS/MRC A PPENDIX
antennas increases. And for given interference power con-
straint at the primary receiver, the more stringent the QoS
requirement, e.g., θ → ∞, the less the effective capacity is.
This is because of the delay limitation. Different selection For simplicity, we use the following notation,
schemes have different amount of advantages at various range
of the delay exponent θ.

VI. C ONCLUSION AND D ISCUSSION −→

X X L−1
X L−1
X
In this paper, we have studied and derived an expression = ···
for the effective capacity of a cognitive-shared channel with n∈θM k1 =0 kM =0
34 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

A. Proof of Theorem 2 D. Proof of Theorem 5

EC(θ)
"Z 
EC(θ) 1 ∞ h 1 1
i+ −α X
−→
"Z  −α = − log 1+ β 1+α z 1+α −1 Kn,k g An,k

1 ∞
Qpk 1 X−→ K θ 0
ps
n,k
= − log 1+ z 
θ N1 B gps An,k +1
0 Bn
   
An,k Γ(An,k + 1)z An,k −1 Bn g ps Γ(An,k + 2)z An,k 
×
A +1 −
A +2 dz 
 An,k Γ(An,k + 1)z An,k −1 Γ(An,k + 2)z An,k   
×  1 + Bn g ps z n,k 1 + Bn g ps z n,k 
An,k +1 −  An,k +2  dz  | {z }
1 1
z + Bn g z + Bn g  
G
ps ps
" Z ∞ −α 

1 (Qpk /N1 B)
−α X−→ Kn,k N1 B 
X −→ Z 1/β
= − log + z 1 
A +1 = − log Kn,k g ps An,k  G dz
θ g ps Bn n,k 0 Qpk θ 
  0
   
 | {z }
I1

 An,k Γ(An,k + 1)z An,k −1 Γ(An,k + 2)z An,k   

×  An,k +1 −  An,k +2  dz  Z ∞ 
z + Bn1g z + Bn1g
α
− 1+α α
− 1+α 
+ β z G dz 
ps ps
1/β 
| {z } 
I2
then we have the result in (29) by implying [24, Eqn. (3.197-
1)] to the above integration. where for An,k 6= 0
 
Γ(An,k + 1) g ps Bn
I1 = 2 F1 A n,k + 1, A n,k ; A n,k + 2; −
β An,k β
B. Proof of Theorem 3 Bn Γ(An,k + 1)g ps
−
β An,k +1
 
g ps Bn
     × 2 F1 An,k + 2, An,k + 1; An,k + 2; −
1 Q
−α log 1+ N pk z β
− log Ezk e 1B
Γ(An,k + 1)β (1 + α)
θ I2 =
A +1
"Z  −α #
∞ g ps Bn n,k (1 + 2α)
1 Qpk z ML−1   
= − log 1+ z M Lρ dz α α β
θ 0 N1 B (z + ρ)ML+1 × An,k 2 F1 An,k + 1, 1 + ;2 + ;−
1+α 1+α g ps Bn
1  
= − log [M L B(M L, α + 1) α α β
θ − (An,k + 1) 2 F1 An,k + 2, 1 + ;2 + ;−
  1+α 1+α g ps Bn
Qpk ρ
× 2 F1 α, M L; α + n + 1; 1 − and for An,k = 0
N1 B  
Bn g g ps Bn
I1 = − 2 F1 2, 1; 2; −
where in the last step, we have used [24, Eqn. (3.197-1)]. β β
β (1 + α)
I2 = −
g ps Bn (1 + 2α)
 
C. Proof of Theorem 4 α α β
×2 F1 2, 1 + ;2 + ;−
1+α 1+α g ps Bn
in the above we have used [24, Eqn. (3.194-1) and (3.194-2)]
1 h  i
− log Ezk e−α log(1+ps gss /N1 B )
(k)
and Γ(1 + x) = xΓ(x).
θ "Z 
1 ∞ h 1 i+ −α
= − log 1+ β 1+α z
1
1+α −1 E. Proof of the expression of Qav given by (40)
θ 0 The proof is on the next page.

M Lρz L−1
× dz
(z + M ρ)L+1 F. Proof of Theorem 6
"Z
1/β
1 M Lρz L−1 The proof is on the next page.
= − log dz
θ 0 (z + M ρ)L+1
Z ∞ # R EFERENCES
α
− 1+α α
− 1+α M Lρz L−1
+ β z dz [1] D. Wu and R. Negi, “Effective capacity: a wireless link model for
1/β (z + M ρ)L+1 support of quality of service,” IEEE Trans. Wireless Commun., Vol. 2,
No. 4, pp. 630–643, Jul. 2003.
[2] C.-S. Chang and J. Thomas, “Effective bandwidth in high-speed digital
We then have the result (38) by using [24, Eqn. (3.194-1) and networks,” IEEE J. Sel. Areas Commun., Vol. 13, No. 6, pp. 1091–1100,
(3.194-2)] to the integrals. Aug. 1995.
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014 35

Proof of the expression of Qav given by (40).

" #+
h i Z ∞ 1
β 1+α 1
(k) (k) (k)
Qav = E gps ps (θ, gps , gss ) = N1 B α − fzk (z) dz
0 z 1+α z
Z " 1
# !
∞
β 1+α 1 X L−1
X L−1
X An,k Γ(An,k + 1)z An,k −1 Bn g ps Γ(An,k + 2)z An,k
= N1 B α − ··· Kn,k g An,k
ps
An,k +1 −
A +2 dz
1/β z 1+α z 1 + Bn g ps z 1 + Bn g ps z n,k
n∈θM k1 =0 kM =0




 Z ∞" 1 #
X L−1
X L−1
X  1 α
Bn g ps (An,k + 1)z An,k + 1+α −1
An,k An,k z An,k − 1+α −1
= N1 B ··· Kn,k g ps β 1+α Γ(An,k + 1)
A +1 −
A +2 dz

 1 1 + Bn g ps z n,k 1 + Bn g ps z n,k
n∈θM k1 =0 kM =0 
 β
| {z }
Q
 1

" Z ∞ #

An,k −2 B g (A + 1)z An,k −1 
An,k z n ps n,k
− Γ(An,k + 1)
An,k +1 −
An,k +2 dz
1/β 1 + Bn g ps z 1 + Bn g ps z 

| {z }


Q2

For An,k 6= 0, we by using [24, Eqn. (3.194-2)] have

  
β 2 Γ(An,k + 1)(1 + α) α α β
Q1 =
A +1 An,k 2 F1 An,k + 1, 1 + ;2 + ;− (43)
g ps Bn n,k (1 + 2α) 1+α 1 + α gps Bn
 
α α β
−(An,k + 1) 2 F1 An,k + 2, 1 + ;2 + ;−
1+α 1 + α gps Bn
    
β 2 Γ(An,k + 1) β β
Q2 =
A +1 An,k 2 F1 An,k + 1, 2; 3; − (An,k + 1) 2 F1 An,k + 2, 2; 3; − (44)
2 g Bn n,k gps Bn gps Bn
ps

and for An,k = 0

 
β 2 (1 + α) α α β
Q1 = − F1 2, 1 + ;2 + ;− (45)
g ps Bn 2 1+α 1 + α gps Bn
 
β2 β
Q2 = − 2 F1 2, 2; 3; − (46)
2gps Bn gps Bn
in the above we have used [24, Eqn. (3.194-2)] and Γ(1 + x) = xΓ(x).

Proof of theorem 6.
EC(θ)
"   !# "Z  #
1
(k)
ps gss
−α log 1+ N 1 ∞ h 1 1
i+ −α M Lρz ML−1
= − log Ezk e 1B
= − log 1 + β 1+α z 1+α − 1 × ML+1
dz
θ θ 0 (z + ρ)
"Z Z ∞ #
1/β
1 M Lρz ML−1 α
− 1+α α
− 1+α M Lρz ML−1
= − log dz + β z dz
θ 0 (z + ρ)ML+1 1/β (z + ρ)ML+1
 
 
1 ML α α 1
= − log    × 2 F1 M L + 1, M L + ; ML + 1 + ;− +M Lρβ 2 F1 (M L + 1, 1; 2; −ρβ) 
θ (ρβ) ML ML + α 1 + α 1 + α ρβ
1+α

We have used [24, Eqn. (3.194-1) and (3.194-2)] to the above integrals.
36 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

[3] J. Tang and X. Zhang, “Quality-of-service driven power and rate

adaptation over wireless links,” IEEE Trans. Wireless Commun., Vol. 6,
No. 8, pp. 3058–3068, Aug. 2007.
[4] ——, “Quality-of-service driven power and rate adaptation for mul-
tichannel communications over wireless links,” IEEE Trans. Wireless
Commun., Vol. 6, No. 12, pp. 4349–4360, Dec. 2007.
[5] S. Akin and M. Gursoy, “Effective capacity analysis of cognitive radio
channels for quality of service provisioning,” IEEE Trans. Wireless
Commun., Vol. 9, No. 11, pp. 3354–3364, Nov. 2010.
[6] L. Musavian and S. Aissa, “Effective capacity of delay-constrained
cognitive radio in Nakagami fading channels,” IEEE Trans. Wireless
Commun., Vol. 9, No. 3, pp. 1054–1062, Mar. 2010.
[7] S. Thoen, L. Van der Perre, B. Gyselinckx, and M. Engels, “Perfor-
mance analysis of combined transmit-SC/receive-MRC,” IEEE Trans.
Commun., Vol. 49, No. 1, pp. 5–8, Jan. 2001.
[8] Z. Chen, J. Yuan, and B. Vucetic, “Analysis of transmit antenna
selection/maximal-ratio combining in Rayleigh fading channels,” IEEE
Trans. Veh. Technol., Vol. 54, No. 4, pp. 1312–1321, Jul. 2005.
[9] Z. Chen, Z. Chi, Y. Li, and B. Vucetic, “Error performance of maximal-
ratio combining with transmit antenna selection in flat Nakagami-m
fading channels,” IEEE Trans. Wireless Commun., Vol. 8, No. 1, pp. 424–
431, Jan. 2009.
[10] C.-C. Hung, C.-T. Chiang, N.-Y. Yen, and R.-C. Wu, “Outage probability
of multiuser transmit antenna selection/maximal-ratio combining sys-
tems over arbitrary nakagami-m fading channels,” IET Commun., Vol. 4,
No. 1, pp. 63–68, Jan. 2010.
[11] V. Blagojevic and P. Ivanis, “Ergodic capacity for TAS/MRC spectrum
sharing cognitive radio,” IEEE Commun. Lett., Vol. 16, No. 3, pp. 321–
323, Mar. 2012.
[12] A. Ghasemi and E. S. Sousa, “Fundamental limits of spectrum-sharing
in fading environments,” IEEE Trans. Wireless Commun., Vol. 6, No. 2,
pp. 649–658, Feb. 2007.
[13] J. Peha, “Sharing spectrum through spectrum policy reform and cogni-
tive radio,” Proc. IEEE, Vol. 97, No. 4, pp. 708–719, Apr. 2009.
[14] A. Jovicic and P. Viswanath, “Cognitive radio: An information-theoretic
perspective,” IEEE Trans. Inf. Theory, Vol. 55, No. 9, pp. 3945–3958,
Sep. 2009.
[15] X. Kang, Y.-C. Liang, A. Nallanathan, H. Garg, and R. Zhang, “Opti-
mal power allocation for fading channels in cognitive radio networks:
Ergodic capacity and outage capacity,” IEEE Trans. Wireless Commun.,
Vol. 8, No. 2, pp. 940–950, Feb. 2009.
[16] L. Musavian and S. Aissa, “Fundamental capacity limits of cognitive
radio in fading environments with imperfect channel information,” IEEE
Trans. Commun., Vol. 57, No. 11, pp. 3472–3480, Nov. 2009.
[17] H. Suraweera, P. Smith, and M. Shafi, “Capacity limits and performance
analysis of cognitive radio with imperfect channel knowledge,” IEEE
Trans. Veh. Technol., Vol. 59, No. 4, pp. 1811–1822, May 2010.
[18] R. Zhang, Y. chang Liang, and S. Cui, “Dynamic resource allocation in
cognitive radio networks,” IEEE Signal Process. Mag., Vol. 27, No. 3,
pp. 102–114, May 2010.
[19] A. Goldsmith, S. A. Jafar, I. Maric, and S. Srinivasa, “Breaking spectrum
gridlock with cognitive radios: An information theoretic perspective,”
Proc. IEEE, Vol. 97, No. 5, pp. 894–914, May 2009.
[20] L. Ozarow, S. Shamai, and A. Wyner, “Information theoretic consider-
ations for cellular mobile radio,” IEEE Trans. Veh. Technol., Vol. 43,
No. 2, pp. 359–378, May 1994.
[21] J. Proakis and M. Salehi, Digital Communications, 5th ed. McGraw-
Hill, 2008.
[22] D. Brennan, “Linear diversity combining techniques,” Proc. IEEE,
Vol. 91, No. 2, pp. 331–356, Feb. 2003.
[23] H. A. David and H. N. Nagaraja, Order Statistics. Hoboken, New
Jersey: John Wiley & Sons, Inc., 2003.
[24] I. Gradshteyn and I. Ryzhik, Table of Integrals, Series, and Products,
Seventh Edition, 7th ed. Academic Press, 2007.
[25] F. Yilmaz, A. Yilmaz, M.-S. Alouini, and O. Kucur, “Transmit antenna
selection based on shadowing side information,” in Proc. IEEE VTC’11
Spring, pp. 1–5, May 2011.
[26] A. Yilmaz, F. Yilmaz, M. Alouini, and O. Kucur, “On the performance of
transmit antenna selection based on shadowing side information,” IEEE
Trans. Veh. Technol., Vol. 62, No. 1, pp. 454–460, Jan. 2013.
[27] M. Abramowitz and I. A. Stegun, Eds., Handbook of Mathematical
Functions, With Formulas, Graphs, and Mathematical Tables. DOVER,
1964, 10th Printing with corrections, Dec. 1972.
[28] R. Zhang, “On peak versus average interference power constraints for
protecting primary users in cognitive radio networks,” IEEE Trans.
Wireless Commun., Vol. 8, No. 4, pp. 2112–2120, Apr. 2009.
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014
Interference Mitigation Using Optimal Successive
Group Decoding for Interference Channels
Omar Abu-Ella and Mohammed Elmusrati
Abstract—In this study we aim to assess the optimal successive
group decoder (OSGD) considering different scenarios and in a
wide range of testing metrics. We investigate the OSGD in the
K−user interference channel (IC) and evaluate its capability
to mitigate interference. We inspect OSGD performance in
terms of its (ergodic and effective) capacity. Also, we evaluate
its minimum required energy per bit, bit error rate (BER),
and outage probability, under different quality of service (QoS)
constraints. This study considers both the spatially correlated
and uncorrelated (Rayleigh and Rician) flat fading channels.
In addition, it explores performance of the OSGD in different
SNR and SIR environments, where we consider both the power-
limited and bandwidth-limited regimes with various cross-talk
values, taking in account several transmit-receive multiple-input
multiple-output (MIMO) antenna configurations. The obtained
numerical results in this work, show that OSGD technique ex-
hibits very efficient performance to cope with interference in the
investigated scenarios, proving by that its competency comparing
to the most developed interference cancellation approaches.1 This
makes OSGD a favorable technique for interference reduction;
especially, if we bear in mind that OSGD is formed on receive-
side processing only. This confirmed also if we contemplate
the reduction of its arithmetic complexity, as a result of its
innate complexity controlling characteristic, comparing to the
large computational complexity of the other optimal interference
cancellation schemes, such as, the maximum likelihood multi-
user detection (ML-MUD) or the other iterative interference
alignment schemes.
Index Terms—Interference Mitigation, Optimal Successive,
Group Decoding, Interference Channels, MIMO, Interference
Alignment, Effective Capacity.
I. I NTRODUCTION
He accelerated expansion of wireless communication is
T foreseen to carry on due to the quick move towards the
next generation technologies. This inspires researchers to make
promising development to realize peak bit rates as high as 1
Gbit/s and more. Achievement of such rate is predicted to be
facilitated by the deployment of distributed broadband wire-
less communications (BWC) systems. This indicates that the
essential importance in the prospect wireless communication
systems is to devise wireless transceivers with the capability
to communicate in a reliable manner in the existence of
interferers [1]. Therefore; studying the interference channel
(IC) and its adopted technologies becomes an indispensable
to keep up and to cope with the demands of such emerging
wireless networks.
Therefore, as a starting point we define the interference
channel as in [2] by the channel of multi pairs of input-
1 Part of the results in this article has been published by Omar Abu-Ella, and
Mohammed Elmusrati, on Optimal Successive Group Decoding to Mitigate
Interference in Wireless Systems. In Proc. 10th IEEE International Conference
on Distributed Computing in Sensor Systems (DCOSS ’14).
37
output terminals, where each input communicates through a
common medium with its respective outputs. Also, authors of
[3] presented the interference channel as a model for studying
networks with two or more (source-destination) pairs where
the signals of the sources interfere with each other at the
destination.
Now, we proceed to introduce one of the most crucial
principles of the interference channel systems, which is how to
interpret the interference effect, traditional wireless transceiver
designers commonly view interference as an augmentation to
the additive Gaussian noise. This assumption is not consis-
tently true; it is valid only if the employed detectors do not
take into consideration the interference formation; however, in
fact, practical systems have some level of knowledge about
interference; because the signals emitted by the interferers
belong to discrete constellations [4]. In the light of this
fact and because of interference management is essential to
attain higher spectral efficiency and thereupon higher peak
bit rates, researchers have evolved and enhanced abundant of
interference-aware mitigation techniques. One of those meth-
ods has lately grabbed much research concentration, known
as (interference alignment) [5]–[7]. Interference alignment
is fundamentally established on the concept of designing
transmitter and receiver to align interfering signals to each
other at receiver side [8]. More precisely, vector interference
alignment divides the dimensions of the receiver observation
space to two subspaces, one of the subspaces is occupied
by the desired signal and all the undesired interference are
aligned to the other subspace. From the theoretical point of
view, interference alignment is claimed to establish optimality
to approach Shannon capacity of interference network at
high SNR [9]. However, the existing interference alignment
schemes in fact are facing very challenging problems when it
comes to practical implementation. Here we state some of the
challenges that considered as strain in practical implementa-
tion of interference alignment algorithms:
• Generally, analytical solution for the interference align-
ment is difficult to obtain. The existing closed form
solutions are only been founded for certain cases with
very limited number of users.
• When the closed form expression exists, global channel
knowledge is required to obtain such a closed form
solution, which is of course requires an overwhelming
feedback overhead. In practice, restricted feedback con-
ditions, can lead to imperfect channel state information at
the transmitter (CSIT), which severely affects the system
efficiency and performance.
• The required signaling dimension of interference align-
38 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014
ment scheme grows exponentially with the total number
of users in the system; consequently, a question rises up
about the impracticality of implementing such a system.
• Although, the distributed interference alignment algo-
rithms require only local channel knowledge at each node,
it is yet require an extensive computational complex-
ity due to the iterative manner of finding its optimum
solution. In addition, the final values and the conver-
gence speed of the kernel iterative algorithm, which the
distributed interference alignment schemes rely on to
optimize their objective function, is very sensitive to the
initialization conditions.
• To avoid feedback some of interference alignment
schemes are build on the assumption of reciprocity of
the wireless network which does not always hold.
To overcome all the previously stated challenges facing in-
terference alignment strategy, we have to look for (no or
very limited) feedback system with tolerable complexity using
non-iterative technique. We try to avoid feedback because
the exhibited overall achieved capacity of the feedback-based
regimes, for instance, in the case of interference alignment
schemes, can be misleading when it is compared to non-
feedback system capacity. In other words, feedback in some
cases can hurt the system more than helping it, by decreasing
the pure throughput of the system. Consequently, our goal is
to have a better strategy to mitigate interference with limited
feedback. So, we should avoid all interference mitigation pro-
cesses at the transmitter side, and consider only the approaches
with the operations at the receiver side, at the same time, we
want to accomplish this task within a feasible computational
complexity.
Looking for another track of research to cope with inter-
ference, we encounter an alternative traditional methodology,
which is based on designing decoders in presence of inter-
ference. This approach employs one of the two following
techniques: successive interference cancelation for small SIR
regimes [10], or, on the other hand, treating interference like
noise for larger SIR regimes. This can be seen in power allo-
cation systems for frequency selective Gaussian interference
channels [11], and in CDMA cellular communication system
design [12]. However, there is still some range of SIR when
both of these techniques are suffering from an error floor [13].
Motivated by the necessary need for an optimal interference
mitigation technique considers all ranges of interference levels,
a novel interference mitigation scheme proposed in [14], where
the authors assumed that each receiver uses a successive
group decoder (SGD) which is considered as an extension
of the conventional successive decoder, however, instead of
decoding only one user at each decoding stage, a subgroup
of users are jointly decoded. Considering an interference
channel system where a fixed power allocated to all users and
assuming predetermined rates for all transmitters, the authors
of [14] obtained the decoding strategy minimizes the outage
probability at every receiver and generates the optimal subset
of interferers that must be decoded along with the desired user
under an imposed complexity constraint.
It is worth to mention that successive group decoders have
a substantial feature, that they can be implemented with
different levels of complexity. They can be implemented with
as low complexity as the conventional successive single-user
decoder to the high-complexity of the maximum likelihood
decoder. Therefore, by imposing a constraint on the decoder
complexity, we can adopt the decoder with an adequate
complexity that each receiver can sustain. In this stream of
research, a constrained partial group decoding (CPGD) scheme
was proposed in [15], [16]. In this technique, each receiver
employs a constrained partial group decoder to decode its
desired message conjointly with a part of the interference.
In other words, this decoder exploits the knowledge about
the interference to determine which interfering signals (with
a constraint on their group size) should jointly decoded along
with the desired signal, while treating the remaining interfering
signals as Gaussian noise.
The bottom line contribution of this work is to explore
the efficiency of using the OSGD in the interference channel
systems, by extensively inspecting its performance in terms of
achieved (ergodic and effective) capacity. In addition to that,
we evaluate its minimum required energy per bit, BER, and
outage probability, under different quality of service (QoS)
constraints. For generality purposes, this study considers both
the spatially correlated and uncorrelated (Rayleigh and Rician)
flat fading channels. Moreover, it investigates the operation of
the OSGD in different SNR and SIR environments, where
we consider both the power-limited and bandwidth-limited
regimes with different cross-talk values, as well as consid-
ering different transmit-receive multiple-input multiple-output
(MIMO) antenna configurations. To make our results more
sensible, we contrast the aforementioned performance with
those of other well known interference mitigation approaches,
such as, maximum likelihood multi-user detection (ML-MUD)
and interference alignment technique.
The remainder of this work is organized as follows: Sec-
tion II defines the used notation in this work. Followed by
Section III describing the model of interference channel which
applied in this study. A brief description of the addressed
concepts and interference mitigation techniques is presented
in Section IV. Complexity issues related to the presented
schemes are discussed in Section V. Numerical results of the
different scenarios in terms of achieved ergodic and effective
capacity, minimum required energy per bit, in addition to the
bit error rate and outage probability performance evaluation
are demonstrated in Section VI. Finally, Section VII concludes
this study.
II. N OTATION
Scalars represented in this study with lowercase italics, vec-
tors, and matrices denoted respectively by lowercase boldface
and uppercase boldface; superscripts ·T and ·H symbolize
the transpose and Hermitian (complex-conjugate) transpose;
[·]i,j stands for the (i, j)th element of a matrix; tr(A) implies
the trace ofPmatrix A; abs(·) designates the absolute value;
Nr PNt H
kHk2 = i=1
2
j=1 |[H]i,j | = tr(H H) is the squared
Frobenius norm of H; E{·} expresses the statistical average; ∼
denotes the distribution equivalence between the left and right
random variables; ≈ means approximately equals; , means
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014 39

is equal by definition to; the operator (x)+ , max(0, x) is

the projection on the nonnegative orthant; ⊗ indicates the
kronecker product; uppercase calligraphic letter (e.g., M)
indicates a finite set of integers; the underlined uppercase
calligraphic letter (e.g., G) represents the ordered partition of H 11
x1 y1
a set; |A| indicates cardinality of the set A; ∈ stands for: is H 12
element of; B ⊆ D indicates that B is a subset of D; Pr(·) is a H 1K
shorthand for the probability of; log(·) expresses the logarithm
term. H 21
H 22
x2 y2
H 2K
III. I NTERFERENCE C HANNEL M ODEL
H K1
We consider only the temporally uncorrelated discrete H K2
model of a flat-fading K-user interference channel (IC) de- H KK
picted in Fig. 1. In this study we assume that each transmitter xK yK
is equipped with Nt ≥ 1 antennas, and intends to commu-
nicate with its designated receiver which is equipped with
Fig. 1: The interference channel.
Nr ≥ 1 antennas. But, due to the broadcasting nature of
the wireless channel the transmitted signal is received by all
the K receivers. The kth receiver is interested only in the
signal transmitted by the kth transmitter; however, it is aware both ends of the channel, i.e., in the transmit and receive
of the coding scheme employed by all other users, (this is sides. Despite its simplicity, Kronecker model seems to be
practical assumption since the security of the information can a reasonable choice when the correlation is quickly vanishing
be maintained by using proper encryption techniques). The with the distance between the transmit and receive ends [19].
receiver may choose to decode some or all of the users only 1 1

if it presumes that will assist the decoding of its intended Hr = Rr2 Hw Rt2 (4)
user. The received signal of the kth receiver at the nth
where Hw is an (Nr × Nt ) normalized complex Gaussian
symbol interval through a K-user interference channel can be
random matrix, Rt , Rr , are the deterministic correlation ma-
expressed by
trices at the transmit and receive end respectively. Because of
√ √ K
X its simplicity and relative accuracy to simulate the realistic
yk [n] = P Hkk xk [n] + αP Hki xi [n] + zk [n], antenna inter-element correlation, we adopted the exponential
i=1,i6=k correlation model to generate the transmit and receive corre-
1≤k≤K (1) lation matrices, where they can be constructed using a single
correlation factor ρ ∈ C and |ρ| ≤ 1 as follows
where yk [n] is the Nrk × 1 received signal vector. further,  abs(j−i)
in this system we assume a quasi static flat fading scenario, ρ ,i 6 j
Rij =
∗ (5)
where the channel gain experienced by the kth receiver from ρabs(j−i) ,i > j
the ith transmitter is described by the Nrk × Nti matrix Hki
consists as in [17], [18] of two components, deterministic (i.e., In this work, Hki is assumed to be perfectly known to the
mean) denoted by Hd and random component represented by kth receiver, but it is unknown to any of the transmitters and
Hr r r any other receiver. zk [n] is the Ntk × 1 complex additive
K 1 white Gaussian noise (AWGN) vector at the receiver k. It
H= Hd + Hr (2) is assumed to have independent and identically distributed
1+K 1+K
q (i.i.d.)∼ CN (0, 1) elements and it is temporally uncorrelated.
K
where 1+K is the LOS component of the channel and P represents the average transmit power used by the ith user.
q
1 α denotes the cross-talk factor between users, i.e., it represents
1+K is the fading component, assuming uncorrelated flat
the relative propagation path loss of the interference channel.
fading. K is the Ricean K-factor of the channel and is defined
xi [n] represents a unit power symbol vector transmitted from
as the ratio of the power in the LOS component of the channel
the ith user during the nth time interval. Hereafter, for the
to the power in the fading component.
sake of simplicity, we omit the symbol interval index n.
kHd k2
K= (3)
E{kHr k2 }
IV. OVERVIEW FOR THE A DDRESSED C ONCEPTS AND
It is clear that for K = 0, the MIMO channel Hki has a pure S CHEMES
Rayleigh fading, also, if 0 < K < ∞, then, Hki has a Rician
fading; while, the case of K = ∞ corresponds to a non-fading This section gives a brief review for the concepts and the
channel scenario. different systems which we address in this study. This section
In this study, we use the Kronecker correlation model is concluded by detailed description of the main idea of the
expressed in (4) to simulate the correlation effect in the OSGD and its employed algorithms.
40 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014
A. Effective Capacity
Modeling the wireless channel regarding to the connection
quality of service (QoS) metrics such as delay, ratio of packet
loss and bandwidth is crucial to promote the QoS in the
future wireless networks. However, the commonly adopted
conventional physical-layer wireless models do not precisely
distinguish these QoS metrics. Intelligently, authors in [20],
proposed and develop a new approach to model the channel
QoS, they introduced a new terminology called (effective
capacity, EC) analogous to the (effective bandwidth), where
they assessed the wireless channel by two EC parameters,
the probability of non-empty buffer, and the connection QoS
exponent.
The first metric, i.e., the non-empty buffer probability is
equivalent in concept to the outage probability of the system,
where the later means that the signal-to-noise ratio (SNR)
at the receive falls below a specific threshold. However, the
probability of non-empty buffer does not equal the outage
probability; because, the former considers the packet accumu-
lation effect on the system, while, the later dose not. Therefore,
the outage probability is less than the probability of non-empty
buffer [20].
Generally, effective capacity is defined as in [21], [22] by
the maximum fixed rate of arrival, that can be supported by
some service process, such that, the requirements specified by
the QoS exponent are statistically guaranteed. More explicitly,
if we define Q as the stationary length of a queue, then, the
decaying rate of the tail of Q distribution, which denoted by
θ is given by
1 by applying Jensen’s inequality after interchanging the the
θ = − lim log Pr(Q ≥ q) (6)
q→∞ q expectation and the logarithm terms in (12).
It is important to notice that θ → 0 refers to a system that has
no delay constraint. On the other hand, θ → ∞ corresponds
to a system with a rigorous delay constraint. In other words,
larger θ implies more restriction on QoS, while smaller θ refers
to looser QoS guarantees.
Equivalently, if we define the buffering delay of the packet
at the steady-state by D, then for a large dmax , Pr(D ≥
dmax ) ≈ e−θδdmax , where δ is a constant depends on the
service and the arrival processes. The effective capacity is
defined in [23]–[25]
n o
Λ(−θ) 1
− = − lim loge E e−θS[t] (7)
θ t→∞ θt
Pt
where S[t] = i=1 R[i] is the time-accumulated service
process, and {R[i], i = 1, 2, . . .} refers to a discrete time sta-
tionary ergodic stochastic service process. Assuming a block
fading scenario with a frame duration T , the formula of the
effective capacity simplifies to
Λ(−θ) 1 n o
− =− loge E e−θT R[i] (8)
θ θT
with a short-time constraint on the total allocated power and
using a normalized input covariance matrix Qx defined as
 N
E xxH
Qx = (9)
P/B
where B stands for the system bandwidth, P is the allocated
power. The stochastic service process in a multiple-input
multiple-out (MIMO) channel system is given by
 
P
B log2 det I + HQx HH =
BN0
 
B log2 det I + Nr SNRHQx HH bps (10)
and the signal-to-noise ratio (SNR) defined as

E kxk2 P
SNR = = . (11)
E {knk }2 Nr BN0
where N0 is the power spectral density of the noise.
Now, we can obtain effective capacity of the closed and
open loop MIMO systems. First, we take into account the
case of closed loop MIMO system, i.e., the system with the
channel state information available at the transmitter (CSIT).
Therefore, the transmitter can adapt its transmitted power and
consequently, Qx according to the channel fading. Hence,
using (8) we can formulate the effective capacity after normal-
izing it by the receiver dimensionality Nr and the bandwidth
B as in (12) in the top of next page.
The trace of Qx conditioned to be less than or equal 1
to maintain the total allocated power condition. From the
formula, one can notice that when θ → 0, QoS constraints
turn to be loose, and the effective capacity reaches the ergodic
capacity as in (13).
However, for θ > 0 the ergodic capacity is generally greater
than the effective capacity. One can readily see that in (14)
Second, we consider the case of the open loop MIMO
system, i.e., the channel state information (CSI) is not available
at the transmitter side. Then, in practice, it is preferred to
allocate power uniformly across the MIMO antennas, and
consequently Qx = N1t I. Therefore, the effective capacity is
given by (15).
where, the subscript (id) refers to the identical distribution
of the power across the antenna elements.
B. Water-filling Power Allocation
As described in [26] for single-user MIMO channel which
can be decomposed to a set of non-interfering parallel sub-
channels, each of them is corrupted by an independent noise,
with a constrained on the total allocated power, such that,
N
X
PT = Pn , (16)
n=1
where N is the number of parallel sub-channels, and Pn is the
allocated power to the nth sub-channel. Then, the maximum
rate of reliable communication using this scheme is
X  
Pn |hn |2
log 1 + (17)
n=1
σ2
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014 41

   
1 H
CE (SNR, θ) = − loge E exp −θT B max log2 det I + Nr SNRHQx H bps/Hz/dimension (12)
θT BNr Qx 0,tr(Qx )≤1
  
1
lim CE (SNR, θ) = E max log2 det I + Nr SNRHQx HH bps/Hz/dimension (13)
θ→0 Nr Qx 0,tr(Qx )≤1

   
1
CE (SNR, θ) = − loge E exp −θT B max log2 det I + Nr SNRHQx HH
θT BNr Qx 0,tr(Qx )≤1
   
1 H
≤− E loge exp −θT B max log2 det I + Nr SNRHQx H
θT BNr Qx 0,tr(Qx )≤1
  
1 H
= E max log2 det I + Nr SNRHQx H (14)
Nr Qx 0,tr(Qx )≤1
   
1 Nr H
CE,id (SNR, θ) = − loge E exp −θT B log2 det I + SNRHH bps/Hz/dimension (15)
θT BNr Nt

It is easy to notice that the sum rate given by (17) can be levels. This is because the function f (SNR) = log2 (1 + SNR)
maximized by choosing the (optimal power allocation) as is a concave function, and it can be approximated using
XN   log2 (1 + x) ≈ x, x→0 (23)
Pn |hn |2
CN := max log 1 + (18) log2 (1 + x) ≈ log2 (x), x≫1 (24)
P1 ,... ,PN
n=1
σ2
Consequently, the system attains a decaying marginal capacity
under the constraint of
gain by adding more power to the the sub-channels with higher
N
X SNR. On the other hand, capacity increases linearly with
Pn = PT , Pn > 0, n = 1, . . . , N. (19) power in the low SNR levels. Therefore, providing some power
n=1 to weaker sub-channels can increase the total sum capacity
Using the Lagrangian method we can solve the problem of the [27]. However, water-filling scheme does not allocate power
concave objective function in (18) as follows to the sub-channels with too low SNR; because, transmitting
  information through such sub-channels is waste of power.
XN XN
Pn |hn |2 Then, capacity of water-filling power allocation algorithm is
L(λ, P1 , . . . , PN ) := log 1 + − λ Pn ,
σ2 given by
n=1 n=1
XN  
(20) Pn∗ |hn |2
C= log2 1 + . (25)
where λ is the Lagrange multiplier. The allocation power σ2
n=1
optimality condition is given by
 The above argument also can be seen from other point of
∂L = 0 if Pn > 0 view, for instance assuming that the system can concentrate its
=
∂Pn 6 0 if Pn = 0 resources, i.e., the transmitted power in this case, in the times
when the channel has high SNR, then the system can achieve
By defining x+ := max(x, 0), then the optimal power alloca- a huge capacity gain. By reflecting this insight into a multi-
tion which satisfies the optimality condition can be expressed user system scenario with large number of users; we can claim
as  + that it is most likely to have at any time instance a sub-group
∗ 1 σ2 of users whose channels are in good condition, then, using
Pn = − , (21)
λ |hn |2 a proper selection scheme the multi-user system capacity can
Then, a numerical algorithm can be employed to compute the be achieved. This form of diversity is called (opportunistic
Lagrange multiplier λ to satisfy the total power constraint beamforming) [28].
Finally, if the SNR of all sub-channels are equal, or the
XN  + transmitter does not know the channel state information, water-
1 σ2
− = PT . (22) filling algorithm transformed to be an equal power allocation
n=1
λ |hn |2
scheme, i.e., Pn = PT /N and capacity in this case is
Given the channel information, noise variance and using (21) XN  
PT
transmitter can determine how to optimally allocate power C= log2 1 + . (26)
N σ2
across the sub-channels. n=1
It is worth to notice that water-filling does not allocate all As one can observe, the slope of the log function is less
the power to the channels with the highest SNR, instead, it is than N ; then, the sum capacity is significantly larger than
still provides some power to the the channels with weaker SNR the capacity of SISO system.
42 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

The Iterative MIL Algorithm

C. Interference alignment using minimization of the interfer- 1 Initialize vk , k = 1, 2, · · · , K to be arbitrary precoding vectors
ence leakage (MIL) 2 Compute the matrix Qk in (30) for k = 1, 2, · · · , K
3 Obtain wk , k = 1, 2, · · · , K using (31)
Fundamentally, interference alignment schemes are based 4 Compute the matrix Rk in (41) for k = 1, 2, · · · , K
on the concept of designing transmitter and receiver to algin 5 Obtain vk , k = 1, 2, · · · , K using (42)
P PK
the interfering signals at the receiver side [8]. In this study 6 Repeat steps 2-5 until convergence of K k=1 Jk and k=1 Lk
we choose the interference alignment scheme which employs
an iterative algorithm to minimize the interference leakage
between users. The basic idea of this scheme as represented From the feasibility condition for perfect interference align-
recently in [29] is to design an iterative interference alignment ment, one requires
algorithm, iterates between two objective functions with a wH j = 1, 2, · · · , K, j 6= k
j Hj vk = 0, (34)
common interference leakage term, to find the locally optimum
vk and wk . The minimization of the interference leakage is Again, a judicious choice for the precoding vectors would
performed subject to a constraint on the dimension of the be to select vk such that the total interference power at the
desired signal subspace. unintended receives due to transmitter k is minimized. The
Considering the design of the wk which is the receiver filter interference power due to transmitter k at receiver j is obtained
vector at the receiver k, for a fixed precoding vectors vk at from the squared Frobenius norm of wH j Hjk vk as
all transmitters. The received interference plus noise at the
receiver k is given by Lkj = tr{Pk vH H
k [wj Hjk ] [wj Hjk ]vk }. (35)

K
X Thus, the total interference power due to the transmitter k is
rk = wH
k Hkj vj xj + wH
k zk . (27) given by
j=1,j6=k Ĺk = tr{vH
k Ŕk vk }. (36)
When the interference
PK signals are aligned, we need to find a where
K
X
wk such that j=1,j6=k wk Hk,j vj = 0. A reasonable choice
Ŕk = Pk [wH H H
j Hjk ] [wj Hjk ]. (37)
for wk is one that minimize the interference leakage power
j=1,j6=k
at receiver k. There is also a constraint on the dimensionality
of the desired signal: rank(wk Hkk vk ) = dk , where dk is the The objective function here is to choose vk to mini-
degree of freedom (DOF) assigned to the kth transmitter. Thus mize Ĺk , subject to the desired signal dimension constraint,
given the channel realization Hkk and the precoding vectors i.e, rank(wHk Hkk vk ) = dk . Including the regularization term
vk , the optimal receive filter wk is designed to minimize the the objective function is modified as
cost function
Ĺk = tr{vH H
k Ŕk vk + vk vk }. (38)
Jk , tr(wH k Qk wk ) (28)
Thus, the constrained optimization is given by
such that
wH min Ĺk = tr{vH
k Hkk vk }, (39)
k Hkk vk = βIdk (29) vk

where Qk is the interference plus noise covariance matrix at such that

receiver k, and it is given by wH
k Hkk vk = γIdk (40)
K
X and γ > 0 s selected such that tr(vH
k vk ) = 1, and
Qk = Pj [Hkj vj ][Hkj vj ]H + INrk (30)
K
X
j=1,j6=k
Rk = Pk [wH H H
j Hjk ] [wj Hjk ] + INtj . (41)
and β > 0 is selected such that tr(wH
k wk )
= 1. Here, Pj is j=1,j6=k
the transmit power of user j. The solution to (30) is given by Notice that R is the reflected covariance matrix of a virtual
wopt = βQ−1 H −1 −1 channel obtained by interchanging the transmitters and re-
k k uk [uk Qk uk ] (31)
ceivers. The optimum solution for vk is given by
where uk = Hkk vk is the desired signal subspace of the kth
vopt −1 H −1 H −1
k = γRk tk [tk Rk tk ] , k = 1, 2, · · · , K (42)
user
1
β= q , (32) where tk = wH
k Hkk
−1 −1 −1 −1
H
tr{[Qk u k Q́k ] [Qk uk Q́k ]} 1
γ=q , (43)
−1 −1
and Q́k = uH −1
k Qk uk tr{[R−1 tH
k k k t́ ] H [R−1 tH t́
k k k ]}
Now consider designing the precoding vectors vk at the
all transmitters, given the receive filtering vectors wk at all and t́k = tk R−1 H
k tk
receivers. The interference signal due to transmitter k at the
unintended receivers is given by The iterative MIL algorithm is summarized as follows

skj = wj Hjk vk xk , j = 1, 2, · · · , K, j 6= k (33)

ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014 43

D. Optimal Maximum Likelihood Multi-User Detection receiver i, the user of interest will be user i, and given the users
(i) √ (i) √ (i)
Assuming that all of the data symbol vectors are equally rates R, channel realization H̃ = [ P1 H1 , · · · , PK HK ],
likely, ML scheme is optimal in the sense of minimizing the M , {1, · · · , K}, and any disjoint subsets A, B of M, It says
probability of error of the detected signal [30]. For our system that an ordered partition G = {G1 , · · · , Gp } of any subset of
model, the ML detector is given by M , {1, · · · , K} for any p ≥ 1 is valid, if:
• Gk 6= φ
x̂ML = arg min{krk − yk k2 } (44)
x∈D • fi (Gk ) = 1, 1≤k≤p
where • i ∈ Gp
√ √ K
X
yk = P Hkk xk + αP Hki xi (45) where, fi (·) is a bounding function whose purpose is to impose
i=1,i6=k
decoding complexity constraint given as

Each x represents an Nt × 1 transmit vector, where x , [x1 , · · 1 |J | ≤ µi
fi (J ) =
·, xNt ]T , the mth data symbol xm is a complex valued, drawn 0 otherwise
from constellation alphabet A, all elements of the vector x are for a specified integer µi ≥ 1.
assumed to be independent and have zero mean unit variance. By defining a rate outage as an event where in a decod-
Also, set D includes all possible transmitted data vectors x, ing stage the rates of the signals to be decoded falls out
and its cardinality |D| = |A|KNt , with an exponential growth of the corresponding achievable rate region, and Ri to be
with KNt . One disadvantage of ML detection is related to the transmission rate of the signal on transmitter i, also,
its optimization problem, where D is in fact not a convex R , [Ri ]1≤i≤M . Then, the rate margin for decoding A while
set. Therefore, the often used numerical convex optimization treating B as noise for two disjoint subsets A, B ⊆ M is
methods are not suitable for such a scheme [30]. Because defined as follows
the conventional exhaustive search method to find the optimal
solution of (44) by evaluating the krk − yk k2 has a complexity
(i) (i)
O(|A|KNt ), it is common to consider the maximum likelihood ε(H̃ , A, B, R) , min {△ (H̃ , D, B, RD )}, A 6= φ
D⊆A,D6=φ
multiuser detection infeasible for most current communica- (46)
tion applications [31]. Recently, authors in [32] introduced (i)
with ε(H̃ , φ, B, R) = 0 and where
a novel promising approach of designing a low-complexity
maximum likelihood multi-user detector employing quantum (i)
△ (H̃ , D, B, RD ) , (47)
search algorithms (QSA) for potential application in wireless    
 (i)H (i) (i)H −1 (i)  X
communication systems. Their presented results show that log I + H̃D I + H̃B H̃B H̃D  − Rj
the employed quantum based MUD search algorithm ties the j∈D
performance of the conventional optimal ML-MUD. However,
(i)
this performance achieved with a significant computational Note that RA ∈ C(H̃ , A, B) if and only if
(i)
complexity reduction compared to the conventional ML-MUD. ε(H̃ , A, B, R) ≥ 0. Now, for any valid ordered partition
This can be seen as an opening for using the joint and group G = {G1 , · · · , Gp } ∈ Qi we define
detection (GD) techniques that can support larger number n  (i) o
(i)
of users and higher order modulation schemes in the next ε(H̃ , G, R) , min ε H̃ , Gk , M\ ∪kj=1 Gj , R
generation of the communication systems. 16k6p
(48)
Now, let Qi to be the set of all valid ordered partitions of
E. Optimal Successive Group Decoding all subsets of M which contain i, then, the outage event O(i)
Successive group decoder (SGD) was introduced as an occurs for the valid ordered partition G = {G1 , · · · , Gp } ∈ Qi
extension of the standard successive decoder in which at each if and only if
stage a subset of users is jointly decoded after treating the (i)
ε(H̃ , G, R) < 0. (49)
transmissions of the remaining users as a Gaussian interfer-
ence. Therefore, an OSGD using the partition G ∈ Qi will attempt
In this system, each receiver employs a successive decoding (i)
to decode user i if and only if ε(H̃ , G, R) ≥ 0.
procedure, in each stage a subset of users are jointly decoded,
To perform this decoding procedure each receiver i where
after subtracting the already decoded users from the received
1 ≤ i ≤ K employs a greedy algorithm (Algorithm 1) that
signal, and by treating the remaining users as AWGN. OSGD
either declare an outage or yields an optimal valid partition of
limits the number of users being jointly decoded at each stage
the OSGD. The involved steps of the successive decoding for
to be at most µ to control the complexity of the decoder.
the OSGD in the ith receiver are as follows.
In the remainder of this section, for the sake of the reader (i)
convenience, we reproduce the basic concepts of the OSGD. 1) Initialize with input H̃ , R
Assuming that the power and the rate of all users are pre- 2) Receiver i runs Algorithm 1
specified and are not dependent on the channel realizations. 3) If there is no outage, Algorithm 1 outputs an optimal
Then, we employ the OSGD proposed in [14], which min- partition, G opt = {G1 , · · · , Gp }, Then
imizes the outage probability of each user. Considering the For 1 ≤ k ≤ p
44 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

 +

• Compute the noise covariance matrix: O |S|min{(|S|−µi ) ,µi } , where (x)+ = max(0, x). Also, the
(i) P (i) iH
Φ Gk = I + j∈M\∪k G H̃ j H̃j and loop consists the steps 4 to 9 is repeated at most |G| times.
j=1 j
jointly decode the users in Gk using the ML Therefore, since Algorithm 1 will invoke Algorithm 2 no more
  − 21 J than K times, the complexity of Algorithm 1 for fixed group
(i) (i) (i)

rule on r̃ [n] = ΦGk y [n] size µi is O K min{K−µi +1,µi +1} , which is polynomial in
n=1
assuming the model K.
 − 12 (i) On the other hand, the MIL iterative interference alignment
(i)
r̃(i) [n] = Φ Gk H̃Gk xGk [n] + scheme requires a huge computational complexity to compute
z̃(i) [n], 1 ≤ n ≤ J the transmit precoding vectors and receiving filters. For in-
with z̃(i) [n] ∼ CN (0, I). stance, in each iteration, the kernel algorithm of this scheme
n oJ 2 2
, requires K(K − 1)(Nr Nt + Nr ) and K(K − 1)(Nt Nr + Nt )
(i) (i) (i) (i)
• Update y [n] ←− y [n] − H̃Gk x̂G [n]
k
n=1 complex multiplications, only to compute the covariance ma-
(i)
where x̂Gk [n] is the decision made trix in the transmitter and the receiver respectively. Also, since
corresponding to the nth symbol interval, it is an iterative based algorithm, it needs (depending on the
through the coherence interval of J symbols, initial conditions) a considerable number of iterations L to
in the codeword of the users in the set Gk , converge.
which are re-encoded and modulated by Finally, the maximum likelihood multi-user detection is the
receiver i post-processing. scheme with the highest computational complexity, where it
End For uses the exhaustive search to find the optimal solution of its
4) Otherwise, Algorithm 1 declares an outage for the objective function as shown in Subsection IV-D. This means
intended user i. the receiver performs its search over all combination of all
possible transmitted vectors. Therefore, overall complexity of
this scheme grows exponentially with the number of users
Algorithm 1 Greedy Partitioning for Fixed Rates and the number of transmit antennas. Thus, if we denote
(i)
1 Initialize S = M, G opt = φ the searching set by D, then, its size is given by |A|KNt ,
2 Determine a group G ∗ ⊆ S using Algorithm 2 after initializing it where |A| is the constellation size of the employed modulation
with user set S and rates RS .
(i) ∗ ∗
scheme.
3 If ε(H̃ , G , S\G , R) < 0 then
4 Declare an outage and Stop.
5 else VI. S IMULATION R ESULTS
(i) (i)
6 Update S ←− S\G ∗ and G opt ←− {G opt , G ∗ }
7 If i ∈ G ∗ In this section we present the numerical results carried out
(i)
8 Output G opt . by Monte Carlo simulation of K-user interference channel
9 Stop. MIMO system. Intentionally, we limit the number of users
10 else
11 Go to Step 2
in the simulated scenarios; because of the fact that in many
12 End if realistic scenarios the number of interfering users which
13 End if significantly affecting the wireless connection between the
access point and its desired user is usually limited. As in the
case of downlink cellular system, where there are only limited
Algorithm 2 Selecting an Optimal Group
1 Initialize user set S and rates RS number of co-channel interfering base stations disturbing the
2 Form S , {G ⊆ S : G 6= φ, |G| = µi or G = S} and set S1 = φ, δ = −∞. connection between the base station of interest and its desired
3 For each G ∈ S
4 Repeat user. Also, in uplink system, assuming that there is no intra-
5 Update S1 ←− {S1 , G}. cell interference, which is a reasonable assumption for the
6 Determine
(i)
ξ = minW⊆G,W6=φ ∆(H̃ , W, S\G, RW )
most of the modern cellular systems, the only source of
and let Ŵ be the minimizing set which among interference comes from the co-channel users in the other
all minimizers has the smallest cardinality neighboring cells and those users must be close enough to the
7 If δ < ξ then set A = G and δ = ξ.
8 Update G ←− G\Ŵ base station of interest to be considered as interferers, thus, it is
9 Until G = φ or G ∈ S1 reasonable to assume that they are limited too. Therefore, in all
10 End For
11 Output G ∗ = A, ε(H̃
(i)
, G ∗ , S\G ∗ , R) = δ and stop.
of the following simulation setups of the interference channel
systems we assume the number of users K is 3 transmit-
receive pairs. One user represented as a desired client and
the other two as interfering sources.
V. C OMPLEXITY I SSUES In this work we aim to investigate the OSGD performance in
In this section, we summarize the computational complexity different scenarios and using a wide range of testing metrics.
issues related to the different used techniques in this study. We inspect the OSGD in the interference channel in terms
Starting with few remarks on the complexity of the OSGD of its achieved (ergodic, effective and outage) capacity, also,
and its employed algorithms. Given the users set S and we assess its minimum required energy per bit, as well, we
the group size constraint µi , we find that the number of evaluate its BER and outage probability performance. This
sub-groups that Algorithm 2 is examining will be at most study considers both the spatially uncorrelated and correlated
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014 45

18 15
OGD, µ = 3, θhat = 0 OGD, µ = 3, θhat = 0
16 OGD, µ = 3, θhat = 5 OGD, µ = 3, θhat = 5
OGD, µ = 3, θhat = 10 OGD, µ = 3, θhat = 10
14 OGD, µ = 2, θhat = 0 OGD, µ = 2, θhat = 0
OGD, µ = 2, θhat = 5 OGD, µ = 2, θhat = 5
12 OGD, µ = 2, θhat = 10 10 OGD, µ = 2, θhat = 10
Rate per user (bps/Hz)

Rate per user (bps/Hz)

MMSE, θhat = 0 MMSE, θhat = 0
hat
10 MMSE, θ =5 MMSE, θhat = 5
hat
MMSE, θ = 10 MMSE, θhat = 10
8

6 5

0 0
−30 −20 −10 0 10 20 30 40 −50 −40 −30 −20 −10 0 10 20 30 40
SNR (dB) SNR (dB)

Fig. 2: OSGD-µ = 2, 3 and MMSE effective rate versus SNR Fig. 3: OSGD-µ = 2, 3 and MMSE effective rate versus SNR
(dB), α = 1, Nt = 3, Nr = 3, ρ = 0 and the normalized QoS (dB), α = 1, Nt = 3, Nr = 3, ρ = 0.9 and the normalized
exponent θ̂ = 0, 5 and 10. QoS exponent θ̂ = 0, 5 and 10.

35
(with various inter-element correlation factors) flat fading OGD, µ = 3, θhat = 0
OGD, µ = 3, θhat = 5
channels, under different QoS constraint values. Also, this 30
OGD, µ = 3, θhat = 10
work explores the OSGD efficiency in different SNR and SIR OGD, µ = 2, θhat = 0
25 OGD, µ = 2, θhat = 5
environments, (where we consider both the power-limited and Rate per user (bps/Hz) OGD, µ = 2, θhat = 10
bandwidth-limited regimes with different cross-talk values). 20
MMSE, θhat = 0
MMSE, θhat = 5
All of aforementioned investigation performed with numerous MMSE, θhat = 10
transmit-receive antenna configurations. 15

10

A. Achieved Capacity 5

In the first subsection we study the achieved ergodic and 0

−50 −40 −30 −20 −10 0 10 20 30 40
effective capacity in (bps/Hz) of the OSGD versus the signal- SNR (dB)

to-noise ratio SNR in (dB). We consider the (3 × 3 and 6 × 6)-

MIMO interference channel, with three different values for Fig. 4: OSGD-µ = 2, 3 and MMSE effective rate versus SNR
the normalized QoS exponents (θ̂ = 0, 5 and 10), also we use (dB), α = 1, Nt = 6, Nr = 6, ρ = 0 and the normalized QoS
two inter-element correlation factors (0, for the uncorrelated exponent θ̂ = 0, 5 and 10.
channel and 0.9 for the highly spatial correlated channel). All
the comparisons carried out here is for a strong interference 30
environment, where the cross-talk parameter (α) is set to be OGD, µ = 3, θhat = 0
OGD, µ = 3, θhat = 5
1. Figures (2, 3, 4 and 5) present the achieved ergodic and 25 OGD, µ = 3, θhat = 10

effective capacity results of MMSE receiver and OSGD with OGD, µ = 2, θhat = 0
OGD, µ = 2, θhat = 5
group-size (µ) equals 2 and 3 respectively. Clearly, it can be 20 OGD, µ = 2, θhat = 10
Rate per user (bps/Hz)

seen from the figures that OSGD has the highest achieved MMSE, θhat = 0
MMSE, θhat = 5
ergodic and effective capacity in all cases. Also, we can notice 15 MMSE, θhat = 10

that the OSGD with µ = 3 is the least affected scheme by

changing the QoS metric θ̂. Another point to mention here is, 10

we can observe the achieved capacity of the OSGD becomes

lower as the channel becomes more correlated. This can be 5

interpreted as a result of decreasing the degrees of freedom

(DoF), where, the channel turns out to be less ranked by 0
−30 −20 −10 0 10 20 30 40
SNR (dB)
increasing the correlation between its elements, consequently,
its eignmodes tend to concentrate in one dominated mode. On
Fig. 5: OSGD-µ = 2, 3 and MMSE effective rate versus SNR
the other hand, since the MMSE receiver is based on extracting
(dB), α = 1, Nt = 6, Nr = 6, ρ = 0.9 and the normalized
the desired user alone, this case can be an advantage for such
QoS exponent θ̂ = 0, 5 and 10.
a scheme, thus, it can help increasing its achieved capacity.
Nevertheless, the achieved capacity of MMSE receiver in all
cases is still far less than that for the OSGD.
46 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

25 20
OGD 4×4 OGD 4×4
OGD 3×3 18 OGD 3×3
MIL 4×4 MIL 4×4
20 MIL 3×3 16 MIL 3×3

14
Rate per user (bps/Hz)

Rate per user (bps/Hz)

15 12

10

10 8

5 4

0 0
−30 −20 −10 0 10 20 30 40 −30 −20 −10 0 10 20 30 40
SNR (dB) SNR (dB)

Fig. 6: OSGD and MIL effective rate versus SNR (dB), α = 1, Fig. 7: OSGD and MIL effective rate versus SNR (dB), α = 1,
(3 × 3, 4 × 4)-MIMO, ρ = 0 and the normalized QoS exponent (3 × 3, 4 × 4)-MIMO, ρ = 0.9 and the normalized QoS
θ̂ = 3. exponent θ̂ = 3.

50
B. Interference Mitigation Capability OGD 8×8
45 OGD 6×6
In the second subsection, we study the OSGD capability to MIL 8×8
40 MIL 6×6
mitigate and suppress interference, so we compare its achieved
effective rate per user measured by (bps/Hz) with the achieved Rate per user (bps/Hz)
35

rate of a well known interference alignment scheme which 30

minimizes the leakage interference between users, we point it 25

here by (MIL). This comparison is done using four different 20

MIMO configurations (3 × 3, 4 × 4, 6 × 6 and 8 × 8), in a high 15
cross-talk environment, where the α parameter set up to be 1,
10
the the normalized QoS exponent chosen to be 3, and for two
different antenna inter-element correlation factors (0, and 0.9). 5

From figures (6, 7, 8 and 9), we notice that in all cases OSGD 0
−30 −20 −10 0 10 20 30 40
has achieved a higher effective capacity comparing with the SNR (dB)

MIL scheme. For instance, when ρ = 0, for the 3 × 3 MIMO

Fig. 8: OSGD and MIL effective rate versus SNR (dB), α = 1,
configuration, the effective rate per user of the OSGD is 70%
(6 × 6, 8 × 8)-MIMO, ρ = 0 and the normalized QoS exponent
higher than that for the MIL. This percentage goes up to 107%
θ̂ = 3.
in case of the 6 × 6 MIMO channel. However, when ρ = 0.9
this percentages become 79% and 119% for the 3 × 3 and
6 × 6 MIMO configurations respectively. This implies that 40
OGD 8×8
OSGD is less affected by the channel inter-element correlation OGD 6×6
35
comparing with the MIL. The overall results in this subsection MIL 8×8
MIL 6×6
reveal that OSGD is more effective to suppress interference 30

comparing with the MIL interference alignment scheme.

Rate per user (bps/Hz)

25

20
C. Minimum Required Energy Per Bit
15
Now, with the same aforementioned simulation setup in the
previous subsection, we study the minimum required energy 10

per bit (Eb /N0 )min for the OSGD and compare it with 5
that for the MIL interference alignment algorithm. Where,
Eb /N0 = RESN R
(SN R) , and RE (SN R) is the effective rate of
0
−30 −20 −10 0 10 20 30 40
the system at the given SN R. Results of this comparison are SNR (dB)

presented in figures (10, 11, 12, and 13). Interestingly, we

Fig. 9: OSGD and MIL effective rate versus SNR (dB), α = 1,
observe that for both OSGD and MIL increasing the inter-
(6 × 6, 8 × 8)-MIMO, ρ = 0.9 and the normalized QoS
element correlation helps the system to work at lower energy
exponent θ̂ = 3.
per bit, because the two systems benefit from the concentration
of the channel eignmodes in only one dominant mode. All the
results show that OSGD requires a lower minimum energy per
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014 47

2 2
10 10

1 1
10 10
Effective rate per user (bps/Hz)

Effective rate per user (bps/Hz)

0 0
10 10

−1 −1
10 10

−2 −2
10 MIL 3×3 10 MIL 6×6
MIL 4×4 MIL 8×8
OGD 3×3 OGD 6×6
OGD 4×4 OGD 8×8
−3 −3
10 10
−15 −10 −5 0 5 −25 −20 −15 −10 −5 0 5
Eb/No (dB) Eb/No (dB)

Fig. 10: OSGD and MIL effective rate versus Eb /No (dB), Fig. 12: OSGD and MIL effective rate versus Eb /No (dB),
α = 1, (3 × 3, 4 × 4)-MIMO, ρ = 0 and the normalized QoS α = 1, (6 × 6, 8 × 8)-MIMO, ρ = 0 and the normalized QoS
exponent θ̂ = 3. exponent θ̂ = 3.

1
10
2
10

0 1
10 10
Effective rate per user (bps/Hz)

Effective rate per user (bps/Hz)

0
10
−1
10

−1
10
−2
10
MIL 3×3
−2
MIL 4×4 10 MIL 6×6
OGD 3×3 MIL 8×8
OGD 4×4 OGD 6×6
−3
10 OGD 8×8
−15 −10 −5 0 5 −3
Eb/No (dB) 10
−25 −20 −15 −10 −5 0 5
Eb/No (dB)

Fig. 11: OSGD and MIL effective rate versus Eb /No (dB),
α = 1, (3 × 3, 4 × 4)-MIMO, ρ = 0.9 and the normalized QoS Fig. 13: OSGD and MIL effective rate versus Eb /No (dB),
exponent θ̂ = 3. α = 1, (6 × 6, 8 × 8)-MIMO, ρ = 0.9 and the normalized QoS
exponent θ̂ = 3.

bit comparing to the MIL scheme. For instance, the OSGD

needs 0.9 dB less energy in case of the highly correlated
(3 × 3) Rayleigh fading channel and this gap increases to be
4.5 dB less required minimum energy per bit in case of (8 × 8)
uncorrelated channel.
D. Capacity Upper Bound
In this part, to have a benchmark for our comparison, we
contrast the achieved ergodic capacity per user of the OSGD
with the achieved capacity of the single user (i.e., interference-
free) (1 × 2, 2 × 2 and 3 × 3)-MIMO uncorrelated channel
systems using the equal and adaptive power allocation. This
comparison is carried out for the band-limited and power-
limited regimes, i.e., in high and low SNR environments.
From the results drawn in figures (14, 15, 16 and 17) which
represent the achieved capacity of the three systems, one can
see that at very low SNR, OSGD capacity is similar to that
of the interference-free with equal power allocation system,
because in such environment the dominant disturbing factor
is the noise not the interference. However, there is still a
gap between the OSGD and the adaptive power allocation
scheme which uses the water-filing algorithm to distribute
the total power across the antennas. This gap arises because
availability of the channel state information at the transmitter
(CSIT) for the water-filing scheme helps it to allocate power
more intelligently across the paths; while, the OSGD lakes
this feature as it is based on a receiver-side processing only,
thus, it assumes an equal power allocation in the transmitter
side. More importantly, we ought to remember that the pure
throughput of the water-filing scheme is hugely affected by
the required feedback. After all, we should also keep in
mind that both of the equal and water-filing power allocation
systems mentioned here are only a single user systems, i.e., no
interference is assumed, and we compare our system, which
is working in a strong interference environment, where α = 1,
with them only because we need to have an upper bound
reference for the possible achieved capacity.
48 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

15 3
SU−MIMO−WF, perfect CSI SU−MIMO−WF, perfect CSI
SU−MIMO −equal allocated power SU−MIMO−WF, quantized CSI , (8 bits)
OGD SU−MIMO−EP
2.5 OGD, strong interference

10 2
Rate per user (bps/Hz)

Rate per user (bps/Hz)

1.5

5 1

0.5

0 0
−50 −40 −30 −20 −10 0 10 20 30 40 −50−45 −40 −35 −30 −25 −20 −15 −10 −5
Power SNR (dB)

Fig. 14: Ergodic rate per user versus SNR (dB) for OSGD, Wa- Fig. 17: Ergodic rate per user versus SNR (dB) for OSGD,
ter filing and Equal power allocation schemes, Nt = 1, Nr = Water filing and Equal power allocation schemes in power-
2, and α = 1, ρ = 0. limited regeim, Nt = 3, Nr = 3, and α = 1, ρ = 0.

30
SU−MIMO−WF, perfect CSI E. Effect of Cross-Talk Factor
SU−MIMO −equal allocated power

25
OGD, with strong interference In this subsection we discuss the effect of changing the
cross talk α on the capacity of the OSGD scheme. It is clear
20 that the cross-talk factor plays a major role in determining
Rate per user (bps/Hz)

the final level of the achieved capacity of any interference

15 channel system, simply because α represents the level of
exchanged interference among users. Thus, it is important for
10 our investigation to evaluate the OSGD effectiveness using
different values of α. However, to gain deeper insight we
5 choose to do this task in different SNR levels. In this part, the
used interference environment is set up to be (4 × 4) MIMO
0
−50 −40 −30 −20 −10 0 10 20 30 40
flat fading channel connecting each transmit-receive pair. To
SNR (dB)
avoid any other effects we adjust both of the normalized QoS
exponent θ̂ and the antenna inter-element correlation factor ρ
Fig. 15: Ergodic rate per user versus SNR (dB) for OSGD, Wa-
to be 0. Figures (18, 19 and 20) present the achieved capacity
ter filing and Equal power allocation schemes, Nt = 2, Nr =
against a wide decibel range of cross-talk intensity for different
2, and α = 1, ρ = 0.
levels of SNR. From the results, it is clear that for high
cross-talk intensities (in the left side of the figures, which
1.8 represents the scenario of the strong interference environment),
SU−MIMO−WF, perfect CSI
1.6
SU−MIMO−WF, quantized CSI, (8 bits) OSGD with group size µ = 3 outperforms both of the OSGD
SU−MIMO−EP
OGD, strong interference with µ = 2 and the MMSE receiver. On the other hand, for
1.4
lower levels of α (on the right side of the figures) there is
1.2 no difference of the performance of the three schemes; this
Rate per user (bps/Hz)

1
is simply because in the very low cross-talk levels OSGD
tends to decode its desired user alone, treating the other weak
0.8
interfering signals as Gaussian noise, i.e., the OSGD converts
0.6 to MMSE decoder.
0.4

0.2 F. Effect of Number of Antennas

0
−50−45 −40 −35 −30
Fig. 16: Ergodic rate per user versus SNR (dB) for OSGD,
Water filing and Equal power allocation schemes in power-
limited regeim, Nt = 2, Nr = 2, and α = 1, ρ = 0.
−25 −20 −15 −10 −5 This subsection focuses on the effect of number of antennas
SNR (dB)
on the achieved capacity of the different systems. Again,
similarly as in the preceding subsection, we set both of θ̂ and
ρ to be 0, then, we evaluate the achieved capacity for MMSE
receiver and OSGD with (µ = 2, and 3) against the SNR levels
using different sizes of MIMO transmit-receive pairs, the con-
sidered MIMO configurations are (1×1, 2×2, 4×4, 6×6, 8×8
and 10×10). Results of the three systems are shown in figures
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014 49

40 8
SNR = 30 dB 10×10
SNR = 20 dB 8×8
35 SNR = 10 dB 7
6×6
SNR = 0 dB
4×4
SNR = −10 dB
30 6 2×2
1×1

Effective Capacity (bps/Hz)

25 5
Capacity (bps/Hz)

20 4

15 3

10 2

5 1

0 0
0 −5 −10 −15 −20 −25 −30 −35 −40 −20 −15 −10 −5 0 5 10 15 20 25 30
Cross talk factor (α) in (dB) SNR (dB)

Fig. 18: MMSE ergodic capacity versus cross-talk factor in Fig. 21: MMSE ergodic capacity versus SNR using different
different SNR levels, Nt = 4, Nr = 4, ρ = 0. sizes of MIMO configuration, ρ = 0.

40 12
SNR = 30 dB 10×10
SNR = 20 dB 8×8
35 SNR = 10 dB 6×6
10
SNR = 0 dB 4×4
SNR = −10 dB
30 2×2
1×1

Effective Capacity (bps/Hz)

8
25
Capacity (bps/Hz)

20 6

15
4

10
2
5

0 0
0 −5 −10 −15 −20 −25 −30 −35 −40 −20 −15 −10 −5 0 5 10 15 20 25 30
Cross talk factor (α) in (dB) SNR (dB)

Fig. 19: OSGD with µ = 2 ergodic capacity versus cross-talk Fig. 22: OSGD with µ = 2 ergodic capacity versus SNR using
factor in different SNR levels, Nt = 4, Nr = 4, ρ = 0. different sizes of MIMO configuration, ρ = 0.

40 it is still far away from the efficiency of the OSGD with group
SNR = 30 dB
SNR = 20 dB size of 3. It is also clear that MMSE is an interference-limited
35 SNR = 10 dB
SNR = 0 dB
scheme no matter what the MIMO size is. Differently, OSGD
30
SNR = −10 dB
with µ = 3 exhibits a noise-limited behavior, revealing its
capability to suppress interference. In between, we perceive
25
Capacity (bps/Hz)

that although OSGD with µ = 2 is not as powerful as the

20 OSGD with µ = 3, but, it succeeds to push its interference-
limited behavior to appear at higher SNR levels, demonstrating
15
a moderate interference mitigation capability; yet, it has the
10 complexity reduction advantage over the OSGD with the full
5
group size.

0
0 −5 −10 −15 −20 −25 −30 −35 −40 G. Effect of Antenna Inter-Element Correlation
Cross talk factor (α) in (dB)
In this part, we aim our attention to study the relationship
Fig. 20: OSGD with µ = 3 ergodic capacity versus cross-talk between the antenna inter-element correlation factor ρ and
factor in different SNR levels, Nt = 4, Nr = 4, ρ = 0. the achieved capacity. In the following scenario, we employ
(4 × 4)-MIMO channel for every transmit-receive pair, with
cross-talk parameter α = 1 and θ̂ = 0. From the results in
(21, 22, and 23). From the outcomes, one can see that although figures (24, 25 and 26), one can notice that for a specific
the OSGD with group size of 2 is more efficient in exploiting SNR level achieved capacity of OSGD with µ = 3 decreases
the increase of MIMO antenna size than the MMSE receiver, monotonically with increasing the correlation factor. However,
50 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

50 7
10×10 γ = 30 dB
45 8×8 γ = 20 dB
6×6 6 γ = 10 dB
40 4×4 γ = 3 dB
2×2 γ = 0 dB
35 1×1 5 γ = −3 dB
Effective Capacity (bps/Hz)

γ = −10 dB

Capacity (bps/Hz)
30 γ = −20 dB
4
25
3
20

15 2

10
1
5

0 0
−20 −15 −10 −5 0 5 10 15 20 25 30 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
SNR (dB) Inter−element Correlation factor (ρ)

Fig. 23: OSGD with µ = 3 ergodic capacity versus SNR using Fig. 25: OSGD with µ = 2 ergodic capacity versus antenna
different sizes of MIMO configuration, ρ = 0. inter-element correlation factor in different SNR levels, using
4 × 4 MIMO for each transmit-receive pair.
6
γ = 20 dB
γ = 10 dB 18
5 γ = 3 dB γ = 30 dB
γ = 0 dB 16 γ = 20 dB
γ = −3 dB γ = 10 dB
γ = −10 dB γ = 3 dB
4 14
γ = −20 dB γ = 0 dB
Capacity (bps/Hz)

γ = −3 dB
12
γ = −10 dB
Capacity (bps/Hz)

3 γ = −20 dB
10

2 8

6
1
4

0 2
0 0.1 0.2
Fig. 24: MMSE ergodic capacity versus antenna inter-element
correlation factor in different SNR levels, using 4 × 4 MIMO
for each transmit-receive pair.
this is not the case for the OSGD with µ = 2 or the MMSE
receiver. They exhibit a monotonic decrease of their attained
capacity with increasing the correlation factor only at low
levels of SNR, but for higher SNR levels, they are interestingly
revealing a sudden increase in their capacities in an overshoot-
like behavior for short range of correlation values, before
they return to steeply decreasing their capacity for higher
correlation. However, apparently, we can recognize that for
all correlation and SNR combination the OSGD is the scheme
with the highest achieved capacity.
H. Bit Error Rate
This subsection presents the results of BER evaluation for
the OSGD scheme in different scenarios, where we conduct
this assessment using various SIMO and MIMO configura-
tions, using several correlation factors. In the first scenario,
we assume a strong interference environment α = 1 with un-
correlated Rayleigh flat fading, where, single transmit antenna
users communicate with their intended receivers which employ
OSGD and are equipped in each case with (2, 3, 4, or 5)
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Inter−element Correlation factor (ρ)
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Inter−element Correlation factor (ρ)
Fig. 26: OSGD with µ = 3 ergodic capacity versus antenna
inter-element correlation factor in different SNR levels, using
4 × 4 MIMO for each transmit-receive pair.
antennas. The modulation scheme used here is the BPSK, and
information decoded in the receiving side symbol by symbol.
The results are depicted in Fig. 27, from which it is obvious
Eb
that BER decreasing linearly with the logarithmic scale of N .
0
Also, we can readily notice the increase of the diversity gain
(which can be inferred from the increase of the absolute slope
of the BER curves) by increasing the number of the receive
antennas. Now, to get a deeper insight on the OSGD efficiency,
we choose to compare its performance with the ML-MUD
scheme, where all the users are jointly decoded in the receiver
side. Fig. 28 presents the BER performance of the OSGD and
ML-MUD in a strong uncorrelated interference environment,
but in this case, all of the transmitters and receivers are
equipped with multiple antennas. At each transmitter the
information bits are modulated and transmitted simultaneously
over all antenna elements of that user, this is of course will
introduce a transmit diversity gain over the previous single
antenna case. The results of this scenarios show that at high
SNR levels BER performance of the OSGD is slightly deviates
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014 51

0
10

−1
10

−2
10

BER
−3
10

OGD, 4×4, ρ = 0.9

ML−MUD, 4×4, ρ = 0.9
−4
10 OGD, 4×4, ρ = 0.5
ML−MUD, 4×4, ρ = 0.5
OGD, 4×4, ρ = 0
ML−MUD, 4×4, ρ = 0

0 1 2 3 4 5 6 7 8 9 10
Eb/No (dB)

Fig. 27: OSGD BER performance versus SNR for Fig. 29: OSGD and ML-MUD BER performance versus SNR
1 × (2, 3, 4, 5) SIMO configurations, with cross-talk fac- for 4 × 4 MIMO configuration, with cross-talk factor α = 1
tor α = 1 and receive antenna inter-element correlation ρ = 0. and antenna inter-element correlation ρ = 0, 0.5 and 0.9.

0
10

0
10
−1
10

−1
10
−2
10

−2
10
BER

−3
10
BER

−3
10
−4
10
OGD, 2×2 OGD, 5×5, ρ = 0.9
−4
10 ML−MUD, 2×2 ML−MUD, 5×5, ρ = 0.9
OGD, 3×3 −5 OGD, 5×5, ρ = 0.5
10
ML−MUD, 3×3 ML−MUD, 5×5, ρ = 0.5
−5 OGD, 4×4 OGD, 5×5, ρ = 0
10
ML−MUD, 4×4 ML−MUD, 5×5, ρ = 0
−6
OGD, 5×5 10
0 1 2 3 4 5 6 7 8
ML−MUD, 5×5
−6 Eb/No (dB)
10
0 1 2 3 4 5 6 7 8 9 10
Eb/No (dB)
Fig. 30: OSGD and ML-MUD BER performance versus SNR
Fig. 28: OSGD and ML-MUD BER performance versus for 5 × 5 MIMO configuration, with cross-talk factor α = 1
SNR for different MIMO configurations, with cross-talk fac- and antenna inter-element correlation ρ = 0, 0.5 and 0.9.
tor α = 1 and antenna inter-element correlation ρ = 0.

from the ML-MUD performance; however, in low SNR levels
OSGD exhibits very enhanced capability to decode its desired
users and matches with the ML-MUD scheme. The reason for
the slight performance drift is that OSGD tends to be a single
decoder at higher SNR levels. On the other hand, Figures (29
and 30) demonstrate the OSGD BER performance for the
(4 × 4 and 5 × 5) MIMO configuration in a strong interference
environment, in this scenario we assume correlation exists
between the elements of each antenna. Results in both figures
present the BER performance of OSGD and ML-MUD for two
different correlation levels, where (ρ = 0.5 and 0.9) along with
the no correlation case, i.e., ρ = 0. All of the shown results
clearly prove that BER performance of OSGD is significantly
matches the performance of the ML-MUD scheme, especially
for the tight specification environments, i.e., for low SNR and
higher correlation levels. demonstrating by that the OSGD
capability to cope with interference.
I. Outage Probability
In this section we present the outage probability results
of the OSGD scheme for several transmit data rates, using
different MIMO configurations, with various correlation set-
tings, including the (uncorrelated, semi-correlated and fully
correlated) scenarios, also for the heterogeneous (Rayleigh and
Racian) flat fading environments. All the results shown in this
part are based on a three-user interference system channel,
with cross-talk factor α = 1, where, simulation results of each
setup are carried out using 10000 channel realizations.
Results depicted in figures (31, 32 and 33) show the outage
probability against the decibel SNR level, using (2 × 2, 3 × 3
and 4×4) Rayleigh MIMO configuration, for the transmit data
rate of (2, 3, and 4) bps/Hz respectively, in this case we assume
no correlation. From the figures we can perceive the increase
of the negative slope of the outage probability (which can be
appreciated as a diversity gain) as the MIMO size increase.
Also, from the multiplexing gain point of view, it can be
seen that OSGD efficiently exploits the MIMO dimension to
52 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

reduce the required power for a specific rate at a predetermined 0

10
outage level. For instance, at 10−3 outage probability for the 2×2, ρ = 0
3×3, ρ = 0
rate of 4 bps/Hz, the required power is reduced by 10 dB 4×4, ρ = 0
when the 2 × 2 MIMO replaced by a 3 × 3 system, further

Outage probability for a rate of: (2) bps/Hz

increase of the MIMO dimension to 4 × 4 yields more 5.5 −1
10
dB reduction in power to sustain the same data rate given the
aforesaid outage probability. From other point of view, Fig. 34
represents the effect of Rician-Rayleigh channel effect on the
outage probability of the OSGD, for this scenario, we assume −2
10
that, the channel between the receiver of interest and its desired
transmitter is modeled by Rician fading with κ factor equals 10
dB, and the channels between the mentioned receiver and the
other two undesired sources are modeled by Rayleigh fading. −3
10
−10 −8 −6 −4 −2 0 2 4 6 8
This assumption is reasonable as indicated in [18] for indoor SNR (dB)

and femtocell systems, where the receiver has a direct line of

sight component from its intended user and non-line of sight Fig. 31: OSGD outage probability performance versus SNR,
components from the interferers. By comparing the results in using rate of 2 bps/Hz, for different MIMO configurations,
Fig. 34 with that in Fig. 33, we do not notice any significant with cross-talk factor α = 1 and antenna inter-element
differences in the outage performance, except for the small correlation ρ = 0.
MIMO dimension, i.e., the (2 × 2), where we notice less than
1 dB power reduction in the case of existing line of sight 0
10
(L.O.S.) path between the receiver and its desired transmitter. 2×2, ρ = 0
3×3, ρ = 0
Now, to evaluate the effect of the antenna inter-element 4×4, ρ = 0
correlation on the outage probability of the OSGD, we assume Outage probability for a rate of: (3) bps/Hz

that antenna in each end has a strong inter-element correlation −1

10
ρ = 0.9, and we present results of the (2 × 2, 3 × 3 and 4 × 4)
Rayleigh fading used in this scenario in Fig. 35. Referring
again to Fig. 33, to compare its results with those in Fig. 35,
one can clearly see that antenna inter-element correlation −2
10
significantly affects the outage probability of the system. Also,
it can be noticed that, the bigger the MIMO dimension the
stronger the effect is. In addition, by contrasting the same two
figures once more, we notice that, when the correlation exists, −3
10
−6 −4 −2 0 2 4 6 8 10 12
it pushes the required power to achieve the predetermined SNR (dB)

outage percentage, say for example, Pout = 10−3 , then,

existing of correlation raises the required power by (6, 7 and
7.5) dB, for the (2 × 2, 3 × 3 and 4×) MIMO configuration re-
spectively. Finally, to show the effect of antenna inter-element
correlation more comprehensively, we choose to look over
this effect on the system from different point of views. Thus,
we evaluate the outage probability in four different cases:
(i)- Non-correlation case, (ii)- Strong correlation between the
elements of the receive antenna, and no correlation in the
transmit side, (iii)- Strong correlation between the elements
of the transmit antenna, and no correlation in the receive side,
(iv)- Strong correlation between the elements of each antenna
in the transmit and receive sides. This comparison is carried
out for two different MIMO dimensions, 2×2 with data rate of
4 bps/Hz and 8 × 8 with data rate of 16 bps/Hz. The results of
these cases are shown in figures (36 and 37). From the depicted
results, we can observe that, as expected, the best performance
is encountered in the non-correlation case, and the worst case
is when each antenna in both side has an inter-element cor-
relation, i.e., the fully correlated case. However, interestingly,
although we assume the same correlation intensity for the case
(ii) and (iii) which represent the semi-correlation cases, we
notice that, performance in case (ii), i.e., when the receive side
antenna elements are correlated, exhibits better performance
Fig. 32: OSGD outage probability performance versus SNR,
using rate of 3 bps/Hz, for different MIMO configurations,
with cross-talk factor α = 1 and antenna inter-element
correlation ρ = 0.
comparing with case (iii), i.e., when the transmit side antenna
elements are correlated. This may reflect the ability of OSGD
to overcome correlation to improve the system performance
as it is a receive-side processing technique.
VII. C ONCLUSION
To summarize our contribution in this study, we say that we
explored the OSGD in the interference channel environment.
In addition, we extensively investigated its performance based
on different criteria, among which, achieved ergodic and effec-
tive capacity. Also, we evaluated its minimum required energy
per bit, we also assessed its BER, and outage probability
performance, under different QoS constraints. For generality
purposes, this study considered the spatially correlated and
uncorrelated (Rayleigh and Rician) flat fading channels. More-
over, it investigated performance of the OSGD in different
SNR and SIR environments, where we considered both the
ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014 53

0 0
10 10
2×2, ρ = 0 2×2, ρ = 0.9
3×3, ρ = 0 3×3, ρ = 0.9
4×4, ρ = 0 4×4, ρ = 0.9
Outage probability for a rate of: (4) bps/Hz

Outage probability for a rate of: (4) bps/Hz

−1 −1
10 10

−2 −2
10 10

−3 −3
10 10
−4 −2 0 2 4 6 8 10 12 14 16 0 5 10 15 20 25
SNR (dB) SNR (dB)

Fig. 33: OSGD outage probability performance versus SNR, Fig. 35: OSGD outage probability performance versus SNR,
using rate of 4 bps/Hz, for different MIMO configurations, using rate of 4 bps/Hz, for different MIMO configurations,
with cross-talk factor α = 1 and antenna inter-element with cross-talk factor α = 1 and antenna inter-element
correlation ρ = 0. correlation ρ = 0.9.

0
0 10
10
2×2, κ = 10 dB, ρ = 0 2×2, ρTx = 0.9, ρRx = 0.9
3×3, κ = 10 dB, ρ = 0 2×2, ρTx = 0.9, ρRx = 0
4×4, κ = 10 dB, ρ = 0 2×2, ρTx = 0, ρRx = 0.9
Outage probability for a rate of: (4) bps/Hz
Outage probability for a rate of: (4) bps/Hz

2×2, ρTx = 0, ρRx = 0

−1
−1 10
10

−2
−2 10
10

−3
−3 10
10
−4 −2 0 2 4 6 8 10 12 14 16 8 10 12 14 16 18 20 22
SNR (dB) SNR (dB)

Fig. 34: OSGD outage probability performance versus SNR, Fig. 36: OSGD outage probability performance versus SNR,
using rate of 4 bps/Hz, for different MIMO configurations, using rate of 4 bps/Hz, for 2 × 2 MIMO configuration, with
with cross-talk factor α = 1 , antenna inter-element corre- cross-talk factor α = 1 and different antenna inter-element
lation ρ = 0, for the Racian-Rayleigh fading scenario with correlation ρ = 0 and 0.9.
κ − factor = 10 dB.

plexity controlling feature; we conclude that OSGD can be

power-limited and bandwidth-limited regimes with different a promising candidate for the detection strategy of the future
cross-talk values, as well as it considered various transmit- communication networks in the interference channel systems.
receive multiple-input multiple-output antenna configurations.
To make our results more sensible, we contrasted the outcomes R EFERENCES
of aforementioned performance assessments with the results
[1] J. Lee, D. Toumpakaris, and W. Yu, “Interference mitigation via joint
of other well known interference mitigation approaches, such detection in a fading environment,” in Information Theory and Applica-
as, the ML-MUD scheme and the MIL interference alignment tions Workshop (ITA), pp. 1 –7, Feb. 2011.
technique. Finally, from the findings of this work, which [2] T. Han and K. Kobayashi, “A new achievable rate region for the inter-
ference channel,” IEEE Transactions on Information Theory, Vol. 27,
clearly demonstrate the OSGD effectiveness to mitigate inter- No. 1, pp. 49–60, Jan. 1981.
ference; besides, being the OSGD a receiver-based processing [3] G. Kramer, “Review of rate regions for interference channels,” in 2006
scheme, thus, it does avoid the huge feedback overhead for International Zurich Seminar on Communications, pp. 162–165, 2006.
[4] X.-H. You, “Future: towards beyond 3G mobile communications,” in
FDD mode or any assumption of network reciprocity in Frontiers of Mobile and Wireless Communication, 2004. Proceedings of
TDD mode which is usually required for other multi-user the IEEE 6th Circuits and Systems Symposium on Emerging Technolo-
detection schemes. Also, because of its ability to adjust the gies:, Vol. 1, pp. 1–5, Jun. 2004, .
[5] V. Cadambe and S. Jafar, “Interference alignment and spatial degrees of
computational complexity to fit with the processing capability freedom for the k user interference channel,” in 2008. ICC ’08. IEEE
of the employed receivers as a result of its built-in com- International Conference on Communications, pp. 971–975, May 2008.
54 ALMADAR JOURNAL FOR COMMUNICATIONS, INFORMATION TECHNOLOGY, AND APPLICATIONS, VOL. 01, NO. 01, JULY 2014

0
support of quality of service,” IEEETransactions on Wireless Commu-
10
nications, Vol. 2, No. 4, pp. 630–643, 2003.
8×8, ρTx = 0.9, ρRx = 0.9
[21] M. Gursoy, “MIMO wireless communications under statistical queueing
8×8, ρTx = 0.9, ρRx = 0
constraints,” in Control, and Computing, 2009. Allerton 2009. 47th
8×8, ρTx = 0, ρRx = 0.9
Outage probability for a rate of: (16) bps/Hz

Annual Allerton Conference on Communication, 2009, pp. 1153–1160.

8×8, ρTx = 0, ρRx = 0
[22] L. Musavian and S. Aissa, “Effective capacity of delay-constrained
−1
10 cognitive radio in nakagami fading channels,” IEEE Transactions on
Wireless Communications, Vol. 9, No. 3, pp. 1054–1062, 2010.
[23] C.-S. Chang, “Stability, queue length, and delay of deterministic and
stochastic queueing networks,” IEEE Transactions on Automatic Con-
trol, Vol. 39, No. 5, pp. 913–931, 1994.
−2
10 [24] C. S. Chang, Performance Guarantees in Communication Networks.
London, UK, UK: Springer-Verlag, 2000.
[25] M. Gursoy, “MIMO wireless communications under statistical queueing
constraints,” IEEE Transactions on Information Theory, Vol. 57, No. 9,
pp. 5897–5917, 2011.
−3
10 [26] D. Tse and P. Viswanath, Fundamentals of Wireless Communication.
4 6 8 10 12 14 16 New York, NY, USA: Cambridge University Press, 2005.
SNR (dB)
[27] R. Adve, “MIMO systems and transmit diversity,” http://www.comm.
utoronto.ca/~rsadve/Notes/DiversityTransmit.pdf, 2007.
Fig. 37: OSGD outage probability performance versus SNR, [28] P. Viswanath, D. Tse, and R. Laroia, “Opportunistic beamforming using
using rate of 16 bps/Hz, for 8 × 8 MIMO configuration, with dumb antennas,” IEEE Transactions on Information Theory, Vol. 48,
cross-talk factor α = 1 and different antenna inter-element No. 6, pp. 1277–1294, 2002.
[29] P. Mohapatra, K. E. Nissar, and C. R. Murthy, “Interference alignment
correlation ρ = 0 and 0.9. algorithms for thek user constant mimo interference channel,” IEEE
Transactions on Signal Processing, Vol. 59, No. 11, pp. 5499–5508,
Nov. 2011.
[30] D. Seethaler, H. Artes, and F. Hlawatsch, “Detection techniques for
[6] R. Tresch, M. Guillaud, and E. Riegler, “On the achievability of interfer- MIMO spatial multiplexing systems,” Elektrotechnik und Information-
ence alignment in the k-user constant MIMO interference channel,” in stechnik (E & I), Vol. 122, No. 3, pp. 91–96, 2005.
2009. SSP ’09. IEEE/SP 15th Workshop on Statistical Signal Processing, [31] H. Dai, A. Molisch, and H. Poor, “Downlink capacity of interferencelim-
pp. 277–280, Sep. 2009. ited MIMO systems with joint detection,” IEEE Transactions on Wireless
[7] M. Yamada and T. Ohtsuki, “Interference alignment in the 2×(1+n) Communications, Vol. 3, No. 2, pp. 442–453, 2004.
MIMO model,” in GLOBECOM Workshops (GC Wkshps), 2010 IEEE, [32] P. Botsinis, S. X. Ng, and L. Hanzo, “Quantum search algorithms,
pp. 1963–1967, Dec. 2010. quantum wireless, and a low-complexity maximum likelihood iterative
[8] J. H. Lee and W. Choi, “Opportunistic interference aligned user selection quantum multi-user detector design,” Access, IEEE, Vol. 1, pp. 94–122,
in multiuser MIMO interference channels,” in GLOBECOM 2010, 2010 2013.
IEEE Global Telecommunications Conference, pp. 1–5, Dec. 2010.
[9] K. Gomadam, V. Cadambe, and S. Jafar, “Approaching the capacity of
wireless networks through distributed interference alignment,” in Global
Telecommunications Conference, 2008. IEEE GLOBECOM 2008. IEEE,
pp. 1–6, Dec. 2008, .
[10] A. Zanella, M. Chiani, and M. Win, “Mmse reception and successive
interference cancellation for MIMO systems with high spectral effi-
ciency,” IEEE Transactions on Wireless Communications, Vol. 4, No. 3,
pp. 1244–1253, May 2005.
[11] W. Yu, G. Ginis, and J . Cioffi , “ Distributed multiuser power
control for digital subscriber lines,” IEEE Journal on Selected Areas
in Communications, Vol. 20, No. 5, pp. 1105–1115, Jun. 2002.
[12] J. Wang and L. Milstein, “CDMA overlay situations for microcellu-
lar mobile communications,” IEEE Transactions on Communications,
Vol. 43, No. 234, pp. 603–614, Feb/Mar/Apr 1995.
[13] J. Lee, D. Toumpakaris, and W. Yu, “Interference mitigation via joint
detection,” IEEE Journal on Selected Areas in Communications, Vol. 29,
No. 6, pp. 1172–1184, Jun. 2011.
[14] N. Prasad and X. Wang, “Outage minimization and rate allocation
for the multiuser gaussian interference channels with successive group
decoding,” IEEE Transactions on Information Theory, Vol. 55, No. 12,
pp. 5540–5557, Dec. 2009.
[15] C. Gong, A. Tajer, and X. Wang, “Interference channel with con-
strained partial group decoding,” IEEE Transactions on Communica-
tions, Vol. 59, No. 11, pp. 3059–3071, 2011.
[16] C. Gong, O. Abu-Ella, X. Wang, and A. Tajer, “Constrained group de-
coder for interference channels,” JOURNAL OF COMMUNICATIONS,
Vol. 7, No. 5, pp. 382–390, May 2012.
[17] E. Biglieri, R. Calderbank, A. Constantinides, A. Goldsmith, A. Paulraj,
and H. V. Poor, MIMO Wireless Communications. New York, NY,
USA: Cambridge University Press, 2007.
[18] C. Siriteanu, S. D. Blostein, A. Takemura, H. Shin, and S. Yousefi,
“Exact performance analysis of MIMO zero forcing detection for
transmit correlated rician rayleigh fading,” CoRR, Vol. abs/1307.2958,
2013.
[19] C.-N. Chuah, D. Tse, J. Kahn, and R. Valenzuela, “Capacity scaling in
MIMO wireless systems under correlated fading,” IEEE Transactions on
Information Theory, Vol. 48, No. 3, pp. 637–650, 2002.
[20] D. Wu and R. Negi, “Effective capacity: a wireless link model for
 ‫مكتب البحث والتطوير‬
‫التعريف بمكتب البحث والتطوير ‪ -‬شركة المدار الجديد‬
‫النقاط البحثية الممولة خالل العام الجامعي ‪3102/3102‬‬
‫مشكلة التشويش بسبب أجهزة التشويش على الهاتف المحمول‬
‫الموجودة بالمساجد ‪(Interference Due to Mobile‬‬
‫)‪Jammers in Mosques‬‬
‫مشكلة صناديق الشفرات (‪)SIM Boxes‬‬
‫مشكلة حساب التكلفة (‪)Service Cost Modeling‬‬
‫تطوير منظومة ذكية إلعداد التقارير ‪(Developing‬‬
‫)‪Business Intelligence Report‬‬
‫تطبيقات المنزل الذكي )‪(Smart Home Applications‬‬
‫شركة المدار الجديد‬
‫‪‬‬ ‫شركة المدار الجديد هي إحدى شركات االتصاالت التابعة للشركة‬
‫الليبية لالتصاالت وتقنية المعلومات القابضة‪ .‬تأسست شركة المدار‬
‫الجديد سنة ‪ ،6991‬وبدأت في تقديم خدماتها للجمهور سنة ‪.6991‬‬
‫تعتبر شركة المدار الجديد أول مشغل للهاتف المحمول في ليبيا‪.‬‬
‫تغطي شبكة المدار الجديد حاليا ً ما يقارب من ‪ %99‬من المناطق‬
‫المأهولة بالسكان في ليبيا‪ ،‬مستعملة ما يقارب من ‪ 6199‬محطة‪،‬‬
‫تعمل أغلب هذه المحطات عند ترددي ‪ 900 MHz‬و ‪1800‬‬
‫‪ ،MHz‬بتقنية ‪ GSM‬وتطويراتها (‪ .)GPRS, EDGE‬يصل عدد‬
‫المشتركين بالشبكة إلى حوالي ‪ 4‬مليون مشترك‪.‬‬
‫‪‬‬ ‫قريبا ً ستشهد شبكة المدار الجديد تطوراً كبيراً‪ ،‬حيث سيتم تحديث‬
‫مكوناتها لدعم تقنيات الجيل الثالث المتطور (‪ )3.7G‬والجاهز لتقبل‬
‫تقنيات الجيل الرابع (‪ .)4G‬مزيد من التفاصيل حول هذا المشروع‬
‫الطموح تظهر في اإلصدار القادم من هذه المجلة‪.‬‬
‫مكتب البحث والتطوير‬
‫سعيا ً من شركة المدار الجديد في مواكبة البحث العلمي وتطوراته‪،‬‬
‫وربط الشركة مع المؤسسات التعليمية والبحثية داخل ليبيا وخارجها‪،‬‬
‫‪‬‬
‫تم استحداث مكتب خاص بالبحث والتطوير داخل الشركة‪ .‬بدأ هذا‬
‫المكتب نشاطاته منذ شهر سبتمبر الماضي‪ ،‬ويعمل المكتب كحلقة‬
‫وصل بين اإلدارات المختلفة بالشركة والجامعات والكليات والمراكز‬
‫البحثية داخل ليبيا وخارجها‪ .‬تشتمل نشاطات المكتب حاليا ً على‬
‫االتي‪:‬‬
‫القيام بأبحاث علمية وعملية في مجال نشاط الشركة‪.‬‬ ‫‪‬‬
‫‪‬‬ ‫عرض واستقبال مقترحات لنقاط بحثية في مجال عمل الشركة‬ ‫‪‬‬
‫وتبني تمويلها واإلشراف عليها‪ .‬يمكن تقسيم هذه النقاط البحثية‬
‫استناداً إلى مستوى العمل المتوقع فيها إلى‪:‬‬
‫‪ o‬نقاط بحثية على شكل مشاريع تخرج لطلبة البكالوريوس‪.‬‬
‫‪ o‬نقاط بحثية على شكل رسائل ماجستير‪.‬‬
‫‪ o‬نقاط بحثية علمية للباحثين وأعضاء هيئة التدريس‬
‫بالجامعات والكليات والمراكز البحثية‪.‬‬
‫اإلشراف على برنامج متكامل للتدريب الصيفي خاص بطلبة‬ ‫‪‬‬
‫‪‬‬ ‫المرحلة الجامعية‪.‬‬
‫اإلشراف على إصدار مجلة علمية محكمة تحت عنوان "مجلة‬ ‫‪‬‬
‫المدار لالتصاالت وتقنية المعلومات وتطبيقاتهما"‪.‬‬
‫لمزيد من التفاصيل حول نشاطات المكتب يمكن االطالع على الموقع‬
‫اإللكتروني لشركة المدار الجديد (‪.)www.almadar.ly‬‬
‫‪4‬‬
 ‫مجلة المدار لالتصاالت وتقنية المعلومات وتطبيقاتهما‬
‫عمله المستقبلي حتى يتم نشر هذه الورقة وبعد ذلك يستخدمها‬
‫كمرجع‪.‬‬
‫يتجنب المراجع مراجعة الورقة إذا كان هناك تضارب في‬
‫المصالح مع أحد الباحثين أو أن موضوع الورقة ال يكون من‬
‫ضمن تخصص المراجع‪.‬‬
‫قراءة الورقة بالتفصيل‪ ،‬والتحقق من المعادالت والخوارزميات‬
‫واألشكال‪ ،‬وتقييم المساهمات‪ ،‬والتحقق من قائمة المراجع‪،‬‬
‫ونسق ووحدة الورقة‪ ،‬والتحقق من لغة وتنظيم الورقة‪.‬‬
‫كتابة تقرير مفصل عن الورقة‪ ،‬ويوصي المراجع بأن تكون‬
‫نتيجة تقييم الورقة أحد الخيارات التالية‪:‬‬
‫‪ o‬قبول‪.‬‬
‫‪ o‬قبول بشروط‪.‬‬
‫‪ o‬طلب تعديالت جوهرية‪.‬‬
‫‪ o‬رفض‪.‬‬
‫حيث يبرر المراجع قراره لكل من هيئة التحرير والباحث في تقرير‬
‫مفصل‪.‬‬
‫ذكر جميع النقاط اإليجابية والسلبية للورقة في تقرير المراجع‪،‬‬
‫كما يتم اإلشارة إلى االحتيال المشبوه أو االنتهاك لحقوق النشر‬
‫في تقرير خاص بهيئة التحرير‪.‬‬
‫لهيئة التحرير الحق في رفض تقرير المراجع إذا لم يقدم تبريرات‬
‫وتعليقات مقنعة في تقريره‪.‬‬
‫تستكمل عملية المراجعة خالل ‪ 6‬أسابيع من تاريخ استالم طلب التقييم‬
‫والمراجعة‪ ،‬ويمكن تمديد هذه المدة أسبوعين إضافيين كحد أقصى إذا‬
‫طلب المراجع ذلك‪ ،‬وسيتم رفض تقرير المراجع عند وصوله بعد‬
‫فترة تزيد عن ‪ 8‬أسابيع‪.‬‬
‫مكافآت تشجيعية‬
‫في إطار نشر ثقافة البحث العلمي‪ ،‬وتشجيع إنتاج أوراق علمية رفيعة‬
‫المستوى تم تخصيص المكافآت التالية‪:‬‬
‫مكافأة مالية قدرها ‪ 011‬د‪.‬ل (خمسمائة دينار ليبي) كمنحة لكل‬
‫ورقة علمية مقبولة للنشر‪.‬‬
‫مكافأة مالية قدرها ‪ 0111‬د‪.‬ل (خمسة آالف دينار ليبي)‬
‫وشهادة رسمية كجائزة ألفضل ورقة سنويا ً‪ ،‬ويتم تحديد أفضل‬
‫ورقة سنويا ً بناء على تقييم المراجعين وهيئة تحرير المجلة‪.‬‬
‫حيث أن عملية تقييم ومراجعة األوراق العلمية المستلمة للنشر‬
‫بالمجلة تستلزم مجهودات كبيرة من قبل السادة المراجعين‪،‬‬
‫وعرفانا ً بهذه المجهودات تصرف مكافأة مالية لكل مراجع عن‬
‫كل مقالة يقوم بمراجعتها للمجلة بحيث ال تتجاوز قيمة هذه‬
‫المكافأة ‪ 401‬دينار ليبي (مائتان وخمسون دينار ليبي)‪ .‬يمكن‬
‫أن تصرف هذه المكافأة على شكل قيمة نقدية أو على شكل هدية‬
‫عينية أو رصيد مكالمات هاتفية من شركة المدار الجديد لرقم‬
‫يحدده المراجع‪.‬‬
‫يجوز لهيئة التحرير إصدار عدداً استثنائيا ً إذا تراكمت‬ ‫‪‬‬
‫المشاركات أو تطلبت أهمية االبحاث سرعة نشرها‪ .‬كما يجوز‬
‫‪‬‬ ‫إصدار عدداً خاصا ً يتناول األبحاث التي تعالج موضوعا ً بعينه‪.‬‬
‫البحث المنشور في المجلة يصبح ملكا ً لها ويؤول إليها حق النشر‪،‬‬
‫‪‬‬ ‫لكن يحق للباحث إعادة نشر بحثه في كتاب وفي هذه الحالة ال بد أن‬
‫يشير إلى المصدر األصلي للنشر‪.‬‬
‫إعداد الورقات‬
‫‪‬‬
‫يمكن للسادة الراغبين في نشر أعمالهم بالمجلة إرسال أعمالهم إلى‬
‫هيئة التحرير على شكل ملف في صيغة ‪ ،pdf‬وإذا تم قبول العمل‬
‫للنشر يجب على الباحث إعادة إرساله في الصيغة التي تطلبها هيئة‬
‫التحرير مع االلتزام بالمواصفات الواردة بموقع المجلة‪.‬‬
‫آلية التقديم‬
‫تكون جميع المراسالت بين الباحثين وهيئة تحرير المجلة عبر البريد‬
‫اإللكتروني ‪.ajcita@almadar.ly‬‬
‫‪‬‬
‫اختيار الورقات المناسبة للنشر‬
‫يتم الفحص األولي لألوراق المقدمة للنشر في المجلة من قبل هيئة‬
‫التحرير للتأكد من أنها تتبع المتطلبات الرئيسية للمجلة من حيث‬
‫المواضيع واألسلوب‪ ،‬وفي حالة تحقق المتطلبات الرئيسية يتم إرسالها‬
‫إلى ثالثة مراجعين على األقل لتقييم الورقة ومراجعتها بعد طمس‬
‫أسماء المؤلفين‪.‬‬
‫رفض الورقات‬
‫يتم رفض الورقة المقدمة إذا توفرت إحدى األسباب التالية‪:‬‬
‫الورقة ال تقدم أي مساهمة قيمة‪ ،‬على سبيل المثال‪ ،‬ورقة تعرض‬ ‫‪‬‬
‫نتائج معروفة إال أن تضيف جديداً إلى المكتبة العربية في‬
‫المجال كما سبق ذكره‪.‬‬
‫‪‬‬ ‫الورقة تحتوي على أخطاء في المفاهيم البديهية أو االستنتاجات أو‬ ‫‪‬‬
‫التحليل‪.‬‬
‫‪‬‬ ‫الورقة تحتوي على احتيال وانتهاك واضح لحقوق النشر‪.‬‬ ‫‪‬‬
‫الورقة تحتوي على الكثير من األخطاء اللغوية والمطبعية‪.‬‬ ‫‪‬‬
‫موضوع الورقة ليس ضمن اهتمامات المجلة‪.‬‬ ‫‪‬‬
‫‪‬‬
‫المراجعون‬
‫تقوم هيئة التحرير بإعداد قاعدة بيانات للمراجعين في كل تخصص‬
‫دقيق‪ ،‬ويتم إرسال الورقة دون ذكر أسماء المؤلفين حفاظا ً على سرية‬
‫التقييم ونزاهته‪ ،‬ومن المتوقع أن يقوم المراجع باآلتي‪:‬‬
‫مراجعة الورقة بسرية تامة‪ ،‬بحيث أن المراجع ال يرسل الورقة‬ ‫‪‬‬
‫ألي طرف ثالث وال يستخدم األفكار الجديدة في هذه الورقة في‬
‫‪3‬‬
 ‫مجلة المدار لالتصاالت وتقنية المعلومات وتطبيقاتهما‬
‫ينبغي أن يستند ترتيب قائمة أسماء الباحثين على نسبة العمل المنجز‪،‬‬
‫وذلك بأن يكون االسم األول للباحث األكثر مساهمة في الورقة‪ ،‬ثم‬
‫الباحث الذي يليه وهكذا‪ ،‬وال يتم قبول ورقة تشتمل على أسماء في‬
‫قائمة الباحثين يثبت أنه ليس لديهم مساهمة في الورقة‪.‬‬
‫يجب ذكر مصادر تمويل البحوث المقدمة للنشر في المجلة (شركة‪،‬‬
‫جامعة‪ ،‬منحة التمويل‪ ،‬الخ) وذلك باإلشارة لها داخل الورقة كنوع من‬
‫الشكر والتقدير‪.‬‬
‫يجب أال تكون الورقات المقدمة للنشر في المجلة قد سبق أن نشرت‬
‫في أي مجلة أو مؤتمر آخر‪ ،‬وأال تكون تحت المراجعة الحالية من‬
‫قبل مؤتمر أو مجلة أخرى‪ ،‬كما أنه سيتم توقيع اتفاق نقل حقوق النشر‬
‫لألوراق المقبولة قبل نشرها‪.‬‬
‫يجوز تقديم ورقات نشرت جزئيا ً من قبل في مجالت أو مؤتمرات‬
‫أخرى‪ ،‬بحيث تكون هناك مساهمة جديدة واضحة في األوراق‬
‫المقدمة‪.‬‬
‫الباحثون في الورقة المقدمة مسئولون عن جميع القضايا والدعاوى‬
‫المتعلقة بحقوق النشر والملكية الفكرية التي تثار من الغير ضد المجلة‬
‫ويتحملون كافة التبعات القانونية المترتبة بالخصوص دون أدنى‬
‫مسئولية من المجلة‪.‬‬
‫يجب على الباحثين احترام أخالقيات الكتابة والنشر وأي احتيال‬
‫جزئي أو كلي لعمل قام به باحثون آخرون سوف يعتبر جريمة علمية‬
‫وفي هذه الحالة سوف يتم اتخاذ اإلجراءات التالية‪:‬‬
‫‪ ‬رفض نشر هذه الورقة‪.‬‬
‫‪ ‬يجب على الباحثين المشاركين في هذه الورقة تقديم تقرير‬
‫توضيحي حول هذه الحالة لمعرفة المسؤول عن هذا العمل‪.‬‬
‫‪ ‬حرمان الباحثين المسؤولين عن عملية االحتيال من التعاون‬
‫واالشتراك مرة أخرى في المجلة‪.‬‬
‫‪ ‬يحق لرئيس تحرير المجلة إبالغ الجمعيات والمؤسسات الوطنية‬
‫والدولية وهيئات تحرير الدوريات العلمية األخرى حول هذه‬
‫الجريمة‪.‬‬
‫‪ ‬إذا تم اكتشاف احتيال في وقت الحق بعد نشر العمل‪ ،‬سوف يتم‬
‫سحب الورقة ونشر مذكرة االحتيال‪.‬‬
‫‪ ‬يجوز للباحثين االعتراض على أي إجراء عند هيئة تحرير‬
‫المجلة عن طريق تقديم اعتراض كتابي موجه إلى رئيس هيئة‬
‫التحرير‪ ،‬وفي هذه الحالة فإنه يجوز لهيئة التحرير استشارة‬
‫خبراء في هذا المجال للحكم في المسألة‪.‬‬
‫يراعى في أولوية النشر ما يلي‪:‬‬
‫‪ ‬تاريخ استالم البحث وتعطى األسبقية للبحوث التي طلب إجراء‬
‫تعديالت عليها‪.‬‬
‫‪ ‬تنوع األبحاث والباحثين لتحقيق التوازن بحيث تنشر المجلة‬
‫ألكبر عدد من الكتاب ومن أكبر عدد ممكن من األقطار في‬
‫العدد الواحد وبأوسع مدى من التنوع‪.‬‬
‫‪ ‬المواضيع التي تخص الشأن الليبي أو اهتمام الشركة‪.‬‬
‫‪2‬‬
‫عمل هيئة تحرير المجلة‬
‫تجتمع هيئة التحرير مرة واحدة كل ثالثة أشهر على األقل أو كلما‬
‫دعت الحاجة إلى ذلك بدعوة من رئيس التحرير أو سكرتيره‪.‬‬
‫يعفى أي عضو من أعضاء هيئة التحرير من عضوية الهيئة إذا تخلف‬
‫عن حضور ثالثة اجتماعات متتالية أو خمسة اجتماعات متفرقة على‬
‫مدار السنة بدون عذر مقبول‪.‬‬
‫تنشر أسماء أعضاء هيئة التحرير والسادة المشاركين في التقييم‬
‫والمراجعة في صدارة كل عدد يصدر من المجلة أو حسب ما تقتضيه‬
‫طبيعة اإلخراج‪.‬‬
‫تمارس هيئة تحرير المجلة االختصاصات التالية‪:‬‬
‫تسلم الدراسات والبحوث المعروضة للنشر وإقرار تحويل ما‬ ‫‪‬‬
‫يصلح منها للتقييم والمراجعة‪.‬‬
‫رسم سياسة المجلة وإقرار المحاور التي تهتم بها‪.‬‬ ‫‪‬‬
‫تحديد سعر المجلة ورسوم االشتراك وطرق توزيعها وإهدائها و‬ ‫‪‬‬
‫كل ما من شأنه تسيير أمورها‪.‬‬
‫تطوير المجلة ومراجعة سياسات التحكيم والنشر وتقييم‬ ‫‪‬‬
‫المستوى العلمي واألكاديمي لها‪.‬‬
‫يمارس رئيس هيئة التحرير االختصاصات التالية‪:‬‬
‫رئاسة اجتماعات هيئة التحرير‪.‬‬ ‫‪‬‬
‫تمثيل المجلة أمام الهيئات الرسمية والقضائية‪.‬‬ ‫‪‬‬
‫اإلشراف على األمور التحريرية بالمجلة‪.‬‬ ‫‪‬‬
‫تلقي المراسالت التي ترد إلى المجلة والرد عليها‪.‬‬ ‫‪‬‬
‫مراجعة كافة مواضيع المجلة قبل تحويلها إلى الطباعة‪.‬‬ ‫‪‬‬
‫يمارس سكرتير التحرير االختصاصات التالية‪:‬‬
‫إعداد جدول اجتماعات هيئة التحرير‪.‬‬ ‫‪‬‬
‫الدعوة الجتماعات هيئة التحرير وتحرير المحاضر وتسجيلها‪.‬‬ ‫‪‬‬
‫مساعدة رئيس التحرير في أعماله والنيابة عنه في حالة غيابه‬ ‫‪‬‬
‫واإلشراف على متابعة قرارات هيئة التحرير‪.‬‬
‫متابعة طباعة المجلة وال سيما ما يتعلق منها بترتيب المواضيع‬ ‫‪‬‬
‫ومواضيع الغالف‪.‬‬
‫قواعد النشر‬
‫تعتمد الموافقة على نشر المقاالت البحثية على الثقة‪ ،‬حيث أنَّ القراء‬
‫يثقون في إجراءات محرري المجلة لتحديد الورقات األكثر أهمية‬
‫ذات المستوى الرفيع‪ ،‬كما أنَّ هيئة التحرير تعطي الثقة للمراجعين‬
‫بأنهم سيعطون الوقت والجهد للتأكد من جميع المعادالت واألشكال‬
‫في كل الورقات‪ ،‬وكذلك الباحثون يثقون في صدق المراجعين‬
‫وعدلهم في اختيار الورقات التي سيتم نشرها‪.‬‬
‫يجب على الباحثين أن يقدموا أفضل ملخص لعملهم األصلي دون‬
‫وجود أي احتيال أو انتهاك لحقوق الطباعة والنشر‪.‬‬
 ‫مجلة المدار لالتصاالت وتقنية المعلومات وتطبيقاتهما‬

‫التعريف بالمجلة‬
‫مجاالت اهتمام المجلة‬ ‫أهداف المجلة‬

‫تغطي المجلة مجموعة واسعة من المواضيع المتعلقة بهندسة‬ ‫مجلة المدار لالتصاالت وتقنية المعلومات وتطبيقاتهما ( ‪ISSN‬‬
‫االتصاالت وتقنية المعلومات‪ ،‬وتطبيقاتهما‪ ،‬فعلى سبيل المثال ال‬ ‫‪( )2313-156X‬رقم إيداع بدار الكتب الوطنية ‪ )4102\963‬هي‬
‫الحصر تقع المواضيع التالية ضمن اهتمام المجلة‪ :‬تقنيات االتصاالت‬ ‫مجلة علمية محكمة متخصصة في مجالي االتصاالت وتقنية‬
‫الالسلكية ونمذجة القناة‪ ،‬الشبكات الالسلكية‪ ،‬االتصاالت المتنقلة‪،‬‬ ‫المعلومات وتطبيقاتهما‪ ،‬تصدر نصف سنويا ً عن مكتب البحث‬
‫الشبكات المتنقلة ذات النطاق العريض‪ ،‬الهوائيات‪ ،‬معالجة اإلشارات‪،‬‬ ‫والتطوير بشركة المدار الجديد بدعم الشركة ورعايتها‪ .‬تطمح هذه‬
‫شبكات الحاسوب‪ ،‬األتمتة الالسلكية‪ ،‬تطبيقات تقنية المعلومات على‬ ‫المجلة إلى تحقيق األهداف التالية‪:‬‬
‫شبكات المحمول‪ ،‬شبكات الحماية‪ ،‬اتصاالت األلياف البصرية‪،‬‬
‫شبكات األقمار االصطناعية‪ ،‬تطبيقات تقنية المعلومات واالتصاالت‬ ‫تشجيع ونشر ثقافة البحث العلمي في مجاالت عمل الشركة‪.‬‬ ‫‪‬‬
‫في نظم الطاقة‪ ،‬المدن الذكية‪ ،‬تحديد المواقع باستخدام شبكات الهاتف‬ ‫توفير مرجع علمي رصين ذو موثوقية عالية يخدم الطالب‬ ‫‪‬‬
‫المحمول‪ ،‬شبكات االستشعار‪ ،‬وغيرها‪.‬‬ ‫الدارسين والمهندسين العاملين والباحثين واألكاديميين‬
‫بالجامعات والمراكز البحثية والصناعية والمؤسسات ذات‬
‫هيئة تحرير المجلة‬ ‫العالقة داخل ليبيا وخارجها‪.‬‬
‫المساهمة في تنمية العلوم الهندسية والتقنية وتطبيقاتها من خالل‬ ‫‪‬‬
‫المشرف العام‬ ‫نشر البحوث النظرية والتطبيقية مع التأكيد على الجودة العالية‬
‫لهذه البحوث وارتباطها بالواقع حاضراً ومستقبالً‪.‬‬
‫عبدهللا علي أعبودة‪ ،‬مدير مكتب البحث والتطوير بشركة المدار‬ ‫إتاحة الفرصة للمبدعين والباحثين في المجاالت المستهدفة بنشر‬ ‫‪‬‬
‫الجديد‪ ،‬ليبيا‪ ،‬البريد اإللكتروني‪a.abouda@almadar.ly :‬‬ ‫نتاج أنشطتهم العلمية والبحثية‪.‬‬
‫المساهمة في إقامة شبكة تعاون علمي بحثي أكاديمي‪.‬‬ ‫‪‬‬
‫رئيس التحرير‬
‫متابعة النشاطات العلمية في مجال هندسة االتصاالت وتقنية‬ ‫‪‬‬
‫محمد سالم المصراتي‪ ،‬أستاذ ورئيس مجموعة االتصاالت والنظم‬ ‫المعلومات في العالم عامة وفي ليبيا خاصة‪.‬‬
‫الهندسية بجامعة فازا‪ ،‬فنلندا‪ ،‬وكذلك أستاذ بقسم الهندسة الكهربائية‬
‫واإللكترونية‪ ،‬جامعة بنغازي‪ ،‬ليبيا‪ ،‬البريد اإللكتروني‪:‬‬
‫محتويات المجلة‬
‫‪mohammed.elmusrati@uwasa.fi‬‬
‫يتم طباعة نسخ ورقية بكمية محدودة لكل إصدار من المجلة مع توفير‬
‫النسخة اإللكترونية منه عبر الموقع اإللكتروني للشركة‪ .‬وتشمل‬
‫أعضاء هيئة التحرير‬
‫محتويات المجلة ما يلي‪:‬‬
‫محمد جمعة الترهوني‪ ،‬أستاذ ورئيس قسم الهندسة الكهربائية‪،‬‬
‫ورقات علمية محكمة من قبل نخبة من أهل االختصاص‪.‬‬ ‫‪‬‬
‫الجامعة األمريكية في الشارقة الشارقة‪ ،‬اإلمارات العربية المتحدة‪،‬‬
‫مقاالت علمية مدعوة لخبراء في مجالي االتصاالت وتقنية‬ ‫‪‬‬
‫البريد اإللكتروني‪mtarhuni@aus.edu :‬‬
‫المعلومات وتطبيقاتهما‪.‬‬
‫علي أحمد قنون‪ ،‬أستاذ مساعد بقسم الهندسة الكهربائية واإللكترونية‪،‬‬ ‫أخبار ومستجدات قطاع االتصاالت وتقنية المعلومات في ليبيا‪.‬‬ ‫‪‬‬
‫جامعة طرابلس‪ ،‬ليبيا‪ ،‬البريد اإللكتروني‪:‬‬ ‫إعالنات تجارية ذات طابع تقني يقع ضمن اهتمام المجلة من‬ ‫‪‬‬
‫مختلف المؤسسات والشركات العامة والخاصة‪.‬‬
‫‪ali.ganoun@ee.edu.ly‬‬ ‫إعالنات ودعايات لنشاطات مختلفة في المجاالت ذات العالقة‬ ‫‪‬‬
‫باهتمامات المجلة كالمؤتمرات وورش العمل والمسابقات‪.‬‬
‫مجدي علي الشيباني‪ ،‬أستاذ مشارك بكلية التقنية الصناعية‪ ،‬ليبيا‪،‬‬
‫البريد اإللكتروني‪mashibani@lptic.ly :‬‬ ‫لغة النشر في المجلة‬
‫محمد مرعي العماري‪ ،‬أستاذ مشارك ورئيس قسم هندسة البرمجيات‪،‬‬
‫تعتمد المجلة لغتان للنشر هما اللغة اإلنجليزية واللغة العربية‪ .‬يفترض‬
‫جامعة بنغازي‪ ،‬ليبيا‪ ،‬البريد اإللكتروني‪:‬‬
‫في المقاالت المقدمة للنشر باللغة اإلنجليزية أن تقدم مشاركة جديدة‬
‫‪mohamed.elammari@uob.edu.ly‬‬ ‫في مجال العمل‪ ،‬أما المقاالت المقدمة للنشر باللغة العربية فيمكن أن‬
‫تكون عبارة عن مقاالت تعريفية بتقنيات حديثة ال يتوفر عنها مصادر‬
‫عبدالقادر الصادق عكي‪ ،‬أستاذ بقسم الهندسة الكهربائية واإللكترونية‪،‬‬ ‫باللغة العربية‪.‬‬
‫جامعة طرابلس‪ ،‬ليبيا‪ ،‬البريد اإللكتروني‪asakki@lttnet.net :‬‬

‫‪1‬‬