IJDIWC

IJDIWC
International Journal of
ISSN 2225-658X (Online)
DIGITAL INFORMATION AND WIRELESS COMMUNICATIONS
Volume 3, Issue 2
2013
TABLE OF CONTENTS
Original Articles
PAPER TITLE
AUTHORS
Brendan Flanagan, Takahiko
PEER KNOWLEDGE ASSISTED SEARCH USING COMMUNITY SEARCH Suzuki, Jun Zeng, Chengjiu Yin,
LOGS
Toshihiko
Sakai,
Kiyota
Hashimoto, Sachio Hirokawa
PAGES
141
MODIFIED PROPORTIONAL FAIR SCHEDULING FOR RESOURCE

REUSE AND INTERFERENCE COORDINATION IN TWO-TIER LTE- Rony Kumer Saha
ADVANCED SYSTEMS
149
A MATLAB BASED
LABORATORY
169
CELLULAR
MOBILE
COMMUNICATION Rony Kumer Saha, A B M

Siddique Hossain
AN OPTIMIZED ENERGY AWARE CHAINING TECHNIQUE FOR DATA Ouadoudi

Zytoune,
GATHERING IN WIRELESS SENSOR NETWORKS
Aboutajdine
Driss
181
Lakshmi
Ramachandran,
AN INTERSECTION BASED TRAFFIC AWARE ROUTING WITH LOW
Sangheethaa Sukumaran, Surya
OVERHEAD IN VANET
Rani Sunny
190
AN ALGORITHMIC APPROACH TO DATABASE NORMALIZATION
197
M. Demba
International Journal of Digital Information and Wireless Communications (IJDIWC) 3(2): 141-148
The Society of Digital Information and Wireless Communications, 2013 (ISSN: 2225-658X)
Peer Knowledge Assisted Search Using Community Search Logs
Brendan Flanaganb, Takahiko Suzukia*, Jun Zengb, Chengjiu Yina, Toshihiko Sakaib,
Kiyota Hashimotoc and Sachio Hirokawaa
a
Research Institute for Information Technology, Kyushu University, Japan
b
Graduate School of Information Science, Kyushu University, Japan
c
School of Humanities and Social Sciences, Osaka Prefecture University, Japan
*6-10-1, Hakozaki, Higashi, Fukuoka, Japan
*suzuki@cc.kyushu-u.ac.jp
ABSTRACT
Search engines have become an increasingly
important educational tool. By just typing a
few keywords into a search engine text-box,
finding information has become an easy
exercise. However, some people cannot use
search engines effectively for various
reasons. Students may not know how to
choose proper keywords. They may have
difficulty in selecting relevant information
from millions of search results. In this paper,
we propose a search engine system that
shares search queries and browsing history
among students. We refer to the search
queries and browsing history as community
search logs. This system analyzes these
community search logs, and shares the
students knowledge and experience with
each other. Our purpose is to help students
to improve their searching efficiency,
especially those who are not good at
searching.
KEYWORDS
Community Search Logs, Search Engine,
Knowledge and Experience Sharing, Peer
Knowledge Assisted Search.
1 INTRODUCTION
Along with the development of the
Internet and search engine technology,
search engines have become an
increasingly important educational tool.
Researchers have indicated that allowing
students to learn as active and self-
directed participants is one of the

greatest benefits of web-based learning
activities,
which
often
involve
information searching tasks [1, 2].
Marchionini [3] and Dias et al. [4] have
reported the difficulty that novice
Internet users often encounter when they
try to search for information on the web
effectively and efficiently. Training
novice users in information searching
skills,
including abstracting and
summarizing the collected information
from the Internet, has become an
important and challenging issue [5].
Search engines are programs that search
documents for specified keywords and
return a list of the documents where the
keywords were found. However, the
term is often used to specifically
describe systems such as Google, Bing
and Yahoo! Search that enable users to
search for documents on the World Wide
Web. Nowadays search engines have
become an increasingly important tool.
Finding information has become an easy
exercise by just typing a few keywords
into a search engine text-box. Some
people can effectively use search
engines and always find the most
suitable keywords to use that return the
most relevant search results. This can be
of great advantage to their work and life.
However, not everyone can use search
engines effectively. Users may not know
how to choose proper keywords. For
example, when looking for science
141
experiments, if School Science Fair
Projects is entered instead of science
projects, this will return more specific
information.
Users may also have difficulty in
selecting the relevant information from a
large number of search results. Although
most search engines employ methods to
rank results in order to provide the
best results first, not all search results
that are ranked high are relevant to the
users needs.
We propose a search engine system that
shares search queries and browsing
history among a group of users. We refer
to the search queries and browsing
history as community search logs.
This system analyzes these community
search logs, and shares the users
knowledge and experience with each
other. Our purpose is to help students to
improve their searching efficiency,
especially those who are not good at
searching.
This paper is structured as follows:
Section 2 is a brief introduction of the
related works. In section 3, we explain
the preparation of data and a preliminary
experiment. Section 4 describes the
proposed system in detail. Section 5
describes our conclusions and future
work.
2 RELATED WORKS
Previous research has been conducted on
using the access logs of Web pages.
Toda et al. [6] focused on Web pages
that were often visited by many users in
order to improve the efficiency of Web
browsing routines as well as the
discovery of new information. In order
to evaluate the importance of the pages
for recommendation, they analyzed the
time and frequency of users who visited
particular pages in access logs. This was
then used to provide collaborative

filtering for Web page recommendation.
Nakao et al. [7] used the web page
access logs of each user to calculate the
similarity of Web pages. They then
proposed a Web page recommendation
method based on similarity. They
considered not only the similarity of
Web pages, but also the similarity of the
sub-trees in which the pages are located.
These two methods were then combined
in a link structure.
So far great importance has been placed
on qualitative analysis. Recently,
learning logs have become popular with
many researchers using them to enable
communication between users. Gotoda
et al. [8] used the activity logs of runners
in creating regional and focused
communities of runners. Any member
can read and comment on the training
plan of other members. This enables
participating members of the community
to share and learn from each others
activity logs even if they are
experienced.
Watanabe et al. [9] used collaborative
learning logs to help participants in a
team to solve the same problem. They
evaluated the number of users who made
the same annotation of their utterances,
and constructed a knowledge graph.
They claimed that the graph is useful to
point out key issues for learners.
Anderson et al. [10] designed and
implemented a montage system to
improve the experience of routine web
browsing that users tend to repeat over
and over in similar situations by
providing a start page. The start page
shows an ensemble of links and content
based on the user's browsing log and
preferences.
Prior study by X Wu et al. [11] proposed
a community based search platform
which combines a search engine and
142
SNS search. By seamless use of various
activities performed while searching and
mutual comments, users can share
information on problems or new
knowledge.
3 DATA PREPARATION AND
PRELIMINARY EXPERIMENT
Two kinds of data are utilized in the
proposed system: a database containing
search targets that we refer to as search
data, and the users search query and
browsing history that we refer to as
community search logs. In this
section, we will introduce the procedure
used to collect both of these kinds of
data.
3.1 Collecting Search Data
Information on papers about mobile
learning was collected as the search
data. The keywords mobile learning
was used to conduct a search using
Scopus 1 , which is the worlds largest
abstract and citation database of peerreviewed literature. As a result, we
collected the information of 13353
papers. Each paper contains the
following components:
Title of the paper
Authors of the paper
Departments, organizations and
countries of the authors
Publisher
Publication date
Keywords
Abstract
A frequency file of the words appearing
in those components was generated.
Finally, this frequency file was then used
to build an index using GETA2, which
1
2
http:// www-scopus-com
http://geta.ex.nii.ac.jp/
is a generic engine for transposable

association.
3.2 Preliminary Experiment
At the initial stage of research no
community search log data was available.
A
preliminary
experiment
was
conducted to create community search
logs by bootstrapping. In this experiment,
we used another system called the
Milky Way search engine, which was
introduced in paper [12]. This system
uses the same experimental data as the
aforementioned search data.
The participants were two sophomore
student classes of a north Taiwan
university. A total of 69 students
participated in the study, including 25
females and 44 males with an average
age of 20 years old. One class was
assigned to be the experimental group,
and the other was the control group. The
experimental group included 36 students
(12 females and 24 males), while the
control group had 32 students (13
females and 20 males). In order to avoid
any effect on the experiment results that
may be caused by different instructors,
the two classes were taught by the same
instructor. The experimental group
learned with the Milky Way search
engine, while those in the control group
learned with a search engine that has a
conventional interface (i.e., displaying
the searched results as a list of document
titles and abstracts). Both groups
received the information searching and
summarizing instructions for analyzing
the research trends of mobile learning
articles before the learning activity. The
details of the Milky Way system are
reported in [12].
In general, the Milky Way could help
users in their learning activities by
providing various search hints. However,
143
some novice users might complain that
they found it difficult to utilize those
various search hints.
3.3 Collecting Community Search
Logs
The search queries and browsing history
of the experimental group were collected
as the log data. As a result, 84 search
queries and 192 browsed web page
histories were collected. Figure 1 shows
the segments of the search queries. Each
search query contains the following:
The number of users who used
the query
The keywords of the query
Figure 2 shows the segments of log data.
Each entry of the log contains the
following:
The IP address of the PC used by
the participants
The date and time when the
participants browsed the search
results
Paper number (1~13353)
Keywords used by participants
22 mobile learning
17 mobile learning y:2010
9 mobile learning device y:2010
7 mobile learning y:2001
7 mobile learning c:Taiwan
Figure 1. Segments of search data
111.249.47.11,201201041255,4230,y:2011 mobile learn

133.5.7.108,201201041256, 5109,mobile learn ubiquitous
163.14.7.115,201201101200,8413,mobile learning c:Japan
163.14.7.82,201201101209,8523,mobile learning
Figure 2. Segments of log data
4 THE PROPOSED SYSTEM

4.1 Design Goal
Figure 3 shows the design goal of the
proposed system. The system shares the
search query history and browsing
activities each user has conducted so far
in order to help other users improve their
searching efficiency. When a user
conducts a search using the search
engine, his/her query and browsing
history are recorded as data reflecting
their search behaviors. The users search
behaviors can be influenced by his/her
knowledge, experience and other factors
[13]. In other words, sharing search
queries and browsing history is also
sharing the users knowledge and
experience.
Figure 3. Design goal of our system
The effect of just one users search

queries and browsing history is limited.
However by using the searching history
from a large number of users can not
only provide other users with
recommendations, but can also provide
basic data to perform statistical analysis
of users search behaviors. When used in
this way, search engines will become
large data warehouse of users
behaviors.
144
However, at the initial stage of the
research only the log data mentioned in
section 3 was utilized, and no new log
data was created while users were using
the proposed system.
4.2 The Functions and User Interface
Of The Proposed System
search without having to

inputting
a
search
query
manually.
There is a back to samples link at the
top of frame A that redirects the user
back to sample list page as shown in
Figure 5.
This section introduces the functions and

user interface of the proposed system.
Figure 4 shows the main user interface
containing three frames.
back to samples
Figure 5. Search query list in frame A
B
C
Figure 4. User interface of the proposed system
1) Frame A displays the search results.

This can also be used to conduct
keyword-based searches. The search
results will be displayed below the query
text-box. Each search result contains the
following parts:
Paper Number (1~13353)
Title of the paper
(stars) indicate how many
users have selected the paper
(circles) indicate how many
other papers were browsed in the
past search of the paper.
This page not only displays the
search queries, but also how
many times the queries have
been used. When the links of
queries are clicked, the system
will automatically conduct a
2) Frame B displays the related papers

and users: When the of a paper in
frame A is clicked, frame B will display
the list of the users who have browsed
the paper. In frame B, the following
items are displayed:
User ID
The number of times that the user
selected the paper.
When the of a paper in frame A is
clicked, frame B will display the papers
that were browsed at the same time as
the paper. In frame B, the following
items are displayed:
Paper Number (1~13353)
The times a paper has been
browsed
Title of the paper
3) Frame C displays the detail of a
paper: When the paper title in frame A is
clicked, frame C will display the details
of the paper. The following items are
displayed in frame C:
Title of the paper
Authors
Departments, organizations and
countries of the authors
145
Publisher
Publication date
Keywords
Abstract
There is a like button at the top of
frame C as shown in Figure 6. Users can
express their interest in the search result
document by clicking the button.
Figure 7. Search results in frame A.
Analyzing the related users: When

one of the associated with paper
8523 is clicked, frame B displays all
the IDs of the users who browsed the
paper. Figure 8 shows a list of the users
in frame B. The number of users in
frame B is equal to the number of
display next to paper 8523 in frame A.
The number in the parentheses following
the user ID indicates how many times
the user browsed the paper.
Figure 6. Details of frame C
4.3 An Example Of Searching Using

The Proposed System
This section will describe an example of
searching using the proposed system.
Conducting a search: Firstly, we use
mobile learning as a search query to
conduct a search. The search results in
frame A are shown in Figure 7.
There are 478 search results for the
query mobile learning. An example of
how to read the results can been seen in
paper 8523 which has 6 and 13 .
This means that 6 users have browsed
the paper and 13 other papers were
browsed at the same time as paper
8523.
Figure 8. User list in frame B
Analyzing the related papers: When

one of the associated with paper
8523 is clicked, frame B displays all
the papers that were browsed at the same
time as this paper. Figure 9 shows the
list of papers in frame B. The number of
the papers in frame B is equal to the
number of for paper 8523 in frame
A. The number in the parentheses
following the paper number indicates
how many times the paper has been
browsed. For example, paper 8523 has
been browsed 7 times. This is because
user#12 browsed this paper twice, and
the other users only browsed this paper
once.
146
3.
Figure 9. A segment of a paper list in frame B
5 CONCLUSIONS AND FUTURE

WORK
When a student conducts a search,
his/her search log, which contains
valuable information, can be a precious
resource to other users. The information
can help students, especially those who
are not good at searching, to improve
their searching efficiency. In this paper,
we proposed a search engine system that
can share search queries and browsing
history among students. We refer to the
search queries and browsing history as
community search logs. This system
analyzes the community search logs, and
shares the students knowledge and
experience with each other.
In the future, we are planning to analyze
the relation between the students who
browsed the same papers and the
common features of the papers that have
been browsed at the same time.
REFERENCES
1.
2.
D. Bilal. Childrens use of the

Yahooligans! web search engine: I.
cognitive, physical, and affective behaviors
on fact-based search tasks, Journal of the
American Society for Information Science,
51, 2000, pp.646665.
G.J. Hwang, P. S. Tsai, C. C. Tsai, & J.
C.R. Tseng, A novel approach for assisting
teachers in analyzing student web-searching
behaviors, Computers & Education, 51(2),
2008, pp.926-938.
G. Marchionini, Information Seeking in

Electronic
Environments.
Cambridge
University Press, New York, 1995.
4. P. Dias, M. J. Gomes, & A. P. Correia.
Disorientation
in
hypermedia
environments: mechanisms to support
navigation,
Journal of Educational
Computing Research, 20(2), 1995, pp.93117.
5. G.J. Hwang & F. R. Kuo, An informationsummarizing instruction strategy for
improving web-based problem solving
abilities of students, Australasian Journal of
Educational Technology, 27(2), 2011,
pp.290-306.
6. T. Toda, T. Mine, & M. Amamiya, Active
Browsing: A Novel Personalized Web
Browsing Support System, Tech. Rep.
IEICE SIG-KBSE 106(473), 2007, pp.1-6
(in Japanese).
7. K. Nakao & T. Mine, A Method for
Relevance Calculation between Browsed
Web
Pages
toward
Personalized
Recommendation System, Tech. Rep.
IEICE SIG-AI 109(51), 2009, pp.31-36 (in
Japanese).
8. N. Gotoda, K. Matsuura, K. Kanenishi, & Y.
Yano, Support of the Running Motivation
based on activity Log, Tech. Rep. IEICE
SIG-ET, 107(536), 2008, pp.65-68 (in
Japanese).
9. Y. Watanabe, T. Kojiri, & T. Watanabe,
Knowledge Organization Framework from
Discussion Records of Collaborative
Learning, Tech. Rep. IEICE SIG-ET,
109(82), 2009, pp.21-26 (in Japanese).
10. C. R. Anderson & E. Horvitz, Web
Montage: A Dynamic Personalized Start
Page, WWW02: Proc. of the 11th
international conference on World Wide
Web, 2002, pp.704-712.
11. X. Wu, J. Zeng, C. Yin, & S. Hirokawa,
Sharing Knowledge and Experience of
Search with SNS, The 17th International
Symposium on Artificial Life and Robotics
2012 (AROB 17th 2012), B-Con Plaza,
Beppu, Oita, Japan, Jan 19-21, 2012,
pp.101-104.
12. B. Flanagan, C. Yin, S. Hirokawa, H. Y.
Sung, & G.J. Hwang, Analyzing Research
Trends of Mobile Learning with the Milky
Way, Proc. 7th International Conference on
Wireless,
Mobile
and
Ubiquitous
Technology in Education (WMUTE2012),
2012, pp.249-253.
147
13. Y. Chuang & L. Wu, User-Based
Evaluations of Search Engines: Hygiene
Factors and Motivation Factors, Proc. of
40th
Annual
Hawaii
International
Conference on System Sciences, 2007, pp.110.
148
Modified Proportional Fair Scheduling for Resource Reuse and

Interference Coordination in Two-Tier LTE-Advanced Systems
Rony Kumer Saha
American International University-Bangladesh (AIUB)
Faculty of Engineering, House 82, Road 14, Block B, Kemal Ataturk Avenue
Banani, Dhaka - 1213, Bangladesh
rony107976@gmail.com
ABSTRACT
In this paper, a two-tier Long Term
Evolution-Advanced
(LTE-Advanced)
system where the second tier is deployed as
femtocells is investigated. Although the
reuse factor of one has been a generic
concept to increase the data rate using
orthogonal frequency division multiplexing
(OFDMA) technology, it is difficult to
achieve high spectral efficiency by
employing this reuse of the same resource
from cell to cell. In addition, high data rate
with minimal transmission energy per bit is
one of the significant issues to reduce
overall energy consumption by mobile
communications. In this paper, we address
these issues using femtocells in the coverage
of macrocells in LTE-Advanced systems. In
line with prevailing high spectral and energy
efficient network requirement, a novel
resource reuse strategy is proposed that
employs the reuse of resources in the tier
level, i.e. femto tier. The same resource
block (RB) can be reused in the femto tier as
long as the interference coordination
condition is satisfied. The proportional fair
(PF) scheduling strategy is adopted on top of
which the proposed resource reuse strategy
is applied resulting Modified PF (MPF)
scheduling strategy. A novel interference
coordination strategy is developed and
investigated in the MPF scheduler. In
addition, a new mechanism called macro
user RB reservation for the system average
fairness performance improvement is also
investigated. A system level resource
scheduling simulator is developed, and its
performance is evaluated based on a number

of scenarios to demonstrate that high
spectral and energy efficiencies can be
achieved in the context of LTE-Advanced
systems.
KEYWORDS
Energy efficiency, femtocell, interference
coordination, LTE-Advanced, resource
reuse, scheduling, spectral efficiency.
1 INTRODUCTION
This research focuses on LTEAdvanced systems and femtocell
technologies, the cutting-edge systems
and technologies of the next generation
mobile communication networks. The
LTE-Advanced system is under the
design and investigation phase, with
several critical issues yet to be addressed
and left as open research problems. One
of the critical issues is to achieve high
data rate to provide heterogeneous
services. However, high data rate
requires highly spectral efficient
networks. The conventional frequency
reuse factor of one in the macrocell is
not sufficient to meet high spectral
efficiency requirements of LTEAdvanced systems. This drives the use
of smaller cells within the coverage of a
macrocell.
Further, high data rate demand of
users for multimedia communication
149
causes increase in energy consumption
of radio access networks [1]. Almost
80% of the total power consumptions in
mobile networks is due to radio access
networks [2]. This results in paying out
considerable attention to energy
consumption of the next generation radio
access networks [3]. Design of energy
efficient radio has now become a new
research direction in both industry and
academia [2]. Several approaches have
been proposed to reduce the base station
(BS) energy [6]. Reducing distance
between transmitter and receiver allows
reduction in transmit power to provide
good received signal quality [2] and can
be implemented by deploying femtocells
as second tier in the existing macrocells.
A femtocell is a low power BS,
typically consumes few watts much less
than that of a macro BS, which covers
small area with good signal quality,
typically deployed as the overlaid
network in the coverage of macrocells
[7]. Hence, by serving sufficient amount
of traffic using femtocells, significant
amount of energy consumption can be
reduced. In addition, because of close in
distance of a femto BS to a mobile
station (MS), the received signal strength
is also improved, which results in
spectral efficiency improvement as well.
Furthermore, scheduling of resources
to a user has significant impact on the
achievable data rate that is normally
performed based on the measurement of
the channel condition between the user
equipment (UE) and the macro BS. The
UE provides the channel information to
the macro BS through uplink
transmissions.
Several
researches
already
addressed
the
resource
scheduling in the macro-femto cell
network architecture based on the
resource reuse among femtocells to
increase the data rates [23]. However,
very few or none of these addressed

explicitly how the spectral efficiency
and energy efficiency can be improved.
This research contribution can be
considered as the first of its kind where
the resource scheduling algorithm based
on the resource reuse in femtocells is
addressed explicitly to fulfil the spectral
efficiency requirements in LTEAdvanced systems along with improving
the energy efficiency.
2
RELATED
CONTRIBUTION
WORK
AND
Now that most of the researches have

emphasized the potentiality of femto
BSs separately in the context of spectral
efficiency [1-8] and energy efficiency
performance [3, 2, 7, and 8]. Also, the
basic principle of schedulers has been
addressed in a number of publications
[17-18].
In specific, [9-10] showed how the
resource allocation could be performed
using PF scheduler in the context of
LTE-Advanced
systems
in
only
macrocells. Apart from that, Ref. [10]
proposed a user grouping based PF
scheduler where users are grouped based
on the number of carriers that can be
scheduled; Ref. [9] proposed crosscomponent carriers (CC) based PF
scheduler to provide high degrees of
fairness between LTE users and LTEAdvanced users; Ref. [11] addressed a
two-tier system considering a partial cochannel assignment approach; Ref. [12]
proposed a hybrid shared and partitioned
spectrum allocation between macro users
and femto users; Ref. [13] proposed a
dynamic resource partitioning to
resources used by the nearby macro
users; Ref. [14] proposed a joint power
and sub-channel allocation scheme in
dense environments; Ref. [15] proposed
150
a decentralized resource allocation
scheme where each femtocell is
restricted to a randomly chosen subset of
resources; and Ref. [16] investigated the
degree of access control of femtocells
and its impact on a mixed macro-femto
cell system based on the co-channel
assignment strategy.
Further, Ref. [2] proposed a new
resource allocation strategy to improve
downlink energy consumption. Ref. [3]
showed that introducing sleep mode in
small cells can improve energy
consumption gain. Ref. [7] emphasized
the
importance
of
interference
management in achieving energy
efficiency.
Furthermore, a number of papers have
identified that resource allocation
techniques have the potentiality of
improving
energy
efficiency
significantly [1, 4, 5, and 6].
However, very few or none researches
have demonstrated explicitly how the
improvement in both spectral efficiency
and energy efficiency can be achieved
with resource scheduling on the same
platform using femtocells. In this paper,
we propose a novel radio resource reuse
and interference coordination technique
employed to the femto tier in order to
improve both energy and spectral
efficiencies.
In doing so, we develop a novel
resource scheduling algorithm on top of
the PF scheduling strategy to minimize
the implementation cost and complexity.
Furthermore, we investigate the effect of
interference on system performances and
develop an interference coordination
strategy to make the scheduler more
realistic. We also investigate the fairness
performance for macro users and femto
users separately and develop a new
mechanism to improve the fairness
performance of the system as a whole.
An extensive system level simulation is

carried out to evaluate the spectral
efficiency and energy efficiency
performances of the proposed scheduling
algorithm.
3 SYSTEM MODEL
A two-tier LTE-Advanced system is
considered, where the second tier is
deployed as femtocells. The system
model for LTE-Advanced is explained in
three major parts: conceptual model,
analytical model, and simulation model.
Each of these parts addresses a number
of aspects of the system and is explained
in the following subsections.
3.1 Chanel Models
We consider a simple channel
propagation model to reduce the
computational runtime that incorporates
the followings.
1) Antenna Patterns: 3-sector cell site
with fixed antenna patterns is considered.
For femtocells and UEs, it is considered
that the azimuth antenna pattern is
omnidirectional.
2) Pathloss Models: Distance dependent
log-normal pathloss model is considered
for all links and is given in table 1.
3) Shadowing Models: Log-normal
shadowing with standard deviation of 8
dB is considered for all links. However,
a 10 dB standard deviation is considered
for links established between serving
femto BSs and serving femto UEs. In
addition, no shadowing correlation is
considered from a macro BS to UEs.
Note that, shadowing correlation is not
applicable for the femto BS since only
one UE per femto BS is considered in a
transmission time interval (TTI).
Table 1. Pathloss models between an UE and a
BS for the macro tier and the femto tier [19].
151
Path
loss:
UE to
macro
BS
Path
loss:
UE to
femto
BS
UE is
outside
PL (dB) =15.3 +
37.6log10R, R in m
UE is
inside an
apartment
PL (dB) =15.3 +
37.6log10R + Low, R in m
Dualstripe
model
PL(dB) =
127+30log10(R/1000), R
in m.
4) Fast Fading Models: In outdoor

environments, Rayleigh fading is
considered where a random fading effect
of -50 dB to 2 dB superimposed on the
average value of the signal at any RB is
assumed.
However,
in
indoor
environments,
Rician
fading
is
considered where a random fading effect
of -60 dB to 5 dB for non line-of-sight
(NLOS) and 5 dB for LOS component
superimposed on the average value of
the signal is assumed. Fast fading effect
is considered as because of motion of
objects between the BS and the UE
irrespective of the BS types. The fading
effect is updated, for example in every
TTI (1 ms) for 150 km/hr, in every 5 TTI
for 30 km/hr, and in every 50 TTI for 3
km/hr and the UE is considered
stationary for a particular simulation
runtime of maximum 1000 ms. Note that,
the carrier frequency is considered 3.5
GHz. The fading effect update for TTIs
corresponding to the velocity of objects
between a BS and an UE is computed as
follows:
Since the fast fades are occurred on an
average of half wavelength distance
apart, the number of TTIs corresponding
to a particular velocity of an object
between a BS and an UE can be
computed as below.
C = f => =
3 1010
3.5 10 9
=> 2 = 4.28cm.
s = vt => 4.28 cm = vt => t =
4.28 cm
833 cm s
t = 5.1ms => t 5TTI for v = 30 km / hr .
Similarly, t 1 TTI for v=150 km/hr and

t 50 TTI for v=3 km/hr.
5) Spatial Correlation, Physical Channel
and Feedback Mechanism: we consider
only physical downlink shared channel
(PDSCH) for all RBs. No physical
downlink control channel (PDCCH) is
considered. In addition, pilot signals for
CQI messages at the BS are not
considered since the channel state
information is assumed to be known at
the BS. The feedback mechanism for
lost packets, i.e. acknowledgement
(ACK) or non acknowledgement
(NACK) is not considered since it is
assumed that all packets that are sent
from the BS received correctly at the UE.
The spatial correlation is not considered
since the antenna configuration is
considered single-input single-output
(SISO) for both BS and MS.
3.2 System Architecture
The system model considers a dense
urban deployment. A single macrocell,
within which a number of macro users
and one femtocell block, is considered.
A dual stripe dense urban femtocell
block is considered [19]. In addition, an
active femto BS in each apartment and
one user per femto BS is assumed. Each
femtocell block is located randomly and
uniformly throughout the coverage of
the macrocell. Figure 1 shows the
considered two-tier macrocell-femtocell
architecture in a dense urban deployment.
Closed subscriber group (CSG), cochannel,
and
adaptive
power
configuration [20] is considered for
femtocells. The channels used by
femtocells are not orthogonal to the
macrocell. Rather, the whole bandwidth
can be shared between the macro users
and the femto users.
152
1
1
~
Rk ,i ( )(1 ) Rk , j ,i ( ), k k i, j ( )
~
tc
tc
Rk ,i ( 1)
1
~
Rk ,i ( )(1 ),
k k i, j ( )
t
c
UE
(2)
th
10
10
10
A Macro
Cell Site
10
10
10
10
A Dual Stripe
Apartment Block
where k is the selected user in the i CC

at the j th RB and tc is the adjusted time
constant or fairness factor.
The PF scheduler provides an optimal
throughput demand while ensuring
optimal fairness performance. The PF
scheduler ensures it by taking into
account the instantaneous channel state
information from the UE through uplink
by a control signalling channel.
10
m
3.4 Resource Reuse Strategy
th
in the i CC at the j th RB in time ;
~
Rk ,i ( ) is the average throughput in the
The PF scheduler always allocates an

RB to a user with the max PM in any
TTI, i.e. it can allocate an RB to only
one user. Hence, the PF scheduler cannot
explore the multi-user diversity at the
RB level in addition to the TTI level.
Table 2 presents the spectral efficiency
requirements of the LTE-advanced
system. Since the peak spectral
efficiency
requirement
of
LTEAdvanced systems is high, it is difficult
to achieve by the PF scheduler without
reusing the same RB by other users in
addition to the user with the max PM. To
facilitate RB reuse, a new resource
scheduling strategy is proposed that
focuses on reusing an RB among femto
users. Unlike PF scheduling, the new
scheduling strategy considers not only a
femto user with the max PM but also
other femto users for assigning an RB in
a TTI, subject to the interference
coordination strategy.
th
past of the user k in the i CC in time .
The throughput of the user ki , j ( ) is
Table 2. Spectral efficiency requirements of

LTE-Advanced systems [22].
Figure 1. Two-tier dense urban macro-femtocell

deployment [19] [21].
3.3 Proportional Fair Scheduling

The proportional fair (PF) scheduling
algorithm is based on maximizing the
scheduling metric, i.e. for any RB j in
any TTI , the PF scheduler schedules
only the user with the maximum
performance
metric
(max
PM).
Mathematically, it can be expressed as
follows [9]:
Rk , j ,i ( )
k i, j ( ) arg max ~
R
(
)
k
k ,i
where ki , j ( ) is the selected
(1)
user in the
i th CC at the j th RB in time ; Rk , j,i ( )

is the estimated throughput of the user k
updated according to the following

condition [27]:
Spectral Efficiency
Peak spectrum
Uplink
15
Downlink
30
153
efficiency (bps/Hz)
Average spectrum
efficiency
(bps/Hz/cell)
Cell-edge spectrum
efficiency
(bps/Hz/cell/user)
1.2 ~ 2.0
2.4 ~ 3.7
0.04 ~ 0.07
0.07 ~
0.12
3.5 Novel Mechanism for Fairness

Improvement
Since in a typical cellular mobile
system, the signal quality in the
macrocell environment is on an average
worse than the femtocell environment,
femto users have advantages from the
scheduling decision perspective because
of the better signal quality.
In order to provide fairness to macro
users, a novel mechanism called macro
user RB reservation is proposed. In the
new mechanism, a certain percentage of
total RBs of the system bandwidth are
considered to be reserved only for macro
users in every TTI. These reserved RBs
not only provide fairness to macro users
as a group but also help improve the
overall system fairness performance.
3.6 Proposed Resource Scheduling
Strategy
The proposed resource scheduling is
based on the PF scheduling strategy. The
resource scheduling strategy for RB
allocation can be stated as follows. An
RB can be shared among several femto
users simultaneously. Once an RB is
assigned to a macro user, the assigned
RB cannot be reassigned to any other
users. However, an RB assigned to a
femto user can be reused by other femto
users simultaneously, subject to the
interference coordination strategy.
Hence, an RB can be reused among
multiple femto users in any TTI.
The impact of the strategy can be

analysed in the view of cell coverage
area. Specifically, a macrocell coverage
area is larger than a femtocell coverage
area, and hence the coverage area can be
a decisive factor for co-channel resource
scheduling. For scheduling purposes,
femtocells can be configured for
resource reuse easily in the frequency
domain because of small coverage areas.
Hence, an RB can be simultaneously
reused among several femtocells through
spatial orthogonality.
In addition, in the view of flexibility,
the proposed scheduling strategy is also
flexible enough since the total
bandwidth of the macro BS can be
allocated to only femto users, if most
traffic is generated from the indoor
environments. This is a feasible scenario,
especially in the night time, when almost
all people work in the indoor
environments. Hence, the scheduling
strategy can provide flexibility to
operators to allocate more resources to
outdoor users in the day time and to
indoor users in the night time.
.
3.7 Interference Coordination Strategy
The interference coordination strategy
is based on the orthogonal frequency
assignment to the macro tier as well as
the femto tier. Both inter-floor modeling
as well as intra-floor modeling is
considered for the femto tier as follows:
1) Inter-tier Interference Coordination
Strategy: RB allocations between the
macro user group and the femto user
group are always orthogonal.
An RB which is assigned to a macro
user by the PF scheduler is not
applicable for reuse, irrespective of
the user category, i.e. macro user or
femto user.
154
An RB is considered to be reused only
when that RB is assigned to a femto
user by the PF scheduler in a TTI.
2) Intra-tier Interference Coordination
Strategy: For macro tier, since RB
allocations among macro users are
always orthogonal to each other, there is
no need for the interference coordination
among macro users. However, for femto
tier, both Inter-floor modelling and Intrafloor modelling is considered.
a) Inter-floor Modelling: RBs are
equally divided into two groups after
reserving a group of RBs for the fairness
improvement of macro user. Each RB
group is assigned to femto users of the
alternate floors (figure 6).
b) Intra-floor Modelling: An RB is
considered to be reused by another femto
user only when that femto users serving
femto BS is non-adjacent to the already
assigned femto users serving femto BS,
irrespective of the number of RB reuses
considered. The resource reuse distance
between femto BSs must be at least 10
meters, irrespective of femto BS
locations in the same floor, either in the
same stripe or in different stripes.
3.8 Resource Scheduling Algorithm
The resource scheduling algorithm
illustrates a systematic and stepwise
method to represent the proposed
resource reuse and scheduling strategy
and incorporates a number of steps as
follows:
Step 1: reserve a certain percentage of
total RBs of the system bandwidth for
macro users. Step 2: estimate and sort
PM of all users at all RBs and do
mapping of sorted PM to corresponding
users.
Step 3: find RBs that are
scheduled to macro users, UMk where k
is any positive integer denotes a user
index. Step 4: find RBs that are
scheduled to femto users, UFk. Step 5:

check second, third, and so forth max
PM at all RBs with femto users
scheduled and apply interference
coordination condition. Step 6: if femto
users at any RB fulfil the interference
coordination condition, allocate that RB
to additional femto users in the order of
their PM values. Step 7: repeat steps 5
and 6 for all RBs, N=Nmax as long as the
max PM, M=Mmax. Mmax is any positive
integer that can be varied based on the
spectral
and
energy
efficiency
requirements, and Nmax is the total
number of RBs. M and N respectively
represents the max PM index and the RB
index.
Figures 2, 3, and 4 illustrate all these
steps in the algorithm and are used in
system level simulation. Since the
resource scheduling strategy uses the PF
scheduling principle as its baseline and
modifies it based on the proposed
resource reuse strategy, it is referred to
as the modified PF (MPF) scheduling
strategy. The
proposed resource
scheduler operation can be explained
with the example given in figure 4.
Assume that the first five RBs, i.e. RB 1,
RB 2,, RB 5 in figure 2 are considered
for the analysis. RB 1, RB 2, and RB 3
are kept reserve for macro users
according to step 1 in the scheduling
algorithm. Hence, these RBs are not
applicable for reuse because of macro
users scheduled to these RBs over the
simulation runtime. Since RB 4 and RB
5 are scheduled to femto users by the PF
scheduler, these RBs can be reused if
the interference coordination condition is
satisfied. According to figure 4, it is
shown that femto users with the 3rd and
4th max PM for RB 4 and the femto user
with the 2nd max PM for RB 5 can
satisfy the interference coordination
condition, i.e. these three femto users are
155
non adjacent to all femto users scheduled
in their respective RBs, irrespective of
whether or not they are in the same floor
in the femtocell apartment block. Hence,
in the proposed scheduler, RB 4 and RB
5 can be reused simultaneously to three
additional femto users on top of the
femto users scheduled by the PF
scheduler.
PM
RB 1
RB 2
RB 3
RB 4
. . . . . . . . . . . .N
. . . . . . . . .
RB 5
Nmax
1st
max
PM
UMk UMk UMk UFk
UFk UMk UFk
UFk
UFk
UFk UMk UFk
2nd
max
PM
UMk UMk UMk UFk
UFk UMk UFk
UFk
UFk
UFk
UFk
3rd
max
PM
UMk UMk UMk UFk UMk UFk
UFk UMk UFk
UFk
UFk UMk
UMk UMk UMk UFk
Mmax
UFk UMk UMk UMk UFk
RB
UFk
UFk UMk UFk
. . . . . . . . . . .
Figure 2. PM vs. RB distributions: Steps 1- 2.

PM
1st
max
PM
RB 1
RB 2
RB 3
RB 4
RB 5
UMk UMk UMk UFk
UFk
. . . . . . . . . . . .N
. . . . . . . . .
UMk UFk
RB
Nmax
UFk
UFk
UFk UMk
UFk
2nd
max
PM
UFk
UFk
UFk
UFk
UFk
UFk
UFk
UFk
3rd
max
PM
UFk
UFk
UFk
UFk
UFk
UFk
Mmax
UFk
UFk
UFk
UFk
UFk
. . . . . . . . . . .
Figure 7 shows the implementation of

the proposed resource scheduler. Most
parts of the proposed MPF scheduler are
also parts of the PF scheduler, except an
additional comparator for user indices, a
decision support system for interference
coordination, and a decision making part
for RB reuse.
The RB reuse decision maker is used
to allocate the same RB to multiple
femto users, as long as the interference
coordination strategy is satisfied. Other
than these three additional parts, the
implementation of the
proposed
scheduler is the same as the PF
scheduler. Because of simple in
implementation on top of the PF
scheduler, the proposed MPF resource
scheduler can be implemented without
significant additional costs for a higher
spectral and energy efficiencies. Note
that, the numbers in circles in figure 7
represent the timing sequence in a TTI
from the channel condition reports in the
uplink to RB allocation for users in the
downlink [21].
3.10 Mathematical Analysis
Figure 3. PM vs. RB distributions: Steps 3-5.
PF
Proposed scheduler
scheduler
An RB reused femto user

PM
1st
max
PM
RB 1
RB 2
RB 3
RB 4
UMk UMk UMk UFk
RB 5
. . . . . . . . . . . . N. . . . . . . . .
Nmax
UFk UMk UFk
UFk
UFk
UFk UMk UFk
2nd
max
PM
UFk
UFk
UFk
UFk
UFk
3rd
max
PM
UFk
UFk
UFk
Mmax
UFk
UFk
UFk
RB
UFk
. . . . . . . . .
An RB reused in a TTI
Figure 4. PM vs. RB distributions: Steps 6-7.
3.9 Implementation and Complexity

Analysis of the Proposed MPF
Resource Scheduler
The resource scheduling algorithm is

exemplified in an analytical model based
on the conceptual model for the system
level simulation to evaluate the
performance of the system.
1) Transmit Power: The transmit power
of a CC is represented by PTCC ( i ) where i
denotes the CC index in the aggregated
system bandwidth. The transmit power
of an RB is represented by PTRB ( j ) where
j represents the RB index in a particular
CC. Assuming all CCs as well as all RBs
have the same transmit power, the
transmit power per RB can be expressed
at any arbitrary i and j as
156
P CC ( i )
PTRB ( j ) = T ALL
X
(3)
where X ALL is the total number of RBs

in a TTI for a CC.
2) Carrier Frequency and Resource
Block: Assume that CCs are aggregated
from a number of bands in the
aggregated bandwidth and represented as
f1, f2,, fB where B denotes the total
number of bands in the aggregated
carrier. Hence, the aggregated spectrum
can be expressed as fagg= {f1, f2 fB}. If
S j represents an RB with an index j, then
S j { 1,2,3,...,X ALL } .
A certain number of RBs is reserved
for macro users and is denoted by X RS .
Since RB allocations to macro users and
femto users are orthogonal, the total
number of RBs that are available for
reuse
can
be
expressed
as
M
R
All
RS
M
X =X
- ( X + X ) where X is the
total number of RBs scheduled to macro
users by the PF scheduler in a TTI.
Hence, the percentage of RBs available
for reuse by the MPF scheduler is given
by,
RB REUSE =
XR
X ALL
100
(4)
3) Spectral Utilization:
The total
number of RBs allocated by the PF
scheduler on a CC in any TTI is given
by X TPF = X All , whereas the total number
of RBs allocated by the MPF scheduler
is given by X TMPF = X All + X TR where X TR
is the number of reuse RBs allocated by
the MPF scheduler in a TTI. Hence,
compared to the PF scheduler, the
spectral utilization improvement factor
of the MPF scheduler is given by,
MPF
U PF
=
X TMPF
X TPF
(5)
4) Spectral Efficiency: The SINR at an

RB j in a TTI can be expressed as [21]
PTRB ( j ).H ( j )
Z
(6)
N ( j ) + I z ( j )
z =1
where I z ( j ) is the interference from

an arbitrary interferer z for a link
established for RB j , and Z is the total
number of interferers for the link. N ( j )
and H ( j ) denote respectively the
thermal noise and the channel transfer
function for RB j .
The transfer function gives the
frequency domain channel gain and
varies with the subcarrier frequency in
every TTI. This gain is referred to as the
frequency domain packet scheduling
(FDPS) gain. The FDPS gain is found to
follow a logarithmic function of the
active number of users k in OFDMA
[24]. The FDPS gain is adopted from [9]
as follows.
Assuming uniform distribution of user
locations throughout the cell, the FDPS
gain is approximated for a single carrier
by the following expression.
1,
k 1
G (k ) 0.11 log e (k ) 1.10, 1 k 13
1.38,
k 13
(7)
The throughput in RB j can be

expressed in terms of the mutual
information
(MI)
by
employing
Shannons formula, as given below [25].
0,
10dB
( dB)
( ) log 2 1 10 10 , 0dB 22dB (8)
4.4,
22dB
where = 0.6 , is the implementation
loss factor that takes into account

modulation and coding schemes (MCS)
and hybrid automatic repeat request
(HARQ) in LTE systems and is the
SINR in dB at RB j .
The mutual information (MI) effective
SINR metric (MIESM) [26] is adopted
for the system level simulation to
157
estimate the average throughput per user
during the course of the simulation
runtime. For the PF scheduler, the
PF
average MI ( ) over all RBs over all
users in a CC is obtained, and then the
PF
effective SINR, eff
per RB per TTI is
estimated from the average MI by
inverse mapping as shown in figure 5.
The inverse mapping function between 10 dB and 22 dB SINR can be
mathematically expressed for the PF
scheduler as
PF
eff
( PF
PF
(2
ee =
(9)
Hence, the system level spectrum
efficiency for the PF scheduler per RB
per TTI can be expressed as
) = 10 log 10
-1)
PF
PF
Rsys
= log2 ( 1 + eff
)
(10)
Similarly for the MPF scheduler,
system level spectrum efficiency per RB
MPF
eff
( MPF ) = 10 log10
MPF
(2
(11)
where the average MI and the
effective SINR per RB per TTI for the
- 1)
MPF scheduler are represented by
MPF
MPF
and eff respectively.

efficiency for the MPF scheduler per RB
MPF
MPF
Rsys
= log2 ( 1 + eff
)
all RBs by any scheduler, the average

transmit power per RB can be expressed
as follows:
RB
(nmacro PTRB
, macro ) ( n femto PT , femto )
RB
PT , ave
RB
nTotal
(13)
RB
where nTotal
is the total number of
RBs in a TTI.
Energy efficiency is defined as the
amount of energy required per bit
transmission and can be expressed as
follows [10] [25]:
PTRB, ave
Rsys
[ J/b]
(14)
where ee is the average energy per

bit transmission in joules per bit (J/b).
6) Fairness Index: Jains fairness index
is adopted for the evaluation of the user
fairness performance to define the
degree of fairness among users.
Mathematically, the fairness index can
be expressed as follows [27]:
2
K
uk
FJ k 1K
2
K uk
(15)
k 1
where K represents the total number

of users and u k represents the total
number of RBs allocated to user k over
the simulation runtime.
(12)
5) Energy Efficiency: Denote PTRB,macro

is the transmit power per RB in
macrocell, and is PTRB, femto the
transmit
power per RB in femtocell.
If nmacro is the number of macro
users, and n femto is the number of
femto users scheduled in any TTI over
Figure 5. Mutual Information versus SINR

curves using Shannons capacity formula and
MIESM.
158
m
en
th
ei
gh
t:
m
ax
i
m
um
10
flo
or
s
Apartment length: maximum 120 m

Floor 5
Floor 4
A
pa
rt
UE
Floor 3
Floor 2
Floor1
3m
Apartment width: maximum 70 m
10 m
10 m
31
32
33
34
35
36
37
38
39
40
21
22
23
24
25
26
27
28
29
30
11
12
18
19
20
10
Macro UE
13
An
UE
3
Serving
femto BS
14
15
16
17
Femto UE
10 m
10 m
A femto UE outside its serving

femto BS connected via macro BS
Co-channel interference zone

A Dual Stripe
Apartment Block
A Macro
Cell Site
Figure 6. Proposed interference coordination in a femtocell block (floor level) [21][23].

Uplink Input
RB
Comparator
CQI
Manager
LA
RB Manager
RB
Directory
Femto RB
Distributor
TD
Scheduler
Instantaneous Throughput
User
Update
RB
Aggregator
2
User Index
Directory
PF Scheduler
PM Generator
Reserved Macro RB
Index Controller
4
User Average
Throughput
Directory
5
Interference Coordinator
Interference
Coordination
Policy Directory
8
7
PM-User Index
Mapping Generator
3/8
User Index
Comparator
OPTION 1:
9
Decision
Support System
RB Reuse
Decision Maker
PF Scheduler
OPTION 2:
MPF Scheduler
RB Allocation
10
Additional Parts for MPF Scheduler

Air Interface
Figure 7. Proposed resource scheduler implementation framework.
159
3.11
Simulation
Parameters,
Assumptions and Scenarios
Simulation parameters and assumptions
are adopted from [28] and are listed in
table 3. Table 4 lists the major scenarios
that are considered for the evaluation of
system performances.
reuses an RB to a femto user only when

that femto users serving femto BS is
non-adjacent to the already assigned
femto users femto BSs. This implies the
greater diversity in reuse of RBs among
femto users and hence improvement in
fairness performance.
4 SIMULATION RESULTS
4.1 Performance of MPF Scheduler
with Interference Coordination
The MPF scheduler can gain advantage
from the multi-user diversity since an RB
can be simultaneously reassigned to a
number of femto users. Figures 8 and 9
respectively present the performance of
the MPF scheduler over the PF scheduler
in terms of average throughput and
fairness. It can be found that the MPF
scheduler outperforms in throughput
performance
whereas
shows
approximately the same in fairness
performance as the PF scheduler.
Note that, in the absence of
interference coordination (figure 8), the
MPF scheduler can obtain higher average
throughput than with considering
interference coordination. However, it
can be obtained only at lower fairness
performance (figure 9). This lower
fairness in the absence of interference
coordination can be explained by figure
10 that shows RB distributions among
users. The distribution of RBs with
interference coordination is more
clustered than in the absence of
interference coordination.
The greater the degree of evenness in
RB allocations among users is achieved,
the higher the value of average fairness
index. Unlike in the absence of
interference coordination, the MPF
scheduler with interference coordination
Figure 8. Average throughput performance of the

MPF scheduler with interference coordination and
in the absence of interference.
Figure 9. Average fairness performance of the

MPF scheduler with interference coordination and
in the absence of interference.
4.2 Role of New Mechanism in MPF

Scheduler
In figure 10, it can be found that the
macro users (user index: 1-20) are
suffered from low RB allocations,
irrespective of whether or not the MPF
scheduler is interference coordinated.
This is because of either the better signal
quality of femto users or the lower gain
160
from multi-user diversity of macro users.
In addition, since the number of macro
users is lower than that of femto users,
this deviation in RB allocations between
the macro user group and the femto user
group becomes high when the system is
fully loaded.
on the systems average throughput

performance (figure 13).
Figure 12. Average fairness performance of the

MPF scheduler: with reserving macro RBs and
without reserving macro RBs.
Figure 10. RB distributions by the MPF

scheduler: with interference coordination and in
the absence of interference.
The role of the proposed new

mechanism macro user RB reservation
can be understood by figure 11 that shows
that macro users are allocated with more
RBs when the new mechanism is
employed. The use of the new mechanism
reduces the gap in RB allocations
between macro users and femto users.
Figure 11. RB distributions of the MPF

scheduler: with reserving macro RBs and without
reserving macro RBs.
Further, figure 12 shows the systems

average
fairness
performance
improvement from employing the new
mechanism with an insignificant impact
Figure 13. Average throughput performance of

the MPF scheduler: with reserving macro RBs and
without reserving macro RBs.
Furthermore, since the number of macro

users and femto users are not equal
especially with the increase in traffic
demand, a user category based
investigation, i.e. the average value of
RBs allocated to each user category is
carried out.
Figure 14 shows the improvement for
evenness in RB allocation since the
deviation between the average number of
RBs allocated to a macro user (among
macro users only) and the average
number of RBs allocated to a femto user
(among femto users only) is reduced.
However, note that the new mechanism
improves fairness in RB allocation since
the deviation between the average RBs
161
allocated to a macro user and the average
number of RBs allocated to a femto user
is reduced. In addition, the mechanism
also improves fairness among macro
users only (figure 15) with little or no
degradation in fairness among femto
users only (figure 16).
Figure 14. Average RB allocated per user with

macro RB reserved and without macro RB
reserved.
Figure 15. Average fairness (among macro users

only) with macro RB reserved and without macro
RB reserved.
4.3 Performance of the MPF Scheduler

for Spectral and Energy Efficiencies in
LTE-Advanced Systems
In order to fulfil the high spectral
efficiency
requirements
of
LTEAdvanced systems, the reuse of resources
in the tier level in addition to the
macrocell level is critical. This in turns
helps not to reduce the macrocell
coverage area. Hence, no cost from

implementing new sites in order to
provide better signal quality for high data
rate.
Figure 16. Average fairness (among femto users

only) with macro RB reserved and without macro
RB reserved.
Peak spectral efficiency is crucial for the

highest data rates that a user can obtain
using the system. In addition, link
reliability over a certain period of time is
an important requirement, especially for
the high data rate links. Link disruption
between an MS and a BS can result in
poor user experience because of service
discontinuity during the course of
interactive sessions between the MS and
the BS. Further, high data with minimal
transmission energy per bit is one of the
significant issues to reduce overall energy
consumption by mobile communications.
Considering these issues, the performance
of the MPF scheduler in terms of spectral
efficiency and energy efficiency is
investigated. It is found that an
appropriate value of the max PM in MPF
scheduler in the simulation can meet both
spectral efficiency and energy efficiency
requirements.
1) Spectral Efficiency: The average
spectral efficiency improves irrespective
of the increase in max PM or the number
of floors. This is because of the fact that
the average spectral efficiency depends
162
upon the aggregated throughput in a cell.
The aggregated throughput in a cell is
improved as long as the number of femto
users scheduled in a TTI is increased
because of good channel quality, and is
not affected negatively by the interference
coordination in the femto tier or the
frequency of resource reuse to a particular
femto user in a TTI.
However, as given in table 5, both peak
and cell-edge spectral efficiencies
degrade with the increase in the number
of floors because of more inter-floor
interference coordination effect, while
keeping max PM fixed. This is because,
as the number of floors increases, interfloor interference coordination effect
becomes significant so that the number of
resources reused to a specific femto user
(the maximum number of resources
reused to a femto user that define the
peak data rate) is decreased, hence peak
spectral efficiency. In addition, this in
turn also reduces the throughput per user,
i.e. cell-edge spectral efficiency.
However, this pitfall can be overcome
by increasing the value of max PM in the
MPF scheduler since increase in max PM
will allow the scheduler to reuse more
resources to a specific user even though
the number of femto users scheduled in a
TTI is the same.
On the contrary, with the increase in
max PM chosen in MPF scheduler, there
is a dramatic rise in both peak and celledge spectral efficiencies since more
reuse of resources to a specific femto user
is now possible, along with having an
opportunity of scheduling more femto
users for resource reuse, because of less
inter-floor interference coordination
effect. Hence, the effect of inter-floor
interference coordination, because of
increase in femtocell user density by
increasing the number of floors, can be
overcome by increasing the number of
max PM in MPF scheduler so that more

resource can be reused to a femto user,
which signifies the self control or
robustness of the proposed scheduler.
2) Energy Efficiency: From the simulation
(table 5), it can be found that irrespective
of the change in max PM or the number
of floors, the energy efficiency improves,
i.e. reduction in energy per bit. This is
because of the fact that in either cases, the
number of femto users in the total number
of users scheduled in a TTI increases.
It is to be noted that the change in the
number of floors has less impact on the
energy efficiency improvement than the
increase in max PM. This can be clarified
by the fact that while max PM is kept
constant, with an increase in the number
of floors, the interference coordination
effect from inter-floor modeling becomes
more significant so that only few number
of femto users can be scheduled for
resource reuse and hence less reduction in
energy per bit requirement.
On the contrary, when the number of
floors is kept fixed, increase in the
number of max PM chosen in the MPF
scheduler, increases the possibility of
reusing the same resource more to a
specific femto user. In addition, it also
gains benefit from less inter-floor
interference coordination effect since the
number of floors considered is fixed.
5 CONCLUSION
In this paper, a novel resource reuse
strategy is proposed to demonstrate
explicitly of how the spectral and energy
efficiencies of LTE-Advanced systems
can be improved. A two-tier LTEAdvanced system where the second tier is
deployed as femtocells has been
investigated. The PF scheduling strategy
is adopted on top of which the proposed
163
Table 3. Simulation parameters and assumptions in two-tier systems.

Parameter
Cell layout1
Number of macrocells considered

Inter site distance1
Urban-dense femto BS modeling parameters
(apartment block model)1
Total users
Channel state information
Physical channels
Number of macro UEs per sector
Number of femto UEs per sector
Minimum distance between UE and macro BS1
Minimum distance from UE to femto BS1
UE speeds of interest
Carrier frequency band
Carrier aggregation
Aggregated carrier spectrum in LTE-Advanced
LTE carrier spectrum
Simulated bandwidth
Per RB bandwidth
Per CC RBs
RB reserved for macro users
Link considered1
Total macro BS transmit power1
Macrocell power control
Macro BS antenna gain, after cable loss1
Femto BS antenna gain1
Femtocell power control1
Noise
UE Noise Figure1
Path loss adjustment constant for 3.5 GHz (from
2 GHz)
Penetration loss (assume: UEs are indoors)1
Traffic model
Link level to system level mapping
Max PM (Mmax)
Fairness factor( tc)
TTI
Window size
Simulation runtime
1
taken from [28].
Value/Assumption
Dense Urban: macro layer has 7 sites (21 sectors)
with wrap-around, 1732 m Inter site distance (ISD),
and 20 % of macro UEs is outdoors.
1
1732 m
N (number of cells per row )
10
M (number of blocks per sector)
1
L (number of floors per block)
2
R (deployment ratio )
1.0
P (activation ratio)
100%
Probability of macro UE being
80%
indoors
100
Known
PDSCH (only data channel)
20
80
>= 35 m
20 cm
UE is stationary
3.5 GHz
Contiguous
80 MHz
20 MHz
20 MHz
180 kHz
100
6% (of total RBs)
Downlink
46 dBm
Fixed transmit power
14 dBi
5 dBi
Adaptive transmit power (0 to 20 dBm)
Thermal noise
9 dB
6 dB (additional)
10dB
Full buffer
MIESM
6
100 ms
1 ms
100 ms to 1000 ms
1000 ms
164
Table 4. Simulation scenarios for the evaluation of system performances.
Aspect
Scenario
objectives
Scenario 1
To investigate the effect of
interference coordination and
reuse of resources to femto
users.
User class
Macro user and femto user
Traffic type
Femtocell
density
Simulation
level
Scenario
analysis
Scenario
findings
Scenario 3
To
investigate
the
performance of the MPF
scheduler for spectral and
energy efficiencies in LTEAdvanced systems.
Macro user and femto user
Full buffer
Fixed
Scenario 2
To
investigate
the
performance of the
MPF scheduler with
employing the proposed
new mechanism.
Macro user and femto
user
Full buffer
Fixed
Link level and system level
System level
System level
1. Average throughput and

fairness performances with the
change of simulation runtime
and employing interference
2. RB distributions among users
with employing interference
Performance of the MPF
scheduler in comparison with the
PF scheduler.
1. Average fairness and

throughput performance
of the schedulers.
1. Performance analysis of
the MPF scheduler as a
function of window size for
the following performance
metrics.
spectral efficiency
energy efficiency
Fairness performances
of macro users as well
as the system as a
whole.
Feasibility of the MPF

scheduler in LTE-Advanced
systems for the performance
improvements in spectral
and energy efficiencies.
resource reuse strategy is applied; the

resultant scheduling strategy is called the
Modified PF (MPF) scheduling strategy.
An extensive system level simulation is
carried out to verify the feasibility of the
MPF resource scheduling simulator in a
number of scenarios that are listed out as
follows:
The effect of interference coordination on
the MPF scheduler performance is
investigated. It is found that with
interference coordination, the scheduler
performs better in fairness, but somewhat
lower in average throughput than in the
absence of coordination. This is obvious
since an RB cannot be allocated blindly,
as with the case in the absence of
interference.
The performance of the new mechanism
called macro user RB reservation is
investigated in terms of the fairness of
macro users as well as the average
fairness performance of the system. The
Full buffer
Variable
simulation result shows that the average

fairness performance of macro users as
well as the system is improved.
The performance of the proposed
resource scheduler in LTE-Advanced
systems is investigated in the context of
spectral and energy efficiencies, varying
the value of maxPM of MPF scheduler.
The simulation results show that the
proposed resource scheduler can meet the
spectral efficiency requirements of LTEAdvanced systems along with ensuring
reduction in energy per bit.
Based on these investigations and
simulation results, we suggest the use of
the proposed resource reuse strategy in
the tier level in addition to the macrocell
level in order to improve both spectral
and energy efficiencies to meet the high
data rate at low transmission energy per
bit requirement of LTE-Advanced
systems.
This
will
help
telecommunication
operators
gain
165
Table 5. Average and peak spectral efficiency performances of the MPF and the PF Schedulers.
Efficiency
Average spectral efficiency
Peak spectral efficiency
max
PM =
3,
floors
=2
Value
of
max
PM
and
no. of
floors
max
PM =
3,
floors
=6
max
PM =
7,
floors
=2
Table 6. Cell-edge and energy spectral efficiency performances of the MPF and the PF Schedulers.
Efficiency
Value
of
max
PM
and
Cell-edge spectral efficiency
Energy efficiency
max
PM =
3,
floors
=2
166
no. of
floors
max
PM =
3,
floors
=6
max
PM =
7,
floors
=2
advantage from efficient spectral and

energy utilizations as well as users from
better experiences while using the system.
ACKNOWLEDGEMENTS
This paper is based on the research that
was carried out at Asian Institute of
Technology (AIT), Thailand, and a part
of this research was presented in the
conference mentioned in [23]. The author
is grateful to Poompat Saengudomlert for
his generous support.
REFERENCES
1.
2.
3.
G. Y. Li, Z. Xu, C. Xiong, C. Yang, S.

Zhang, Y. Chen, and S. Xu, EnergyEfficient Wireless Communications: Tutorial,
Survey, and Open Issues, IEEE Wireless
Communications, pp. 28-35, December 2011.
W. Cheng, H. Zhang, L. Zhao and Y. Li,
Energy Efficient Spectrum Allocation for
Green Radio in Two-Tier Cellular Networks,
Proceedings of IEEE GLOBECOM, pp. 1-5,
2010.
B. Badic, T. O'Farrrell, P. Loskot, and J. He,
Energy Efficient Radio Access Architectures
for Green Radio: Large versus Small Cell
Size Deployment, IEEE VTC-Fall, pp. 1-5,
2009.
4.
Y. Chen, S. Zhang, S. Xu and G. Y. Li,

Fundamental Trade-offs on Green Wireless
Networks,
IEEE
Communications
Magazine, pp. 30-37, June 2011.
5. C. Han, T. Harrold, S. Armour, I. Krikidis, S.
Videv, P. M. Grant, H. Haas, J. S. Thompson,
I. Ku, C. Wang, T. A. Le, M. R. Nakhai, J.
Zhang and L. Hanzo, Green Radio: Radio
Techniques to Enable Energy-Efficient
Wireless Networks, IEEE Communications
Magazine, pp. 46-54, May 2011.
6. L. M. Correia, D. Zeller, O. Blume, D.
Ferling,Y. Jading, I. G dor, G. Auer and L.
Van Der Perre, Challenges and Enabling
Technologies for Energy Aware Mobile
Radio Networks, IEEE Communications
Magazine, pp. 66-72, Nov. 2010.
7. K. Zheng, Y. Wang, W. Wang, M. Dohler
and J. Wang, Energy-Efficient Wireless InHome: The Need for Interference-Controlled
Femtocells,
IEEE
Wireless
Communications, pp. 36-44, Dec. 2011.
8. J. Hoydis et al., Green Small-Cell NetworksCost- and Energy-Efficient Way of Meeting
the Future Traffic Demands, IEEE Vehicular
Technology Magazine, pp. 37-43, Mar. 2011.
9. Y. Wang, K. I. Pedersen, T. B. Sorensen and
P. E. Mogensen, Carrier Load Balancing and
Packet
Scheduling
for
Multi-Carrier
Systems, IEEE Transactions on Wireless
Communications, Vol. 9, No. 5, pp. 17801789, 2010.
10. S. Songsong, F. Chunyan and G. Caili, A
Resource Scheduling Algorithm Based on
User Grouping for LTE-Advanced System
with Carrier Aggregation, International
Symposium on Computer Network and
167
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
Multimedia Technology, CNMT, Wuhan, pp.

1-4, 2009.
Y. Wu, D. Zhang, H. Jiang and Y. Wu,A
Novel Spectrum Arrangement Scheme for
Femto Cell Deployment in LTE Macrocells,
2009 IEEE 20th International Symposium on
Personal, Indoor and Mobile Radio
Communications, Tokyo, pp. 6-11, Sept.
2009.
Y. Bai, J. Zhou and L. Chen, Hybrid
Spectrum Usage for Overlaying LTE
Macrocell and Femtocell, IEEE Global
Telecommunications
Conference,
GLOBECOM, Honolulu, pp. 1-6, 2009.
Z. Bharucha, H. Haas, G. Auer, and I.
Cosovic, Femto-Cell Resource Partitioning,
2009 IEEE GLOBECOM Workshops, pp. 16, 2009.
J. Kim and D. H. Cho, A Joint Power and
Subchannel Allocation Scheme Maximizing
System Capacity in Dense Femtocell
Downlink Systems, 2009 IEEE 20th
International Symposium on Personal, Indoor
and Mobile Radio Communications, Tokyo,
pp. 1381-1385, Sept. 2009.
X. Chu, Y. Wu, L. Benmesbah, and W. K.
Ling,Resource Allocation in Hybrid
Macro/Femto Networks, 2010 IEEE
Wireless Communications and Networking
Conference Workshops (WCNCW), Sydney,
pp. 1-5, Apr. 2010.
D. Das, and V. Ramaswamy, Co-channel
Femtocell-Macrocell
Deployments-Access
Control, 2009 IEEE 70th Vehicular
Technology Conference Fall (VTC 2009Fall), Anchorage, pp. 1-6, Sept. 2009.
J. Y. Kwak, J. S. Nam and D. H. Kim, A
Modified Dynamic Weighted Round Robin
Cell Scheduling Algorithm, ETRI Journal,
Vol. 24 No.5, pp. 360-372, 2002.
J. M. Holtzman, Asymptotic Analysis of
Proportional Fair Algorithm, Personal,
Indoor and Mobile Radio Communications,
Vol. 2, pp. F33-F37, Sep/Oct 2001.
Alcatel-Lucent, picoChip Designs and
Vodafone, Simulation Assumptions and
Parameters
for
FDD
HeNB
RF
Requirements, R4-092042, May 4-8, 2009.
Retrieved
in
2010,
from
http://www.ftp.3gpp.org/.
3GPP, 3rd Generation Partnership Project;
Technical Specification Group Radio Access
Networks; Home Node B radio Frequency
(RF) requirements (FDD) (Release 9), 3GPP
TR 25.967, 9.0.0. March 2009. Retrieved in
2010, from http://www.3gpp.org/.
R. K. Saha and P. Saengudomlert, Resource
Reuse and Scheduling in Femtocell Deployed
Two-Tier LTE-Advanced Systems (Masters
thesis study submitted to Asian Institute of
Technology),
Asian
Institute
of
Technology, Bangkok, 2010.
G. Yaun, X. Zhang, W. Wang and Y. Yang,
Carrier Aggregation for LTE-Advanced
23.
24.
25.
26.
27.
28.
Mobile Communication Systems, IEEE

Communications Magazine, pp. 88-93,
February 2010.
R. K. Saha and P. Saengudomlert, Novel
Resource Scheduling for Spectral Efficiency
in LTE-Advanced Systems with Macrocells
and Femtocells, The 8th Electrical
Engineering/
Electronics,
Computer,
Telecommunications
and
Information
Technology (ECTI) Association of Thailand
Conference, pp. 340-343, 2011.
A. Pokhariyal, T. E. Kolding, and P. E.
Mogensen,Performance
of
Downlink
Frequency Domain Packet Scheduling for the
Long Term Evolution, The 17th Annual
IEEE International Symposium on Personal,
Indoor and Mobile Radio Communications
(PIMRC06), Helsinki, pp. 1-5 Sept. 2006.
J. Ellenbeck, J. Schmidt, U. Korger, and C.
Hartmann, A Concept for Efficient SystemLevel Simulations of OFDMA Systems with
Proportional Fair Fast Scheduling, 2009
IEEE GLOBECOM Workshops, Honolulu,
pp. 1-6, Nov. 30 2009-Dec. 4, 2009.
K. Brueninghaus, D. Astely, T.Salzer, S.
Visuri, A. Alexiou, S. Karger and G.-A.
Seraji, Link Performance Models for System
Level Simulations of Broadband Radio
Access Systems, 2005 IEEE 16th
International Symposium on Personal, Indoor
and Mobile Radio Communications, 2005.
H. J. Bang, T. Ekman, and D. Gesbert,
Channel Predictive Proportional Fair
Scheduling, IEEE Transaction on Wireless
Communications, Vol. 7, No. 2, pp. 482-487,
February 2008.
3GPP, 3rd Generation Partnership Project;
Technical Specification Group Radio Access
Network; Evolved Universal Terrestrial radio
Access (E-UTRA); FDD Home eNode B
(HeNB) Radio Frequency (RF) requirements
analysis (Release 9), 3GPP TR 36.921,
9.0.0., March 2010. Retrieved in 2010, from
http://www.3gpp.org/.
168
A MATLAB Based Cellular Mobile Communication Laboratory

Rony Kumer Saha1 and A B M Siddique Hossain2
American International University-Bangladesh (AIUB)
Faculty of Engineering, House 82, Road 14, Block B, Kemal Ataturk Avenue
Banani, Dhaka - 1213, Bangladesh
1
rony107976@gmail.com, 2siddique@aiub.edu
ABSTRACT
In this paper, we present a novel
MATLAB
based
Cellular
mobile
communication (CMC) laboratory course
that comprises a total of ten experiments,
covering
the
fundamental
design
parameters,
considerations,
and
estimations of CMC both in the radio
interface and core network levels. In the
radio interface side, estimation of path
loss, fading, power delay profile, and
received bit error probability, while in the
core network side, estimation of link
budget
of
earth-satellite-earth
communication,
inland
microwave
communication, and radio resource
allocation and scheduling are included. In
addition, a fundamental to the MATLAB
is also introduced at the very beginning
for allowing students to understand of
how to code in MATLAB. The laboratory
was offered in fall 2012 at American
International
University-Bangladesh
(AIUB) for three sections each with a
group of 40 students. The students
comments on understanding, achieved
knowledge,
overall
satisfaction,
importance of the laboratory course, and
any changes on the existing course
content are documented and assessed.
KEYWORDS
Cellular mobile communication, experiment,
laboratory, MATLAB, software tool.
1 INTRODUCTION
CMC is one of the fastest growing

sectors in telecommunication industry
worldwide. The users of CMC have
increased rapidly in the last two decades
globally. User demands for rich
multimedia services at high data rates
are ever increasing. Telecommunication
vendors and operators have been
consistently putting significant efforts to
fulfill the user needs.
To address the high demand of CMC
services,
hands-on
experienced
workforce is a prerequisite that are
primarily supposed to be provided by the
universities. As mentioned in the table I
of reference [1], a number of universities
have already offered laboratory courses
related to the mobile wireless
communication with emphasizing on
hardware, software, or a combination of
hardware
and
software.
Several
publications, for example [1], addressed
the
importance
of
wireless
communications
laboratory
to
complement the course while few others
addressed with specific areas such as
cellular network planning [2], wireless
networks and mobile systems [3],
wireless and mobile embedded systems
[4]. However, because most universities
lack heavily from sufficient funding,
resources, and facilities, it is difficult to
setup physical laboratory of CMC to
provide
students
with
hands-on
experiences. This huge investment for
physical setup can be saved by
developing a virtual environment using
169
MATLAB software tools. With a
software based CMC laboratory,
students can model many features of
CMC, analyze and evaluate the
performance in both link and system
level. However, since a CMC laboratory
based on MATLAB tool is not obvious,
in this paper we present a novel
MATLAB based CMC laboratory.
The paper is organized as follows. In
section 2, the methodology used to
develop each experiment is discussed.
All experiments of the laboratory are
briefly described in section 3. The
laboratory assessment is incorporated in
section 4. We finish the paper with a
conclusion in section 5.
2 LABORATORY DEVELOPMENT
METHODOLOGY
In this section, we describe how the
development
of
each
laboratory
experiment is carried out. We consider a
stepwise approach to develop each
experiment
as
follows.
Step01:
introduction into which relevant
background,
problem
statement,
objective, and significance of the
experiment is described. Step02: system
or link model that incorporates three
parts: conceptual model, analytical
model, and simulation model. The
conceptual model incorporates the
system or link architecture and
configuration. The analytical model
incorporates necessary mathematical
expressions
that
transform
the
conceptual model into a methodical
demonstration. The simulation model
incorporates
typical
simulation
parameters, assumptions, and scenarios
which are used to simulate the system or
link behavior based on the analytical
model for performance evaluation. In
addition,
simulation
experimental
procedure of each experiment such as

simulation algorithm is included.
Step03: performance analysis and
evaluation where the simulation result at
system or link level is analyzed and
performance is evaluated with regard to
realistic results. Step04: predefined
experimental
question-answer
into
which a number of predefined questions
are included, and students are asked to
answer these questions using MATLAB
simulator to investigate the degree of
change in performance with the variation
of system or link parameters. Step05:
references section cites all the materials
that are used for developing the
experiment and recommended for further
study. Step06: appendices include
mainly help documents, e.g. relevant
MATLAB functions and simulation mfiles including MATLAB codes for
performance measures of an experiment.
3 LABORATORY EXPERIMENTS
In this section, we describe all the
experiments of CMC laboratory
concisely in terms of major objectives
and
significance,
necessary
mathematical
expressions
and
illustrations, and relevant performance
analysis and evaluations. A detailed
description of each experiment can be
found emailing to either of the authors.
3.1 Experiment 1: Introduction to
MATLAB
Because the CMC laboratory is
MATLAB simulator based, it is essential
for the students to get familiarized with
the use of this tool. In line with so, this
experiment is designed to give students
an overview on how to define and use
operators, functions, variables, etc. in
MATLAB. More specifically, we
170
emphasize considerably on common
mathematical
functions,
complex
numbers, elementary matrices, vector
and matrix calculations, numerical
operations and transformations of
matrices, operator precedence, general
and logical functions, data manipulation
commands, and graphics. However,
because of MATLABs wide spread
applications, we limit our focus on using
those instructions that are more relevant
to simulate the objectives of this
laboratory experiments only.
3.2 Experiment 2: Propagation
Models and Path Loss Estimation in
Cellular Mobile Communication
In this experiment, we primarily carry
out the impact of carrier frequency (f)
and distance (d) on path loss. In
addition, a sensitivity analysis is carried
out that provides critical parameters such
as base station (BS) antenna height (hb),
mobile station (MS) antenna height (hm)
is incorporated for the system design and
planning purpose.
We consider the very optimistic Freespace model, the very pessimistic
International Telecommunication Union
Radiocommunication Sector (ITU-R)
model, and the more realistic Hata
model. All models are conceptually and
analytically described, followed by
respective
simulation
performance
evaluation. We finish this experiment
with a comparison of these path loss
model simulation results as shown in
figure 1.
Figure 1. Path loss versus distance estimation

(f(MHz)=1800; hm (m)=3, hb (m)=50, d (km)=1
to 10).
In this experiment, students can

understand how the received signal
strength varies with the physical
parameters of the environment. The
received signal and noise strengths
respectively are the function of energy
per bit (Eb), and noise spectral density
(N0). Hence, estimating the required
received energy per bit over N0 (
) is
inevitably a crucial need for designing

reliable
wireless
communication
systems. This estimation provides
students with information regarding
fundamental trade-offs between received
power
and
channel
bandwidth
requirements, for example, for a given
bandwidth (BW), the channel capacity
depends directly on
3.3 Experiment 3: Estimation of

Received Bit Energy for Data Rates in
Wireless Communication
Eb
N0
Eb
N0
[5].
Denote data rate by R (bits/s), the

bandwidth utilization of a radio channel
link can be mathematically expressed
as follows.
(1)
= R BW
171
Hence, by plotting
Eb
N0
as a function
of , students are able to define regions

that are constrained either by received
power or available channel bandwidth as
given in [5].
3.4 Experiment 4: Multipath Fading
in Cellular Mobile Communication
In this experiment, students simulate
the small-scale multipath fading effect
on the transmitted signal. For non lineof-sight (NLOS) case, the received
signal envelope follows Rayleigh
distribution, and hence, the fading effect
is simulated using built-in Rayleigh
channel fading function in MATLAB.
However, when at least one LOS
component is present, the received signal
envelope follows Rician distribution,
and hence, the fading effect is simulated
using built-in Rician channel fading
function in MATLAB.
The fading effect is investigated under
both frequency-flat and frequencyselective channel conditions in order to
understand the effect of multipath
propagation over the single path. The
effect of change in Doppler spread,
symbol duration, and the Rician K-factor
on the channel response is also analyzed.
Figure 2 shows an example of Rayleigh
fading as well as Rician fading channel
responses for frequency flat as well as
frequency selective fading.
(a)
(b)
(c)
(d)
Figure 2. Frequency flat (a) Rayleigh fading and
(b) Rician fading channel responses and
frequency selective (c) Rayleigh fading and (d)
Rician fading channel responses.
3.5 Experiment 5: Power-Delay

Profile and Doppler Spectrum for
Channel Classification in Cellular
Mobile Communication
This experiment provides students
with how to classify channels by
evaluating power-delay profile and
Doppler spectrum in cellular mobile
communication. Power-delay profile and
Doppler spectrum are major indicators
for designing channel bandwidth (Bs)
and transmitted symbol duration (Ts).
172
In multipath channels, the transmitted
signal propagates over various paths, and
each path is characterized with relative
power and delay. The power-delay
profile of the channel is an indication of
the expected degree of dispersion of the
transmitted signal that defines the
maximum
possible
transmission
bandwidth (Bc) by evaluating delay
spread of the channel for proper
reception.
Similarly, the spectral broadening or
Doppler spread (fd) of the transmitted
signal, caused by the degree of relative
motion between BS and MS, defines the
maximum possible transmitted symbol
duration (Tc). Based on the relative
magnitude of Bc over Bs and Tc over Ts,
the channel can be classified as given in
table 1.
Table 1. Channel classification in mobile
communication.
Parameter
Delay
spread
Doppler
spread
Channel classification
Condition
Channel type
Frequency flat
Bc >> Bs
Frequency
Bc < Bs
selective
Slow fading
Tc >> Ts
Fast fading
Tc < Ts
3.6 Experiment 6: Design of Cellular

Mobile System
This experiment gives the students an
overview on cellular mobile system
design. With changing system scenario,
students are able to understand how the
system
design
parameters
and
requirements change for optimal
performance.
The design of cellular mobile system
depends on several issues. Hence, we
restrict the scope of system scenario
considerations to carrier-to-interference
level, cell structure, system and channel
bandwidths, environmental profile, cell
splitting and sectorization strategy,

receiver filter characteristics, traffic
distribution, trunked system nature, and
coverage area.
Given such an explicit system
scenario, students are able to estimate
system design requirements and
parameters such as frequency reuse
factor, minimum co-channel cell reuse
ratio, spectral efficiency, system
capacity, minimum adjacent channel
frequency separation, number of cells
required for the coverage area, channel
allocations with or without sectorization,
new transmit power after cell splitting,
traffic intensity, bit transfer capacity,
and trunking efficiency [6].
Figure 3 shows an example scenario
of the effect of sectorization on channel,
a total of 125 channels, allocation for
fixed channel assignment in the 7-cell
cluster. Denote cells by A, B, C, D, E, F,
and G, figure 3 shows the channel
allocation to each cell with no
sectorization.
However,
when
sectorization is considered, the channel
allocation only for the sector 1 of each
cell is shown. Similarly, the channel
allocation to other sectors such as 2 and
3 for 3-sectored cell and sectors 2, 3, 4,
5, and 6 for 6-sectored cell can be
estimated. Note that with sectorization,
the capacity is improved by a factor of 3
and 6 respectively for 3-sectored cell and
6-sectored cell as compared with the
capacity with no sectorization, however,
results in an increase in call blocking
rate.
Cell
index
A
B
C
D
No sectorization
1 8 15 22 29 36 43 50 57 64 71 78
85 92 99 106 113 120
2 9 16 23 30 37 44 51 58 65 72 79
86 93 100 107 114 121
3 10 17 24 31 38 45 52 59 66 73
80 87 94 101 108 115 122
4 11 18 25 32 39 46 53 60 67 74
81 88 95 102 109 116 123
173
E
F
G
Sector
index
A1
B1
C1
D1
E1
F1
G1
5 12 19 26 33 40 47 54 61 68 75
82 89 96 103 110 117 124
6 13 20 27 34 41 48 55 62 69 76
83 90 97 104 111 118 125
7 14 21 28 35 42 49 56 63 70 77
84 91 98 105 112 119 0
With sectorization
3-sectored cell
6-sectored
cell
1 22 43 64 85 106
1 43 85
2 23 44 65 86 107
2 44 86
3 24 45 66 87 108
3 45 87
4 25 46 67 88 109
4 46 88
5 26 47 68 89 110
5 47 89
6 27 48 69 90 111
6 48 90
7 28 49 70 91 112
7 49 91
increase in the degree of modulation

results in more bit errors for PAM, PSK,
and QAM.
Figure 3. Fixed channel assignment to cells in a

7-cell cluster with no sectorization, 3-sectored
cell, and 6-sectored cell.
3.7 Experiment 7: Estimation of Bit

Error Probability of Modulation
Schemes
In this experiment, we are primarily
concerned with finding an appropriate
modulation scheme at the transmitting
side using bit error probability
measurement. We consider modulation
schemes such as M-point pulse
amplitude modulation (M-PAM), Mpoint phase shift keying (M-PSK), Mpoint quadrature amplitude modulation
(M-QAM) and M-point orthogonal
signal sets. We then draw a conclusion,
based upon the results obtained from
these considered schemes, on selecting
an appropriate modulation scheme that
can be well suited in power-limited
region and bandwidth-limited region.
The followings (figures 4 and 5) are
example line graphs for bit error
probability versus
Eb
, and it can be
N0
found that irrespective of the degree of

modulation, PAM is susceptible more to
bit errors than PSK, QAM and
orthogonal signal sets. In addition,
Figure 4. Bit error probability for 4-PAM,

2 2 QAM and 4-point orthogonal signal sets.
Figure 5. Bit error probability for 16-PAM,

4 4 QAM and 16-point orthogonal signal
sets.
However, the opposite is the case for

orthogonal signal sets. Overall, PAM and
QAM schemes are bandwidth-efficient,
and orthogonal signal sets are powerefficient [7].
174
3.8 Experiment 8: Radio Resource
Allocations and Scheduling in Cellular
Mobile Communication
Radio resource allocation and
scheduling is one of the important areas
in mobile communication that needs
considerable attention to address several
significant issues such as spectral
efficiency, energy efficiency, and quality
of service. Scheduler defines the specific
usage of physical resources such as time,
frequency, and power. In this
experiment, students evaluate the
performance of the generic schedulers
such as Round Robin (RR), Max-SNR,
and Proportional Fair (PF) in terms of
average cell spectral efficiency and
Jains fairness index in a macrocell of
LTE-Advanced systems.
The system model considers a dense
urban deployment. A single macrocell is
considered. A number of macro users
(per cell for a 3-sector site) are assumed
within the macrocell coverage (figure 6).
Figure 6 shows the cell layout in the
system architecture, where macro users
are distributed randomly and uniformly
throughout the macrocell.
The throughput in any arbitrary
resource block (RB) can be expressed in
terms of the mutual information (MI) by
employing Shannons formula as given
below [8] [9].
0,
10dB
( dB)
( ) log 2 1 10 10 , 0dB 22dB
4.4,
22dB
(2)
where is the SINR at any RB in a

particular transmission time interval
(TTI); = 0.6 is the implementation loss
factor.
The average MI ( ) over all RBs over
all users in a component carrier (CC) is
obtained, and then the effective
ISD
3R
A single macrocell
site
Figure 6. Cell layout in the system architecture

[8].
SINR, eff per RB per TTI is estimated

from the average MI by inverse
mapping. The inverse mapping function
can be mathematically expressed for as
eff ( ) 10 log10 (2 1)
(3)
efficiency per RB per TTI can be
expressed as
(4)
Rsys = log 2 ( 1 + eff )
Jains fairness index is adopted for the
evaluation of the user fairness
performance to define the degree of
fairness among users. Mathematically,
the fairness index can be expressed as
follows [10].
2
uk
FJ k 1K
2
K uk
(5)
k 1
where K represents the total number of

user and uk represents the total number
of RB allocated to user k over the
simulation runtime.
Simulation
parameters
and
assumptions are adopted mainly from
[11], which are based on the system
architecture explained. The spectral
efficiency performance and the fairness
performance of these schedulers are
shown in figure 7. From figure 7, it can
be found that RR scheduler provides the
best fairness but the least spectral
175
efficiency performance. On the contrary,
Max-SNR scheduler provides the best
spectral efficiency but the worst fairness
performances.
However,
the
PF
scheduler provides a trade-off between
fairness
and
spectral
efficiency
performances by taking the user
previous resource allocation history into
account.
transmitter and receiver with gain and

loss in different parts.
Given that transmit power is 0 dBW,
total cable loss is 4 dB, carrier frequency
is 6 GHz, intermediate frequency
bandwidth of the receiver is 10 MHz,
receiver is at room temperature, and
90% radio link reliability, students can
estimate the effective isotropic radiated
power (EIRP), free-space path loss of
the hop, power flux density at the
receiver front end, receiver noise
threshold, antenna aperture diameter,
and the received power. Figure 9 is a
simulated result for a single hop
microwave link. All gains and losses are
shown along with the signal propagation
from transmitter to receiver.
3.10 Experiment
Satellite Link
Figure 7. Spectral efficiency and fairness

performances of generic resource schedulers.
3.9 Experiment 9: Design of Inland

Digital Microwave Link
In this experiment, students study the
basic concepts of high frequency wave
propagation, optical and radio horizon,
ducting phenomenon, Fresnel zones,
earth bulge, and factor k: a ratio of
effective earth radius to the true earth
radius and its effect on the link range.
Figure 8 shows an example point-topoint microwave backhaul between
10:
Design
of
Satellite backhaul in mobile systems is

a solution where other alternative
backhaul technologies are not feasible
such as islands, remote locations [12].
Typically, a long-distance network
through satellite link is designed and
established through geostationary earth
orbit (GEO) satellites at frequencies in C
band (6-4 GHz) and in Ku band (14-12
GHz). In this experiment, students study
the fundamental issues and constraints in
designing
earth-satellite-earth
communication link as given in figure
10.
Refer to figure 10, given the
parameters such as transmit power,
satellite antenna gain, uplink and
downlink slant range, and satellite
antenna gain-to-noise temperature ratio,
we can estimate the earth stationsatellite-earth station link budget. In
addition, uplink carrier-to-noise ratio,
(C/N)u, downlink carrier-to-noise ratio,
(C/N)d, and the total link carrier-to-noise
176
Antenna
gain
Antenna
gain
LOS-RRL Free-space loss
Diameter
Fade margin
transmitter
NF
Cable
loss
Cable
loss
receiver
received
power
Transmit
power
A single-hop microwave link with no repeater
Distance
Figure 8. A typical single hop microwave communication link.
Figure 9. Typical link budget estimation of a LOS microwave link.
nu (t)
su (t)
Receiver
Down
converter
TWTA
Gu/Tu
EIRPs
su (t ) + n u (t )
s(t)
EIRP = Pt Gt
nu (t)
G/T
HPA
Pt
LNA
Gt
st nt
Figure 10. Basic earth-satellite-earth communication link [13].
177
ratio, (C/N)total can also be estimated. If
(C/N)u >>(C/N)d, then C/N (C/N)d. In
this case the satellite link is said to be
downlink-limited.
When
(C/N)u
<<(C/N)d, then C/N (C/N)u, and the
satellite link is said to be uplink-limited.
We consider both single carrier and
multi-carrier per transponder for the link
budget estimation and carryout an
analysis of how multi-carrier scenario
affects the link performance. Followings
as given in table 2 are parameters that
are considered as an example scenario to
estimate the satellite link performance
for the operation of single carrier per
transponder.
Table 2. Default simulation parameters and
scenarios for satellite link [13].
Parameter
Transponder operation band
Noise bandwidth
Satellite antenna gain-to-noise
ratio
Satellite saturation EIRPs
Satellite TWTA BOi
Satellite TWTA BOo
Earth station transmit antenna
gain
Earth station receive antenna
gain
Earth station carrier power fed to
antenna
Maximum uplink and downlink
slant range
System noise temperature Ts
Uplink tracking loss
Downlink tracking loss
Interference into or from
adjacent satellite, terrestrial
interference, etc. and any
influence by the atmosphere
Value
14/12 GHz
46 MHz
-1.6dB/K
44 dBW
11 dB
6 dB
57.6 dB
56.3 dB
174 W
37,506 km
160 K
1.2 dB
0.9 dB
Negligible
The simulation output gives the

desired link budget along with other
measures as given in table 3 for single
carrier per transponder operation.
Table 3. Output simulation parameters and
values for satellite link.
Parameter
Carrier EIRP
Uplink free space loss
Uplink carrier-to-noise ratio
Satellite EIRP
Downlink free space loss
Downlink carrier-to-noise ratio
Total satellite link (earthsatellite-earth) carrier-to-noise
ratio
Satellite characteristics
Value
80.0055
dBW
206.8463 dB
23.3591 dB
44 dBW
205.5074 dB
24.8926 dB
127.2986 dB
uplink
limited
4 LABORATORY ASSESSMENT
In order to assess the significance of
the CMC laboratory, we carried out a
student survey within the registered
students of the CMC laboratory of fall
2012 semester at AIUB. The total
number of students participated in the
survey was ninety one. The survey
questions were categorized into two
parts: part I incorporates general
questions and part II incorporates query
based questions, with five questions in
each part (table 4). Students were asked
to put a tick on either Yes or No, with an
option for making comments, explaining
the reasons for their opinions.
Figure 11 shows the response of the
students for each question of the survey
(table 4). From figures 11 (a) and (b), it
can be found that overall most of the
students
Table 4. Student survey questions of CMC
laboratory assessment.
No
Q01
Q02
Q03
Question
Part I: General category
The laboratory experiments are relevant
and helpful for understanding the
contents of the course cellular mobile
communication.
The experiments of the lab are easy to
understand and have been written
properly with sufficient details.
The lab provides sufficient realization
of the practical aspects of cellular
178
Q04
Q05
Q06
Q07
Q08
Q09
Q10
mobile communication using MATLAB

software tool.
The lab is helpful for further research
and professional career development in
mobile communication.
The lab is equipped with sufficient
resources for simulation.
Query based
Is the laboratory time of three hours
duration sufficient for the experiments?
Is there any prerequisite the students
should complete before they enroll for
the lab?
Should there be any changes in the
content, strategy, methodology, etc. of
the lab?
Is there any deficiency in the lab
experiments that should be addressed?
Is there any comment on improving the
lab?
responded to Yes. Particularly, majority

have agreed upon questions 01 through
05, i.e. the laboratory is relevant and
helpful for understanding the CMC
course, provides sufficient realization of
the practical aspects of CMC, and is
helpful for further research and
professional career development in
mobile communication. In addition,
according to the majority of the students,
the experiments have been written
properly with easy to understand
features, and the lab is equipped with
sufficient resources for simulation.
From figure 11 (b), for query based
questions, 06 to 10, majority of the
students responded in question 06 to 07
to Yes, meaning specifically, the
laboratory time is sufficient, and the
students should complete prerequisites
such as digital signal processing,
telecommunications engineering, and
how to work with MATLAB. In
question 08, students responded almost
equally, meaning, they were confused. In
question 09, majority opinioned there
was no significant deficiency in
laboratory experiments. Moreover, in the
final question 10, students were asked to
make comments on improving the
laboratory. Selected few comments

made by a considerable number of
students are as follows: arrangement of
few field visits to telecommunication
industry may facilitate students gaining
some real life experiences, few
practical types of equipment can be
introduced, and understanding of
MATLAB coding should be a
prerequisite.
(a)
(b)
Figure11. Response of the students to the
survey questions (a) general category (b) query
based.
5 CONCLUSION
In this paper, we introduced a
MATLAB based cellular mobile
communication
(CMC)
laboratory
course. The laboratory course comprises
of ten experiments, covering the
fundamental
design
parameters,
179
considerations, and estimations of CMC
both in the radio interface and core
network levels. A survey over ninety one
students was carried out to assess the
importance and impact of the laboratory.
Almost all students found the
laboratory experiments relevant and
helpful for understanding the CMC
theory course and for further research
and professional career development.
Since the laboratory is just MATLAB
software based, the laboratory can be
offered at affordable cost to provide
students to gain practical realization of
the CMC theory course concepts at the
university. The authors would be pleased
to share the laboratory course materials
with any individual or institution
interested in it and can be reached at
either of the authors email addresses.
5.
6.
7.
8.
9.
ACKNOWLEDGEMENT
The authors are thankful to Abdur
Rahman, Head of Undergraduate
Program and Farhadur Arifin, an
Assistant Professor of the EEE
department, AIUB, Bangladesh for their
generous support to develop the
laboratory.
10.
11.
REFERENCES
1.
2.
3.
4.
F. Cassara, Wireless Communications

Laboratory,
IEEE
Transaction
on
Education, Vol.51, no.3, pp. 132-140, Feb.
2006.
Z. Dawy, A. Husseini, E. Yaacoub, and L.
Al-Kanj, A Wireless Communications
Laboratory on Cellular Network Planning,
IEEE Transaction on Education, Vol.53,
no.4, pp. 653-661, Nov. 2010.
S. F. Midkiff, An Experiential Course in
Wireless Networks and Mobile Systems,
IEEE CS and IEEE ComSoc, pp. 9-13, Jan.Mar. 2005.
J. Chenard, Z. Zilic, and M. Prokic, A
Laboratory
Setup
and
Teaching
Methodology for Wireless and Mobile
12.
13.
Embedded Systems, IEEE Transaction on

Education, Vol.49, no.1, pp. 378-384, Aug.
2008.
E. Dahlman, S. Parkvall & J. Skld, 4G
LTE/LTE-Advanced
for
Mobile
Broadband, 1st ed., Oxford, UK: Academic
Press, pp. 15-26, 2011.
T. S. Rappaport, Wireless Communications
Principles and Practice, 2nd ed.,
NJ:
Pearson Education International, pp. 25-67,
2002.
P.
Saengudomlert,
Digital
Communications, Asian Institute of
Technology (AIT). Bangkok, Thailand, pp.
77-96, January 2011. [online] Available:
http:// www.tc.ait.ac.th/faculty/poompat/.
3GPP; Technical
Specification Group
Radio
Access
Networks,
Evolved
Universal Terrestrial Radio Access (EUTRA); Radio Frequency (RF) System
Scenarios (Release 9), 3rd Generation
Partnership Project, TR 36.942, Vol. 9.0.1,
pp. 12-13, April 2010.
J. Ellenbeck, J. Schmidt, U. Korger, and C.
Hartmann, A Concept for Efficient SystemLevel Simulations of OFDMA Systems with
Proportional Fair Fast Scheduling, IEEE
GLOBECOM Workshops 2009, USA, pp. 16, Nov. 30 2009-Dec. 4 2009.
H. J. Bang, T. Ekman and D. Gesbert,
Channel Predictive Proportional Fair
Scheduling, IEEE Trans. on Wireless
Commun., Vol. 7, no. 2, pp. 482-487, Feb.,
2008.
3GPP; Technical specification Group Radio
Access Network, Evolved Universal
Terrestrial radio Access (E-UTRA); FDD
Home eNode B (HeNB) Radio Frequency
(RF) requirements analysis (Release 9), 3rd
Generation Partnership Project, TR 36.921,
Vol. 9.0.0, pp. 32-36, Mar. 2010.
O. Tipmongkolsilp, S. Zaghloul, A. Jukan,
The Evolution of Cellular Backhaul
Technologies: Current Issues and Future
Trends, IEEE Commun. Surveys &
Tutorials, Vol. 13, no. 1, pp. 97-113, First
Quart. 2011.
K. M. Ahmed, Satellite Communications,
Asian Institute of Technology (AIT).
Bangkok, Thailand, lecture 05, Aug. 2010.
180
An Optimized Energy Aware Chaining Technique for Data Gathering

in Wireless Sensor Networks
Ouadoudi Zytoune1 And Driss Aboutajdine2
1
:ENCG, Universit Ibn tofail, Knitra, Morocco. Membre associ au LRIT
2
:Faculty of Sciences, LRIT unit associe au CNRST(URAC29),
University Mohamed V-Agdal. Rabat, Morocco
zytoune@univ-ibntofail.ac.ma, aboutaj@fsr.ac.ma
ABSTRACT
Wireless Sensor Network (WSN) is a
collection of small sensor nodes with
aptitude to sense, compute and transmit data
that are deployed to observe a physical
environment. The sensor node has limited
capabilities, especially its energy reserve, its
processing ability and its memory storage.
Data dissemination and gathering protocols
design for WSN are crucial challenges since
those protocols should be easy, energyefficient, and robust to deal with a very large
number of nodes. Also, they should be selfconfigurable to node failures and dynamic
changes of the network topology. In this
paper, we present a new algorithm for
gathering sensor reading based on chain
forming using Ant Colony Optimization
(ACO) technique. To allow network lifetime
extension, the ACO provides the shortest
network nodes chaining instead of starting
from the furthest node and using Greedy
algorithm as PEGASIS do. The leader role
duration is defined for each node based on
its required energy to do this role in the
established chain. Which avoids fast nodes
energy depletion and then, the network
lifetime would be extended. Through
simulation, it is proved that the proposed
algorithm allows network stability extension
compared to the most known chaining
algorithm.
KEYWORDS
Wireless Sensor Networks; Data Gathering;

ACO;
PEGASIS;
Network
lifetime
extension.
1 INTRODUCTION
A Wireless Sensor Network (WSN) is
composed of a collection of tiny and
lightweight sensor nodes deployed in
large number to monitor the surrounding
conditions [1]. Those nodes cooperate to
collect, process and transmit the
surrounding
conditions
such
as
temperature, humidity, vibration The
WSN have many application domains
such as environmental survey, smart
home,
medical
and
agricultural
monitoring, etc.
Some of the early literature works on
WSNs have discussed the benefits of this
kind of networks compared to MANETs
[2, 3, 4]. WSNs have main advantages
over the conventional networks deployed
for the same purpose such as greater
coverage, accuracy, reliability and all of
the above at a possibly lower cost.
Since the sensor node has small size, and
maybe deployed in hostile region, its
available energy is considered as the
main constraint. Therefore, much
attention must be given to the energy
consumption when designing protocols
for this kind of networks [5, 6].
WSN research domain is much active so,
in the last few years, a variety of
protocols have been proposed for
prolonging the WSN duration service
181
when gathering the sensor reading to the
sink. The most of those protocols can be
divided as either flat or hierarchical
based.
In densely deployed sensor network, an
event is often detected by more than one
sensor and duplicated data are generated.
This data redundancy is often
eliminated, which cannot only decrease
the global data to be transmitted and
localized most traffic within individual
groups, but also, reduces the traffic and
consequently, contention in a wireless
network, which allows decreasing
energy consumption.
Data aggregation is a way to reduce
energy consumption. This aggregation
consists of suppressing redundancy in
different data messages. This is the key
idea for the most hierarchical routing
protocols.
In addition, scalability is one of the key
design attributes of sensor networks.
Since a single-tier network can conduct
the gateway to overload with the
increase in sensors density, the main
objective of hierarchical routing is to
efficiently
preserve
the
energy
consumption of sensor nodes by
involving
them
in
multi-hop
communication within a particular
cluster and by performing data
aggregation and fusion that decrease the
number of data messages that would be
transmitted to the sink.
In literature, many research works have
explored hierarchical routing in WSN
from different perspectives. Some of the
hierarchical protocols are LEACH [7, 8],
PEGASIS [9], TEEN [10,11], SEP [12],
DEEC [13] and APTEEN [14].
LEACH is the most popular energyefficient
hierarchical
clustering
algorithm that was proposed for
reducing power consumption in WSN by
doing data aggregation. It uses clustering
technique to prolong the lifetime of the

WSN where a cluster-head (CH) collects
the data from all nodes in its
neighborhood, fuses and sends the
information, on one data packet, to the
Base Station (BS). In order to reduce the
data quantity that would be transmitted,
the CH uses an aggregation technique
that combines the original data into a
smaller size of data that carry only
meaningful information to all individual
sensors. Thus, LEACH reduces the
number of nodes communicating directly
with BS and allows better network
lifetime extension.
PEGASIS [9] is a chain-based protocol
that allows minimizing the energy
consumption at each sensor node. The
main idea of this protocol is doing data
aggregation over all the network nodes
that are chained. This protocol is
considered as an optimization of the
LEACH since that rather than
classifying nodes in clusters, the
algorithm forms chains of all the
network sensor nodes. Based on this
arrangement, each node transmits to and
receives from only one closest node of
its neighbors in the chain. Unlike
LEACH that uses hierarchical clustering,
PEGASIS uses a flat topology that
permits to avoid the overhead of
dynamic cluster formation as in LEACH.
In PEGASIS all sensor nodes are
organized to form a chain for data
transmission and reception. Each node of
the formed chain take turns being the
leader for communication to the BS.
Data gathering round starts from each
endpoint of the chain. Data are
aggregated along the path to the
designated leader node that transmits the
aggregated data to the BS as depicted in
Figure 1. Accordingly, PEGASIS
achieves a best reduction in energy
consumption as compared to LEACH
182
because it requires only one designated
node to send one combined data to the
BS.
A greedy algorithm is proposed to be
used in PEGASIS to form the chain;
each node selects the closest neighbor
that is not chained yet and so as until all
network nodes are chained. This
algorithm is executed before starting the
first round of data transmission. To
construct the chain, PEGASIS starts with
the furthest node from the BS. This is to
make sure that nodes farther from the BS
have close neighbors, since in the greedy
algorithm the neighbor distances will
augment gradually because nodes
already on the chain cannot be revisited
again. After chain formation, data
transmission phase will start. In this
phase, node can deplete its residual
energy, and then the chain will be
reconstructed in the same manner to
avoid the dead node.
proposed in [15] is to use the Ant

Colony Optimization to form the
network chain. Both the two techniques
start the chain from the farthest node to
the BS to ensure that this node has a
neighbor. By the provided assessment, a
slit enhancement of the network lifetime
is observed.
In this paper we present a new technique
for gathering data in order to extend the
network lifetime. The main object is to
reduce the energy consumed by network
nodes and then extends the network
service duration by choosing the shortest
chain. The remainder of the paper is
arranged as follows. Section-2 provides
the problem statement. The detail of the
proposed technique has been discussed
in section-3. Simulation parameters and
results have been given in section-4.
Based upon the simulation results,
conclusions have been drawn and some
recommendations for future work have
been proposed in section-5.
2 PROBLEM STATEMENT
Figure 1: Chain in PEGASIS
To enhance PEGASIS, the paper [15]

presented two techniques; The main idea
of the first technique consists to allow
each node to become leader for
Xi=(d2Bref/d2Bi).Xref times, where dBref is a
distance reference to BS, dBi is the
distance between node i and the BS and
Xref is an arbitrary coefficient to
overcome the error by rounding Xi to the
nearest integer. The second technique
2.1 Problem Definition

In this work we are interested in WSN
composed from N wireless nodes
deployed randomly in the monitored
area. The main object of this work is to
extend the network service duration until
the first node dies so called stable
network lifetime. Which means that the
first node has its residual energy
depleted and then is failed to play its
function in the network. We assume the
BS is fixed at a far distance from the
sensor nodes that are not mobile.
In this paper we consider a flat network,
that the data gathering is based on chain.
The network nodes are organized in
chain and at any transmission round, one
node is selected as leader to collect data
in the chain and transmit it to the BS.
183
PEGASIS [9] and Seetharam et al. [15]
propose that the chain forming starts
from the farthest node to the BS. As the
chain length will not be the shortest one,
the network lifetime is not optimized.
The radios have power control and can

expend the minimum required energy to
reach the intended recipients. The radios
can be turned off to avoid receiving
unintended transmissions.
2.2 Energy Dissipation Model
To receive an l-bits message, the radio

expends (2):
In this work, we assume a simple

scheme for modeling the radio hardware
energy dissipation as discussed in [7].
The transmitter dissipates energy when
running the radio electronics and the
power amplifier. The receiver dissipates
energy to run the radio electronics, as

shown in Figure 2. For the experiments
described here, the free space channel
model is used. Consequently, to transmit
an l-bits message over a distance d, the
radio expends (1).
Figure 2. Radio energy dissipation model.
ETX l.E elec l.E amp .d 2
(1)
Where Eelec is the energy dissipated per

bit in the transmitter circuitry (to run the
transmitter or receiver circuitry in case
of receiving) and Eamp.d2 is the energy
dissipated for transmission of a single bit
over a distance d.
The electronics energy (Eelec) depends on
many factors such as the digital coding
operation, the signal modulation and
demodulation, filtering, and spreading,
whereas the amplifier energy, Eamp.d2,
depends on the distance to the receiver
and the acceptable bit-error rate.
E RX l.E elec
(2)
It is also assumed that the radio channel

is symmetric, which means the cost of
transmitting a message from node A to
node B is the same as the cost of
transmitting a message from B to A.
3 OPTIMIZED ENERGY
EFFICIENT DATA GATHERING
TECHNIQUE FOR WIRELESS
SENSOR NETWORK
In this work we aim at developing a
data-gathering scheme that extends the
network lifetime by reducing energy
dissipation of all the network nodes.
Also, we attempt to equitably distribute
the energy load between network nodes.
As we can consider the nodes chaining
problem analog to the traveling
Salesman one, we can use the Ant
Colony Optimization to form the
shortest data gathering chain. Since
PEGASIS starts chaining from the
farthest node to the sink, it cannot
provide the shortest chain as Figure 3
gives an illustration. Ten nodes are
scattered randomly in the network area.
We use the greedy algorithm to
determine the network data gathering
chain. Thus as depicted in this Figure,
the PEGASIS chain is formed as {1, 7,
6, 4, 3, 9, 8, 10, 2, 5} but the shortest
chain is {5, 2, 3, 9, 8, 10, 4, 6, 7, 1}.
Comparing the chains length, PEGASIS
uses a chain 18.21% long than the short
chain. That means that the energy
consumption is greater.
184
the edge (i,j) is updated as follow:
m
(5)
ij
ij .(1 ) ijk
k
is the pheromone evaporation rate. ijk
Figure.3. Nodes chaining in PEGASIS is not the

energy efficient.
It is so difficult to determine the shortest

chain since the problem is NP hardness.
We propose a meta-heuristic method
based on Ant colony optimization [16].
The problem is then considered as the
Traveling Salesman Problem where the
optimization
searches
to
find
heuristically the shortest path to visit all
nodes one time.
The ant colony
optimization works as follow; an ant n
starts from a random node i and selects
the next node j randomly based on the
probability Pij given in Equation (3):
ij .ij

if
k Vi
Pij ik .ik
(3)
k
0 Otherwise
Where ij is the pheromone value of
edge (i-j), Vi contains neighbor nodes of
node i that are not yet visited by ant n.
The parameters and control the
relative importance of the pheromone.
And ij is the heuristic information,
which is given by (4):
1
ij
(4)
dij
Where the distance between nodes i and
j is dij.
When an ant passes through an edge, it
deposits a pheromone value. In order to
avoid local optima this algorithm uses
the evaporation, which periodically
reduces the pheromone value deposited
on a trail. Thus, the pheromone ij over
is the pheromone quantity deposed by

ant k on this edge(i,j) and m is the ants
number that are traveled by this edge.
j)
Q/Lk if ant k traveled by
edge(i,
(6)
ijk
Otherwise
Where Q is a constant and Lk is the

length of the path traversed by ant k.
After many tours, the algorithm
convergences to the short path since
each edge of this path would have the
maximal pheromone.
As the chain is formed, we aim at
developing a system that would provide
that the total energy dissipation is
distributed equally among all the
network nodes to ensure that the network
nodes die equitably.
In every data-gathering round, a leader
node is selected to receive data from the
chain and transmit it to the BS. During
every round, each node receives a data
packet from its neighboring node in the
chain, aggregates it with its own data
packet and transmits it to its other
neighbor in the chain. A simple token
passing approach initiated by the leader
is used to organize the data transmission
in the chain. Often, the data transmission
starts from the chain end-nodes to its
next nodes in the chain, which do data
aggregation and so that until the leader.
The leader elected in a particular round
receives the fused data packets of the
nodes in the chain from its two
neighbors, aggregates it with its own
data packet and finally this single data
packet is transmitted to the base station.
Because leader role is energy greedy
consuming, PEGASIS ensure that at
each transmission round a new node is
selected as leader in order to distribute
energy load. Since, the nodes consume
185
different energy when being leader
depending on its location in the chain
and its distance to the BS, which means
that a certain nodes would die quickly.
So, let network node plays leader
depending on its energy capacity can
ensure best energy load balancing.
Let consider a network formed by N
chained nodes and Ei0 is the initial
battery energy of the node i. Let EiBS is
the required energy for the node i to
transmit to the base station, Eij is the
energy to transmit from the node i to
node j and Erec is the energy cost
required for receiving a packet.
The node i deployed in the monitored
area can do Ti transmissions in its
lifetime (until the depletion of its
residual energy) in which it is the leader.
For notation simplicity, we suppose that
the chain nodes are organized as follow;
{1, 2...N} and then the end chain nodes
are 1 and N. For the node 1, that will be
leader for T1 times, will do T1
transmission
to
the
BS
and
T2+T3+...+TN transmissions to the next
node in the chain, lets node 2. Then, the
equation (7) must be respected.
E10 T1.(E1BS E rec ) (T2 T3 ... TN ).E12 (7)
For a not end chain node i, this node will

do Ti transmission to BS (node i will be
leader for Ti times) and 2.Ti.Erec from the
two chain elements when it is a leader
and consumes the energy (T1+T2+...+Ti1).(E(i)(i-1)+Erec) for transmitting to the
left element of chain and (TN+TN1+...+Ti+1).(E(i)-(i+1)+Erec) to transmit to
the right element of the chain. E(i)(i-1) and
E(i)-(i+1) correspond respectively to the
transmission energy from node i to its
predecessor and to its successor in the
chain. The equation (8) gives the relation
between the numbers of becoming
leaders for the entire network nodes.
E1BS E rec
E12
...
E12 T1 E10

... T2 E 20
E 21 E rec E 2BS 2.E rec E 23 E rec
.
(8)
...
...
...
... ...

E 21
...
E NBS E rec TN E N 0
E 21
As constraint, all Ti must be positives.

We solve the equation above to
determine how many times a node can
be a leader for the chain. If the solution
gives negative results, a linear
optimization can be used. With this
manner the nodes energies would be
consumed equitably.
Since Ti is given, the node i when it is
selected as leader it plays this role for Ti
consecutive transmission rounds rather
than one transmission round as in
PEGASIS.
In the following, the proposed technique
is described.
3.1 Set-up Phase
The ant colony optimization algorithm is
used the find the shortest chain. Thus, N
ants are randomly deployed in the N
network nodes. Each ant starting from an
arbitrary node selects the next node to
visit using Equation (3), and so on until
it makes a tour. Thus, the pheromones
are updated. The same algorithm is run
till convergence to the optimal chain.
3.2 Steady State Phase
After the shortest chain is formed, data
gathering will start. In every datagathering round, a leader node is
selected to receive data from the chain
and transmit it to the sink. During a data
gathering round each node in the
network receives a data packet from its
neighboring node (in the chain),
aggregates it with its own data packet
and transmits it to its other neighbor in
the chain. A simple token passing
approach initiated by the leader is used
to organize the data transmission. The
data transmission starts from the chain
end-nodes to its next nodes in the chain,
186
which do data aggregation and so that
until reaching the leader. This leader
aggregates the received data with its own
and transmits it the sink.
Since the leader consumes considerable
energy, and in order to distribute energy
load, each node i that becomes leader
plays this role for Ti consecutive
transmissions rounds.
4 SIMULATION PARAMETERS
AND RESULTS
To evaluate the performance of our
technique, several MATLAB simulations
were performed and the represented
results are an average. We consider a
square network with N nodes deployed
randomly in the field. The used
parameter values in our work are given
in Table1.
Table1: Simulation Parameter Values.
Description
Symbol
Network dimension
Xm*Ym
Number of network N
nodes
Data packet length
L
Electronic Energy
Eelec
Amplifier Energy
Eamp
location from (0.5Xm, 0.5Ym) to (0.5Xm,

3Ym). The simulation results are shown
in Figure.5. As depicted, the network
lifetime of booth the protocols decrease
when the BS is far from the network
because the needed energy to reach the
sink increases with the distance. The
proposed technique allows extension of
the network lifetime for every base
station position.
Figure.4. Network lifetime vs. Network nodes

number.
Value
100m*100m
10-200
2500 Bits
50nJ/bit
100pJ/bit.m2
First we run simulation of our scheme

called Energy Efficient Data Gathering
technique (EEDG) and PEGASIS
varying the network nodes number from
100 to 300. All the network nodes have
the same initial energy that is 0.5J. The
base station is located at (50m, 200m).
We are interested at the network lifetime
until the first node run out its residual
energy. The results are represented in
Figure.4. As depicted, our scheme
performs better than PEGASIS since the
lifetime extension is up to 20%.
In the second situation we intend to
investigate the effect of the Base Station
location on the performance of the
proposed algorithm. We consider a
network of 100 nodes with initial energy
of 0.5J, and we vary the base station
Figure.5 Network lifetime for different BS

location.
Figure.6 represents the network lifetime

for different node initial energy. As we
can observe, the network lifetime is
extended for all the considered node
initial energy. The relative lifetime
extension is between 22,96% and
24,61%.
Figure.6 Network lifetime for different nodes

initial energy.
187
The simulation results presented above
illustrate that the proposed scheme
exploits judiciously the network energy
since it selects the short network chain
so that the transmission energy is
optimized. It exploits, also, the capacity
to play leader role for each node to
ensure a best energy load repartition
over chain nodes.
5 CONCLUSION
Network nodes chaining is a better solution
to extend network lifetime since only the
leader node that do long transmission to the
BS. In this paper, we proposed an Ant
colony optimizing technique to form
network nodes chain ensuring that this chain
is the short one. Chain nodes become leader
for a lot of time based on its required energy
to transmit to the BS and to its two chain
neighbors. The proposed technique allows
extra network transmission rounds as
compared to PEGASIS. This extension is
guaranteed for different number of network
nodes, for different BS location and for
different nodes initial energies. The
simulation results show that our technique
outperforms to the well-known protocol for
chaining in wireless sensor networks
(PEGASIS protocol). In future, we will
continue the work investigating the effect of
data correlation on the network performance
for this kind of routing protocols.
6 REFERENCES
1.
2.
3.
Akyildiz I. F., Su W., Sankarasubramaniam

Y., Cayirci E.: A survey on sensor networks.
IEEE Communications Magazine, vol. 40,
no. 8, pp. 102-114, 2002.
Chandrakasan A., Amirtharajah R., Cho S.,
Goodman J., Konduri G., Kulik J., Rabiner
W., Wang A.: Design considerations for
distributed micro-sensor systems. In Proc.
IEEE 1999 Custom Integrated Circuits
Conference (CICC99), San Diego, CA,
USA, May 1999, pp. 279-286.
Estrin D., Girod L., Pottie G., Srivastava M.:
Instrumenting the world with wireless
sensor
networks.
In:
Proc.
IEEE
International Conference on Acoustics,

Speech and Signal Processing (ICASSP
2001), Salt Lake City, Utah, USA, May
2001, vol. 4, pp. 2033-2036.
4. Estrin D., Govindan R., Heidemann J.,
Kumar S.: Next century challenges: Scalable
coordination in sensor networks: In: Proc.
5th Annual ACM/IEEE International
Conference on Mobile Computing and
Networking
(MobiCom'99),
Seattle
Washington, USA, August 1999, pp. 263270.
5. Nguyen L. T., Defago X., Beuran R.,
Shinoda Y.: An Energy Efficient Routing
Scheme for Mobile Wireless sensor
Networks. In: Proc. IEEE ISWCS 2008, pp.
568-572.
6. Yrjola J.: Summary of Energy Efficiency
Communication Protocol for Wireless Micro
Sensor Networks", 13th March 2005.
7. Heinzelman W.R., Chandrakasan A.,
Balakrishnan
H.:
Energy
efficient
Communication Protocol for Wireless
Microsensor Networks. In: Proc. IEEE
Computer Society of the Thirty Third
Hawaii International Conference on System
Sciences (HICSS'00), Washington, DC,
USA, Jan. 2000, vol. 8, pp. 8020.
8. Heinzelman W.R., Chandrakasan A.,
Balakrishnan H.: An Application-Specific
Protocol
Architecture
for
Wireless
Microsensor Networks. IEEE Transactions
on Wireless Communications (October
2002), vol. 1(4), pp. 660-670.
9. Lindsey S., Raghavendra C.S.: PEGASIS:
Power efficient Gathering in Sensor
Information System. In: Proc. IEEE
Aerospace Conference, vol. 3, Big Sky, MT,
Mar. 2002, pp. 1125-1130.
10. Manjeshwar A., Agrawal D. P.: TEEN: A
Protocol for Enhanced Efficiency in
Wireless Sensor Networks. In: Proc. 1st
International Workshop on Parallel and
Distributed Computing Issues in Wireless
Networks and Mobile Computing, San
Francisco, CA, April 2001.
11. Lou W.: An Efficient N-to-1 Multipath
Routing Protocol in Wireless Sensor
Networks. In: Proc. IEEE MASS05,
Washington DC, Nov. 2005, pp. 1-8.
12. Bestavros A., Smaragdakis G., Matta I.:
Sep: A stable election protocol for clustered
heterogeneous wireless sensor networks. In
Second International Workshop on Sensor
and Actor Network Protocols and
Applications (SANPA 2004), 2004
188
13. Wang M.W., Qing L., Zhu Q.X.: Design of
a distributed energy-efficient clustering
algorithm for heterogeneous wireless sensor
networks.
Computer
Communication
(Elsevier), 29(12): 2230-2237, 2006.
14. Manjeshwar A., Agrawal D. P.: APTEEN: A
Hybrid Protocol for Efficient Routing and
Comprehensive Information Retrieval in
Wireless Sensor Networks. In: Proc. 2nd
International Workshop on Parallel and
Distributed Computing Issues in Wireless
Networks and Mobile computing, San
Francisco CA, April 2001, pp. 2009-1015.
15. Seetharam A., Acharya A., Bhattacharyya
A., Naskar M. K.: An Energy Efficient Data
Gathering Protocol for Wireless Sensor
Networks. Journal of Applied Computer
Science, no. 1(2). 2008 pp. 30-34.
16. Dorigo, M., Maniezzo, V. and Colorni, A.:
The ant system: Optimization by a colony of
cooperating agents, IEEE Transaction on
System Man, and Cybernetics-Part B
26(1996), pp. 2941.
189
An Intersection Based Traffic Aware Routing With Low Overhead in

VANET
Lakshmi Ramachandran1, Sangheethaa Sukumaran2, Surya Rani Sunny3

1,3
M.Tech Computer Science & Engineering, Calicut University, India
2
Associate Professor, Calicut University, India
1
lakshmiramachandran@hotmail.com
ABSTRACT
Vehicular Ad-hoc network is an emerging
class of wireless network that provide
efficient communication between vehicles
and between vehicles and road side units.
VANETs are gaining attention from both
industrial and academic point of view due to
its applications ranging from safety
applications to infotainment applications.
The main component for success of VANET
applications is its routing. Design of an
efficient routing protocol is important for
smart Intelligent Transport System (ITS). In
this paper, we consider how traffic lights
affect the design of routing protocol in urban
vehicular
communication.
In
urban
environment the roads are partitioned to
several segments due to the traffic lights.
The design of a routing protocol in such an
environment is difficult. This paper proposes
a new intersection based routing protocol
which takes traffic lights in to consideration
and with low overhead.
KEYWORDS
VANET, ITS, Intersection based routing,
broadcast storm problem.
1 INTRODUCTION
During the last few years vehicular
communication is attracting growing
attention from both academic and
industrial point of view. This is because
of applications ranging from road safety

to traffic control and up to infotainment.
Vehicular ad-hoc networks (VANETs)
are self organized networks built up
from moving vehicles. Fig. 1 gives an
overview of VANET. VANETs are
instantiation
of
Mobile
Ad-hoc
Networks (MANETs). As in MANETs
packet forwarding in VANET takes
place through multi hop relaying. There
are certain features that distinguish
VANETs from MANETs. These include
high mobility of nodes, frequent network
partition; constraints on roadways etc.
These characteristics pose technical
challenges to implement a high
performance VANET.
Vehicular networks can be used to
facilitate the service customization to the
needs of individual nodes. Possible
applications [1] for such network can be
generally classified as safety and non
safety applications. Safety applications
include cooperative driving, accident
avoidance etc. Non-safety applications
include traffic information, toll service,
internet access, games, entertainment
etc.
The main component for success of
VANET applications is its routing. The
history of VANET routing protocols
starts with MANET routing protocols
190
such as Ad-hoc On Demand Distance
Vector routing (AODV) [2], Dynamic
Source Routing [3] etc. But these
protocols fail to address certain peculiar
characteristics of VANET. Various
routing protocols especially for VANET
were proposed to make routing more
efficient and reliable.
In this article we consider the vehicular
communication in highly dense urban
environment. Intersection based routing
protocols are highly reliable in such
environment. In intersection based
routing, when vehicles move on straight
road, they forward by greedy
forwarding. When they reach an
intersection a decision is made whether
to forward in same direction or to
perpendicular direction.
Many intersection based routing
protocols have been proposed to carry
efficient routing in VANET. But only
few protocols consider traffic lights. The
existing protocols that consider traffic
lights are having high overhead. This
paper proposes a new routing protocol
having low overhead. The remaining of
this paper is organized in to following
sections: Section II gives related works,
Section III describe the problem with
traffic lights, Section IV gives the new
approach and finally Section V
concludes the paper.
Figure 1 Overview of VANET
2 RELATED WORKS
VANET routing protocols can be
categorized in to two: topology based
and geographic (position based) routing
protocols. In topology based routing
protocols each node is expected to know
the entire network topology. In
Geographic or position based routing
protocols the decision on routing is
based on the position of the sender,
position of the destination and position
of the senders one hop neighbors using
GPS. It is assumed that each node knows
its position and the position of the
destination. The position of its one hop
neighbors is obtained from periodically
exchanged beacons. In geographic
routing the messages can be forwarded
to destination without knowing the
topology and without prior route
discovery. This section briefly describes
some prominent geographic routing
protocols.
2.1 Greedy Perimeter Coordinator
Routing (GPCR)
Greedy Perimeter Coordinator Routing
[4] makes use of street and junctions to
forward packets. GPCR consists of 2
parts: a restricted greedy forwarding
procedure and a repair strategy.
In all intersection based routing
protocols the actual routing decision is
made on junctions. In GPCR nodes on
junctions are called coordinators.
Packets are always forwarded to these
coordinators rather than to nodes across
the junction. If coordinators are not
present, the packets are forwarded to
node with the largest distance from the
forwarding node. If more than one
coordinator is present then one is taken
randomly and packet is forwarded to this
191
coordinator. In this protocol a decision is
made greedily so that a node with great
progress to destination is always
selected. When packet reaches a local
optimum, Right Hand Rule is used to see
to which street the packet should follow
as the repair strategy.
GPCR is effective as it does not need
any external information and graph
planarization algorithms. But first part of
GPCR fails on curve road and second
approach fails on sparse road. It selects
street without understanding whether
there are sufficient nodes to carry the
packet.
2.2 Improved Greedy Traffic Aware
Routing (GyTAR)
GYTAR [5] is an intersection based
routing protocol. It consists of 2
modules: dynamic junction selection
through which a packet must reach
destination and an improved greedy
strategy used for forwarding. A
forwarding node looks for the
neighboring junctions in digital map.
Then it assigns a score to each of these
junctions considering the traffic density
and curve metric distance to destination.
The score is calculated as in Eq. 1.
Ncon: constant that gives ideal

connectivity within a cell.
, : weighting factors for distance and
vehicular traffic ( + =1).
The junction with highest score is
selected to forward the packet. Once
junction is selected the next phase is to
forward the packet between these
junctions. Each vehicle maintains a
neighbor table which is updated using
periodically
exchanged
HELLO
messages. When a node receives a
packet it computes the new predicted
position of each neighbor using the
information recorded and then selects
the next hop. In case of local optimum
recovery strategy used in GyTAR is
Carry and forward. In carry and forward
recovery mechanism the node carries the
packet until next junction or another
vehicle closer to the destination comes to
its transmission range.
Advantages of GyTAR are: it reduces
control message overhead and efficiently
handle
often
occurring
network
partitions. The disadvantage is higher
neighbor table overhead in highly
dynamic environment.
2.3 Vehicle Assisted Data Delivery
(VADD)
ScoreJ=[1-Dp]+[min(Navg/Ncon,1)] (1)
Dj: curve metric distance from candidate

junction j to destination
Di: curve metric distance from junction I
to destination.
Dp: (Dj / Di) closeness of candidate
junction to destination point
Between junction I and J:
Nv: number of vehicles between I and J
Nc : no of cell between I and J
Navg: average number of vehicles per cell
(Nv/Nc)
VADD [6] aims at improving routing in

disconnected networks by using the idea
of carry and forward together with
vehicular mobility prediction. This
protocol relies mainly on 3 basic
principles: First, whenever possible
transmit through wireless link. Second,
roads with highest speed should be
chosen if the packet need to be
forwarded along road. And at last
dynamic path selection is carried during
the entire packet forwarding process.
192
VADD defines three packet models:
Straight
road,
Intersection
and
Destination.
in stime. When packet reaches sj it

calculates delay as in Eq. 2.
d (si, sj) = current time - stime
In intersection mode it uses protocols

like L-VADD, D-VADD, and H-VADD.
Location First Probe VADD (L-VADD)
selects a node closest to the next
forwarding path. Direction First Probe
VADD (D-VADD) selects a node that is
going towards the next forwarding path.
Hybrid VADD (H-VADD) combines LVADD and D-VADD. Out of these, HVADD has higher performance.
Advantage of VADD is its high delivery
ratio compared to other routing
protocols. At each intersection VADD
takes the best path available. But when
density is low optimal path may not be
always available. VADD estimates delay
based on statistical data. As vehicle
density changes with time, the shortest
delay may not be the real optimal one.
2. 4 Static Node Assisted Adaptive
Routing Protocol In Vehicular
Networks (SADV)
Static Node Assisted Adaptive Routing
Protocol [7] in Vehicular Networks
deploys a static node at intersections.
This static node stores packet and wait
until there are vehicles within
communication. The static notes have a
digital street map. SADV have 3
modules: Static Node Assisted Routing
(SNAR), Link Delay Update (LDU) and
Multi Path Data Dissemination (MPDD).
SNAR: SNAR utilizes the static nodes
deployed at intersections to store and
forward through optimal paths.
LDU measures delay from Static node si
to static node sj by inserting a single
field stime in to packet head. When si
receives a packet, it records current time
(2)
MPDD: When the load in network is not

too high multi path routing is used to
decrease delay by hitting a faster path
than single path.
The advantages of SADV are: SNAR
module stores and forwards packet
through the best available path, LDU
module effectively calculates the real
time delay and MPDD reduces the delay.
But practically, deployment of static
module at each intersection is not
feasible.
2. 5 Virtual Vertex Routing (VVR)
Virtual Vertex Routing [8] introduces a
new concept called virtual vertex that is
proximity of a vertex and uses the
information about lines. Proximity of a
vertex refers to the area within the circle
with vertex as centre and radius is half
the radio range. An intermediate node in
the proximity uses Floyd Algorithm to
forward packet to the destination. In
VVR edge connectivity is maintained by
exchanging HELLO messages.
The basic VVR mechanism includes two
parts: Initialization and Vertex Change.
VVR model consists of vertices
(junctions), edges (straight-line between
2 junctions) and nodes (vehicles).
During initialization the shortest path
between all pairs of vertices is calculated
using Floyds algorithm. Vertex change:
Packets in VVR is forwarded from one
vertex to another vertex. When a packet
arrives at a vertex, intermediate
destination vertex is selected. Here
comes the proximity of vertex. That is
any node in the proximity of
193
geographical location of vertex acts as a
vertex. This process continues until the
packet reaches the proximity of last
vertex. When the last vertex proximity is
reached the packet is greedily forwarded
to the destination.
intersection broadcasts its connectivity

information. All nodes hearing this,
rebroadcasts the information. This is
repeated until the vehicles on the other
segment
hear
this
connectivity
information.
The advantage of VVR is that, it rarely

falls in to routing holes. This is because
of prior knowledge to distribution of
nodes. The disadvantage is its high
overhead.
3 PROBLEMS WITH TRAFFIC

LIGHTS
2. 6 Shortest Path Based Traffic

Aware Routing (STAR)
Shortest path based Traffic Aware
Routing [10] is an intersection based
routing protocol proposed for urban
areas where traffic density is high and
intersections are deployed with traffic
lights. In urban areas with traffic lights,
the routing path will be alternating with
red and green light segments. The
vehicles will be moving in a stop-and-go
pattern. Due to red light the vehicles
may be gathered at the intersections and
as a result of this the segment
connectivity is disconnected. The
segment is reconnected by vehicles
moving from green light to red light
segment. STAR makes use of these
connected red light segments.
In STAR, before reaching an intersection
the packets are forwarded greedily with
carry and forward recovery mechanism.
When a packet reaches an intersection it
checks whether the red light segment
close to the destination is connected. If
connected, the packet is forwarded to
this red light segment. If no red light
segments are available the packet is
forwarded to green light segment. The
segment connectivity is learned by
broadcasting connectivity information.
Each
vehicle
approaching
the
In urban areas the intersections are

employed with traffic lights. These
traffic lights divide the road in to
different segments. The vehicles move
smoothly on a green light segment and
when light turns to red they slow down
and wait at intersections until the light
turns green. Thus the vehicles tend to
cluster at both directions in a red light
segment. The figure 2 shows a
disconnected segment. In the figure
vertical line shows green light segment
and horizontal line shows red light
segment. Thus traffic lights frequently
create disconnected links on a red light
segment.
In this paper the above problem is
tackled by the vehicles turning right
from a green light segment to red light
segment.
These
vehicles
ensure
connectivity on a red light segment.
Disconnected link
Figure 2 Disconnected link
194
4 RED LIGHT FIRST
FORWARDING
Shortest Path Based Traffic Aware
Routing (STAR) protocol is an
intersection based routing protocol that
takes traffic light in to consideration.
STAR also makes use of the vehicles
turning right from green light segment to
red light segment to ensure end to end
connectivity on a red light segment. In
STAR the segment connectivity is
measured by broadcasting connectivity
information. This broadcast mechanism
leads to serious problems like
redundancy, contention and collision of
messages. This problem is referred as
broadcast storm problem [11]. Thus
STAR protocol introduces much
overhead in the network.
In this paper we tackle this problem by
introducing a new intersection based
routing protocol with low overhead. This
paper proposes a protocol Red Light
First Forwarding (RLFF) which avoids
blind flooding to measure connectivity.
In Red Light First Forwarding, it is
assumed that each vehicle is employed
with GPS and each node knows its
position, position of its neighbours and
position of destination.
On a straight road, the vehicles forward
packets by greedy forwarding together
with carry and forward recovery
mechanism. In greedy forwarding when
a node wants to send a message to
destination, it checks whether the
destination is reachable. If the
destination is in its transmission range
the message is forwarded directly. If
destination is not reachable, it forwards
the packet to an intermediate node which
is near to the destination than itself.
When it reaches an intersection, it
checks whether the red light segment
near to destination is available. The

connected red light segment is given
prior importance than green light
segment. Thus the packets always stay
on connected links.
In RLFF segment connectivity is
measured by broadcasting connectivity
information. Each vehicle approaching
an intersection sends out broadcast
messages. In RLFF only vehicles having
maximum progress from broadcasting
source is allowed to rebroadcast.
In new approach relative distance
between hosts is used to decide whether
to drop a rebroadcast. This approach
eliminates the broadcast storm problem.
Suppose a host Hhears a broadcast
message from a source Sfor the first
time. Saydbe the distance between H
and S. If the distance d is small, the
rebroadcast can provide only a small
additional coverage. If d is large it
provides larger additional coverage area.
This idea is used in this approach.
Let dmin measures the distance between
host and the source from which the
broadcast message is heard and D be the
threshold distance.
Step 1: When a message is heard for
first time, initialize dmin as distance to the
broadcasting host. If dmin < D go to step
4.
Step 2: Wait for random number of
slots. Then submit the message for
retransmission
and
wait
for
retransmission to start. If the message is
heard again, go to step 3. Otherwise the
message is retransmitted and exits the
procedure.
Step 3: Update dmin if distance to host
from which message is heard is smaller.
195
If dmin < D go to step 4. Otherwise
resume the interrupted waiting.
Step 4: Cancel retransmission. Exit the
procedure.
RLFF is designed to have low network
overhead, minimum delay and high
delivery ratio.
5 CONCLUSION
VANET have recently been the topic of
extensive research due to its wide range
of application. Routing is the main
component for success of VANET
applications. There are lots of routing
protocols designed especially for
VANET. But most promising for urban
scenario is intersection based routing
protocols. This paper deals with an
intersection based routing protocol that
consider traffic lights and having low
overhead. Designing of routing protocols
that handle all applications is not at all
practical.
This approach will be simulated in ns2
[12] by generating mobility models
using
VanetMobiSim
[13].
The
performance of new protocol will be
measured in terms of network overhead,
delivery ratio and delay and will be
compared with existing protocols like
STAR and GyTAR.
6 REFERENCES
1. Yasser Toor, Paul Muhlethaler, Anis
Laoutti and Arnaud De La Fortelle
Vehicular
Ad-hoc
Networks:
Applications and related technical
issues,
IEEE
Communications
Surveys, 3rd Quarter 2008, Vol 10,
No.3.
2. Perkins, C.; Belding Royer, E.; Das,
SAd-hoc On Demand Distance Vector
Routing, IETF RFC, 3561, July 2003.
3. D. B. Johnson and D. A. Maltz,

Dynamic Source Routing In Ad-hoc
Wireless
Networks,
in
Mobile
Computing, chapter 5, 1996.
4. Christian Lochert, Martin Mauve,
Holger
Fubler
and
Hannes
HartensteinGeographic Routing in
City Scenarios MobiCom 2004 Poster
Abstract.
5. Moer Jerbi, Sidi-Mohammed Senouci,
Rabah Meraihi and Yacine GhamriDoudane, An Improved Vehicular AdHoc Routing Protocol for City
Environments,
IEEE
ICC07,
pp.3972-79.
6. Jing Zhao and Guohong Cao, VADD:
Vehicle-Assisted Data Delivery in
Vehicular Ad-Hoc Networks, IEEE
Transactions on Vehicular Technology,
Vol. 57, No. 3, May 2008.
7. H.Lee, Youndo Lee, Taekyoung Kwon
and Yanghee Choi Virtual Vertex
Routing (VVR) for Course Based
Vehicular Ad-Hoc Networks, IEEE
WCNC07, pp.4405-10.
8. Y.Ding, C.Wang and L.Xiao A Static
Node Assisted Adaptive Routing
Protocol in Vehicular Networks,
ACM VANET07, pp.59-68.
9. Rupesh Kumar and S.V Rao
Directional Greedy Routing Protocol
(DGPR)
in
Mobile
Ad-hoc
Networks,International Conference on
Information Technology, 2008.
10. Jin-Jia-Chang, Yi-Hua Li, Wanjiun
Liao, Ing-Chau Chang Intersection
Based Routing in Urban Vehicular
Communication with Traffic Light
Considerations,
IEEE
Wireless
Communication, Feb 2012.
11. Yu Chee Tseng, Sze Yao Ni, Yuh
Shyan Chen ang Jang ping Sheu, The
Broadcast storm problem in Mobile
Ad-hoc Network, wireless networks
2002.
12. Ns-2. http:// www.isi.edu/nsnam /ns.
13. M. Fiore and J. Hrri, The
Networking Shape of Vehicular
Mobility, ACM MobiHoc 08, pp.
26172.
196
An Algorithmic Approach to Database Normalization

M. Demba
College of Computer Science and Information
Aljouf University, Kingdom of Saudi Arabia
bah.demba@ju.edu.sa
ABSTRACT
When an attempt is made to modify
tables that have not been sufficiently
normalized undesirable side-effects may
follow. This can be further specified as
an update, insertion or deletion anomaly
depending on whether the action that
causes the error is a row update,
insertion or deletion respectively. Most
of the recent works on database
normalization use a restricted definition
of normal forms where only the primary
key is taken into account and ignoring
the rest of candidate keys.
In this paper, we propose an algorithmic
approach for database normalization up
to third normal form by taking into
account all candidate keys, including the
primary key. The effectiveness of the
proposed approach is evaluated on many
real world examples.
KEYWORDS
Relational database, Normalization, Normal
forms, functional dependency, redundancy.
1 INTRODUCTION
Normalization is, in relational database
design, the process of organizing data to
minimize redundancy. It usually
involves dividing a database into two or
more tables and defining relationships
between the tables. The objective is to
minimize modification problems that
could arise after modification of a table
field. Edgar Codd, the inventor of the
relational model, also introduced the
concept of normalization and Normal

Forms (NF). In general, normalization
requires additional tables and some
designers find this first difficult and then
cumbersome.
Violating one of the first three rules of
normalization, make the application
anticipates any problems that could
occur, such as redundant data and
inconsistent dependencies.
When using the general definitions of
the second and third normal forms (2NF
and 3NF for short) we must be aware of
partial and transitive dependencies on
all candidate keys and not just the
primary key. Of course, this can make
the process of normalization more
complex;
however,
the
general
definitions place additional constraints
on the relations and may identify hidden
redundancy in relations that could be
missed [1].
A functional dependency XA is
partial if some attribute BX can be
removed from X and the dependency
still holds. Let A, B, and C be attributes
of a relation R, AB and BC be two
dependencies that hold in R. Then C is
transitively dependent on A via B
(provided that A is not functionally
dependent on B or C) [1].
Most of the recent works on database
normalization are web-based tools
without presenting algorithms and define
the two notions, 2NF and 3NF, with
respect to primary keys only, ignoring
the other candidate keys as in [2], [3, [4],
[5]. However, the original definitions of
197
the two notions as given in [6], [7] , [1],

[8] consider all candidate keys. To see
the difference between the two
approaches, suppose we are given the
schema relation R(A,B,C,D,E,F)
together with the dependencies
section 5 we present the algorithms for

the decomposition into 2NF and 3NF
respectively. In section 6 we conclude
the paper.
F={ABCDEF; BCADEF; B D;
EF }.
A super key is a set of attributes which

will uniquely identify each tuple in a
relation.
A candidate key is a minimal super
key.
A primary key is a chosen candidate
key.
An attribute A is a key attribute in a
relation R, if A is part of some
candidate key. Otherwise it is a nonkey attribute in R, i.e., A is not a
component of any candidate key.
Given a relation R, a set of
attributes X in R is said to functionally
determine another set of attributes Y,
also in R, written X Y, if and only if
each X value is associated with
precisely one Y value; R is then said
to satisfy the functional dependency
X Y.
A functional dependency XY is
called trivial if Y is a subset of X.
A set of functional dependencies F is
in canonical form if each functional
dependency XA in F, A is a
singleton attribute.
We have two candidate keys A and BC.

If A is selected as the primary key of the
relation, and ignoring BC then there is
no partial dependencies on A, implying
that the relation is in 2NF. But if all
candidate keys are taken into account
(that is considered here), although there
are no partial dependencies on A, we
have a partial dependency BD on
candidate key BC, implying that the
relation is not in 2NF.
In this paper, we propose an algorithmic
approach for database normalization that
uses the original and general definitions
of normal forms. The general definitions
take into account all candidate keys of a
schema relation. For each synthesized
relation, a primary key is also generated.
The algorithms are presented step-bystep so designers can learn and
implement them easily.
Throughout the paper, R represents a
relational schema; A, B, C,... denote
attributes; X, Y and Z denote set of
attributes, and F a set of functional
dependencies.
The rest of the paper is organized as
follows: section 1 presents some basic
concepts and notations, in section 2 we
present a procedure for removing
redundant attributes, in section 3 we
present an algorithm for removing
redundant attributes and another for
redundant dependencies. Section 4
presents an algorithm for classifying
dependencies into full and partial and in
2 BASIC DEFINITIONS
Hereafter, all functional dependencies

are supposed in canonical form and input
schema relations are at least in first
normal form (1NF).
3 REMOVING REDUNDANT
ATTRIBUTES and DEPENDENCIES
Before determining partial dependencies,
some redundant functional dependencies
could be removed. To do that, many
algorithms have been proposed for
198
removing
redundant dependencies,
called minimal cover [9], [10]. To
achieve this goal, one needs to compute
the closure of a given set of attributes
and then remove redundant attributes.
Let X be a set attributes and F be a set of
functional dependencies. Let X+ be the
set of all attributes that depend on a
subset of X with respect to F, i.e., X+ is
the set of attributes Z, such that XZ
F. X+ is called the closure of X w.r.t F.
The Algorithm 1 computes the closure of
a given set of attributes w.r.t F:
Algorithm 1: computes X+.
Input: A relation R, a set of functional
dependencies F and a set X of attributes.
Output: X+, the closure of X.
X+:= X
While there is a fd Y A F
If YX+ and A X+ then
X+:= X+ A
End if
End while
To check if a functional dependency

YA holds in a set of dependencies F,
we compute the closure of Y and check
If AY+. This test is particularly useful,
as we will see later in the next
algorithm.
We can use attribute closure to remove
redundant dependencies from a set of
functional dependencies. We can do this
by examining each dependency in turn to
see if it is redundant. A dependency is
redundant if it can be inferred from the
other dependencies, and can thus be
removed.
Given a set of dependencies F, an
attribute B is extraneous in XA with
respect to F if A (X-B)+. If B(X-B)+

then B is called an implied extraneous
attribute. If B is extraneous, but not
implied, then it is nonimplied. For
example,
suppose
we have
the
functional dependencies A,BC and
AC in F. Then B is a nonimplied
extraneous attribute in A,BC. As
another example, suppose we have the
dependencies A,BC and AB in F.
Then B is an implied extraneous
attribute in A,BC.
F is called partially left-reduced if no
attributes are
implied extraneous
attribute. The elimination of nonimplied
extraneous attribute is postponed until
the time that redundant dependencies are
eliminated, using the Algorithm 3. Many
algorithms, sometimes difficult to reuse,
have been proposed in [11], [12] for
removing extraneous attributes.
In our approach, we propose the
following algorithm
to eliminate
implied extraneous attributes by
minimizing
attribute
closure
computations. Algorithm 3 removes any
implied extraneous attributes.
Algorithm 2: Removes implied
extraneous attributes.
Input: F, a set of FDs.
Output: F', a partially leftreduced set of FDs.
Let F':=F
For each fd XAF' do
Let Y:=X
For each attribute BY
If A(Y-B)+ w.r.t F' then
Y:=Y-B
End if
End for
F':=F'-(XA)(YA)
End for
199
Remove all duplicated dependencies, if

any.
Example 1:
Suppose we want to normalize the 1NF
relation ClientRental up to 3NF. This
example is taken from [12] with minor
modifications.
This relation contains 9 attributes and
18 dependencies:
Clientrental(clientNo, propertyNo,
clientName, pAddress, rentStart,
rentFinish, rent, ownerNo,
ownerName)
with the following dependencies F:
f 1:
f 2:
f 3:
f 4:
f 5:
f 6:
f 7:
f 8:
f 9:
f10:
f11:
f12:
f13:
f14:
f15:
f16:
f17:
f18
propertyNo,rentStart,
ownerNo
clNo,propertyNo,
clName
clNo,rentStart,
clName
clNo
propertyNo,rentStart
clNo,propertyNo
ownerNo
propertyNo
clNo,rentStart
clNo,rentStart
propertyNo
propertyNo
clNo,rentStart
propertyNo
clNo,rentStart
clNo,rentStart
propertyNo, clNo
rentFinish
rentFinish
rentFinish
clName
clName
rentStart
oName
oName
oName
pAddress
pAddress
rent
rent
ownerNo
ownerNo
propertyNo
clNo
ownerNo
Initialization: F':=F
For the fd f1, we have
+
rentFinish (propertyNo,rentStart) ;
then F' is changed by removing the

attribute ownerNo at the rhs of f1.
For the fd f2 and f3, we have
+
rentFinish ( clNo,propertyNo) and
+
rentFinish ( clNo,rentStart)
respectively; then F' is changed by
removing the attribute clName at the
rhs of f2 and f3.
The final partially left-reduced
dependencies F' of F is then:
f 1:
f 2:
f 3:
f 4:
f 5:
f 6:
f 7:
f 8:
f 9:
f10:
f11:
f12:
f13:
f14:
f15:
f16:
f17:
f18:
clNo,propertyNo
clNo,rentStart
clNo
clNo,propertyNo
ownerNo
propertyNo
clNo,rentStart
clNo,rentStart
propertyNo
propertyNo
clNo,rentStart
propertyNo
clNo,rentStart
clNo,rentStart
propertyNo, clNo
rentFinish
rentFinish
rentFinish
clName
clName
rentStart
oName
oName
oName
pAddress
pAddress
rent
rent
ownerNo
ownerNo
propertyNo
clNo
ownerNo
As one can see there is no more

extraneous implied attributes in F'.
However, in f18 the attribute clNo is a
nonimplied extraneous one because of
the dependency f14 and propertyNo did
not functionally determines clNo.
Nonimplied extraneous attribute will be
removed by the next procedure.
The next step consists to remove
redundant dependencies. The main idea
is to compare the lhs of dependencies
that have the same rhs.
200
Definition. A dependency f1: YA is

redundant in a set of dependencies F if
and only if there is a dependency f2:
XA in F such that YXF+, i.e.
XY+. Algorithmically this can be
expressed as follows:
Algorithm 3: Removes redundant
dependencies.
Input: F, a set of partially
left-reduced dependencies.
Output: Fm, a minimal cover of F
1. Fm:= F
2. For each XAFm do
while there exists a YAF
G:= Fm-(YA)
If XYG+ then
Fm:= G
End if
End while
End for
3. Fm is a minimal cover of F.
To continue with our previous example,
let lhs(fj)lhs(fi)+ means that the lefthand-side of the dependency fj is in the
closure of the left-hand-side of the
dependency fi, i.e. lhs(fi) functionally
determines lhs(fj).
f1 is redundant because
lhs(f2)lhs(f1)+
lhs(f2)lhs(f3)+
lhs(f4)lhs(f5)+
lhs(f7)lhs(f8)+
lhs(f7)lhs(f9)+
lhs(f11)lhs(f10)+

lhs(f12)lhs(f13)+
lhs(f14)lhs(f15)+
lhs(f14)lhs(f18)+
After removing the redundant
dependencies (10), we get a minimal
cover Fm :
f 2:
f 4:
f 6:
f 7:
f11:
f12:
f14:
f16:
f17:
clNo,propertyNo
clNo
clNo,propertyNo
ownerNo
propertyNo
propertyNo
propertyNo
clNo,rentStart
rentFinish
clName
rentStart
oName
pAddress
rent
ownerNo
propertyNo
clNo
Unfortunately, there can be several

minimal covers for a set of functional
dependencies. We can always find at
least one minimal cover Fm for any set of
dependencies F using Algorithm 3.
Theorem 1. Let F be a set of partially

left-reduced dependencies, and Fm its
minimal cover obtained by the
Algorithm 3. Then we have F+ Fm+.
4 FULL and PARTIAL
DEPENDENCIES
A functional dependency X A is a
full functional dependency if removal of
any attribute B from X means that the
dependency does not hold any more; that
is, for any attribute BX, (X - B) does
not functionally determine A. A
functional dependency XA is a partial
dependency if A is not a key-attribute
and some attribute BX can be removed
201
from X and the dependency still holds;

that is, for some BX, (X - B) A [7].
Full and partial dependencies are
generated as follows:
Algorithm 4: Determines full and
partial dependencies.
Input: Fm, a minimal cover set
of F.
Output: Fp, Ff, sets of partial
and full dependencies in
Fm respectively.
Let Fp:= and Ff:=Fm.
For each dependency XAFf
If X is a proper subset of a
candidate key and A is not a
key attribute then
Fp:=Fp{XA}
Ff:=Ff-(XA)
While there is a fd ZBFf
s.t. ZA+ do
Fp:=Fp{ZB}
Ff:=Ff-(Z B)
End while
End if
End for
In the If-statement, we make condition
on the candidate keys and not the
primary key, because if a relation has
many keys, all have to be considered
(see the definition of 2NF in the next
section).
We can continue our running example
by initializing Ff to Fm:
f 2:
f 4:
f 6:
f 7:
f11:
f12:
clNo,propertyNo
clNo
clNo,propertyNo
ownerNo
propertyNo
propertyNo
rentFinish
clName
rentStart
oName
pAddress
rent
f14:
f16:
f17:
propertyNo
clNo,rentStart
ownerNo
propertyNo
clNo
We have three candidate keys (clNo,

propertyNo), (clNo, rentStart) and
(propertyNo, rentStart). No matter is the
choice of the primary key, we can
determine full and partial dependencies.
Fp={ clNoclName;
propertyNopAddress;
propertyNorent;
propertyNoownerNo;
ownerNooName
}
As Fp the relation ClentRental is not
in 2NF. In fact, the attribute clName is
partially dependent on the candidate
keys (clNo, propertyNo) and (clNo,
rentStart). Also, the attributes pAddress,
ownerNo and
oName are partially
dependent on the candidate keys (clNo,
propertyNo) and (rentStart, propertyNo).
The set of full dependencies is:
Ff={ clNo,propertyNorentStart;
clNo,propertyNorentFinish;
clNo,rentStartpropertyNo;
propertyNo,rentStartclNo }
Lemma 1. Fm= Ff FP.
Lemma 2. Let R be a relation schema
and Fm its minimal cover dependency
set. If there is no partial dependency in R
then Fp=.
5 NORMALIZATION
5.1 The Second Normal Form
A relation is in second normal form
(abbreviated 2NF) if it is in 1NF and no
202
non-key attribute is partially dependent

on any candidate key [7], [1]. In other
words, no XY where X is a strict
subset of a candidate key and Y is a nonkey attribute.
The following algorithm is used to
decompose a 1NF relation schema R into
2NF:
Algorithm 5: Decomposes into 2NF.
Input: R, Ff and Fp.
Output: relations into 2NF.
Let G:=Fp and Xf the set of
attributes in Ff.
For each YAG do
If Y is a key attribute then
- create RY, and add any
attribute in YG+.
- remove from G any
dependency whose lhs is in YG+
- choose Y as the
primary key of RY.
- if AXf, Xf:=Xf-{A}
End if
End for
Let KXf be the (chosen)
primary key of R. Create a new
relation schema RK(Xf).
If Y is a key-attribute that means that Y

is causing a partial dependency.
Theorem 2. Let R be a 1NF relation. We
have the following results:
a. If Fp= then R is automatically in
2NF.
b. If Fp then R is not in 2NF.
Proof.
a. If Fp=, i.e., there is no partial
dependency in R, that means also
that the first if-condition of the
algorithm will never hold. Then RK
is the only relation created by the

Algorithm 5. As RK is partial
dependency free (no partial
dependency exists in Ff), then R=
RK is in 2NF.
b. If Fp, i.e., the first if-condition
of the algorithm holds at least
once. Therefore, R is not in 2NF
as RRK.
Let's assess our example:
Ff={ clNo,propertyNorentStart;
clNo,propertyNorentFinish;
clNo,rentStartpropertyNo;
propertyNo,rentStartclNo }
Fp={ clNoclName;
propertyNopAddress;
propertyNorent;
propertyNoownerNo;
ownerNooName
}
and Fm=FfFp.
As we can see all attributes in Ff depend
fully on the candidate keys (clNo,
propertyNo), (clNo, rentStart) and
(propertyNo, rentStart) and all the
attributes in Fp depend directly/indirectly
on the key-attributes
clNo
or
propertyNo, i.e., the two key-attributes
are causing partial dependencies.
For clNoclName
clNo is a key-attribute, a new relation
RclNo(clNo, clName) is then created
For propertyNopAddress
propertyNo is a key-attribute, a new
relation
RpropertyNo(propertyNo, pAddress,
rent,ownerNo, oName)
203
is then created with all attributes in

propertyNo+.
If we choose arbitrarily
(clNo,
propertyNo) as the primary key, but
no special consideration will be given
to this key over the other candidate
keys, then a new relation is needed:
R(clNo,propertyNo)(clNo, propertyNo,
rentStar, rentFinish)
Note that when a 1NF relation has no
composite candidate keys (candidate
keys consisting of more than one
attribute), the relation is automatically in
2NF. If there is no composite candidate
keys, then Fp= and by theorem 2 the
relation is in 2NF.
Algorithm 6: Decomposes into 3NF.

Input: R1,...,Rn relations in 2NF, and the
set Ft.
Output: a set of relations into 3NF.
Let G:=Ft
For each dependency XAG do
- create a new relation RX, if
not already created,
- add X and A in RX, if not
already added,
- consider X as the primary key
of RX,
- remove XA from G, and
- remove the attribute A from
the relations R1,...,Rn.
Fnd for
5.2 Third Normal Form

A relation is in third normal form
(abbreviated 3NF) if it is in 2NF and
none of its non-key attributes are
transitively dependent upon any
candidate key [6], [1], [7]. An alternative
(simpler) definition is a relation is in
3NF if in every non-trivial dependency
XA either X is a super key or A is a
key attribute (i.e., A is contained within
a candidate key).
The transitive dependencies set on 2NF
relations is defined by:
Ft={ XAFm/ X is not a subset of any
candidate key and A is not a keyattribute }
Note that partial dependencies, even if
violate the 3NF, are not considered in Ft,
because the input relations are supposed
already in 2NF using the previous
algorithm. The following procedure is
used to decompose 2NF relations into
3NF.
Now, we can complete our example

given the relations:
RclNo(clNo, clName)
rent,ownerNo, oName)
R(clNo,propertyNo)(clNo, propertyNo,
Fm={
f 2:
f 4:
f 6:
f 7:
f11:
f12:
f14:
f16:
f17:
clNo,propertyNo
clNo
clNo,propertyNo
ownerNo
propertyNo
propertyNo
propertyNo
clNo,rentStart
rentFinish
clName
rentStart
oName
pAddress
rent
ownerNo
propertyNo
clNo
}
We can say that the set transitive
dependencies is
Ft={ ownerNooName } as ownerNo
is not a subset of a candidate key and
oName is not a key-attribute.
204
The only relation that contains the

transitive dependency Ft is RpropertyNo. We
then decompose RpropertyNo into two
relations and remove the attribute
oName from that relation to get all
relations into 3NF:
R(clNo,propertyNo)(clNo,propertyNo,
RclNo(clNo,clName)
rent, ownerNo)
RownerNo(ownerNo, oName)
The
algorithms
are
dependency
preserving as the original ClientRental
relation can be recreated by joining
the 3NF relations R(clNo,propertyNo), RclNo,
RpropertyNo
and RownerNo through the
primary key/foreign key mechanism.
6 CONCLUSION
In this paper we have presented
algorithms for relational database
normalization into 2NF and 3NF using
their general definitions in a step-by-step
feature. The first step before performing
the procedure is to make a preprocessing
on the set of dependencies to remove
redundant dependencies. We have tested
our algorithms on many realistic
examples with multiple candidate keys
taken from different sources.
This work has mainly the following
major advantages: (i) the general and
original definition of normal forms is
used, (ii) the removal of redundant
dependencies, (iii) in all phases, the
computation of attributes closure are
minimized compared to other algorithms
although using a restricted definition of
normal forms, and (iv) a primary key is
determined for any generated relation.
8 REFERENCES
1.
Thomas, C., Carolyn, B.:

Database
Systems, A Practical Approach to Design,
Implementation, and Management, Pearson
Fourth edition (2005).
2. Bahmani A., Naghibzadeh, M. and
Bahmani, B.:
Automatic database
normalization and primary key generation,
Niagara Falls Canada IEEE (2008).
3. Beynon-Davies, P.:
Database systems,
Palgrave Macmillan, Third edition, ISBN 1
4039--16012 (2004).
4. Dongare,Y. V., Dhabe,P. S. and Deshmukh,
S. V.: RDBNorma: A semi-automated tool
for relational database schema normalization
up to third normal form, Interrnational
Journal of Database Management Systems,
Vol.3, No.1 (2011).
5. Vangipuram, R., Velputa, R., Sravya, V.: A
Web Based Relational database design Tool
to Perform Normalization, International
Journal of Wisdom Based Computing,
Vol.1(3) (2011).
6. Codd, E.F.: Further normalization of the
data base relational model. In Data Base
Systems, Courant Inst. Comptr. Sci. Symp.
6, R. Rustin, Ed., Prentice-Hall, Englewood
Cliffs, pp. 33--64 (1972).
7. Elmasri, R., Navathe, S.B.,: Fundamentals
of Database Systems, Addison Wesley,
fourth Edition (2003).
8. Date, C.J.:
Introduction to Database
Systems (8th ed.). Boston:
AddisonWesley. ISBN 978-0-321-19784-9 (2004).
9. Ullman, J.D.:
Principe of
Database
Systems.
Computer
Science
Press,
Rockville, Md. (1982).
10. Maier, D.:
The Theory of relational
Databases. Computer Science Press,
Rockville, Md. (1983).
11. Bernstein, P.A.: Synthesizing third normal
form
relations
from
functional
dependencies. ACM Transactions on
database Systems, vol.1, No.4, pp.277--298
(1976).
12. Diederich, J., Milton, J.: New Methods and
Fast
Algorithms
for
Database
Normalization, ACM Transactions on
database Systems, Vol.13, No.3, pp. 339-365 (1988).
205

IJDIWC - Volume 3, Issue 2

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

IJDIWC - Volume 3, Issue 2

Uploaded by

Copyright:

Available Formats

MODIFIED PROPORTIONAL FAIR SCHEDULING FOR RESOURCE

COMMUNICATION Rony Kumer Saha, A B M

AN OPTIMIZED ENERGY AWARE CHAINING TECHNIQUE FOR DATA Ouadoudi

AN ALGORITHMIC APPROACH TO DATABASE NORMALIZATION

Peer Knowledge Assisted Search Using Community Search Logs

directed participants is one of the

then used to provide collaborative

is a generic engine for transposable

Figure 1. Segments of search data

111.249.47.11,201201041255,4230,y:2011 mobile learn

Figure 2. Segments of log data

4 THE PROPOSED SYSTEM

Figure 3. Design goal of our system

The effect of just one users search

search without having to

This section introduces the functions and

Figure 5. Search query list in frame A

Figure 4. User interface of the proposed system

1) Frame A displays the search results.

2) Frame B displays the related papers

Figure 7. Search results in frame A.

Analyzing the related users: When

Figure 6. Details of frame C

4.3 An Example Of Searching Using

Figure 8. User list in frame B

Analyzing the related papers: When

Figure 9. A segment of a paper list in frame B

5 CONCLUSIONS AND FUTURE

D. Bilal. Childrens use of the

G. Marchionini, Information Seeking in

Modified Proportional Fair Scheduling for Resource Reuse and

performance is evaluated based on a number

very few or none of these addressed

Now that most of the researches have

An extensive system level simulation is

4) Fast Fading Models: In outdoor

s = vt => 4.28 cm = vt => t =

t = 5.1ms => t 5TTI for v = 30 km / hr .

Similarly, t 1 TTI for v=150 km/hr and

where k is the selected user in the i CC

3.4 Resource Reuse Strategy

The PF scheduler always allocates an

Table 2. Spectral efficiency requirements of

Figure 1. Two-tier dense urban macro-femtocell

3.3 Proportional Fair Scheduling

where ki , j ( ) is the selected

i th CC at the j th RB in time ; Rk , j,i ( )

updated according to the following

3.5 Novel Mechanism for Fairness

The impact of the strategy can be

scheduled to femto users, UFk. Step 5:

UMk UMk UMk UFk

UFk UMk UFk

UFk UMk UFk

UMk UMk UMk UFk

UFk UMk UFk

UMk UMk UMk UFk UMk UFk

UFk UMk UFk

UMk UMk UMk UFk

UFk UMk UMk UMk UFk

UFk UMk UFk

Figure 2. PM vs. RB distributions: Steps 1- 2.