You are on page 1of 7

Representing and Organizing Contextual

Data in Context Aware Environments


U. Mahmud, U. Farooq, M. Y. Javed and N. A. Malik
Abstract Context aware systems visualize smart service adaptations in pervasive computing environments. The context
aware process starts by gathering the contextual data through sensors present in the environment. The contextual data is
represented as an ontology using a platform independent language like Web Ontology Language (OWL). This activity is carried
out in a Context Gathering and Representation Module (CGRM) that is part of our context aware system. To provide more
power to the ontological representation, allied components should be provided with this module. These components include
freshness and confidence calculators. This paper presents a context representation module with freshness and confidence
modules.
Index TermsContextual Data, Knowledge Modeling, Ontology Design, Pervasive Computing



1 INTRODUCTION
ONTEXT aware systems are characterized by the
awareness of their environment to provide improved
and intelligent services to the users [1]. Context
aware systems offer smart service adaptation on the basis
of contextual data. The context includes information
about the people, computation devices, activity and the
environment within a given spatial and temporal frame
[2].
Gathering the contextual data and its subsequent sto-
rage is the first process in a context aware system. Differ-
ent techniques exist to model contextual data. Contextual
data is viewed as an attribute-value pair at this level.
Models exist for both representations of contextual data
as well as its organizations. The representation of contex-
tual data is concerned with how the storage system can be
constructed. The organization is concerned with how the
data is structured.
We propose an OWL based representation scheme and
a User context and Service context based organization for
CGRM. In addition we also propose allied components
that are required to perform pre-processing on contextual
data for its subsequent usage.
Section 2 provides a brief on context organization and
representation followed by a review on the modeling
techniques for both representation and organization in
Sections 3 and 4 respectively. Sections 5 and 6 are con-
cerned with the characteristics of contextual data and its
gathering modules. While sections 7 and 8 highlight the
limitations of OWL and provide solutions to the limita-
tions of representation models, respectively. Sections 9
and 10 present the architecture as well as the taxonomy of
contextual data as developed using Protg [3]. Section 11
discusses evaluation of the test results and concludes in
Section 12.
2 CONTEXT ORGANIZATION AND REPRESENTATION
The collection of data from the sensors is the first task in
the development of context aware system [1]. The data is
gathered from the sensors present in the environment as
well as though the user devices. The gathered data is raw,
acquired through wireless or wired links and may require
post-processing to remove noise. The contextual data is
dynamic and corresponds to the mobility of the nodes in
the smart space. This data can also be used to deduce
more information based on the supporting axioms. Con-
textual data is found in both atomic and composite forms
[4].
Gathering the context requires a representation tech-
nique that allows us to maintain the diverse data acquired
from the environment. The collected data is represented
and stored in the system. The representation technique
should conform to the heterogeneity across multiple plat-
forms as pervasive environments are loosely coupled.
Two research questions are highlighted in context gather-
ing. First: What is the suitable technique for context data re-
presentation? And Second: how is the data organized?
3 MODELING TECHNIQUES FOR CONTEXT
REPRESENTATION
Among various modeling techniques for representation
of contextual data use of attribute value pairs using Set
Theory, mereotopology, Fuzzy values and OWL are
promising. OWL is the widely used representation stan-
dard of contextual data. OWL is a tagged modeling
scheme based on eXtensible Markup Language (XML). In
[5] sets are created that provide attribute value pairs of
contextual data. In [6] researchers, identify that by using
set theory the diversity in contextual data and its compo-

U. Mahmud is with National University of Sciences and Technology,
(NUST), Islamabad, Pakistan
U. Farooq is with National University of Sciences and Technology,
(NUST), Islamabad, Pakistan
M. Y. Javed is with National University of Sciences and Technology,
(NUST), Islamabad, Pakistan
N. A. Malik is with National University of Sciences and Technology,
(NUST), Islamabad, Pakistan


C
JOURNAL OF COMPUTING, VOLUME 4, ISSUE 3, MARCH 2012, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 61

sition cannot be realized to its extent. They propose me-
reotopology to identify part-of and boundary-of relations
in addition to composition as provided by set theory.
Sensors provide a range of values for a single attribute
thus, highlighting the need for fuzzy variables. This de-
mands a fuzzy logic based system to interpret data [7].
Form Table 1 it is clear that OWL is the most suitable
representation language for context aware systems. OWL
is an ontology representation language used for publish-
ing and sharing contextual information [3]. OWL is XML
based and represents concepts and relationships. Though
OWL lacks the power offered by mereotopology and set
theory, which can be provided as stereotypes in OWL
with allied modules to manage them.
4 ORGANIZING CONTEXTUAL DATA
To answer the second question, set forth in Section 2, dif-
ferent organizations for contextual data have been pro-
posed in the literature. Dey considers context as Primary
and Secondary information [8]. The Primary information
involves spatial, behavioural and temporal information
while all other information is Secondary. Gwizdka di-
vides the information as Internal and External to the user
[9]. Petrelli considers Material and Social contexts where
Material context includes location and the available infra-
structure and Social context comprises of the communal
and personal traits of a user [10]. Hofer has identified
context to be Physical; which is measureable and Logical;
which is derivable [11]. Riaz considers context to be com-
posed of Services context, the User context and the Con-
text aware systems context [12]. Schilit distinguishes
among Device context; of all the devices, User context;
including user profile and preferences and Physical con-
text; of the environment [13]. Another division based on
Service context and User context is given in [14]. Sheng
argues that context can be classified as Atomic; having
unit value and Composite; combination of atomic and
composite values [4]. Roussaki divides context as Geo-
graphical; pertaining to time and space and Logical; all
the rest [15]. Badii describes context as Device; of a data
source, Semantic; high level context including location,
environment and entity, and Application; where reason-
ing can be inserted [16].
5 ORGANIZATION FACETS
5.1 Dynamism
Contextual data can be divided as static or dynamic i.e.,
on the basis of how rapidly the data changes. For exam-
ple, name and temporary address of a user can be consi-
dered as static. Similarly, dynamic context data can be
current time. An interesting question is where to place
boundary between static and dynamic context. This may
be different for different systems and user preferences
may have an effect on its selection. Moreover, freshness of
data must also be identified so as to discard old values.
This is more important for highly dynamic data where
decision taken on old values may not produce the desired
result.
5.2 Composition
Composition has been addressed in [4] and is arranged as
Atomic; unit and indecomposable context and Composite;
that is a combination of atomic and composite context.
For composite context to be useful, the context must be
constructed prior to its usage. For example, location may
be constructed as a composition of room, floor and block.
5.3 Deducibility
Deducibility refers to the fact that some contextual values
can be derived from others [11]. The context may be di-
vided as Sensed; that can be measured directly from the
environment e.g., temperature, and Deduced; that can be
deduced on the basis of other context. The deduced con-
text is constructed as IF-THEN-ELSE rules. For example
IF serviceQueue != EMPTY, THEN serviceState = BUSY.
5.4 Ownership and Point of Reference
Context can also be organized on the basis of ownership.
The context maybe organized with reference to the user
i.e. all data owned by and referring to the user vs. all the
rest. Roussaki considers context to be organized as con-
forming to Spatial and Temporal category and Logical
TABLE 1
COMPARISON OF MODELING TECHNIQUES
ModelingTech
nique
Set
Theory
Mereo
topolgy
Fuzzy
Sets
OWL
Complex Com
position

Object Orienta
tion

Platform Inde
pendence

FuzzySupport
Relationship
AmongConcepts

AxiomSupport
Mathematical
Function

Deduction Sup
port


Fig. 1. Composite Contextual Data
JOURNAL OF COMPUTING, VOLUME 4, ISSUE 3, MARCH 2012, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 62

category [15]. Similarly, context can be arranged on the
basis of Service context and User context as in [12], [13],
and [14].
5.5 Uncertainty
Due to the noisy nature of the sensors, a degree of certain-
ty must be present with each value. The context can be
organized as having high confidence and low confidence.
The boundary among the two must be flexible and adapt-
able to a users preferences. Confidence measure can be
given a numerical worth with each value. The confidence
value can be dependent on reliability of the sensor, confi-
dence on the value as provided by the sensor.
6 CHARACTERISTICS OF CONTEXTUAL DATA
6.1 Context Sharing
The contextual data is acquired through different sensors
present in the environment. Some sensors maybe shared
among different participants. Similarly, different partici-
pants may have the same context.
6.2 Easy Access
The contextual data must so be organized so as to provide
easy access for updating and recovery. Tree based struc-
tures provide efficient retrieval and storage in the form of
a hierarchy.
6.3 Interoperability
The context is provided by different sensors having hete-
rogeneous platforms. To share the contextual data among
sensors and participants as well as among several partici-
pants, the model should mask the underlying heterogene-
ity of the distributed system as well as provide a mechan-
ism that provides predicted information in the event of
missing or imperfect contextual data [18]. This is pro-
vided through OWL which is an XML based language.
6.4 Equality
It is possible that two contexts have the same value for all
attributes. This leads us to identify both as equal contexts.
Similarly, two distinct contexts maybe interpreted as be-
longing to the same context with a degree of similarity.
6.5 Uncertainty and Freshness
The system must allow identifying latest values and also
providing confidence on the values provided through
sensors. This also helps in categorizing long-term and
short-term contexts.
7 LIMITATIONS IN OWL BASED REPRESENTATION
7.1 Axiom Support for Deduced Context
As discussed in Section 5.3 the deduced context is a con-
text that is given as an implication. To identify the value
of the deduced context a rule must fire. These rules must
be provided through PELLET, FACT++ etc, or may be
implemented as an external module
7.2 Freshness and Uncertainty Calculator
The freshness and uncertainty values are easily stored as
timestamp and numerical values for each variable. To
calculate the values for context an external calculator
must be present that calculates and then updates the val-
ues in OWL.
7.3 Complex Relationships
The complex relationships of part-of and boundary-of can
only be presented as stereotypes with every context class.
OWL does not support mereotopology.
8 REALIZING CONTEXT CHARACTERISTICS AS
EXTERNAL MODULES
8.1 Composition
A Context C is composed of different values Vs gathered
from different sensors. The values may have numerical or
string evaluations. A Context C is represented in (1). A
Context C
I
maybe composed of another Context C
j
as
shown in (2). Fig. 1 highlights composition.
C = {I
1
, I
2
, I
3
, I
k
] (1)
C

C
]
(2)
8.2 Freshness
A variable V is associated with a freshness indicator f that
highlights the difference of the current time with the time
at which the value was measured in (3). This freshness
value is applicable to only measured data and for the dy-
namic data. For immeasurable or static data the freshness
maybe considered as zero. The value of f is given in milli-
seconds. The larger the value of f, the older is the data
value. The value of f cannot be negative highlighting the
measured time precedes the current time.
= currcntIimc -timc0otoHcosurcJ (3)
Since a Context C is composed of numerous Variables
having different freshness value, the freshness of context

c
is the maximum freshness value of all the composed
variables as in (4). This highlights that a Context C has no

Fig. 2. Architecture of Context Gatherer and Representation Module
(CGRM)
JOURNAL OF COMPUTING, VOLUME 4, ISSUE 3, MARCH 2012, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 63

variable measured earlier than value
c
.

c
= Nax
n

n
(4)
8.3 Uncertainty and Confidence
The uncertainty factor identifies the confidence on the
measured value. The confidence is measured at context
gatherer where the confidence d is a value associated
with each variable. The value of d lies in the interval [0-1].
A lower value of d identifies more uncertainty. The con-
fidence is dependent on the certainty value c provided by
the sensor on its measurement, the reliability r of that
sensor as viewed by the system and the freshness of the
value (5). The certainty c and the reliability r are both in
the interval [0-1]. Reliability is of the sensor and certainty
is provided by the sensor.
Reliability can be provided by the user manually or
predicted through a history. For example, a sensor may
not be reliable to the system but the sensor may consider
its output to be highly reliable. In the absence of certainty
provided by the sensor, the reliability is taken as the con-
fidence. The fresher the value the confidence would be
high.
J =
c
]
(5)
The confidence on the context C is reflected by the va-
riables it comprises as shown in (6). The confidence of the
context u
c
cannot be given the maximum or minimum
value, but is considered to be the arithmetic mean of all
the confidence values of the variables involved. For the
static data items with zero freshness the confidence is set
to maximum.
J
c=
d
n
i=0
n
, u J
c
1 (6)
8.4 Deduced Context
To represent deduced context as variables, supporting
rules must be provided as derivations. The total number
of rules should not be large. Large number of rules will
result in wastage of time and space [17]. The set of
axioms are kept in a separate table using IF-THEN-ELSE
statements.

8.5 Point of Reference
To solve the ownership problem, the contextual data can
be organized under different possessors. We recommend
a service based and user based organization that allows
us to view the problem as a relation between user and
services [14]. So a context C will either belong to a service
or a user as shown in Fig. 1.
8.6 Similarity of Contexts
Given a set of contextual data, any possible combination
of values is regarded as a distinct Context C. C
I
is identic-
al to C
j
if all the values of corresponding attributes are
same. It follows that an appropriate action for C
I
is ap-
propriate for C
j
if both are identical. A C
I
maybe similar to
C
k
with degree of similarity D based on the mutual dis-
tance. If C
I
is similar to C
k
then an action appropriate for
C
I
may be applied to C
k
with confidence d.
8.7 Minimal Context
The set of contextual data being comprehensive tends to
be large since everything is part of context. Also, no guar-
antees can be made for the availability of the complete
data set in every situation. Hence, a minimal context set is
considered which is sufficient to effectively draw infe-
rences. This minimal set is dependent on the sensors
available in the environment.
9 ARCHITECTURE OF CGRM
A hybrid approach that is based on OWL and stereotypes
that provide the organization of contextual data is pro-
posed. The stereotypes are kept as data properties in
OWL. The context is organized as Service context and
User context. The functions that must be provided by a
contextual data repository are implemented as Java
classes. These include confidence and freshness calcula-
tions and axioms for deduced context. OWL 2.0 provides
an abstraction to provide axioms support in addition to
organization and data representation. These axioms facili-
tate rule based reasoning within OWL [19]. We represent
our context as a set of variables with different typecast
values. The proposed module is shown in Fig. 2. The Sen-
sors are external to the system and are present in the en-

Fig. 3. Context Catalog
TABLE 2
DESCRIPTION OF CONTEXT CLASSES
ContextClass Description
ServiceTemporal Currenttimeanddateoftheservice.
ServiceSpatial Locationoftheservice.
ServiceEnvironment EnvironmentoftheService
ServiceActivity Currentstatusoftheservice.
ServiceIdentification Name,ownerandthetypeofservice.
UserTemporal Currenttimeoftheuserdevice.
UserSpatial Currentlocationoftheuser.
UserEnvironment Characterizestheuserssurroundings.
UserActivity Thecurrentactivityoftheuser
UserIdentification Identificationparametersoftheuser.
UserHealth Healthconstraintsforeachuser.
UserDemand Tasksoftheuser
JOURNAL OF COMPUTING, VOLUME 4, ISSUE 3, MARCH 2012, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 64

vironment.
9.1 Dispatcher Module
The dispatcher is responsible for invoking sensor devices
and receiving their result. The dispatcher is also responsi-
ble for maintaining and tracking sessions.
9.2 Confidence and Freshness Calculator Module
This module calculates the confidence as well as freshness
of each value. Moreover, the confidence and freshness of
the context is also calculated. This module is allied to the
OWL and is also responsible for storing the values with
confidence and freshness in individuals of ontology.
9.3 Axiom Firing Module
The axiom firing module fires rules to identify deduced
context. These rules are fired when the data is being
stored in OWL repository. The purpose of this module is
to identify context values for deduced context items.
9.4 Ontology (OWL) Store
The OWL based ontology is maintained here. The OWL
file has been developed using protg tool. The data val-
ues coupled with freshness and confidence are stored
here in the form of OWL individuals.
9.5 Query Handler Module
The query handler acts as an interface between CGRM
and the Context Inference Engine.
10 CONTEXT CATALOG IN CGRM
We have selected a limited set of data items based on the
requirements of everyday interactions in a computer la-
boratory of our department. The context catalog is gener-
ated using Protg tool [3]. The context is broadly catego-
rized into the Services context and the Users context. Ser-
vice context defines the current situation of the services
including the activity, location, identification and envi-
ronment. User context maintains information about the
user and its activity as well as its demand. This also in-
cludes the information about the user device. It is as-
sumed that there is only one user and only one service in
the shared environment. Fig. 3 shows the context catalog
as developed using protg. Table 2 gives a description of
all the context classes identified. Table 3 enlists some of
the variables and their values with the freshness f and
confidence d.
The OWL representation has been constructed using
protg. Protg is a free ontology editor developed by
Stanford University. Protg tool works by making the
concepts and then identifying the object properties. The
values are given as individuals. Fig. 4 shows a snapshot
of an individual in OWL as developed using Protg.
Service2001 is a Service context and the Bit rate value is
identified with confidence 0.6 and freshness 60 seconds.
11 EVALUATION
Among the issues in our system we find that context
aware systems consider everything as contextual data
and hence the attributes in the system tend to be large. In
addition each attribute has freshness as well as confi-
dence values associated with it. The metrics of the ontol-
ogy as created in protg highlight that there are 90 data
properties. To gather all these values and store them with
calculation of additional properties is a time consuming
task especially for hand held devices where time and
space are constrained.
11.1 Time Analysis
To perform time analysis for the CGRM a flow model is
envisaged as shown in Fig. 5. It is assumed that there are
no deduced data attributes in the system and hence no
time is spent in firing the rules of deduction. Assume for
any arbitrary attribute the time required to gather the
value from the sensor is t
g
, time required to calculate
freshness is t
f
, time required to calculate confidence is t
c

and the time required to store the value is t
s
as in (7).
I = t
g
+t
]
+ t
c
+t
s
(7)
The time to collect data from the sensor is inherently a
function of the bandwidth of the connection and is
beyond our system. Similarly, the time to store data is
constant as this is a function dependent on the storage
device. For the evaluation time to calculate freshness and
time to calculate confidence are considered as in (8).
I = t
]
+t
c
(8)


Fig. 4. An Individual in OWL. Developed in Protg
TABLE 3
CONTEXTUAL DATA AND VALUES
Variable Context
Class
f(sec) d
(01)
Values
ServiceBat
tery
Service
Activity
30 0.7 Alarm
ServiceBit
Rate
Service
Activity
60 0.6 High
UserRole UserID 0 0.9 Admin
UserLoca
tion
UserSpa
tial
120 0.5 Lab1
UserDate UserTem
poral
600 0.9 01Jan 2012
UserTime UserTem
poral
30 0.9 1035hrs

JOURNAL OF COMPUTING, VOLUME 4, ISSUE 3, MARCH 2012, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 65


11.2 Data Set
We have selected the data set as provided by Petzold that
contains the movements of different participants [20]. The
entries have timestamps that allow us to calculate fresh-
ness. The entries made are confirmed by a human user
and hence the reliability is high. The entries do not have
certainty values hence; mock values for reliability and
certainty are used.
11.3 Test result
The test was performed on a machine with Intel 2.0 GHz
Core 2 Duo CPU, 2.0 GB of RAM and on a Windows XP
SP 3 operating system. The test was developed in Java
and executed on JRE 1.6. The time was recorded for the
calculation of freshness as well as confidence for each
entry and Mean time was calculated. The time was rec-
orded in nanoseconds. Fig. 6 shows the result. The curve
(with squares) shows that the time is a monotonically
increasing function of number of entries and has a linear
behavior as in (8). The line (with triangles) in Fig. 6 shows
the least squares line that minimizes the root mean square
error as shown in (9). This equation is derived through
the normal forms of least square line [21].
y = 68.12x + 4981u.47 (9)
The more the number of variables the more time
would be spent in calculating freshness and confidence.
Since for a comprehensive contextual data repository the
total attributes are bound to be in a large number, we
propose the notion of minimal context as discussed in
Section 8.7. The minimal context is more useful when
working in mobile hand held platforms where space and
time are constrained.


12 CONCLUSION
Context aware system promise smart service interactions
for pervasive environments. Context awareness is rea-
lized by gathering contextual data from the environment
and then inferring it to facilitate the everyday interac-
tions between users and services.
This paper presents the Context Gatherer and Repre-
sentation Module (CGRM) that acquires contextual data
from the environment and stores it using OWL. The de-
sign of this module is in two steps. In first step the mod-
eling technique is selected as OWL which is the de jure
standard. OWL allows us to represent data in platform
independent, XML based tagged language. The second
step involves the organization of contextual data. A Ser-
vice context vs. User context based organization is fa-
vored as day to day interactions are between services and
users. The organization in OWL has been developed us-
ing protg. In addition freshness and confidence calcula-
tor module and axiom firing module is developed that
allows us to perform allied processing.
The system is tested by performing time analysis of
the allied components that are external to the OWL file.
The associated processing exhibits a linear behavior
where in we suggest the use of a minimal context to re-
duce the total number of variables in the system.
ACKNOWLEDGMENT
Thanks to Instut fur Pervasive Computing and Jan Pet-
zold for providing online free access to context database
and test data. Special thanks to Col (R) Nasir Mahmud for
his interest in this work.
REFERENCES
[1]T.StrangandC.LinhoffPopien,AContextModellingSurvey,
Proc. of First International Workshop on Advanced Context
Modelling,ReasoningandManagement,2004
[2] S.B. Mokhtar, D. Fournier, N. Georgantas and V. Issarny,
ContextAware Service Composition in Pervasive Computing
Environments,inLectureNotesinComputerScience(LNCS),pp.
129140,2006
[3]M.Horridge,H.Knublach,A.Rector,R.StevensandC.Wroe,A
Practical Guide to buliding OWL Ontologies Using The
ProtegeOWLPluginandCOODEToolsEdition1.0.Protg,
[Online]http://protege.stanford.edu.
[4]Q.Z.Sheng,S.Pohlenz,H.S.Wong,A.H.H.NguandZ.Maamar,
CONTEXTSERV: A Platform for Rapid And Flexible
DevelopmentofContextAwareWebServices,Proc.IEEE31st
InternationalConferenceonSoftwareEngineering,(ICSE2009),pp.
619622,2009.
[5] S. W. Loke, Incremental Awareness And Compositionality: A
Design Philosophy For ContextAware Pervasive System,
JournalofPervasiveandMobileComputing,pp.239253,2010.
[6]H.R.SchmidtkeandW.Woo,TowardsOntologyBasedFormal
Verification Methods for Context Aware Systems, in Lecture
NotesinComputerScience(LNCS),pp.309326,2009.
[7]U.Mahmud,N.A.MalikandM.Y.Javed,FutureChallengesin
Context Aware Computing, Proc. WWW/Internet 2007, Villa
Real,Portugal,pp.306310,2007.
[8] A.K. Dey, Understanding and Using Context, in Personal and

JOURNAL OF COMPUTING, VOLUME 4, ISSUE 3, MARCH 2012, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 66

UbiquitousComputing,pp.47,2001.
[9] J.Gwizdka, Whats In The Context? Proc. Conference on
Human Factors in Computing Systems (CHI 00), The Hague,
Netherlands,(ACMSIGCHI),2000
[10] D. Petrelli, E. Not, M. Zancanaro, C. Strapparava and O. Stock,
ModellingandAdaptingtoContext,inPersonalandUbiquitous
Computing,pp.2024,2001.
[11] T. Hofer, W. Schwinger, M. Pichler, G. Leonhartsberger and J.
Altmann, ContextAwareness On Mobile Devices The
Hydrogen Approach, Proc. 36th Annual Hawaii International
ConferenceonSystemsSciences,Hawaii,pp.292302,2002.
[12] M. Riaz, S.L. Kiani, S. Han and Y. Lee, Service Delivery in
Context Aware Environments: Lookup And Access Control
Issues,Proc.ofRTSCA05,2005.
[13] B.N. Schilit and M.M. Theimer, Disseminating Active Map
InfrastructuretoMobileHost,inIEEENetwork,pp.2232,1994.
[14] U. Mahmud, N. Iltaf, A. Rehman and F. Kamran, Context
Aware Paradigm for a Pervasive Computing Environment
(CAPP), Proc. WWW/Internet 2007, Villa Real, Portugal, pp.
337346,2007.
[15] I. Roussaki, M. Strimpakou, C. Pils, N. Kalatzis and N.
Liampotis, Optimizing Context Data Dissemination and
Storage in Distributed Pervasive Computing Systems, in
JournalofPervasiveandMobileComputing,pp.218238,2010.
[16] A. Badii, M. Crouch and C. Lallah, a ContextAwareness
Framework for Intelligent Networked Embedded Systems,
Proc. Third International Conference on Advances in Human
Oriented and Personalized Mechanisms, Technologies and Services
(CENTRIC),pp.105110,2010.
[17] E.J.Y. Wei and A.T.S. Chan, CAMPUS: A Middleware for
AutomatedContextAwareAdaptationDecisionMakingatRun
Time, in Pervasive and Mobile Computing, preprint Nov,
2011.
[18] U. Mahmud, N. Iltaf and F. Kamran, Context Congregator:
Gathering Contextual Information in CAPP, Proc. 5th
International Conference on Frontiers of Information
Technology,(FIT07),Islamabad,pp.134141,2007.
[19]D.RiboniandC.Bettini,OWL2ModelingAndReasoningWith
Complex Human Activities, in Journal of Pervasive and Mobile
Computing,pp.379395,2011.
[20] J. Petzold, Augsburg Indoor Location Tracking Benchmarks,
InstutfurPervasiveComputing.[Online]January17,2005.[Cited:
August 18, 2011.]
http://www.pervasive.jku.at/Research/Context_Database/viewS
ubmission.php?key=9&action=View&keyname=b4a4&table=c0b
0ad8dc8d0&db=9e8f8772.
[21] J.H. Mathews and K.D. Fink,Numerical Methods Using
Matlab,PearsonEducation,p.265,Ed.4,2005.

Engr. Umar Mahmud did his BE in Software Engineering from Na-
tional University of Sciences and Technologies (NUST), Islamabad,
Pakistan, in 2003 and MS in Software Engineering from NUST, in
2006. He is presently pursuing PhD degree in Software Engineering
at, NUST. He has more than 6 years of teaching at UG level. He is
presently serving as Engineer in Department of Computer Software
Engineering, Military College of Signals (MCS), NUST. His areas of
interest are context-awareness, pervasive computing and machine
learning.

Prof. Umar Farooq did his BSc from Punjab University, Lahore,
Pakistan in 1973, and MSc from Quaid-e-Azam University, Islama-
bad, Pakistan, in 1975. He is currently serving as a Prof. of Mathe-
matics with Department of Humanities and Basic Sciences in Military
College of Signals (MCS), NUST. He has more than 36 years of
teaching experience.

Prof. Dr. Muhammad Younus Javed did his BE Electrical Engineer-
ing from UET Lahore, Pakistan, in 1982. He completed his MS in
Predictive Systems from the same university in 1988 and PhD in
Adaptive Communication Systems from University of Dundee, Scot-
land, United Kingdom in 1991. He is presently serving as DEAN in
the College of Electrical and Mechanical Engineering (CE&ME,
NUST. His areas of interest are biometrics, parallel systems, operat-
ing systems, computer networks, digital image processing, database
systems and design & application of algorithms. He has more than
200 national/international publications to his credit.

Nazir Ahmed Malik did his BSc (Hons) in Naval Sciences from Uni-
versity of Karachi, Pakistan in 1996 and then M.Sc. in Computer
Science from UAF, Faisalabad, Pakistan, in 2000. He is pursuing his
PhD degree in Information Security from College of Electrical and
Mechanical Engineering, NUST. He is presently serving as Head of
Distance Learning at Pakistan Naval Engineering College (PNEC),
NUST. His areas of interest are Security, Privacy, Trusted Computing
and Pervasive Computing.


Fig. 6. Time analysis of freshness and confidence calculator with
least squares line
40000
50000
60000
70000
80000
90000
100000
110000
120000
102 159 352 433 449 849 912 983
NumberofEntries
CalculationTime LeastSquaresLine
JOURNAL OF COMPUTING, VOLUME 4, ISSUE 3, MARCH 2012, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 67

You might also like