You are on page 1of 12

TEMPORAL INFORMATION SYSTEMS AND THEIR APPLICATIONS

TO MOBILE AD HOC ROUTING

V. Mary Anita Rajam, V.Uma Maheswari and Arul Siromoney


Department of Computer Science and Engineering,
Anna University, Chennai -- 600 025, India

Contact email: asiro@vsnl.com

ABSTRACT
A Temporal Information System, that incorporates temporal information into the
traditional Information System of Rough Set Theory (RST) and Variable Precision
Rough Sets (VPRS), is presented. Mobile Ad hoc Networks (MANETs)
dynamically form a network without an existing infrastructure. The Dynamic
Source Routing (DSR) protocol of MANETs is modified in this paper to use recent
routes. Weighted elementary sets are introduced in temporal information systems
and used to route packets in mobile ad hoc networks. Notions from VPRS are also
brought into weighted temporal information systems and used in routing. The
performance of these proposed routing protocols is studied.

Keywords: Rough Set Theory, Mobile ad hoc networks, Temporal Information


System, Routing.

1 INTRODUCTION symptoms and tests; and the decisions (or decision


attribute) may be diseases.
Rough Set Theory is a mathematical tool
that deals with vagueness and uncertainty. Mobile ad In an information system, elements that
hoc networks are a collection of wireless mobile have the same value for each attribute are
nodes that can dynamically form a network. The indiscernible and are called elementary sets. Subsets
state of each link in a mobile ad hoc network of the universe with the same value of the decision
changes with time. attribute are called concepts. A positive element is an
element of the universe that belongs to the concept.
This paper introduces temporal information
systems that are then applied to mobile ad hoc For each concept, the greatest union of
routing. elementary sets contained in the concept is called the
lower approximation of the concept and the least
Each mobile node maintains a route cache union of elementary sets containing the concept is
of known routes. It is shown in this paper that giving called the upper approximation of the concept. The
more importance to the recent routes in the route set containing the elements from the upper
cache is useful. This has led to the notion of approximation of the concept that are not members
weighted elementary sets in temporal information of the lower approximation is called the boundary
systems, where more recent elementary sets are region. The lower approximation of the concept is
given more importance. also known as the positive region.

1.1 Rough Set Theory A set is said to be rough if the boundary


region is non-empty. A set is said to be crisp if the
In Rough Set Theory (RST) [21], boundary region is empty.
introduced by Zdzislaw Pawlak, a data set is
represented as a table, where each row represents an Variable Precision Rough Sets (VPRS)
event or an object or an example or an entity or an [31], proposed by Ziarko, is a generalization of the
element. Each column represents an attribute that can rough set model, aimed at modelling classification
be measured for an element. This table is called an problems involving uncertain or imprecise
information system. The set of all elements is known information. Classification with a controlled degree
as the universe. For example, if the information of uncertainty is possible with this model. It is
system describes a hospital, the elements may be possible to generalize conclusions obtained from a
patients; the attributes (condition attributes) may be smaller set of observations to a larger population.

Ubiquitous Computing and Communication Journal 1


In RST, the lower approximation of a stamp is introduced as an attribute, in addition to the
concept is defined using an inclusion relation. Here condition attributes and the decision attribute of the
in VPRS, the lower approximation is defined using a traditional information system.
majority inclusion relation. The β-positive region is
the union of elementary sets which are either A temporal information system is
completely contained in the concept or are almost introduced by Bjorvand [29]. The temporal
contained in the concept, with a maximum error of information system has a sequence attribute in
1 – β. addition to the condition attributes and the decision
attribute present in the traditional information
The conditional probability of an element system. The value of the sequence attribute is an
being positive in an elementary set is the probability integer, based on the time of occurrence of the
that the element is positive, given that the element objects in the information system. Methods are
belongs to that elementary set. It is the ratio of the proposed in [29] and [28] to convert this temporal
number of positive elements in that elementary set to information system into the traditional information
the number of elements in that elementary set. When system, so that rough set techniques can be applied.
this conditional probability is greater than a threshold The method proposed in [29] depends on time
β (0.5 < β ≤ 1) the elementary set is said to fall in the intervals that must be fixed and defined in advance.
β-positive region. Trend expressions [28] are added to transform the
traditional method of translating a temporal
1.2 Rough Sets in Temporal Contexts information system into an information system. So, a
new attribute is added to the information system, for
A temporal system is a time based system which values are set based on the trend. A real time
which shows the temporal variation of some specific temporal information system is proposed in [29], in
data or attribute. Time-series data are a kind of which the difference between the time of occurrence
temporal data and are results of some observations of the current row and the previous row (if rows are
usually ordered in time. Time-series data often sorted according to time) is also stored in the
possess content that is conflicting and redundant. information table.
The data may be imprecise; therefore a precise
understanding of information cannot be derived from Having a linearly ordered universe, based
the data. Rough set theory offers a powerful toolset on time, is also used in bringing in the notion of time
for confronting this situation. into a rough set information system. This temporal
information system with a linearly ordered universe
Analysis of time-series data and [19], [8], [1], [3], [20], [2] provides information
constructing suitable data from the time-series that about the behaviour of objects in time and state of an
can be used by rough sets are investigated in [15], object is described by some attributes. The elements
[18], [11]. Reducts can be found from the original can be the behaviour over time of the same object,
data and rules can be generated from the acquired multiple objects, or objects independent of each
reducts using rough sets [15]. For constructing other.
suitable data from the time-series, different methods
are tried. In the mobile window method [4], a Temporal templates [19], which are
window is moved along the time-series; the data homogeneous patterns occurring in some periods, are
points falling into the window are transferred into a extracted from temporal information systems. The
rough sets object. The window method poses temporal templates are then used to discover
restrictions on how far back in time dependencies behaviour of temporal features of objects.
can be traced. In the columnizing method [18], the
time series are organized in columns, such that each Using temporal templates [20], a temporal
row represents an object where each column is an multiple information system [19] is introduced that
economic indicator, and each row represents a describes many objects along the time axis e.g.
different point in time. several users visiting a website. For each object, a
sequence of temporal templates is found. Collections
In [11], time-series is represented using a of patterns, called episodes, appearing in sequences
series of events or states. An event is something that frequently are found. New attributes are generated
occurs, and is associated with a time. The values of from the found frequent episodes.
attributes are trends, rather than values measured at
points in time, so that dependencies can be traced Temporal patterns that can be potentially
back as long as required. specific for one musical instrument or a group of
instruments are searched [8]. Such patterns are used
A decision table for time-series called a as new descriptors. A time window based technique
time series decision table, is proposed in [16]. A time is used to measure the values of the descriptors in

Ubiquitous Computing and Communication Journal 2


time. Optimal temporal templates that respond to 1.3 Application of Rough Sets to Computer
temporal patterns are then determined. From the Networks
temporal templates, episodes (collections of
templates that occur together) are found. Very little work has been done in the
application of rough set theory to Mobile Ad Hoc
Information maps can be constructed from Networks and Mobile Ad Hoc Routing. A few papers
data represented by temporal information systems have applied rough set theory to networks. Rough set
[1]. The temporal information system at the current theory is used in intrusion detection in computer
time t is also viewed as a family of decision systems networks [6], [14] and [32]. Rough set approach is
[1], where the universe of the decision system at time also applied to the flow control of a UDP-based file
t1, is a subset of the decision system at time t2, for transfer protocol [33] to accomplish a real time data-
t1 < t2. transferring rate adjustment task; to network fault
diagnosis [24]; and to achieve the rules for object
Some work is based on how the data varied recognition and classification for mobile robots [12].
over a particular duration of time (say, two years, Rough set logics is combined with artificial neural
two months etc.). Different attributes are assigned networks for failure domain exploration of
for each time period in [30]. telecommunication network [7]. The basic concepts
of rough set theory are illustrated using an example
A dynamic information system model concerning churn modeling in telecommunications
based on time sequence is proposed in [13]. The [23].
attributes in the information system can have
different values at different time points. 1.4 Overview of the Paper

1.3 Routing in Mobile Ad Hoc Networks This paper presents a Temporal Information
System that brings temporal information into the
Sending data from a source mobile node to traditional Information System of Rough Set Theory
a destination mobile node through a route (or and Variable Precision Rough Sets. The DSR
sequence of intermediate mobile nodes) is routing. protocol is modified to study the use of recent routes.
Routing is one of the most difficult issues in mobile The paper then introduces the notion of weighted
ad hoc networks. Each node in an ad hoc network is elementary sets in temporal information systems and
responsible for routing. Hence each node should uses this to route packets in mobile ad hoc networks.
maintain the information necessary for routing, that This paper then uses VPRS in weighted temporal
is, the next hop or the path through which the data information systems and uses this in routing. The
has to be routed to the destination. This information performance of these proposed routing protocols are
is either available even before it is needed studied.
(proactive) or is got only when necessary (reactive).
2 TEMPORAL INFORMATION SYSTEMS
Proactive routing is usually done using table
driven protocols where tables are formed initially, 2.1 Information Systems and Decision Systems
and are updated either periodically or when some
change is known. Consider a universe U of elements. An
information system I is defined as I = (U, A, V, ρ),
Reactive routing protocols are also known where
as on-demand routing protocols. In this kind of A is a non-empty, finite set of attributes;
protocols, a route is found for a destination only is the set of attribute values
when there is a need to send information to the of all attributes,
destination. Each node does not have knowledge of where is the set of possible values of
the location of all other nodes in the network. All attribute ;
nodes when they learn or use routes, store the routes is an information function,
to the destinations in a routing table or a cache. such that for every element ,
is the value of attribute for
Most of the on-demand routing protocols element .
[5], [25], [9], discover routes when needed by The information system can also be viewed as an
flooding route request packets in the network. A information table, where each element
route reply is sent back to the source node by either corresponds to a row, and each attribute
an intermediate node that has a route to the corresponds to a column.
destination or by the destination.
I = (U, A, V, ρ), is known as a decision
system, when an attribute is specified as the

Ubiquitous Computing and Communication Journal 3


decision attribute. A decision system is used for elementary sets whose conditional probability is
predicting the value of the decision attribute. greater than or equal to where 0.5. The -
is known as the set of condition attributes. negative region is the union of the elementary sets
The concept is the set of elements of that whose conditional probability is less than where
have a particular value (say, ) of the decision 0.5. These are based on the definitions in [34].
attribute . That is, When , we denote it as , and note that
Normally, is a boolean attribute that takes one of .
two possible values. When is known as
a multi-valued decision attribute. The range of is (0.5,1] in the original
VPRS definition. This indicates probabilistically that
These definitions are based on the definition the elementary set is positive, when the decision
of Rough Set Information System in [21, 22, 17]. attribute is boolean. It appears that when the decision
attribute is multi-valued with as the number of
2.2 Regions of the Universe possible values, the range of is (1/k,1].

An equivalence relation , called 2.3 Temporal Extensions


indiscernibility relation, is defined on the universe
as A Generic Temporal Information System
(GTIS) is defined as a set of information tables,
with each information table
located at a time on the time
In the information system , the elementary axis. A time interval can also be considered instead
set containing the element with respect to the of a time instance.
indiscernibility relation , is
A special case is when , that is, the
same universe of elements that appears in each
information table. A particular element , for
The lower approximation of the concept each attribute , would have a value
, with respect to and equivalence relation at each time . For example, patient $x$
on , is the union of the elementary sets of with has fever at , and does not have fever at . This
respect to that are contained in , and is denoted special case is when each .
as
In this paper, this is treated as a single
The upper approximation of is the union Temporal Information System defined as
of the elementary sets of with respect to that , where is the time attribute with
have a non-zero intersection with , and is denoted as a set of pairs with
as as a sequence of time instances,

The lower approximation of is also


known as the Positive region of . The set For each elementary set that is formed from
is called the Boundary region the set of attributes (ie the set of attributes without
of . The set is called the Negative region of the time information), there are now elementary
. sets, where the first elementary set consists of
elements that occurred between time instance and
The conditional probability that an element , the second elementary set of elements between
in an elementary set is positive is and , and the last elementary set of elements
between and . This can be pictured as vertical
blocks of elementary sets along a time axis.

The conditional probability that the element 2.4 Information System in a Mobile Node
in the elementary set is negative is
The use of an information system in mobile
ad hoc routing was introduced in [26]. The
information system was modified in [27] to represent
When the context is clear, the conditional probability the route better, by using the link information rather
of an elementary set is taken to be . than the node information. A threshold was used in
the identification of a good next hop.
The -positive region is the union of the

Ubiquitous Computing and Communication Journal 4


Let be a set of mobile nodes. A route is a hop. The decision attribute is taken as the next hop,
path through mobile nodes in and is denoted as a and the predicted next hop is also known as the
sequence of mobile nodes predicted value of the decision attribute.
Each mobile node maintains a
route cache that stores all the routes that knows. The destination can possibly be reached
Any route in the route cache is a path starting from through several different sequences of intermediary
that mobile node , and so , the first node in the nodes. In other words, several different
route, is itself. Any route in the route cache is a combinations of attribute values make it possible for
simple path, where no node repeats, that is, the destination to be reached. That is, elements in
for in the path, . So, . several different elementary sets correspond to routes
that lead to this particular destination. Thus several
Each mobile node has an elementary sets play a role in identifying the best
information table associated next hop for a particular destination. So, it is not
with it. Each row in the information table possible to use a single elementary set to predict the
corresponds to a route in the route cache maintained value of the decision attribute. The union of these
by that mobile node . elementary sets is used. In other words, for a
particular destination, the union is taken of all the
Let be the set of all possible links elementary sets that correspond to valid routes from
between the nodes. Each condition attribute the current mobile node to the destination.
corresponds to a particular link in the set of all
possible links between the nodes. So, is the same A stringent method of predicting the next
hop is when all the elements in this union of
for any and is denoted as . Each condition
elementary sets have the same value of the decision
attribute is a boolean attribute, with , and
attribute, then this value is taken as the predicted
is set to 1 or 0 depending on whether or not that link
next hop. In other words, all known routes to this
is present in the route corresponding to that element.
destination should have this particular node as the
So, since for every , and
next hop. This can also be considered as that value of
is denoted as since it is the same in each mobile
the decision attribute for which all these elementary
node .
sets are in its lower approximation. It is to be
remembered that the decision attribute is a multi-
A mobile node knows a route either because
valued attribute, and so the lower approximation is
the route is in a packet that passes through this
with respect to a value of the decision attribute.
mobile node, or because this mobile node is in
promiscuous mode, and this route is in a packet that
Another method is to have the predicted
passes between two nodes that are within range of
next hop as that value of the decision attribute where
this mobile node. When a mobile node knows a route
the union of these elementary sets is in the -positive
it is added to the route cache only if it is not identical
region. The conditional probability is determined
to a path or a sub-path of any other route already
using the union of elementary sets, and not a single
present in the route cache. However, every time a
elementary set. The probability that a particular next
mobile node knows a route, a row corresponding to
hop occurs given that the route leads to a particular
this route is always added to its information table.
destination is taken as the conditional probability.
This conditional probability should be greater than a
Consider an element x corresponding to
threshold . In other words, a large number of
a route . When a
known routes to this destination have this particular
row is added to the information table, the values of
node as the next hop.
the condition attributes corresponding to the links
are set as 1.
2.6 Temporal Decision System in a Mobile Node
2.5 Decision System in a Mobile Node
In a Temporal Information System (TIS) for
a mobile node, each element (corresponding to a
In traditional Rough Set Theory and VPRS,
route) has a particular value of the time attribute, that
the value of the decision attribute of a new element
is, each element falls in a particular time interval.
(or unknown element or test case) is predicted, based
This is determined by the time stamp of the next hop
on which elementary set it falls into. The elementary
of the route that corresponds to this element. The
set into which it falls is determined by the values of
Temporal Decision System (TDS) in a mobile node
the attributes of that element.
is used to predict the next hop.
The decision system in a mobile node is
An appropriate method (as described in the
used to predict the next hop for a particular
previous section) is used to determine the predicted
destination. This next hop is called the predicted next

Ubiquitous Computing and Communication Journal 5


next hop in each time interval. The predicted next route in the route cache, it initiates a route discovery,
hop for the TDS is then determined based on the and gets back a route reply with the route to the
number of time intervals in which it is the predicted destination. This source route is placed in the data
next hop. packet and the data packet is sent to the next hop in
the route.
The predicted value of the decision attribute
is determined from the TDS based on the probability When a data packet reaches an intermediate
of a particular value of the decision attribute being node, the source route in the data packet is used to
the predicted value in the different time intervals. forward the data packet to the next hop. If a node
This probability is the number of time intervals in while sending the data packet to the next hop, finds
which that value of the decision attribute is the that the link does not exist, it uses the shortest route
predicted value divided by the total number of time present in its route cache to that destination. If the
intervals. The predicted value of the decision route is not found in the route cache, route discovery
attribute is the value for which this probability is is done.
greater than a threshold . In other words, that
particular next hop has been the predicted next hop When a route discovery is required, the
in most of the time intervals. node broadcasts route request packets. If any
intermediate node receiving the route request has a
In Weighted Temporal Information Systems route to the required destination in its route cache, it
(WTIS) and Weighted Temporal Decision Systems sends a route reply to the initiator of the route
(WTDS), weights are assigned to the discovery. Else, the node appends its own address to
elementary sets between time instances and , the the route in the route request and re-broadcasts the
elementary sets between and , and the route request. If the route request reaches the
elementary sets between and , respectively. destination, the destination reverses the route and
The predicted value of the decision attribute is sends back a route reply to the initiator of the route
determined after associating weights with the time discovery.
intervals. The predicted value of the decision
attribute is that value of the decision attribute where When a path is to be added to the cache, the
the probability is greater than a threshold . The following are done. If a prefix of the path to be
probability is the sum of the weights of time intervals added is present in the cache, the rest of this path is
in which that value of the decision attribute is the appended to the path present in the cache. If the
predicted next hop divided by the sum of the weights whole of the new path to be added is not present in
of all the time intervals. the cache, the path is added to the cache. If there is
no free space in the cache, then a victim is picked up
When the more recent time intervals play a and the route to be added is put in the victim's place.
more important role, the weight of a more recent If there is any path or a subpath in the cache entry
time interval is higher than the weight of a less that is a prefix or the same as that of the path that is
recent time interval. added, then the time stamps of those links are set to
be equal to the time stamps of the links in the path
3 MOBILE AD HOC ROUTING USING that is added.
RECENT ROUTES
When a link error occurs, a route error with
This section first describes the original information about the dead link is sent to the original
Dynamic Source Routing (DSR) protocol, an on- sender of the data packet. Paths, or subpaths starting
demand routing protocol. The DSR protocol is then with the given dead link, are removed from the cache
modified so that the most recent applicable route in in nodes that receive the route error.
the route cache is used at each intermediate node.
The performance of the modified protocol is 3.2 DSRrecent
evaluated.
This section describes the proposed
3.1 Dynamic Source Routing DSRrecent protocol and the modifications made to
the existing DSR protocol.
In Dynamic Source Routing (DSR) [5],
each mobile node has a route cache to store the In the source node, in DSR, the shortest
routes that are known to that mobile node. route in the route cache is used, whereas in
DSRrecent, the route to the destination in the route
The source node is the node that wants to cache, that has the most recent next hop, is used
send a data packet. It uses the shortest route present (using algorithm findRecentRoute).
in its route cache to the destination. If there is no

Ubiquitous Computing and Communication Journal 6


In intermediate nodes, in DSR, the source (iv) Average hop count: The average
route is used to determine the next hop. Only if there number of hops from the source node to the
is a link error, the route cache is used. The shortest destination node.
route to the destination in the route cache is then
placed in the data packet, instead of the original The network simulator ns2 [10] is used for
source route. the experiments. The following parameters are ones
However, in DSRrecent, the best route is that have been often used in such studies. The
determined from the route cache (using algorithm random waypoint mobility model is used in a
findRecentRoute). The route in the data packet is rectangular field. Constant bit rate traffic sources are
then modified such that the route from the route used. The radio model in the simulator is based on
cache replaces the subpath (from the current node) in the Lucent Technologies WaveLAN 802.11.
the route in the data packet. providing a 2Mbps transmission rate. A transmission
range of 250 m is used. The link layer modeled is the
If no route is found (to the destination) in Distributed Coordination Function (DCF) of the
the route cache of the intermediate node, or if the IEEE 802.11 wireless LAN standard. The source-
found route will result in a loop, the data packet is destination pairs (connections) are spread randomly
forwarded according to the existing route in the data over the network. 512 byte data packets are used.
packet.
Nodes move in a field with dimensions
If a link error occurs in any node, the best 1500 m X 300 m with a maximum speed of 2 m/sec.
route found in the route cache (using algorithm The pause time is 20 seconds. The number of nodes
findRecentRoute) is used. If there is no route in the is kept fixed as 50. The number of communicating
route cache, route discovery is done. source-destination pairs is varied from 5 to 40.
Simulations are run for 1000 simulated seconds.

findRecentRoute() {
besttime = 0;
Packet Delivery Ratio (%)

bestroute = NULL; 120


foreach possible-route do 100
t = time of next hop in possible-route; 80
DSR
if t > besttime then 60
besttime = t; DSRrec
40
bestroute = possible-route; 20
end 0
end 5 10 20 30 40
return bestroute;
Number of Connections
}

3.3 Performance Evaluation Figure 1: Packet delivery ratio vs. number of


connections for DSR and DSRrecent
The performance of DSRrecent is evaluated
using the following metrics, that are normally used in
such studies:
6
Normalized control

(i) Packet delivery ratio: The ratio of the 5


overhead

data packets delivered to the application layer of the 4


DSR
destination to those sent by the application layer of 3
DSRrec
the source node. 2
1
(ii) Normalized control overhead : The ratio 0
of the number of control packets sent to the number 5 10 20 30 40
of data packets received in the application layer.
Number of connections

(iii) Average end-to-end delay: The average


delay from when a packet is sent by the source node
until it is received by the destination node. Figure 2: Normalized control overhead vs. number of
connections for DSR and DSRrecent

Ubiquitous Computing and Communication Journal 7


The packet delivery ratios for DSR and more recent time intervals are assigned higher
DSRrecent are very similar for 5 and 10 connections. weights than the less recent time intervals.
When the number of connections is increased from
20 to 40 there is an improvement in the packet The use of the WTIS to predict the value of
delivery ratio from 3% to 11% (Fig. 1). The the decision attribute has already been described in
normalized control overhead for DSRrecent is more section 2.6. This uses the predicted value of the
than that for DSR when the number of sources is 5. decision attribute in different time intervals. The
With increase in the number of connections from 10 experiment described here uses a simple approach to
to 40, there is an improvement of 4% to 25% over determine the predicted value of the decision
DSR (Fig. 2). In average hop length and average attribute in a particular time interval. A predicted
end-to-end delay, DSRrecent is seen to perform value of the decision attribute in a time interval has
worse than DSR as the number of connections is atleast one element, with that value of the decision
increased (Fig. 3, Fig. 4). attribute, in the union of elementary sets. That is, the
union is in the upper approximation for that value of
the decision attribute.

18
Average hop count

16
14 4.1 Routing Based on WTIS
12
10 DSR
8
Here, the route cache of the mobile node is
DSRrec used as the WTIS. Routes that are learnt and used are
6
4 added to the cache of the mobile node. When routes
2
0 are added, the time stamp of each link is added along
with the routes. However, unlike DSR, even if the
5 10 20 30 40
same route is present in the cache earlier, the new
Number of connections route is added with the new stamp stamps. So, the
cache now has the same route multiple times, but
with different time stamps.
Figure 3: Average hop count vs. number of
connections for DSR and DSRrecent In the source node, initially, as in DSR, a
shortest route in the route cache, if available, is
placed as the source route in the data packet. If not
available, route discovery is done.
Average end-to-end delay

10
9
8 Then in the source node, and in any
7 intermediate forwarding node, the WTIS is used to
6 DSR
5 determine the best next hop (using algorithm
4 DSRrec
3 findWeightBasedHop). If the next hop is found, and
2 does not result in a loop, the data packet will be
1
0 forwarded to this next hop. If this next hop is
different from the one in the source route that is
5 10 20 30 40
already in the data packet, this new next hop is
Number of Connections appended to the source route in the data packet at the
current node and the route is invalidated by setting a
flag in the data packet.
Figure 4: Average end-to-end delay vs. number of
connections for DSR and DSRrecent If a next hop cannot be determined from the
WTIS, or if the next hop results in a loop, if the
source route in the data packet has not been
4 MOBILE AD HOC ROUTING USING invalidated earlier, the data packet is forwarded
WEIGHTED TEMPORAL INFORMATION according to the source route. Else, a route discovery
SYSTEMS (WTIS) is done.

In Temporal Information Systems, each The total time is divided into time intervals.
elementary set is associated with a particular time The list of next hops to the destination that are
interval. In Weighted Temporal Information present in the route cache is found. For each possible
Systems, elementary sets in different time intervals next hop, from the current time interval till the initial
have weights. Since it is seen in the previous section time interval, a weighted sum of the number of times
that the recent route in the route cache is useful, that the particular next hop is used is found. More

Ubiquitous Computing and Communication Journal 8


weight is assigned if the next hop has been used in The average hop length and the average
the recent past. That is, the weights assigned end-to-end delay for TIME_WT are more than that
decrease for earlier time intervals. for DSR when the number of sources is 5 and 10.
But when the number of connections is increased
from 20 to 40, it is seen that there is a slight
improvement of about 2% in average hop length and
findWeightBasedHop(){ of about 5\% in average end-to-end delay over DSR
Find all possible next hops that will lead to the (Fig.7, Fig.8).
destination from this node;
foreach possible next hop nh
timeInterval = currentInterval;

Packet Delivery Ratio (%)


weightedSum = 0; 120
weight = maxWeight; 100
totalWeight = 0; 80
DSR
while timeInterval >= 0 do 60
if nh is used as a nexthop in TIME_WT
40
timeInterval then 20
weightedSum = weightedSum + 0
weight; 5 10 20 30 40
end
Number of Connections
timeInterval = timeInterval -1;
//previous timeInterval
totalWeight = totalWeight + weight;
weight = weight -1; Figure 5: Packet delivery ratio vs. number of
end connections for DSR and TIME_WT
ratio[nh] = weightedSum / totalWeight;
end
Find the nexthop nh for which the value of ratio
6
Normalized control

is the maximum and return


} 5
overhead

4
DSR
3
TIME_WT
The ratio of the weighted sum of the usage 2
of the node to the total weight is found. The node for 1
which the ratio is greater than a threshold $\beta'$ is 0
chosen as the next hop. 5 10 20 30 40
Number of connections
4.2 Performance Evaluation

The parameters used are the same as those


given in section 3.3. The size of the time interval is Figure 6: Normalized control overhead vs. number of
taken as 40 seconds. The value of $\beta'$ used is connections for DSR and TIME_WT
0.5.

The proposed protocol used in this section 16


Average hop count

is referred to as TIME_WT. The packet delivery 14


ratios for DSR and TIME_WT are nearly similar for 12
5 and 10 connections. When the number of 10 DSR
connections is increased from 20 to 40 there is a 8
6 TIME_WT
slight improvement in the packet delivery ratio from 4
5% to 7% (Fig. 5). 2
0
The normalized control overhead for 5 10 20 30 40
TIME_WT is more than that for DSR when the Number of connections
number of sources is 5 and 10. With increase in the
number of connections from 20 to 40, there is an
average improvement of 14% to 22% over DSR
(Fig.6).
Figure 7: Average hop count vs. number of
connections for DSR and TIME_WT

Ubiquitous Computing and Communication Journal 9


Average end-to-end delay The routing protocol is similar to that of the
7 previous section. The next hop is chosen using the
6 notion of threshold $\beta$ ($\beta$--positive
5 regions) as described in algorithm
4 DSR
findVPRSWeightBasedHop().
3 TIME_WT
2
1
5.2 Performance Evaluation
0
The parameters used are the same as that
5 10 20 30 40
given in section 3.3. The size of the time interval is
Number of Connections taken as 40 seconds. The value of $\beta$, $\beta'$
used are 0.6,0.5 respectively.

Figure 8: Average end-to-end delay vs. number of The proposed protocol used in this section
connections for DSR and TIME_WT is referred to as VPRS_WT. The packet delivery
ratio for VPRS_WT is less than that for DSR for 5
and 10 connections. When the number of
5 MOBILE AD HOC ROUTING USING connections is increased from 20 to 40 there is a
BETA-POSITIVE REGIONS IN WTIS slight improvement in the packet delivery ratio from
2% to 6% (Fig. 9).
5.1 Routing Based on Beta-Positive Regions
The normalized control overhead for
The experiment described in this section VPRS_WT is more than that for DSR when the
determines the predicted next hop as that value of the number of sources is 5, 10. With increase in the
decision attribute where the union of these number of connections from 20 to 40, there is an
elementary sets is in the $\beta$-positive region, as average improvement of 14% to 19% over DSR (Fig.
described in section 2.5. 10).

findVPRSWeightBasedHop(){ The average hop length and the average


Find all possible next hops that will lead to the end-to-end delay for VPRS_WT is more than that for
destination from this node; DSR when the number of sources is 5, 10. But when
Foreach possible next hop nh do the number of connections is increased from 20 to
TimeInterval = currentInterval - 1; 40, it is seen that there is a slight improvement of
weightedSum = 0; about 4% in average hop length over DSR (Fig. 11).
weight = maxWeight;
totalWeight = 0; The average end-to-end delay for
while timeInterval >= currentInterval – k VPRS_WT is similar to that of DSR when the
do number of sources is 5. But when the number of
nhopCount = the number of routes with connections is increased from 10 to 40, it is seen that
next hop nh and willlead to the destination there is an improvement of about 6% in average hop
in timeInterval ; length over DSR (Fig. 12).
totalCount = the number of routes that will
lead to the destination in timeInterval
ratio1 = nhopCount/totalCount
Packet Delivery Ratio(%)

if ratio1 > $\beta$ then 120


weightedSum = weightedSum + 100
weight; 80
DSR
end 60
timeInterval = timeInterval -1; //previous VPRS_WT
40
timeInterval 20
totalWeight = totalWeight + weight;
0
weight = weight -1;
5 10 20 30 40
end
ratio[nh] = weightedSum / totalWeight; Number of Connections
end
Find the nexthop nh for which the value of ratio is
greater than $\beta'$ Figure 9: Packet delivery ratio vs. number of
} connections for DSR and VPRS_WT

Ubiquitous Computing and Communication Journal 10


Using recent routes (DSRrecent) was found to
Normalized control 6 improve packet delivery ratio and normalized control
5 overhead. Temporal information was brought into
overhead

4 information systems. Recent elementary sets were


DSR
3 given more importance in the two proposed methods,
VPRS_WT TIME_WT and VPRS_WT. The VPRS_WT method
2
1 uses notions from VPRS. It was seen that the control
0
overhead is much better, while the packet delivery
ratio, average hop length and average end-to-end
5 10 20 30 40
delay are slightly better than that of DSR. It was also
Number of connections seen that the improvement in performance increases
with the number of connections.

Figure 10: Normalized control overhead vs. number 7 REFERENCES


of connections for DSR and VPRS_WT
[1] A. Skowron and P. Synak. Patterns in
Information Maps. In Rough Sets and Current
Trends in Computing, volume 2475 of Lecture Notes
16 in Artificial Intelligence, pages 453 – 460. Springer,
Average hop count

14
12
002.
10 DSR
[2] A. Skowron and P. Synak. Reasoning Based on
8 Information Changes in Information Maps. In Rough
6 VPRS_WT
Sets, Fuzzy Sets, Data Mining, and Granular
4
Computing, volume 2639 of Lecture Notes in
2
0 Artificial Intelligence, pages 229 – 236. Springer
2003.
5 10 20 30 40
[3] A. Wieczorkowska, J. Wroblewski, P. Synak and
Number of connections D. Slezak. Application of temporal descriptors to
musical sound recognition. Journal of Intelligent
Information Systems, 21(1):71 – 93, 2003.
Figure 11: Average hop count vs. number of [4] J.K. Baltzersen. An attempt to predict stock
connections for DSR and VPRS_WT market data: a rough sets approach. Master’s thesis.
1996.
[5] David B.Johnson and David A. Maltz. Dynamic
source routing in ad hoc wireless networks. In
Average end-to-end delay

7
Imielinski and Korth, editors, Mobile Computing,
6
volume 353. Kluwer Academic Publishers, 1996.
5
4 DSR [6] Zhongmin Cai, Xiaohong Guan, Ping Shaoa, and
3
Guoji Sun. A rough set theory based method for
VPRS_WT
2 anomaly intrusion detection in computer network
1 systems. In Expert Systems, volume 20, pages 251 –
0 259, 2003.
5 10 20 30 40 [7] Frank Chiang and Robin Braun. Intelligent
failure domain prediction in complex
Number of Connections
telecommunication networks with hybrid rough sets
and adaptive neural nets. In 3rd International
Information and Telecommunication Technologies
Figure 12: Average end-to-end delay vs. number of Symposium, 2004.
connections for DSR and VPRS_WT [8] Slezak D., Synak P., Wieczorkowska A., and
Wroblewski J. Kdd-based approach to musical
instrument sound recognition. In Proc. of the 13th
6 CONCLUSIONS International Symposium on Foundations of
Intelligent Systems, Vol. 2366, Lecture Notes in
This paper presents temporal extensions to Artificial Intelligence, Springer, pages 28 – 36, 2002.
Rough Set Theory and Variable Precision Rough [9] Rohit Dube, Cynthia D. Rais, Kuang-Yeh Wang,
Sets. These extensions are applied to Mobile Ad hoc and Satish K. Tripathi. Signal stability-based
routing. Illustrative experiments are described and adaptive routing (SSA) for ad hoc mobile networks.
the results are presented. IEEE Personal Communications, 4(1):36{45, 1997.
[10] K. Fall and K. Varadhan. The ns manual

Ubiquitous Computing and Communication Journal 11


(formerly ns notes and documentation), 2002. Hengshan Geng. Application of rough set theory in
http://www.isi.edu/nsnam/ns/doc/index.html. network fault diagnosis. In Proceedings of the Third
[11] Roed G. Knowledge extraction from process International Conference on Information Technology
data: A rough set approach to data mining on time and Applications (ICITA 05), pages 556 – 559 vol. 2.
series. In Mater's thesis, 1999. [25] C. Perkins and E. Royer. Ad-hoc on demand
[12] Wang Haijun and Chen Yimin. Sensor data distance-vector routing for mobile computers. In
fusion using rough set for mobile robots system. In Proceedings of the Second international workshop on
Proceedings of the 2nd IEEE/ASME International Mobile Computing Systems and applications, pages
Conference on Mechatronic and Embedded Systems 90 - 100, 1999.
and Applications, pages 1 – 5, 2006. [26] V. Mary Anita Rajam, V. Uma Maheswari, and
[13] Xiaowei He, Liming Xu, and Wenzhong Shen. Arul Siromoney. Mobile ad hoc routing using rough
Dynamic information system and its rough set model set theory. In 2006 International Conference on
based on time sequence. In Proc. of 2006 IEEE Hybrid Information Technology -Vol2 (ICHIT'06),
International Conference on Granular Computing, pages 80 - 83, November 2006.
pages 542 – 545, 2006. [27] V. Mary Anita Rajam, V. Uma Maheswari, and
[14] Peng Hong, Dongna Zhang, and Tiefeng Wu. Arul Siromoney. Extensions in mobile ad hoc
An intrusion detection method based on rough set routing using variable precision rough sets. In IEEE
and svm algorithm. In 2004 International Conference International Conference on Granular Computing,
on Communications, Circuits and Systems, ICCCAS pages 237 – 240, November 2007.
2004, pages 1127 – 1130 vol. 2. [28] Lin S., Chen S., and Ning Z. Several techniques
[15] Herbert J. and Yao J. T. Time-series data usable in translating a TIS into IS. 30(5), 2003.
analysis with rough sets. In Proc. of 4th International [29] Bjorvand A. T. Mining time series using rough
Conference on Computational Intelligence in sets -a case study. In Komorowski H. J. and Zytkow
Economics and Finance, pages 908 – 911, 2005. J. M., editors, Principles of Data Mining and
[16] Li J., Xia G., and Shi X. Association rules Knowledge Discovery, Vol. 1263, Lecture Notes in
mining from time series based on rough set. In Proc. Computer Science, Springer, pages 351 - 358, 1997.
of the Sixth International Conference on Intelligent [30] Kowalczyk W. and Slisser F. Analyzing
Systems Design and Applications, pages 509 – 516, customer retention with rough data models. In
2006. Komorowski H. J. and Zytkow J. M., editors,
[17] J. Komorowski, Z. Pawlak, L. Polkowski, and A. Principles of Data Mining and Knowledge Discovery,
Skowron. Rough sets: A tutorial. In S. K. Pal and A. Vol. 1263, Lecture Notes in Computer Science,
Skowron, editors, Rough Fuzzy Hybridization: A Springer, pages 4 – 13, 1997.
New Trend in Decision-Making, pages 3-98. [31] W.Ziarko. Variable precision rough set model.
Springer-Verlag, 1999. Journal of Computer and Systems Sciences, 46(1):39
[18] Shen L. and Loh H. T. Applying rough sets to – 59, 1993.
market timing decisions. Decision Support Systems, [32] Wang Xuren, He Famei, and Xu Rongsheng.
Special Issue: Data Mining for Financial Decision Modeling intrusion detection system by discovering
Making, 37(4):583 – 597, 2004. association rule in rough set theory framework. In
[19] Synak P. Temporal templates and analysis of Proceedings of the International Conference on
time related data. In Rough Sets and Current Trends Computational Intelligence for Modelling Control
in Computing, volume 2005 of Lecture Notes in and Automation and International Conference on
Computer Science, pages 420 – 427. Springer, 2000. Intelligent Agents, Web Technologies and Internet
[20] Synak P. Temporal feature extraction from Commerce, page 24, 2006.
temporal information systems. In Ning Zhong, [33] J. Zhang and R. D McLeod. A udp-based file
Zbigniew W. Ras, Shusaku Tsumoto, Einoshin transfer protocol (uftp) with flow control using a
Suzuki, editors, Foundations of Intelligent Systems, rough set approach. 2005.
14th International Symposium, ISMIS 2003, Vol. [34] W. Ziarko. Set approximation quality measures
2871, Lecture Notes in Computer Science, Springer, in the variable precision rough set model. In Proc. of
pages 270 – 278, 2003. 2nd Intl. Conference on Hybrid Intelligent Systems,
[21] Z. Pawlak. Rough sets. International Journal of Santiago, Chile, 2002.
Computer and Information Sciences, 11(5):341– 356,
1982.
[22] Z. Pawlak. Rough Sets — Theoretical Aspects
of Reasoning about Data. Kluwer Academic
Publishers, Dordrecht, The Netherlands, 1991.
[23] Zdzislaw Pawlak. Rough set theory and its
applications. In Journal of Telecommunications and
Information Technology, 2002.
[24] Yuqing Peng, Gengqian Liu, Tao Lin, and

Ubiquitous Computing and Communication Journal 12

You might also like