Content uploaded by Naveen Gupta
Author content
All content in this area was uploaded by Naveen Gupta on Apr 26, 2019
Content may be subject to copyright.
Journal of Network and Innovative Computing.
ISSN 2160-2174 Volume 4 (2016) pp. 209–219
c
MIR Labs, www.mirlabs.net/jnic/index.html
Ant Pheromone Evaluation Models Based Gateway
Selection in MANET
Naveen Kumar Gupta1, Deepak Gupta2, Rakesh Kumar3and Amit Kumar Gupta4
1Department of Computer Science and Engineering,
Motilal Nehru National Institute of Technology, Allahabad, India
naveenkumar.gpt@gmail.com
2Mathematics Department,
SR Group of Institutions, Jhansi, India
deepakjhansi2007@rediffmail.com
3Department of Computer Science and Engineering,
Madan Mohan Malaviya University of Technology, Gorakhpur, India
rkiitr@gmail.com
4Image Enterprises, Gurgaon, Haryana India
amitguptaedu053@gmail.com
Abstract:A mobile ad hoc network (MANET) is an infrastruc-
ture less, low cost, small range autonomous network where mo-
bile devices can share the data as well as resources. If any mo-
bile node in MANET wants to communicate with fixed host, it
requires discovering an appropriate Internet gateway. A novel
adaptive gateway discovery scheme using ant like mobile agent
has been proposed in this paper. Ant releases a chemical sub-
stance called Pheromone. In this paper, different pheromone
evaluation models are used to calculate this pheromone value.
On the basis of pheromone value, traffic and route stability has
been analyzed. To calculate this pheromone value, different
pheromone evaluation models are discussed in this paper. These
models provide an adaptive route to a gateway in different con-
ditions. A discovery value is computed and used in selecting an
optimal gateway in case of multiple gateways. The analytical
study carried out in this paper validate that ant scheme with
different model provides a better route discovery than existing
ones. The proposed scheme also provides a stable and an opti-
mal route between a mobile node and particular gateway.
Keywords:Ant-like mobile agent, Pheromone value, pheromone,
evaluation model, Discovery value.
I. Introduction
A mobile ad hoc network is a type of ad hoc network which
can be established as per requirement in small range. It
is autonomous, self configured, infrastructure less in nature
where each mobile node can move arbitrary within a par-
ticular range. A node in MANET can use the resources of
another node. A MANET can be a Wi-Fi connection, or any
other medium, such as satellite or cellular connection. This
network doesnt require any central administration. Nodes
can be mobile devices also. So node can move within the
range. An arbitrary node within the range can join network
and communicate with other nodes of the network. But there
is a limitation with MANET that a node in MANET can-
not communicate outside the network or Internet. To com-
municate with the Internet, it requires an Internet gateway.
Internet gateway provides communication between the local
network and global network i.e. Internet. So gateway discov-
ery techniques are needed as discussed in this paper. If more
than one gateways are discovered then need to select the best
one on the basis of different parameters e.g., hop counts, traf-
fic, congestion, load etc. Selecting best gateway is known as
adaptive gateway discovery. So this paper is presenting an
adaptive gateway discovery in MANET.
Ant like mobile agents (ALMA) helps in discover the gate-
way. The ants provide an optimal route to Internet gateways.
Ants use routing algorithms as discussed in [14,10,9] and
provide a route to the destination. Here destination is an In-
ternet gateway. Ant uses the concept ant colony optimization
for searching the food source. When the ant moves toward-
s the food source then they release a special chemical sub-
stance called pheromone. Other ants follow the same path
based on pheromone concentration. Pheromone concentra-
tion decreases with time and this decrement is known as E-
vaporation represented by ε. If any ant found a shorter path
then this path will have higher pheromone concentration. So
other ants will follow a path with maximum pheromone con-
centration. So in this way, an optimal path to the destination
is discovered. If any other ant passes through the same path
then pheromone concentration increases. This increment is
known as reinforcement represented by ρ.
There are some problems with ant based routing:
•Discovered path may not fit for the situation.
MIR Labs, USA
Ant Pheromone Evaluation Models Based Gateway Selection in MANET 210
•In a MANET, bandwidth is limited resource and high
traffic in route may create congestion.
•Optimal route should be identified as early as possible.
The pheromone values are network state indicator. If any
variation occurs because of link breakage or congestion (bot-
tleneck) then pheromone values will change which results
traffic flow variation. This variation can be helpful in routing
procedure to load balancing though some other path or start
to search for a new path. A new parameter Discovery Value
(DV) is also proposed in this paper. If a routing algorithm
discovered more than one gateway for any particular mobile
node in MANET then selection of gateway will depend on
DV. A gateway with minimum discovery value will be se-
lected. This discovery value is a calculation of hop counts,
stability, load, traffic and security parameters, etc. In this
way selecting a suitable gateway is known as adaptive gate-
way discovery.
The remainder paper is organized as follows: Related work
and literature reviews are discussed in section 2. In section
3 different ant pheromone evaluation models are discussed.
The Proposed gateway selection scheme is described in sec-
tion 4. Section 5 describes the Analytical model. Analytical
proof of proposed scheme is given in section 6. Finally, sec-
tion 7 gives the conclusion about the paper.
II. Related Work
Dorigo et al. [6] proposed the concept of ant system. They
used ant as an agent and solved some problems like travel-
ling salesman problem (TSP), the quadratic assignment prob-
lem and job scheduling problems, etc. They use the concept
of pheromone concentration to find the path from source to
destination. Yi et al. [14] showed that the ant system can be
utilized for multicast routing in the MANET and proposed
an improved ant based routing algorithm for MANET and
compared with other routing protocols. After simulation he
analyzed that ant based routing algorithm works better than
proactive and reactive routing algorithms in some situations.
Fernando Correia et al. [5] studied ants behavior based on
pheromone concentration and simulate with MANET rout-
ing problems. After studying, he proposed some models for
pheromone value evaluation in MANET. In this paper these
models are used for gateway discovery. We have studied the
evaluation of ant system and analyzed that ant related algo-
rithms can easily solve any routing related problems.
Ayyadurai et al. [1] proposed a hybrid routing scheme which
combines Ant and AODV, where AODV finds the path in
local MANET and ALMA finds the path towards gateway.
In this scheme, they used AODV which suffers from initial
delay and they did not consider network load and stability
factor to select appropriate gateway. Bin et al. [2] proposed
an adaptive Gateway discovery scheme. In this approach,
TTL values of gateway advertisement messages are adjust-
ed dynamically according to the Internet traffic generated
by the mobile nodes and their relative location from Inter-
net Gateways with to which they are registered but they did
not consider the stability factor. Gupta et al. [8] analyzed
different pheromone evaluation models and used progres-
sive pheromone reduction with maximum value (PPR-MV)
in proposed gateway discovery approach. In case of multiple
gateways, they used a Discovery Value to select an optimal
gateway.
An efficient load aware gateway discovery approach was pro-
posed by Srivastava et al. [13] where load is considered but
path may be unstable and continuously changing intermedi-
ate routes. Pandey et al. [12] proposed congestion avoid-
ing gateway selection scheme and used proxies to reduce the
network overhead but path with less congestion may not be
optimal path. Bo et al. [3] uses mobile IP to discover opti-
mal gateway but mobile IP suffers from frequent handovers
which is undesirable process. Yuste et al. [15] proposed an
adaptive gateway discovery scheme where interval of emis-
sion of gateway advertisement is dynamically adjusted. This
technique is shown to give better performance than conven-
tional proactive gateway discovery scheme but it is not effi-
cient.
Bouk et al. [13] proposed a gateway discovery algorithm on
the basis of multiple QoS path parameters. They proposed
two algorithms: first algorithm for reactive zone and second
algorithm for proactive zone. In this case, we must identi-
fy the suitable zone in advance which is very difficult and
complex. Zaman et al. [16] proposed an adaptive gateway
discovery approach based on path load balanced. This ap-
proach primarily focuses on maximal source coverage. To
cover maximum geographical area, this technique can pro-
vide better end-to-end delay and packet delivery ratio and in
some specific cases only. Zhanyang et al. [17] proposed a
simplified scheme for Internet connectivity where they used
the concept of virtual MANET and consider that all Internet
gateways are at fixed position but moving (dynamic) gateway
node may exist.
III. Pheromone Evaluation Models
When the ants go to search food, they always want to search
for a better (short) route. So initially they go in differen-
t (arbitrary) direction and release pheromones. Pheromones
concentration (value) decrease with time and long route takes
long time. So a route with the higher pheromone concentra-
tion is a better path. Other ants attract on higher pheromones.
Finally best route is selected and other ants follow the same
route for food collection. If this overall process will com-
pare with a networking environment then food source is an
Internet gateway and ants are data packets. More than one In-
ternet gateways (like food sources) are also possible. There
may be more than one path to reach the destination. In the
networking environment virtual ants are considered so ants
will release pheromone trail in the networks. Pheromone e-
vaporates with time and reaches to 0 and if any other ant pass-
es through the same route then it will reinforces (increases).
In the Ant System (AS) model, route selection from source
to destination (or next hop selection) is probabilistic. It us-
es heuristic selection based on pheromone values present in
links. These pheromone values are calculated by formula:
Φj,k,t =ε.Φj,k,t−1+
n
X
i=1
∆Φi
j,k,t (1)
where ∆Φi
j,k,t =(1
dj,k if (j, k)∈path use by ant i.
0otherwise.
211 Gupta et al.
where Φj,k,t is pheromone value of link between node jto
node kat time t.∆Φi
j,k,t is the pheromone evaporation re-
inforcement value and ε(0 <ε<1) is evaporation rate.
Reinforcement value is calculated upon the distance from n-
ode jto node k traveled by ant i. The total numbers of ants
are ’n’. This paper intended to identify the problems related
to MANET, like frequent change in topology, broken link,
congestion etc, on an initial stage and helpful in selecting the
best path available. To achieve this goal, we will use four
pheromone evaluation models for evaluating evaporation or
reinforcement on the optimum network situation. These four
pheromone evaluation models are given by Fernando Correia
et al. [2] as follows:
•Ant system pheromone model
•Temporal active pheromone model
•Progressive pheromone reduction model
•Progressive pheromone reduction with maximum value
model
A. Ant System Pheromone Model
In the equation (1), evaluation of pheromone value in AS
model is presented. In this equation pheromone value is not
depending on the previous network state, but in network en-
vironment routing always depends on the previous state. So
the equation (1) is modified to:
Φj,k,t =ε.Φj,k,t−1+
n
X
i=1
∆Φi
j,k,t ∧Φj,k,t−1(2)
This model identifies pheromone values on three following
different phases:
•Learning phase: In this phase, the pheromone value
increases in a logarithmic manner until reaching a sta-
tionary value. The reinforcement of the pheromone val-
ue depends on the evaporation value and the traffic rate
which flows through the path.
•Maintenance phase: In this phase, the pheromone val-
ue remains constant with small variations. Here, the re-
inforcement rate is equal to evaporation rate.
•Evaporation phase: When data transfer through the
link is over then pheromone value decreases rapidly un-
til reaches 0. This condition may occur due to broken
link or a traffic jam.
The Ant System pheromone model increases its pheromone
intensity until reaches a stationary value when the link capac-
ity has enough resources to transport the packets. It presents
a fast growth rate to reach a stable value. When the data
transfer session ends, or when in a presence of a bottleneck
due link problem, the AS model will react to this and present
a fast decrease of the pheromones. This model can be use-
ful when fast response is required. However, on a network,
changes on traffic transfer rate like packet jitter or a burst of
packets could create a response similar to link bottleneck and
give wrong information about the network state.
B. Temporal Active Pheromone (TAP) Model
The ant system pheromone model suffers from incorrect re-
sult in case bottleneck, because pheromone value decreas-
es very fast when the path is idle. So the temporal active
pheromones model, tries to solve this problem and manage
a real behaviour of the pheromones with MANET routing
problems. When real ants deposit a pheromone in the trail,
the pheromone is active for a particular time. Similarly, in
this model, even if the path is idle, pheromone value will ac-
tive for a certain time. In this model, when the pheromone
activity decrease, it will have a value equals to 0.
Φt=(1(active)t set < t < t set +δ
0(evaporation)t > t set +δ(3)
Tt=PΦt
where Φtand δrepresents the single pheromone state
at time tand the duration of activity respectively. The
pheromone intensity of a link Ttis the addition of all de-
posited pheromones on the link. The evaporation phase is the
time duration needed to stop the pheromone activity; hence
it is equal to the pheromone life time. This model needs to
simulate at each node which requires more memory and com-
putational capacity at each node. Memory will store updated
tables with pheromone values and timers deal with compu-
tational capacity is responsible to define the pheromone ac-
tivity. The timers must be set to a well specified value ac-
cording to the pheromone activity. On completion of data
transfer, the reinforcement rate decreases to 0. So network
state can be analysed during the maintenance phase. If the
network has sufficient resources to transfer the packet then
pheromone variation will be about to zero. In case of bro-
ken links or congestion, the pheromone refreshes value will
change.
The jitter and burst of data also affect the pheromone value
because they related to the pheromone activity decrease rate.
In this model, decrease rate is linear, so these effects can be
neglected and the link can get its pheromone value when lost.
Due to these reasons, the temporal active pheromone model
is more stable to the network state changes than the Ant Sys-
tem pheromone model. This model will be slower to detect
network problems than the AS model.
C. Progressive Pheromone Reduction (PPR) Model
The Progressive Pheromone Reduction model uses relative-
ly less node resources in computation of pheromone inten-
sity. The pheromone value depends on traffic flow of link.
On increment of traffic flow, pheromone value also increases.
But the increase rate directly depends on the link’s capacity
(bandwidth). Traffic variation can be analysed by pheromone
increase rate. This model presents two valid phases: the
maintenance phase and the evaporation phase. The main-
tenance phase is observed as long as the traffic flows through
the link. The evaporation phase is progressive and happen-
s when the traffic ceases its activity. There enforcement and
evaporation formulas of PPR model are described as follows:
Ant Pheromone Evaluation Models Based Gateway Selection in MANET 212
Φt=(Φt−1+ρ
Φt−1−ε(4)
where ε=(1with link activity.
ε∗2without link activity.
Where ρis the increase rate and Φis pheromone value.When
data transfer rate is constant, the pheromone value increases
with same increase rate. When this increase rate changes, it
means that the network is suffering from some problem. In
PPR model, increase rate ρis set to 1.The pheromone value is
evaporated after some constant time periodically. Every time
when evaporation procedure is called, the pheromone value
is decreased by evaporation factor ε. The value εdepend-
s on traffic passing through link. If the packets are passing
through the link then it represents the activeness of the link
and εis set to 1, however, in case of traffic jam, εupdates
its value every time the evaporation procedure is called and
it represents the idle state of that link. Thus, we can say that
minor traffic problems like packet burst, jitter or route repair
procedure cannot affects the pheromone value, but if any link
breaks during data transfer session then it will cause progres-
sive pheromone evaporation and without reinforcement, the
pheromone value decreases quickly to 0.
The PPR model is a simple probabilistic model to evaluate
the pheromone values. It represents the pheromone intensity
variation to identify and classify the state of network, with a
fixed scale of probability. The evaporation procedure, allows
that small variations in the traffic rate, doesn’t be notice in
the pheromone global value and the path will continue to be
marked as valid. When the data transfer through a route stops
in the network, the pheromone activity through the route also
stops and pheromone value periodically decreased to 0. On
stopping the pheromone activity, the route can be released.
D. Progressive Pheromone Reduction with Maximum Value
(PPR-MV) Model
This model is about equivalent to the progressive pheromone
reduction model, but there is a minor difference of dead-
line of pheromone value. In PPR-MV model, a max-
imum pheromone value is defined so it cant be greater
than MAX VAL. This model considers the two levels of
pheromone evaluation. The first level is attached data session
and second level is associated a link between nodes. The link
state indicator is represented by the variation of the sum of all
pheromones assign to the sessions present on that link. Link
has sufficient resources to begin a new data session then the
sum of pheromones will grow on a direct proportion of the
number of sessions.
In a highly dynamic network like MANET, selection of ap-
propriate neighbours is very important to data transfer. The
selection of appropriate neighbour depends on pheromone
value. Setting a limit on pheromone up to MAX VAL, the
variations in the pheromone intensity would not be sufficient
to recognize the jitter, packet burst or congestion situations,
but in case of traffic jam, the pheromone intensity will pro-
gressively decrease and returns to 0. The refresh and evapo-
ration formulas of PPR-MV models are defined as follows:
Φt=(Φt−1+ρif Φt−1< M AX V AL
MAX V AL otherwise (5)
Φt= Φt−1−ε
where ε=(1with link activity.
ε∗2without link activity.
PPR-MV model considers three phases to evaluate the
pheromone intensity.These are learning phase, maintenance
phase and evaporation phase. The learning phase is related
to the session packet rate in the path and it ends when the
pheromone value reach the maximum that has been selected.
Maintenance phase represents that data transfer session is ac-
tive and the maximum pheromone value is maintained. Sim-
ilar to PPR model, small traffic rate variations are absorbed
by the progressive evaporation procedure and the pheromone
value shouldn’t be almost the same. The evaporation phase
is related to the end of data session, traffic jam and periodic
decrement of pheromone value. When the data transfer com-
pletes its session, the pheromone progressive reduces to 0. In
this model, network state identification is difficult because of
slow reaction to the network state changes. This model has
ability to differentiate between active routes and idle routes.
IV. Proposed Gateway Discovery Approach
with Ant
In our proposed scheme, each mobile node is considered as
home of ants, where each data packet is considered as a sin-
gle ant. The gateway node is equivalent to the food source of
ants, where the ultimate goal of ants is to discover the opti-
mal food source with optimal route. Similarly, in MANET,
the ultimate goal is to discover the optimal gateway with op-
timal route. In the conventional algorithms minimum hop
counts and load / congestion on route were considered to s-
elect optimal gateway. In the proposed algorithm we have
used proper balance of hop count, load factor and stability
factor. So we can get an optimal gateway with suitable route.
A. ANT based Routing Protocol
Ant-based routing algorithms have been explored by S.
Marwaha, L. Hwang and D. Karthikeyan [11,9,10]. Ants
are simply agents having the ability to move across the
network. They move node to node randomly and update
the routing tables of visiting nodes. Routing ants contain a
history of visited nodes. On reaching a node, an ant updates
the routing table of visited node and updates its history.
When history size increases then overhead also increase,
so history size should be well defined. Other than history
size, the population of ants also affects routing overhead. In
ant-based routing algorithm, each ant works independently
and they cannot communicate with each other directly. In
the conventional ant algorithms the next hop is selected
randomly, but this paper implements no return rule [10]
while selecting the next hop at a node because if the next
hop selected is the same as the previous node then selected
link will not provide an optimal route. The General behavior
of ant is shown in Figure 1
213 Gupta et al.
Figure. 1: Ant traversing the network and providing routing
information to nodes
In our gateway discovery scheme, discovery time is reduced
because ALMA takes the responsibility to discover the gate-
way and provides a highly dynamic approach to discover an
optimal gateway in different conditions. It also provides au-
tomatic handover to another gateway while the running gate-
way becomes unavailable due to power off or any other prob-
lem. Each node frequently broadcasts ADV messages to its
neighbour nodes. The ADV message helps in maintaining
the neighbor list. This used for selecting the next neighbour
node by the ants. Hybrid MANET follows dynamic topolo-
gy of the network so connectivity of mobile nodes with other
active mobile nodes and with gateway nodes is not certain.
The dynamic topology of mobile nodes causes more delay in
searching the route to a destination either in MANET orIn-
ternet.
The Ant colony optimization approach, i.e., ALMA is used to
find the route to the gateway. The hybrid mechanism operates
simultaneously when a route to the destination is needed by
the source mobile node. In proposed scheme, the routing
table of the mobile node is updated dynamically as shown
in Table 1. The Internet gateway contains a routing table to
update the information from the Internet.
Table 1: Entries in Routing Table of mobile node
Destination Address
Next Hop Address
Number of Hops
Pheromone Value
Pheromone Decay Time
Balance Index
B. Gateway Discovery Scheme
When a mobile node requires an Internet connection, it must
discover an Internet gateway first. Initially mobile node uses
a default prefix address to discover the gateway. The source
mobile node initiates a gateway discovery process by send-
ing forward ant(FANT) message to its neighbours. FANT
maintains a history field of visited intermediate nodes. Each
neighbor multicasts FANT until it reaches gateway node.
This message passes through multiple intermediate nodes.
The intermediate node receives FANT, update ants history
and resend to a neighbour with maximum probability. In-
termediate node uses FANT to update its routing table with
pheromone value and decay time. On reaching FANT to In-
ternet gateway, it checks whether it is a first ant with the same
source. If any other ant reaches to the same gateway previ-
ously then drop this ant otherwise it will update the routing
table of gateway with source node, next hop, hop count and
pheromone value and then gets converted into backward ant
(BANT) message.
The BANT packet contains the global IPv6 address of the
gateway, pheromone value and pheromone decay time with
the balance index, the network prefix of the gateway, the pre-
fix length and lifetime. The BANT packet is unicast from
the gateway to the source mobile node will pass through the
list of intermediate nodes visited by the FANT packet. When
BANT reaches to its source then update the source routing ta-
ble with information generated from the link traversed. The
link probability is calculated using pheromone value and the
pheromone decay time given by ant. The balance index is
a ratio of current gateway load and gateway capacity. The
source mobile node receiving more than one BANT packet
selects the adaptive gateway and a route by receiving max-
imum link probability, minimum hop count with lesser bal-
ance index. The source mobile node auto-configures its IP
address after selecting the gateway using the prefix of the
gateway with the suffix wireless interface address. The ac-
tive intermediate nodes also preserve the updated adaptive
information set by the BANT packet before the link prob-
ability reaches a minimum threshold value over time. On
reaching the threshold value, the active intermediate mobile
node generates a new FANT packet with the source mobile
node address available from its routing table. The generat-
ed FANT packet updates the route to the gateway with either
the new active node or the old active node for forwarding
the data packets. The FANT and BANT packets explore and
quickly reinforce the paths to the gateway. They also en- sure
the previously discovered paths do not get saturated when the
actual link fails.
The proposed algorithm uses mobile agent to reduce the
route discovery latency with less convergence time. There
are three different scenarios are considered to discover the
gateway.
Scenario 1: All mobile nodes in the MANET have very lim-
ited resources i.e. memory, computational capacity, etc. In
this scenario ant system pheromone model is the best model
to discover the gateway. In the ant system pheromone model,
when data trans- mission session ends then pheromone value
rapidly decreases to zero. If data transmis- sion suffers from
traffic congestion or delayed due to some other problem then
this model assumes that the data transmission session is over
and pheromone value de- creases to zero, hence this model
suffers from incorrect result. This is the disadvantage of the
ant system pheromone model.
Scenario 2: If all mobile nodes have sufficient resources,
e.g. memory, computational capacity, timer and sufficient
bandwidth. This model removes the problems of ant sys-
tem pheromone model and pheromone value does not change
rapidly. This model is more stable than the ant system
pheromone model, but it is slower to detect the network prob-
lems.
Scenario 3: If some nodes have relatively less resources
and gateway node is much far from the mobile node then
PPR model and the PPR-MV model works efficiently.
But the problem with PPR model is that, it doesnt have
maximum pheromone threshold value so pheromone value
may reach to infinite. PPR-MV model remove this problem
Ant Pheromone Evaluation Models Based Gateway Selection in MANET 214
by limiting maximum pheromone value to MAX VAL. In
the proposed algorithm, the PPR-MV model is used for
calculating the pheromone values. This model gives the best
result in normal conditions. The proposed algorithms for
this scenario are as follows:
Algorithm: Gateway Discovery & Route Selection
Pre-assumption: Following assumptions have been made
•no return rule is considered in the algorithm.
•Every node has a table which contains the destination n-
ode, next hop, pheromone value and hop count for each
neighbour.
•index represents the Index of last neighbor detail insert-
ed in the table.
Main function for gateway discovery is given as follows:
Algorithm 1 Main Function
1: procedure
2: Initially set pheromone value for each link Φi,j = 0
3: if (S wants to send/receive data to/from internet host)
then
4: set Sas a source
5: GW Discovery(source)
6: end if
7: Select a gateway with minimum DV value
8: end procedure
The above algorithm is a main algorithm for gateway dis-
covery. In this algorithm pheromone values of all links are
initialized to zero. If any mobile node wants to use Inter-
net then it calls GW Discovery(source) algorithm where re-
questing node is source node. This GW Discovery (source)
algorithm will call further algorithms as per requirement and
return DV value for each discovered algorithm. The main al-
gorithm will select a gateway with minimum DV value. The
main function/algorithm of the gateway discovery uses the
following functions/algorithms:
Algorithm 2 GW Discovery (source)
1: procedure
2: initiate F ANT, f ant id, index ←0
3: F AN T [index + +] ←f ant id
4: F AN T [index + +] ←source
5: for (each neighbour Ns∈neighbour of source) do
6: if (Ns == GW )then
7: GW Discovered(N s, index, F AN T )
8: else
9: GW F Routing (Ns, index, F AN T )
10: end if
11: end for
12: end procedure
The source node initiates ant packet. This ant packet is in
form of array. The first index of this ant contains ant ID and
second index contains source ID of ant. After initialization
it is broadcasted to all of its neighbours. When it reaches
to any neighbour first of all it checks weather this node is
gateway node or any intermediate node. If current node is
gateway node then this algorithm will IGW Discovered (Ns,
j, FANT) otherwise IGW FRouting (Ns, j, FANT) algorithm
will be initiated. Here Nsis neighbour of source mobile node
and FANT is forward ant.
Algorithm 3 GW F Routing (MN , index, F AN T )
1: procedure
2: if (MN ∈F AN T )then
3: drop packet and exit
4: end if
5: F AN T [index + +] ←M N
6: for (each neighbour Nmn ∈neighbour of MN)do
7: if (Nmn == GW )then
8: GW Discovered(N mn, index, F AN T )
9: else if (ΦMN ,NMN ,t−1< MAX V AL)then
10: ΦMN ,NMN ,t = ΦM N,NM N ,t−1+ρ
11: else
12: ΦMN ,NMN ,t =M AX V AL
13: end if
14: if (t%maxtime == 0) then
15: ΦMN ,NMN ,t = ΦM N,NM N ,t−1−ε
16: if (link is active) then
17: ε= 1
18: else
19: ε=ε∗2
20: end if
21: end if
22: update entries of Table of node
23: GW F Routing (Nmn, index, F AN T )
24: end for
25: end procedure
Each mobile mode contains a table which has details of each
visited ant. The above algorithm checks the table of interme-
diate mobile node. If it already contains the detail of visiting
ant it means this ant packet is visited previously and ant pack-
et will be dropped. Otherwise ant packet will update node ID
in next empty slot and table of node will store detail about
this ant. The first entry of table contains ID of source node,
second slot will empty for gateway id, third slot holds select-
ed neighbour node ID of mobile node and finally fourth entry
stores the pheromone value for this neighbour mobile nose.
The pointer pos points the next empty slot in table at interme-
diate mobile node. In this protocol, ant going towards gate-
way is known as forward ant. The current intermediate node
will again broadcast the FANT to its neighbors until gateway
node is discovered. If a particular link is activated and not
used for a particular time t max then pheromone value will
be evaporated by given formula. If link is using again and
again then pheromone value of link will reinforce its value
until reached at MAX VAL.
Algorithm 4 GW BRouting(M N, N, index, BANT )
1: procedure
2: if (index==1) then
3: print(path information is present in BANT)
4: return BANT
215 Gupta et al.
5: else if (ΦBAN T [index],BAN T [index+1],t−1<
MAX V AL)then
6: ΦBAN T [index],BAN T [index+1],t =
ΦBAN T [index],BAN T [index+1],t−1+ρ
7: else
8: ΦBAN T [index],BAN T [index+1],t =MAX V AL
9: end if
10: if (t%maxtime == 0) then
11: ΦBAN T [index],BAN T [index+1],t =
ΦBAN T [index],BAN T [index+1],t−1−ε
12: if (link is active) then
13: ε= 1
14: else
15: ε=ε∗2
16: end if
17: end if
18: update entries of Table of node
19: index =index −1
20: GW BRouting(BAN T [index], N, index, B ANT )
21: end procedure
The above algorithm is also applied on intermediate mobile
node but this algorithm is applies when gateway is discov-
ered. This procedure applies in reverse direction i.e. from
gateway node to source node so this ant packet is also known
as BANT. The ant packet follows the route according to in-
termediate node entry present in BANT packet. The selected
route is present in BANT packet. The BANT packet contains
reverse path. The pointer pos points the current intermedi-
ate mobile node. The pheromone value reinforcement and
evaporation is same as IGW FRouting (MN, N, j, BANT[])
algorithm.
Algorithm 5 GW Discovered(GW, index, F ANT )
1: procedure
2: if (GW ∈F AN T )then
3: drop packet and exit
4: end if
5: initialize BANT
6: balance index ←
(current gateway load/gateway capacity)
7: DV ←H op ∗FH+balance index ∗FL+FS/RS
8: F AN T [index]←DV
9: for (p= 0 to index)do
10: BAN T [p]←F AN T [p]
11: end for
12: N←index
13: if (ΦGW,F ANT [index−1],t−1< MAX V AL)then
14: ΦGW,F ANT [index−1],t =
ΦGW,F ANT [index−1],t−1+ρ
15: else
16: ΦGW,F ANT [index−1],t =M AX V AL
17: end if
18: if (t%maxtime == 0) then
19: ΦGW,F ANT [index−1],t =
ΦGW,F ANT [index−1],t−1−ε
20: if (link is active) then
21: ε= 1
22: else
23: ε=ε∗2
24: end if
25: end if
26: update entries of Table of node
27: index =index −1
28: GW BRouting(BAN T [index], N, index, B ANT )
29: end procedure
The above algorithm IGW Discovered (GW, j, FANT []) used
when gateway is discovered by FANT packet. In the first
stage it also checks for duplicate discovery. If FANT is vis-
ited previously then simply drop this packet otherwise ini-
tialize BANT packet. Gateway node calculates balance index
and Discovery Value. Discovery value is calculated using
stability factor in terms of pheromone value, hop count in
terms of number of intermediate visited node and balance in-
dex. We decided fraction of each parameter in simulation
part. So, gateway node will calculate discovery value and
put it into last slot of FANT packet. When multiple BANT
packets received at source node then source node selects on-
ly one path which has minimum DV. After calculating DV
at gateway node, FANT will convert to BANT and all values
with index copied to BANT packet. The pheromone value
reinforcement and evaporation is same as IGW FRouting al-
gorithm. Gateway node also contains a routing table and s-
tores all information same as intermediate mobile node. The
gateway node table has one extra field i.e. information about
discovery value. At the end, it calls IGW BRouting algorithm
until BANT reached at source mobile node.
V. Analytical Model
Through our analytical model, the proposed approach is
proved analytically. Each node has a communication range
R, where it can communicate with other node i.e. neighbour
node. In the proposed scheme, a pheromone value Φi,j is
assigned to every link between node iand node j. When ant
passes through this link, the pheromone value is increased
by reinforcement factor. The pheromone value decreases
with some value after every ttime. The ultimate goal of this
scheme is reaching at the gateway node. The ant scheme is
a probabilistic approach so next neighbour should be chosen
with highest probability. The Equation (6) computes the to-
tal probability of the link towards the gateway g. The mobile
agent located at particular node uses pheromone value Φi,j,g
to calculate the link probability Pi,j,g towards the gateway g.
The specific next hop neighbour is chosen according to the
probability distribution in each link.
Pi,j,g =
(Φi,j,g+T)S
PNi
j=1(∆Φi
i,j,g+T)Sif j∈Ni
0otherwise
(6)
In the above equation, Sand Tare pheromone sensitivity val-
ue and threshold value respectively. S≥0, used to modulate
the differences between pheromone amounts present in link
probability. The value S < 1evaporate the link, while S > 1
will reinforce the differences between links. The value Se-
qual to 1gives the normal form. If T≥0larger then large
amounts of pheromone will be present before an appreciable
Ant Pheromone Evaluation Models Based Gateway Selection in MANET 216
effect will be seen in the link probability. The link probability
Pi,j,g of the node i fulfills the constraint in giving transition
probabilities as in Equation 7.
X
j∈Ni
Pi,j,g = 1, i ∈[1...N](7)
During the route finding process to the gateway, mobile a-
gents deposit pheromone on the edges. The pheromone con-
centration is varying from one edge to another between the
nodes connecting them by an amount ∆Φ. The pheromone
value changes at the edge e(i, j) when mobile agents moving
from node ito node jare given in Equation 8.
Φi,j,g = Φi,j,g + ∆Φ (8)
The selection decision gives the fact that the path length be-
tween the nodes connecting the gateway is less than R and
the route selected is stable. The various symbols used in this
analytical model are briefly described in Table 2.
Table 2: Description of Symbols used
Symbol Description
Φi,j Pheromone value of edge e(i, j)∈E
Pi,j Probability value of edge e(i, j)∈E
V s Source node (vertex)
R Mobile Nodes communication range
T Pheromone Threshold value (constant)
S Pheromone Sensitivity value (constant)
E Set of all edges (links) in wireless network
N i Set of Neighbors of node i
Nn=|V|number of nodes in the network
V g Gateway node (vertex)
A. Route Maintenance
The route maintenance algorithm is responsible for maintain-
ing the same established route to the gateway or establishing
any other better route. Ants continuously move from one
node to another node and pheromone value evaporate after
time max time. When a data packet passes through node i
toward the gateway gto a neighbour node j, it increases the
pheromone value if link e(i,j) by ∆Φ i.e. represented in equa-
tion 8. The evaporation process is represented in equation 9
given below:
Φi,j = (1 −q)Φi,j , q ∈[0...1] (9)
In both the cases pheromone value of the route keeps
on changing. In this case new route with minimum D-
Vmay discover but continuously route cannot be changed
because it will represent the instability of the route. The
new pheromone value will be compared with the existing
pheromone value and if new value is at least 20 percent less
than the old pheromone value then the route will be updat-
ed. The proposed algorithm for route maintenance is given
as follows:
Algorithm 6 Gateway and Route maintenance
Pre-assumption: Time twill increase at constant rate
1: procedure
2: set t= 0
3: for (each link Li,j,g ∈network) do
4: if (ANT passes through Li,j,g)then
5: Φi,j,g ←Φi,j,g + ∆Φ
6: end if
7: if (max time%t== 0) then
8: Φi,j,g ←(1 −q)Φi,j,g
9: set t= 0
10: end if
11: end for
12: calculate DV periodically based on information giv-
en by ANT
13: if (DV new < (DV old0.2∗DV old)) then
14: update route and Internet gateway according DV
new and ANT
15: end if
16: end procedure
The algorithm represents the route maintenance procedure
from source node to gateway node. If any ant passes through
the same route then pheromone value of this route increas-
es by reinforcement constant on other hand if no ant fol-
lows the same route for a particular time then pheromone
value decreases with evaporation factor. The DV depends on
pheromone value and other factors. During data transfer if
ant finds new route with minimum discovery value then we
should change the route but in this way discovery process
will be unstable and we need to handover the connection a-
gain and again. To manage the stability of discovery pro-
cess we considered a new logic, if newly discovered route
has 20% less DV then only route will change otherwise route
will remain same. If any node receives a duplicate packet,
it sets the DUPLICATE ERROR flag and sends the packet
back to the previous node. The previous node deactivates the
link to this node then any data packet will not sent to this
direction.
B. Connection Recovery
In the proposed scheme, network is considered as dynamic
in nature. So any node may join or leave the network any
time. If node moves from the network then connection to the
gateway may be lost. This may happen due to link failure,
bandwidth limitation, node movement or power failure. In
this case connection recovery mechanism is required. When
connection is lost, link probability of related nodes goes to
zero, i.e. identified by neighbour nodes. Then neighbour
node again establishes the connection with maximum link
probability and connection to the gateway is recovered.
VI. Result and Discussion
To evaluate the proposed scheme, we have considered an ad
hoc network. In this network, we have evaluated some results
on the basis of variable number of mobile nodes. On the basis
of result of analytical study, we observed that ant scheme
works better than proactive and reactive gateway discovery
schemes. Figure 2 and Figure 3 represent the comparative
analysis of the packet delivery ratio and end to end delay
respectively.
217 Gupta et al.
A. Packet Delivery Ratio (PDR)
It is the ratio of total data send from source and received at
the destination. In this case, the mobile node is the source
and gateway is the destination when data is uploading. In
case of downloading, the mobile node is the destination and
gateway node is the source. So we have to consider both
the cases. When simulation starts then FANT packets broad-
casted in all directions. In this case, less number of FANT
packets are received at proper destination. So packet delivery
ratio is relatively lesser than proactive and reactive gateway
discovery scheme. When first BANT message received at
source node then connection has been established. After es-
tablishing the connection, the probability of losing the pack-
ets gets reduced. Figure 2 represents the result of the com-
parison of ant scheme with proactive and reactive scheme
and proved that ant scheme works better than proactive and
reactive scheme, in terms of packet delivery ratio.
Figure. 2: Packet Delivery Ratio
B. End-to-End Packet Delivery Latency
End to end delay is the time difference between data send
from source and received at the gateway. This includes al-
l delays due to buffering during the route discovery time,
queuing at the interface queue, and retransmission latency
at the MAC layer, as well as propagation and transmission
time. In initial stage, proper path is not established. So, ants
pass through random path. In this case end to end delay in-
creased, but after finding the optimal path, packet directly re-
ceived as destination without much delay. As time increases,
source node knows about all paths. So, in case of connection
breakage, it establishes another optimal path rapidly. Figure
3 describes the comparative analysis of end to end delay with
proactive and reactive gateway discovery schemes and shows
that on increasing time, end to end delay get decreased.
Figure. 3: End-to-End delay
VII. Conclusions and Future Directions
In this paper, four models for pheromone value computation
are discussed. Each model has different characteristics, so
each model can be used in some specific condition. The
Pheromone variation, used in this paper is an indicator of
network state. The proposed approach overcomes the short-
comings of other existing gateway discovery protocols and
provides the optimal route to the Internet gateway with less
discovery time, high packet delivery ratio and also minimizes
end to end delay. These improvements are obtained at the
cost of slightly higher average overhead. This is possible
with the help of a new metric i.e. discovery value. This new
metric combines hop count, stability factor, congestion and
load on the gateway. A route with minimum discovery value
has been selected.
The future work can involve generation of ants in the net-
work, as per requirement. Loss of ants should be implement-
ed in certain conditions. There is a need of adding inter-
agent communication and intelligence to the ant agents dur-
ing route discovery.
References
[1] V. Ayyadurai, R. Ramasamy. Internet Connectivity for
MANET Using Hybrid Adaptive Mobile Agent Proto-
col, International Arab journal of information technol-
ogy, V (1), pp. 25-33, 2008.
[2] S. Bin, S. Bingxin, L. Bo, H. Zhonggong, Z. Li. Adap-
tive gateway Discovery Scheme for Connecting Mobile
Ad Hoc Networks to the Internet. In Proceedings of the
IEEE International conference on wireless communica-
tions, networking and mobile computing, pp. 795-799,
2005.
[3] L. Bo, Y. Bin, S. Bin. Adaptive Discovery of Internet
Gateways in Mobile Ad Hoc Networks With Mobile IP-
based Internet Connectivity. In Proceedings of the 5 th
IEEE International conference on wireless communica-
tions, networking and mobile computing, pp. 1-5, 2009.
[4] S. H. Bouk, I. Sasase, S. H. Ahmed, N. Javaid. Gate-
way Discovery Algorithm Based on Multiple QoS Path
Parameters between Mobile Node and Gateway Node.
In Journal ofCommunication and Networks, XIV, pp.
434-442, 2012.
[5] F. Correia, V. J. Lobo, T. Vazao. Models for pheromone
evaluation in ant system for mobile ad hoc networks. In
Proceedings of the First IEEE International Conference
on Emerging Network Intelligence, pp. 85-90, 2009.
[6] M. Dorigo, A. Colorny, V. Maniezzo. Ant system: op-
timization by a colony of cooperating agents. In IEEE
Transactions on Systems, Man, and Cybernetics, Part
B, XXVI (1), pp. 29-41, 1996.
[7] N. K. Gupta, A. K. Gupta, R. Kumar. An Adaptive
Gateway Discovery using Mobile Agent in MANET. In
International Journal of advanced Research in Com-
puter Science and Software Engineering, IV (2), pp. 4-
10, 2014.
Ant Pheromone Evaluation Models Based Gateway Selection in MANET 218
[8] N. K. Gupta, R. Kumar, A. K. Gupta, P. Srivastava.
Ant Pheromone Models Based Gateway Selection in
MANET. In Proceedings of the 6 th International con-
ference on Innovations on Bio-Inspired Computing and
Applications (IBICA), CCCXXIV, pp. 297-311, 2014.
[9] L. Huang, H. Han, J. Hou. Multicast Routing Based on
Ant System. In Applied Mathematical Sciences, I, pp.
2827-2838, 2007.
[10] D. Karthikeyan, M. Dharmalingam. Ant based intelli-
gent routing protocol for MANET. In Proceedings of
the IEEE International Conference on Pattern Recog-
nition, Informatics and Mobile Engineering (PRIME),
pp. 11-16, 2013.
[11] S. Marwaha, C. K. Tham, D. Srinivasan. A Novel Rout-
ing Protocol Using Mobile Agents and Reactive Route
Discovery For Ad Hoc Wireless Networks. In Proceed-
ings of the IEEE International Conference on Network-
s, pp. 311-316, 2002.
[12] A. Pandey, N. Gupta, A. Agarwal, R. Kumar, J.
Prakash. Congestion Avoiding Gateway Discovery and
Selection Scheme for MANET. In Proceedings of the
Eighth International Conference on Computer Commu-
nication Network (ICCN), pp. 39-47, 2014.
[13] P. Srivastava, R. Kumar. An Efficient Load-Aware
Gateway Discovery for Internet Connectivity in
MANET. In Proceedings of the Eighth International
Conference on Computer Communication Network (IC-
CN), pp. 87-95, 2014.
[14] Z. Yi, L. Y. Chun. An improved ant colony optimisation
and its application on muticast routing problem. In In-
ternational Journal of Wireless and Mobile Computing,
V (1) pp. 18-23, 2011.
[15] J. Yuste, A. Trivio, F. D. Trujillo, E. Casilari. Improved
Scheme for Adaptive Gateway Discovery in Hybrid
MANET. In Proceedings of the 30 th IEEE Internation-
al Conference on Distributed Computing System Work-
shop (ICDCSW), pp. 270-275, 2010.
[16] R. U. Zaman, K. U. R. Khan, A. V. Reddy. Path load
balanced Adaptive Gateway Discovery in integrated
Internet-MANET. In Proceedings of the Fourth IEEE
International Conference on Communication Systems
and Network Technologies (CSNT), pp. 203-206, 2014.
[17] X. Zhanyang, H. Xiaoxuan, Z. Shunyi. A Simplified
Scheme of Internet Gatyeway Discovery and Selection
for MANET. In Proceedings of the 5 th IEEE Inter-
national Conference on Wireless Communication, Net-
working and Mobile Computing, pp. 1-4, 2009.
[18] T. Zhi, Z. Hui. An Improved Ant Colony Routing Algo-
rithm for WSNs. In Journal of Sensors, MMXV Article
ID 438290 pp. 4, 2015.
[19] R. U. Zaman, K. U. R. Khan, A. V. Reddy. Weighted
Load Balanced Adaptive Gateway Discovery in Inte-
grated Internet MANET, in Distributed Computing and
Internet Technology, R.Natrajan, G. Barua and M. R.
Patra (eds.), Springer International Publishing, Switzer-
land, 2015.
Author Biographies
Naveen Kumar Gupta received his B.
Tech degree in Computer Science and
Engineering from Inderprastha Engi-
neering College Ghaziabad (U.P.), In-
dia in 2011 and M. Tech in Infor-
mation Technology from Madan Mo-
han Malaviya University of Technolo-
gy Gorakhpur (U.P.), India in 2014. P-
resently he is pursuing PhD from Moti-
lal Nehru National Institute of Technol-
ogy Allahabad (U.P.), India since July,
2015. He also worked as Assistant Professor at SR Group
of Institutions Jhansi (U.P.), India from August-2014 to July-
2015. He has published five research papers in International
and one in National Journal/ conferences. He has completed
his M. Tech thesis on the topic Adaptive Gateway Discovery
in Mobile Ad hoc Networks. His main research interest lies
in Wireless Sensor Networks and Ad hoc Routing protocols.
Dr. Deepak Gupta is workingas As-
sistant Professor in the Department of
mathematics at SR Group Of Insti-
tutions, Jhansi since 2006. He was
awarded Ph.D in Mathematics from
Bundelkhand University Jhansi in the
year 2013 under the guidance of An-
jana Solanki. He work in the area
of Non-Markovian queueing model and
Fuzzy logic.
Dr. Rakesh Kumar received his B.E.
degree in Computer Engineering from
M.M.M. Engineering College Gorakh-
pur (U.P.), India in 1990, M. E. In
Computer Engineering from S.G.S. In-
stitute of Technology and Science, In-
dore, India in 1994 and Ph D in Com-
puter Engineering from Indian Institute
of Technology, Roorkee, India in June
2011. Presently he is an Associate Professor in the depart-
ment of Computer Science and Engineering at Madan Mohan
Malaviya University of Technology Gorakhpur since January
01, 2006. He has supervised a large number of M Tech Dis-
sertations and also supervising four Ph D students. He has
published a large number of research papers in International
and National Journals and conferences of high repute. He is
a recipient of best research paper award and also principal
investigator of a major research project ongoing since June
2013 sanctioned for three years from University Grant Com-
mission, India. He is a life member of CSI, ISTE and also a
Fellow of IETE and IE (India). His main research interests
lies in Mobile Ad hoc Routing, Quality of Service Provision-
ing, MANET-Internet Integration and Performance Evalua-
tion.
219 Gupta et al.
Amit Kumar Gupta received his B.
Tech degree in Computer Science and
Engineering from Institute of Technol-
ogy and Management Meerut (U.P.),
India in 2012. Presently he is a Soft-
ware Engineer in Image Enterprises,
Gurgaon, (Haryana), India. He has
published four research papers in Inter-
national and one in National journals /
conferences. His main research interest
lies in Mobile Ad hoc Routing protocols and security issues.
