Cost Effective Routing Protocols Based on Two Hop Neighborhood Information (2NI) in Mobile Ad Hoc Networks

Int. J. Advanced Networking and Applications
Volume: 07 Issue: 03 Pages: 2771-2778 (2015) ISSN: 0975-0290
2771
Cost Effective Routing Protocols Based on Two
Hop Neighborhood Information (2NI) in Mobile
Ad Hoc Networks
Dr. Anuradha Banerjee, Dr. Paramartha Dutta
----------------------------------------------------------------------ABSTRACT-----------------------------------------------------------
Ad hoc networks are collections of mobile nodes communicating with each other using wireless media without any
fixed infrastructure. During both route discovery and traversal of route-reply packets from destination to source,
broadcast of packets is required which incurs huge message cost. The present article deals with the message cost
reduction during transmission of route-reply from destination to source. Also the redundancy that is visible within
the 2-hop neighborhood of a node is minimized during broadcasting of route-reply. This improves the average
lifetime of network nodes by decreasing the possibility of network partition. The scheme of 2NI can be used with
any reactive routing protocol in MANETs.
Keywords – Ad hoc networks, Broadcast Balloon, Cost-Effective, Route-request, Route-reply, Two-hop
neighborhood information.
--------------------------------------------------------------------------------------------------------------------------------------------------
Date of Submission: Nov 30, 2015 Date of Acceptance: Dec 04, 2015
--------------------------------------------------------------------------------------------------------------------------------------------------
I. INTRODUCTION
AD hoc networks are multi-hop wireless networks
consisting of radio-equipped nodes that may be stationary
or mobile. They communicate with each other in a
collaborative way mostly over multi-hop wireless links
without the help of any fixed network infrastructure or
centralized administration. These are deployed mainly in
battlefields and disaster situations such as earthquakes,
floods etc. Many routing protocols have been proposed for
ad hoc networks. They can be categorized as proactive and
reactive routing protocols.
Among proactive routing protocols, destination-
sequenced distance vector (DSDV) [1], wireless routing
protocol (WRP) [2], global state routing (GSR) [3] and
cluster-based gateway switch routing (CGSR) [4] are well
known. In all proactive routing protocols the nodes
proactively store rout information to every other node in
the network. In general, the proactive routing protocols
suffer from extremely huge storage overhead because they
store information both about active and non-active routes.
This inculcates the unnecessary complexity of discovering
routes to the destinations with which a node rarely
communicates. Reactive or on-demand routing protocols
are designed to reduce this overhead. In reactive routing
protocols, when a source node needs to communicate with
a destination, it floods route-request packets throughout
the network to discover a suitable route to the destination.
Dynamic source routing (DSR) [5], ad hoc on-demand
distance vector routing (AODV) [7], adaptive
communication aware routing (ACR) [8], flow-oriented
routing protocol (FORP) [9] and associativity-based
routing (ABR) [10] are well-known among the reactive
routing protocols. AODV builds routes using a route-
request, route reply query cycle. When a source node
desires to send packets to a destination for which it does
not already have a route, it broadcasts a route-request
(RREQ) packet across the network. Nodes receiving this
packet update their information for the source node and set
up pointers backward to the source node in their routing
tables. A node receiving the route-request (RREQ) packet
sends a route-reply (RREP) if it is either the destination or
has a recently established route to the destination with.
Dynamic source routing (DSR) is similar to AODV in that
it forms a route on-demand when a source node requests
one. It uses source routing instead of relying on the routing
table at each device. Determining source routes require
accumulating the address of each router in the route-
request message. In FAIR [11], the source node transmits
RREQ packets that arrive at the destination through
multiple paths. Depending upon the locations, residual
energy, velocity etc. various characteristics of the routers,
the destination node evaluates performance of the paths by
considering their stability and agility. Then
communication from source to destination begins through
one of the best paths. FORP and ABR are link stability
based routing protocols that also rely on the flooding of
RREQ packets for route discovery. So, if the number of
RREQ packets can be reduced then much lesser number of
routers will be involved in the route discovery process. As
a result, network throughput or data packet delivery ratio
enhances with decrease in energy consumption in nodes.
This part is taken care of by efficient reactive routing
protocols like FORP, ABR, FAIR etc. However these
protocols do not focus on the message cost incurred by the
broadcast of route-reply packets while discovering a route
from destination to source. If reuse of a hop is possible
during the transmission of both RREQ and RREP packets,
then that will increase weight of the hop in selection of a
path from source to destination.
In the present article, a cost effective scheme 2NI is
proposed that can be used with any reactive routing
protocol for performance enhancement. In 2NI we have
rigorously studied the eligibility of each hop in the path

2772
options from source to destination, in the context of
traversal of route-reply from destination to source. Weight
of a hop increases if that can be used for route-reply also.
Otherwise, a broadcast operation will be required to
discover a route from that router to source. During this
broadcast of route-reply, redundancy within two-hop
downlink neighborhood of a node is examined and
eliminated. Each node within the two hop neighborhood of
a router receives RREP packet from exactly one node
among the 1-hop and 2-hop downlink neighbors of the
same router. This significantly reduces message cost in the
network increasing packet delivery ratio and the number
of alive nodes preserving network connectivity.
Simulation results firmly illustrate these advancements in
favor of 2NI embedded reactive routing protocols
compared to the ordinary versions of those protocols.
II. PROBLEM DESCRIPTION AND OVERVIEW OF
2NI
During route discovery process, source node broadcasts
RREQ packets towards the destination node according to a
balloon broadcast structure (described in section III).
These packets reach the destination through multiple
paths. Among those paths, exactly one is elected for
communication depending on different criteria in different
protocols. Generally, all the links in a communication
route in ad hoc networks are not bidirectional. Therefore
in most of the cases, the path that is used for
communication from source to destination cannot be
completely reused for transmission of acknowledgements
or RREP packets from destination to source. In the routing
protocols existing in literature, the destination node floods
RREP packets towards the source. This incurs huge
message overhead. Automatically the energy consumption
in the network nodes becomes high. As a result, the
number of alive nodes reduce giving rise to network
partition. Hence, nodes residing in one part of the network
become unreachable for the others decreasing packet
delivery ratio up to a great extent.
2NI extracts and utilizes some structural benefits of ad
hoc networks arising from advantageous node positions. It
encourages the inclusion of bidirectional links in the
communication route from source to destination provided
it is energy-efficient (remaining energy is high in battery
of the two nodes involved in the link) and stable (nodes in
the link are presently close enough and have high
possibility of staying close in future due to small relative
velocity). Sometimes it is also seen that a node cannot
reach its immediate predecessor in 1-hop but can directly
reach some of the predecessors of its immediate
predecessor in the communication route from source to
destination.
This can be seen in figure 1 where nm and nm+2 are 1-hop
downlink neighbors of ni although the link from np to ni is
not bidirectional. Selection of this kind of links in the
route from source to destination enhances the reusability
of the route during transmission of RREP from destination
to source. Sometimes these cannot be achieved in 1-hop
but in 2 hops. For example, in figure 2 it can be seen that
nk acts as the router from ni to nm+1; Similarly, nl acts the
bridge from ni to np.
If no such advantage can be extracted, then RREP packet
has to be broadcasted. For example, consider figure 1. The
node nm+1 does not have any 1-hop or 2-hop downlink
neighbor which is a predecessor of nm+1 in the
communication route from source to destination. Then 2NI
will advise inclusion of the hop from nm to nm+1 in the
communication route from source to destination provided
nm+1 has high residual battery power and huge number of
downlink neighbors towards source ns to which if RREP is
forwarded then RREP packets will traverse a great
distance to the source. Cost of messages incurred here is
worth the advancement towards the source. Moreover this
waives out the requirement of more intermediate nodes to
reach ns, saving message cost and energy consumption in
network nodes. Weight of the hop from nm to nm+1 increase
if nm+1 has less number of downlink neighbors that move
RREP away from ns. These movements are mere
distractions for RREP packets and even if RREP reach ns
through these downlink neighbors it will have to traverse
an unnecessary long path.
III. ROUTE DISCOVERY PROCEDURE OF 2NI
Assume that, in a route S, the link from a node na to nb is
broken where the destination is nd. In order to repair the
broken link, na initiates a route discovery session to nd.
The last known location of nd to na is at time t1 and its
value is (xd(t1),yd(t1)). Maximum velocity of nd is vmd.
Since Rmax is the maximum possible radio-range in the
network and H is the maximum possible number of hops,
so the maximum possible distance a route-request packet
can traverse in its entire lifetime, is HRmax. The time
required to traverse this distance, is (HRmax)/vs where vs is
the signal speed. Let MQ be the maximum possible size of
message queue in a node and tm be the smallest possible
time required to forward a message. Then maximum
waiting time of a route request packet in a path is H.tm.
(MQ-1). So, the total lifetime TL of a route-request packet
is formulated as,
TL=H[Rmax/vs+ tm (MQ-1)] (1)
The movement of nd by this time, is bound by a circle
Cd(t1) with center (xd(t1),yd(t1)) and radius R=(TL . vmd).
2NI suggests that, instead of broadcasting route-request
packets in every direction in the network, it is broadcast in
a balloon kind of structure. The route-request packets are
bound by the straight lines which are tangents to the circle
Cd(t1) drawn from location of na at time t i.e. (xa(t),ya(t)),
and the rest portion of the circle itself not covered by those
tangents. For example, consider figure 3.
As seen from figure 3, route-requests are broadcast from
point A; they remain within the two straight lines AC and
AB where tangents from point A touch Cd(t) at points B
and C. As B or C is reached, route-request packets traverse

2773
within the sector CDB towards O which is center of the
circle. Coordinate of O is (xd(t1),yd(t1)). So, the broadcast
balloon is ACDBA.
Area of the broadcast balloon = area of the ABC +
area of Cd(t1) bound by the line CB and arc CDB (2)
On the other hand, area of Cd(t1) bound by the line CB and
arc CDB = area of Cd(t1) – area of region CEB
area of region CEB = area of OCEB–area of OCB (3)
Let the coordinate of C be (h,z). Then,
(xa(t)-h)2
+(ya(t)-z)2
= C2
(4)
And
(xd(t1)-h)2
+(yd(t1)-z)2
=R2
(5)
Where C2
=(xd(t1)-xa(t))2
+ (yd(t1)-ya(t))2
(6)
From (4),
h2
+z2
-2hxa(t)-2zya(t)+z1(t)=0 (7)
Similarly, from (5),
h2
+z2
-2hxd(t1)-2zyd(t1)+z2(t)=0 (8)
here z1(t)=x2
a(t)+ y2
a(t) – C2
(9)
and
z2(t)=x2
d(t1)+y2
d(t1) – R2
(10)
Subtracting (8) from (7),
h=(z2(t)–z1(t)-2z(yd(t1)-ya(t)))/{2(xd(t1)-xa(t))} (11)
Simplifying from (11),
h=p– rz (12)
where p=(z2(t)– z1(t)) /{2(xd(t1)-xa(t))} (13)
r=(yd(t1)-ya(t))/{(xd(t1)-xa(t))} (14)
Putting (12) in (4),
z2
(r2
+1)+2zM+ V = 0 (15)
M = r(xa(t) – p) – ya(t) and
V = -2rzp+ (xa(t) – p)2
+ y2
a(t) – C2
From (15),
z = (-2M±(4M2
-4(r2
+1)V))/{2(r2
+1)}
Let z_st = (-M+(M2
-(r2
+1)V))/{(r2
+1)} and
z_en = (-M-(M2
-(r2
+1)V))/{(r2
+1)}
Corresponding to these possible values z_st and z_en of z,
the values of h are h_st and h_en, s.t.
h_st = p – r z_st and h_en = p – r z_en
Length of CO is denoted as l_co and defined as,
l_co = {(h_st-xd(t1))2
+(z_st-yd(t1))2
}
Similarly, length of BO is denoted as l_bo and defined as,
l_bo = {(h_en-xd(t1))2
+(z_en-yd(t1))2
}
Length of CB is denoted as l_cb and defined as,
l_cb = {(h_st-h_en)2
+(z_st-z_en)2
}
Let  be the angle COB. Then,
 = cos -1
{(l_co2
+l_bo2
-l_cb2
) / (2 l_co l_bo)}
So, area of the region OCEB=(/360) l_co2
(16)
Area AR of OCB ={Q(Q-l_co)(Q-l_bo)(Q-l_cb)} (17)
Where Q = (l_co+l_bo+l_cb)/2
So, area AR_1 of region CBE = (/360)  l_co2
- AR
From figure 3, it is evident that area AR_2 of CDB is
given by,
AR_2=l_co2
- AR_1 (18)
Area of the ABC is denoted by AR_3 and defined by,
AR_3={Q1(Q1-l_ca)(Q1-l_cb)(Q1-l_ab)} (19)
where l_ca, l_bc and l_ab denote the lengths of the arms
CA, BC and AB, respectively, of the triangle ABC in
figure 3.
So, l_ca = {(h_st-xa(t))2
+(z_st-ya(t))2
}
l_ab = {(h_en-xa(t))2
+(z_en-ya(t))2
}
Hence, area AR_4 of the broadcast balloon ACDBA in
figure 3 is given by,
AR_4 = AR_2 + AR_3
IV. HOP SELECTION IN 2NI
Let the hop from np to ni be a part of the route S from
source ns to destination nd. If np is a direct downlink
neighbor of ni, then it is easiest for ni to send the RREP
from ni to np. If np is a downlink neighbor of some
downlink neighbor of ni (i.e. 2-hop downlink neighbor)
then also it is cost-effective to include the hop from np to
ni during communication from source ns to destination nd.

2774
Below we illustrate different situations that can take place
during computation of weight of each hop.
Case – 1
If np or any predecessor of np is a direct downlink neighbor
of ni (illustrated in fig. 1), then weight wpi(t) of the hop
from np to ni at time t is given by,
wpi(t)=MAX[frpij(S,t) exp sdpi(S,t)] (20)
 njMi(S,t)
where frpij(S,t)=Wpi(S,t){(ei(t)/Ei)βpi(t) βij(t)} (21)
and sdpi(S,t) = (1/(4×pi(S,t)))
Where Mi(S,t) is the set of downlink neighbors of ni that
have been included in S. Wip(S,t) is weight of the hop from
np to ni in route S at time t according to the underlying
routing protocol. If the routing protocol does not assign
any weight to the hop, Wpi(S,t) acquires the value 1. ei(t)
and Ei denote the residual charge of ni at time t and the
maximum battery power of the same node, respectively.
pi(S,t) is the number of direct downlink neighbors of ni in
the route S at time t. βpi(t) specifies the affinity between
the two nodes np and ni in terms of relative velocity,
distance radio-range, their history of communication till
time t etc. Similarly the significance of βij(t) can also be
understood.
wpi(t) will acquire a high value (between 0 and 1) if ni is
not exhausted much already i.e. (ei(t) / Ei) is high (np is
already included in the communication route, so its energy
need not be considered now; similar case with nj, it is a
predecessor of np in the route S), np and ni share a strong
wireless bond (βpi(t) is high), similarly ni and nj are
strongly connected (βij(t) is high) and np has a significant
number of downlink neighbors who have already been
included in the route S (i.e. pi(S,t) is high; this is a
measure of reusability of the hop during transmission of
RREP packets).
βij(t) denotes affinity of the node nj w.r.t. ni at time t and
NEi(t) is the set of 1-hop downlink neighbors of ni at time
t. In (22), nj has been continuously residing within the 1-
hop downlink neighborhood of ni from (t - ϖij(t)) to current
time t in the present session.
(f1ij(t) exp f2ij(t)) if this is the first time the
link
from ni to nj has been
established or | 1ij(t)| = 0
βij(t) = (22)
{ φij(t) pl_rtij(t) (f1ij(t) exp f2ij(t))} 1/3
Where f1ij(t)={1– (|νi(t) - νj(t)| + 1) -1
} (23)
f2ij(t)={dij(t) / (Ri + 1)} (24)
φij(t)=[|ψ1ij(t)|f3ij(t)]1/2
(25)
f3ij(t)=((Tij(σ))/|ψ1ij(t)|-ϖij(t))/(ψ2ij(t)×Γmaxij(t)) (26)
σ  ψ1ij(t)
pl_rtij(t)=1– tot_pkt_lostij(t) / (tot_pkt_sentij(t) + 1) (27)
Relative velocity of ni w.r.t. nj at time t is given by (νi(t)
- νj(t)). Its effect on βij(t) is modeled as f1ij(t). Please note
that f1ij(t) always takes a fractional value between 0 and 1,
even when νi(t) = νj(t). As the magnitude of relative
velocity of ni w.r.t. nj at time t increases, it leads to the
reduction in the value of f1ij(t), which in turn, contributes
to increase the link stability. f2ij(t) expresses the
dependence of βij(t) on the distance between the nodes ni
& nj at time t. As ni is the predecessor of nj at time t, nj
must be within the transmission range (or radio-range) of
ni at that time. Since Ri denotes the radio-range of ni,
upper limit of the distance d′ij(t) between ni and nj at time t
is Ri. As per the expression of f2ij(t), it also acquires a
fractional value between 0 and 1. As d′ij(t) decreases, f2ij(t)
decreases enhancing the link performance. As per the
history of communication between ni and nj till time t is
concerned, Γmaxij(t) indicates the maximum duration of
the link from ni to nj in earlier communication sessions
occurring between the two nodes till time t. In the
formulation of φij(t), ψ1ij(t) is the set of communication
sessions till time t in which nj resided within the radio-
range of ni for a time duration higher than ϖij(t). ψ2ij(t) is
the total number of communication sessions that took
place between ni and nj till time t. Tij(σ) is the duration of
the link between ni and nj in the communication session σ.
It is evident that Tij(σ) ≤ Γmaxij(t) and |ψ1ij(t)| is less than
or equal to ψ2ij(t). So, φij(t) ranges between 0 and 1. If
φij(t) is high then it signifies that according to the history
of communication between ni and nj till time t, (Tij(σ) -
ϖij(t)) is high in a large number of communication sessions
that took place between the two nodes till time t. Hence,
the remaining life of the present link between those two
nodes is also expected to be high.
tot_pkt_lostij(t) and tot_pkt_sentij(t) denote the total
number of packets lost in the link from ni to nj till time t
and the total number of packets actually sent in that link
till time t. So, (tot_pkt_lostij(t) / tot_pkt_sentij(t)) is the
packet loss rate in the link from ni to nj till time t. Lesser is
the packet loss rate greater will be reliability of the link. 1
is added to tot_pkt_sentij(t) in (27) to avoid 0 value in the
denominator when the link is completely new i.e.
tot_pkt_sentij(t) is equal to 0. Since all of φij(t), f1ij(t),
f2ij(t) and pl_rtij(t) take positive fractional values, βij(t) is
between 0 and 1. This kind of formulation inculcates
energy and relative velocity consciousness in route
selection even if the underlying protocol does not consider
these important factors.
Case – 2
If np or any predecessor of np in route S is a direct
downlink neighbor of any downlink neighbor of ni
(consider fig. 2), then weight wpi(t) of the hop from np to ni
at time t is given by,
wpi(t)=MAX[fr1pijk(S,t)exp sdpi(S,t)] (28)
 njMi(S,t),
nkNEi(t) UPj(t)

2775
fr1pijk(S,t) = Wpi(S,t){(ei(t)/Ei) (ek(t)/Ek) βpi(t) βik(t) βkj(t)}
sdpi(S,t) = (1/(6×1pi(S,t)))
Mi(S,t) is the set of 2-hop downlink neighbors of ni that
have been included in the route S. UPj(t) is the set of
uplink neighbors of nj at time t. 1pi(S,t) is the number of
two hop downlink neighbors of ni in S. Significance of all
other symbols are explained earlier.
wpi(t) acquires a high value if both ni and nk have
sufficiently high residual battery power (battery power of
nj has no role to play here since it is already included in
the communication route S), stable bonds exist between
the pairs np and ni, ni and nk, and nk and nj. The effect
increases if ni has a high number of 2-hop downlink
neighbors included in S.
Case – 3
If both case 1 and case 2 are applicable (illustrated in
figure 4), the one that produces more weight is taken into
account.
Case – 4
If none of the above cases is applicable, then ni broadcasts
the RREP packet mentioning ns as destination. If the
broadcasted RREP packet arrives at a node na which has
ns as one of its direct (1-hop) or 2-hop downlink
neighbors, then na stops further broadcasting and sends the
RREP to ns if ns is its 1-hop downlink neighbor or to an
uplink neighbor of ns which is a downlink neighbor of na.
If any node receives same RREP from more than one
node, it forwards RREP on the first occasion and discards
it in subsequent cases.
Let ST(i,t) denote the set of all 1-hop and 2-hop
downlink neighbors of ni at time t including ni. Also
assume that q-th member of the set (ST(i,t) – ni) is denoted
as nq and its distance from ns at time t is dqs(t). If dqs(t) <
dis(t), then movement from ni to nq (1 or 2 hop whatever)
is advantageous because the movement is an advancement
towards ns which is the ultimate goal of RREP packets.
The movement will be more cost effective if the
movement from ni to nq is stable enough in terms of
relative velocity, distance and history of communication
between the nodes.
wpi(t)=(Wpi(S,t)× a1pi(t) × (ei(t)/Ei) ) / (a2pi(t) × bpi(t)) (29)
[  {(1-dqs(t) / dis(t)) βiq(t)}1/2
]1/|AS(i,t)|]
nq AS(i,t) if | AS(i,t)| > 0 and nq is a
1- hop downlink neighbor of ni
[{(1-dqs(t)/dis(t)) (βim(t) βmq(t)) 1/2
}1/2
]1/|AS(i,t)|]
nq AS(i,t) if | AS(i,t)| > 0 and nq is a
a1pi(t) = 2- hop downlink neighbor of ni
where nm is the intermediate
node that produces maximum
value of (βim(t) βmq(t)) 1/2
among all the bridge nodes
from ni to nq
0.01 Otherwise
a2pi(t) = 1 –[{(1- dis(t) / (1+dqs(t)))]1/|(ST(i,t) – n
i - AS(i,t))|
nq (ST(i,t) – ni - AS(i,t))
bpi(t) = ( |q,i(t)| / (|ST(i,t)|-1)) 1/|(ST(i,t))|
nq ST(i,t)
Let AS(i,t) denote the set of such nodes to which
movement from ni is advantageous. Higher is the value of
(dis(t) - dqs(t)) and |AS(i,t)|, higher will be weight of the
hop from np to ni. So, |(ST(i,t) – ni) - AS(i,t)| is the set of
nodes to which movement from ni is not advantageous. So
if nq ( ST(i,t) – ni - AS(i,t)) then dqs(t) ≥ dis(t). Lesser is
the value of (dqs(t) - dis(t)) and |(ST(i,t) – ni - AS(i,t))|
better for weight of the hop from np to ni. Keeping in mind
all these dependencies, wpi(t) for this case is formulated
below: bpi(t) deals with the redundancy in two hop
downlink neighborhood of ni. q,i(t) denotes the set of
members of ST(i,t) of which nq is a direct downlink
neighbor at time t. Therefore, q,i(t)  (ST(i,t) – nq).
Hence, |q,i(t)|  (|ST(i,t)|-1). If redundancy in two hop
downlink neighborhood of ni is high, then it will decrease
weight of the hop from np to ni.
V. SIMULATION RESULTS
Experimental Setup
Simulation of the mobile network has been carried out
using ns-2 [15] simulator on 800 MHz Pentium IV
processor, 40 GB hard disk capacity and Red Hat Linux
version 6.2 operating system. Graphs appear in figures 5 to
12 showing emphatic improvements in favor of cost
effective route discovery. Number of nodes has been taken
as 20, 50, 100, 300 and 500 in different independent
simulation studies. Speed of a node is chosen as 5m/s, 10
m/s, 25 m/s, 35 m/s and 50 m/s in different simulation
runs. Transmission range varied between 10m and 50m.
Used network area is 500m 500m. Used traffic type is
constant bit rate. Mobility models used in various runs are
random waypoint, random walk and Gaussian.
Performance of the protocols AODV, ABR and FAIR are
compared with their 2NI embedded versions 2NI-AODV,
2NI-ABR and 2NI-FAIR respectively. In order to maintain
uniformity of the implementation platform, we have used
ns-2 simulator for all the above-mentioned communication
protocols. The simulation matrices are data packet
delivery ratio (total no. of data packets delivered100/total
no. of data packets transmitted), message overhead (total
number of message packets transmitted including data and
control packets) and per node delay in seconds in tracking
destination (total delay in tracking the destination in
different communication sessions / total number of nodes).
Simulation time was 1000 sec. for each run.

2776
Experimental Results
Figure 5 shows that the initially the data packet delivery
ratio improves for all the protocols with increase in
number of nodes and then it starts reducing at least for
those protocols that are not immune much to link
breakages generated by great increase in number of nodes
or increased node speed. For eg, ABR and FAIR are much
more immune to this, compared to AODV, as per the
underlying functioning logic of the protocol. The reason is
that the network connectivity improves with increase in
number of nodes, until the network gets saturated or
overloaded with nodes. When the overloading occurs, cost
of messages become very huge and the packets hinder one
another from reaching their destinations by colliding.
Figure 7 shows that for all the protocols cost of messages
increase with increase in number of nodes. This is quite
self-explanatory. Both the route discovery procedure and
communication saves a lot of messages. From figure 11 it
may be seen that as the number of nodes increase, the
communication The reason is that more number of
communications is initiated with increased number of
nodes and due to better network connectivity more
destinations can be tracked now which are far apart. Also
the phenomenon of more packet collision increases the
delay in tracking destinations. Figures 6,8,10 and 12 are
concerned with the influence of node speed on these
metrics. As the node speed increases, many links break
increasing the network congestion and message collision.
Those links need to be repaired using link repair
mechanism. This requires a new route discovery session to
find a route to nd from nb where the broken link is from nb
to na. Route discovery once again means the injection of a
huge number of route-request packets into the network.
Colliding messages are unable to reach their respective
destinations; hence they need to be retransmitted. This
causes additional delay in the process and injects some
more messages. As a result, packet delivery ratio
decreases with increased cost and delay.
2NI reduces the injection of route-request packets to a
great extent since an intermediate node that has recently
communicated with the destination, broadcasts the route-
request only to those downlink neighbors from which it is
possible to drive the RREQ to the actual destination within
the lifetime of RREQ packet generated by the source of
the communication session. This increases the node
lifetime and reduces the packet collision. The
improvements are evident from figures 7, 8, 9 and 10. As
far as delay in tracking the destination is concerned, 2NI
embedded versions show significant improvement. The
reason is that RREQ packets in 2NI embedded versions
face much less hindrances due to lesser amount of packet
collisions compared to the ordinary versions of those
protocols. Therefore, those RREQ packets are driven to
their respective destinations much sooner in protocols with
2NI facility.
Please note that the improvement produced by 2NI-
AODV over ordinary AODV is more than those produced
by 2NI-ABR over ordinary ABR and 2NI-FAIR over
ordinary FAIR. The reason is that in AODV, among all
discovered routes from source to destination, the one with
minimum hop count is elected for communication, without
considering stability of the links (stability is expressed
mainly in terms of relative velocities between the two
nodes forming a link). On the other hand, in ABR, the
route with maximum number of stable links is elected as
optimal. FAIR is even more conscious on link stability as
well as agility. Hence, the phenomenon of link breakage is
more frequent in AODV than ABR as well as FAIR. In
order to repair the broken link, more RREQ messages are
injected into the neighborhood of the broken link in case
of ABR and FAIR whereas in AODV a new route
discovery session is initiated altogether which requires
generation of a huge number of RREQ packets once again.
Actually, link breakage in all protocols increases message
overhead decreasing the network throughput with different
intensity determined by the logic of the protocol itself.
Note that, the phenomenon like route discovery and link
repair are less devastating in ABR and FAIR than in
AODV. So, performance enhancement of 2NI-AODV
over AODV is more than that produced by 2NI-ABR over
ABR and 2NI-FAIR over FAIR. Conclusion
2NI proposes a two hop neighborhood information
based technique of route discovery as well as route
selection, among all available options. This technique can
be applied along with any routing protocol for enhancing
it’s performance i.e. increasing data packet delivery ratio,
reducing message cost, energy consumption and
communication delay. It gives weight particularly to the
bidirectional links so that a route that is used for
transmitting data packets to the destination, can be used to
transmit acknowledgement packets back to the source.
This saves the message cost that would otherwise have
been required for discovering a route from destination to
source.
REFERENCES
[1] C.E. Perkind & P. Bhagat,”Highly Dynamic
Destination Sequenced Distance Vector Routing
(DSDV) For Mobile Computers” Computer
Communications Review, vol. 24 no. 4 1994, pp.
234–244.
[2] S. Murphy, J.J. Garcia-Luna-Aceves, “An Efficient
Routing Protocol for Wireless Networks”, ACM
Mobile Networks and Applications Journal, pages
183-197, Nov.1996.
http://citeseer.nj.nec.com/10238.html
[3] T.W. Chen, M. Gerla, “Global State Routing: A new
routing scheme for wireless ad hoc networks” , IEEE
Conference On Information, Communication and
Control, 1998
[4] C.C. Chiang et. Al.,”Routing in clustered multi-hop
wireless networks with fading channel”, IEEE
Conference on Innovative Systems. 1997, Singapore
[5] Broch, J., Maltz, D., Johnson, D., Hu, Y., , And
Jetcheva, J. A performance comparison of multi-hop
wireless ad hoc network routing protocols. In

2777
Proceedings of the Fourth Annual ACM/IEEE
International Conference on Mobile Computing and
Networking (MobiCom ’98) (Dallas, Texas, USA,
Aug. 1998).
[6] A. Kodole, P.M. Agarkar, “A survey of routing
protocols in mobile ad hoc networks”, Multi-
disciplinary Journal of Research in Engg. & Tech,
vol.2 no. 1, pg – 336-341, 2015
[7] C.E. Perkins, E.M. Royer, “Ad Hoc On-demand
Distance Vector Routing” IEEE Wokshop on Mobile
Computing Systems and Applications., 1999
[8] T. Ueda, “ACR: An adaptive communication aware
routing through maximally zone-disjoint shortest
paths in ad hoc wireless networks with directional
antenna”, Journal of Wireless Communication and
Mobile Computing, vol. 16 no. 3,2007
[9] T.X. Brown, H.N. Babow & Q. Zhang, “Maximum
Flow Life Curve for Wireless Ad Hoc networks”
ACM Symposium on Mobile Ad Hoc Networking and
Computing, USA, 2001
[10]C.K. Toh & P. Bhagwat, “Associativity based routing
for mobile ad hoc networks”, Wireless Personal
Communications, 1997, vol. 4 no. 2., pp. 1-36
[11]A. Banerjee, P. Dutta, “Fuzzy-controlled adaptive and
intelligent route selection (FAIR) in ad hoc
networks”, European Journal of Scientific Research,
vol. 45 no. 3, 2010, pp. 367-382
[12]S. Corson,”Ad Hoc Networks: Routing, Performance
Issues and Evaluations”, Internet draft, IETF Mobile
Ad Hoc Networking Group. 1999
[13]R. Parasher, Y. Rathi, “A-AODV: A modern routing
algorithm for mobile ad hoc networks”, International
Research Journal of Engg. And Technology, vol. 2
issue 1, 2015
[14]S. Prakash, J.P. Saini, S.C. Gupta, “Methodologies
and Applications of Wireless Mobile Ad Hoc
Networks Routing Protocols”, International Journal of
Applied Information Systems, vol. 1, no. 2, 2012
[15]http://www.isi.edu/nsnam/ns/tutorial/
Figures:
S: ns  …nm nm+1 nm+2 … npni…nd
Figure 1: Some predecessors of ni in S are its 1-hop downlink neighbors
nl
S: ns  …nm nm+1 nm+2 … npni…nd
nk
Figure 2: Some predecessors of ni in S are its 2-hop downlink neighbors
Figure 3: Illustration of broadcast balloon
nl
S:ns  …nm nm+1 nm+2 … npni…nd
Figure 4: Some predecessors of ni in S are its 1-hop or 2-hop downlink
neighbors
Figure 5: packet delivery ratio vs number of nodes
Fig 6: packet delivery ratio vs node speed
0
20
40
60
80
100
120
Packet delivery ratio (y-axis)vs node
speed (x-axis)
AODV
2NI-AODV
ABR
2NI-ABR
FAIR
2NI-FAIR
5 10 25 30 50
O
BC
A
D
E

2778
Fig 7: Message cost vs number of nodes
Fig 8: Message cost vs node speed
Fig 9: Energy consumption vs number of nodes
Fig 10: Energy consumption vs node speed
Fig 11: Communication delay vs number of nodes
Fig 12: Communication delay vs node speed

Cost Effective Routing Protocols Based on Two Hop Neighborhood Information (2NI) in Mobile Ad Hoc Networks

More Related Content

What's hot (18)

Viewers also liked (14)

Similar to Cost Effective Routing Protocols Based on Two Hop Neighborhood Information (2NI) in Mobile Ad Hoc Networks (20)

More from Eswar Publications (20)

Recently uploaded (20)

Cost Effective Routing Protocols Based on Two Hop Neighborhood Information (2NI) in Mobile Ad Hoc Networks