You are on page 1of 37

25

Bird Flight-Inspired Routing Protocol for Mobile Ad Hoc Networks


SUDIP MISRA and GOPIDI RAJESH, Indian Institute of Technology
One of the major challenges in the research of mobile ad hoc networks is designing dynamic, scalable, and
low cost (in terms of utilization of resources) routing protocols usable in real-world applications. Routing
in ad hoc networks has been explored to a large extent over the past decade and different protocols have
been proposed. They are based on a two-dimensional view of the ad hoc network geographical region, and
are not always realistic. In this article, we propose a bird ight-inspired, highly scalable, dynamic, energy-
efcient, and position-based routing protocol called Bird Flight-Inspired Routing Protocol (BFIRP). The
proposed protocol is inspired by the navigation of birds over long distances following the great circle arc,
the shortest arc connecting two points on the surface of a sphere. This sheds light on how birds save their
energy while navigating over thousands of miles. The proposed algorithm can be readily applied in many
real-world applications, as it is designed with a realistic three-dimensional view of the networks geographic
region. In the proposed algorithm, each node obtains its location coordinates (X, Y, Z), and speed from the
GPS (Global Positioning System); whereas, the destinations location coordinates (X, Y, Z), and speed are
obtained from any other distributed localized service. Based on the location information, the source and
each intermediate node choose their immediate neighbor as the next hop that has the maximum priority.
The priority is calculated by taking into consideration the energy of the node, the distance between the node
and the destination and the degree of closeness of the node to the trajectory of the great circle arc between
the current node and the destination. The proposed algorithm is simulated in J-SIM and compared with
the algorithms of Ad Hoc On Demand Distance Vector (AODV), and Most Forward Within Distance R (MFR)
routing protocols. The results of the simulations show that the proposed BFIRP algorithm is highly scalable,
and has low end-to-end delay compared to AODV. The algorithm is also simulated in various scenarios, and
the results demonstrate that BFIRP is more efcient than AODV in energy and throughput by 20% and
15% respectively. It also shows satisfactory improvement over MFR in terms of throughput and routing
overhead.
Categories and Subject Descriptors: C.2.2 [Computer-Communication Networks]: Network Protocols
Routing protocols
General Terms: Design, Performance
Additional Key Words and Phrases: Mobile ad hoc networks, bird ight, throughput, great circle,
energy-efciency, J-SIM, bandwidth
ACM Reference Format:
Misra, S. and Rajesh, G. 2011. Bird ight-inspired routing protocol for mobile ad hoc networks. ACM Trans.
Auton. Adapt. Syst. 6, 4, Article 25 (October 2011), 37 pages.
DOI = 10.1145/2019591.2019594 http://doi.acm.org/10.1145/2019591.2019594
The work of S. Misra was partly supported by a Grant from the Department of Science and Technology,
Government of India, Grant No. SR/FTP/ETA-36/08, which the author gratefully acknowledges.
This work was done when G. Rajesh was a visiting summer student doing internship at IIT Kharagpur.
However, during the period of this work, the author was a full time student at Indian Institute of Information
Technology, Allahabad, India.
Authors address: S. Misra; email: smisra@sit.iitkgp.ernet.in.
Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted
without fee provided that copies are not made or distributed for prot or commercial advantage and that
copies show this notice on the rst page or initial screen of a display along with the full citation. Copyrights
for components of this work owned by others than ACM must be honored. Abstracting with credit is per-
mitted. To copy otherwise, to republish, to post on servers, to redistribute to lists, or to use any component
of this work in other works requires prior specic permission and/or a fee. Permission may be requested
from Publications Dept., ACM, Inc., 2 Penn Plaza, Suite 701, New York, NY 10121-0701, USA, fax +1 (212)
869-0481, or permissions@acm.org.
c 2011 ACM 1556-4665/2011/10-ART25 $10.00
DOI 10.1145/2019591.2019594 http://doi.acm.org/10.1145/2019591.2019594
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
25:2 S. Misra and G. Rajesh
1. INTRODUCTION
A wireless ad hoc network is an aggregation of mobile nodes that autonomously com-
municate with each other without the presence of a centralized entity deployed in the
network. This particular feature of these networks makes them more suitable for dis-
tributed applications such as providing fast, efcient, and dynamic communication
for public safety, search and rescue operations, and military networks, where there
are no xed infrastructures provided for the spontaneous deployment of the network.
Due to the decentralized nature of these networks, they are typically characterized
by dynamic changes in the network topology attributed to the mobility of nodes, lim-
ited battery power, limited radio transmission and receiving range, and high bit error
rate (BER). These typical characteristics therefore present a challenge in designing a
dynamic and energy-efcient routing protocol.
Over the past decade, a large number of protocols have been proposed, but none
of them has taken a realistic view of the ad hoc networks geographical region into
account. Also, most of them have low scalability and efciency with increase in the
size of the network. This has been subject to extensive research over the past decade
and it is the most challenging area in the eld of ad hoc networks.
Ad hoc network routing protocols can be classied into two categories: topology-
based and position-based. In topology-based routing protocols, routing of the packets
is done using information about the existing links in the network. These protocols
are further divided into three subcategories: proactive, reactive, and hybrid protocols
based on how and when the state of the link is updated. In a proactive routing pro-
tocol, every node in the network shares its routing table with its neighbors at regular
intervals, which helps the other nodes to discover new routes for a particular destina-
tion and develop a global view of the entire network. The main advantage of proactive
protocols is the short duration of time required to nd the shortest path to the desti-
nation, but it comes at the expense of valuable bandwidth required for maintaining a
global view of the entire network at each node. Proactive protocols have low scalability
with increase in size of the network, because of the routing information broadcast by
all the nodes throughout the entire network. Also, protocols that fall into this cate-
gory are not dynamic because of the large communication overhead required to nd
an efcient path to a destination when the mobility of nodes is high. The lifetime
of the network is also less due to broadcasting of routing tables at regular intervals,
and the energy drain for construction of paths that may not be useful. The proposed
protocols in this category are DSDV [Perkins and Bhagwat 1994], WRP [Murthy and
Garcia-Luna-Aceves 1996], and Fisheye [Pei et al. 2000].
On the other hand, reactive protocols reduce the bandwidth consumption by estab-
lishing a path to the destination only when it is required. However, these protocols
have the drawback of high end-to-end delay due to the fact that a node has to wait for
a long time before selecting an immediate neighbor to which the packet has to be for-
warded because of the dynamic network topology. Reactive protocols can be deployed in
large size networks as they do not broadcast routing information at regular intervals.
They increase the routing overhead considerably due to the ooding of the network
with route request packets (RREQ) which results in poor performance in terms of rout-
ing overhead and lifetime of the network. Consequently, reactive protocols have a low
degree of applicability in dynamic networks, where mobility of the nodes is high. Some
of the proposed protocols in this category are AODV [Perkins and Bhagwat 1994], DSR
[Johnson and Maltz 1996], and ABR [Toh 1997]. Hybrid protocols are a combination of
both proactive and reactive protocols aiming to reduce the overheads incurred in these
protocols. But, the protocols in this class are not applicable for networks of capacity
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
Bird Flight-Inspired Routing Protocol for Mobile Ad Hoc Networks 25:3
larger than 100 nodes, because of large overlapping of zones, as in the case of ZRP
[Haas et al. 2002].
The second type of routing protocol is the class of position-based routing protocols,
which make routing decisions based on the geographical location of the nodes. Based
on the location information, different protocols followdifferent criteria to forward pack-
ets. For example MFR [Takagi and Kleinrock 1984] uses the greedy technique, whereas
DREAM [Basagni et al. 1998] uses directional ooding, and GRID [Liao et al. 2001]
uses the dominating set concept. The most popular technique is greedy forwarding
[Cao and Xie 2003; Karp and Kung 2000; Takagi and Kleinrock 1984; Qabajeh et al.
2009], which does not explore and maintain paths between source and destination;
instead the source piggybacks the location of the destination and chooses a neighbor
as the next hop that is closest to the destination. Whenever a packet reaches a dead
end, these techniques may not be able to nd an optimum path; instead they use re-
covery routing techniques [Bose et al. 1999; Karp and Kung 2000; Kuhn et al. 2003]
to route data packets around dead ends, which increases the throughput. However,
these selected paths may end up as detouring even though there are better paths in
other directions. Position-based routing protocols are similar to reactive routing pro-
tocols because the path to the destination is explored only when a packet is to be
forwarded. Although they are similar to the reactive routing protocols, they are quite
opposite in the case of scalability. This is due to the fact that position-based routing
protocols will not ood the entire network with packets to explore paths to the des-
tination. Position-based routing protocols are dynamic and energy-efcient compared
to the topology based routing protocols, because these protocols do not share routing
tables at regular intervals over the entire network, resulting in an increase in the
lifetime of the network. One of the major limitations of the position-based routing pro-
tocols is the periodic broadcast of BEACONS, which are small packets that annunciate
their presence to their immediate neighbors. This process increases trafc in the chan-
nel, thereby leading to a decrease in the packet delivery ratio. Some of the proposed
protocols in the class of position-based routing are MFR [Takagi and Kleinrock 1984],
DREAM [Basagni et al. 1998], GPSR [Karp and Kung 2000], LAR [Ko and Vaidya
2000], and GLR [Na and Kim 2006]. All the protocols proposed in this class are only
applicable to applications that assume a two-dimensional network region. They are
not suitable for applications that require a three-dimensional network region. Re-
cently, researchers have proposed a three-dimensional location-aware routing protocol
called 3D-GPR [Day et al. 2008], which is the only protocol to the best of our knowl-
edge that considers nodes present at different altitudes. But in this algorithm, paths
to each destination are maintained in a routing table, which reduces protocol scalabil-
ity due to the broadcasting of next-hop-request (NHREQ) packets. Thus, the protocol
has a low degree of scalability.
In this article, we propose a bird ight-inspired routing protocol, which is highly
scalable, dynamic, energy-efcient, and is designed with a three-dimensional view of
the ad hoc networks geographical region.
1.1. Our Contribution
We have taken a biologically/ecologically inspired approach to solving the problem,
since most naturally occurring processes are well known to be extremely efcient. We
have drawn inspiration from exploring how birds (mostly the migratory birds), which
are typically small in size and possess very limited intelligence and energy, navigate
long distances by expending small amounts of energy. It was discovered that birds
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
25:4 S. Misra and G. Rajesh
either knowingly or unknowingly follow the great circle arc, the shortest path between
two points on the globe, as stated in the literature [Qiu 2005].
In this article, we have come up with three ideas to design a routing protocol applica-
ble in real-world applications with high scalability, energy efciency, and throughput,
and low latency. First is the realistic three-dimensional view of the ad hoc networks
geographical region. We consider nodes to be deployed over the surface of a sphere
that is more or less similar to that of the real world. We assume that each node ob-
tains its location coordinates (X, Y, Z) and speed information from the GPS (Global
Positioning System) and the destinations location coordinates (X, Y, Z) and speed in-
formation from any other distributed localized service [Li et al. 2000]. The trajectory
of the great circle arc, which is the shortest path between two points on the surface of
a sphere is used to route data packets. This process does not involve any routing over-
head except for periodic broadcast of the beacons by a node to announce its presence to
the nodes within its transmission range. A beacon is a packet of small size containing
the location, energy, and speed information of a node. The second idea is to vary the
fundamental period of beacon broadcast with the mobility of the nodes. Our proposed
protocol has high adaptability to rapid changes in the topology of the network. This is
achieved by changing the period of the beacon broadcast in accordance with the mo-
bility of the node. But this has the drawback of additional bandwidth utilization. The
nal idea is the local repair time-out, to solve the problem of voids (which is the region
in the neighborhood of the node in the direction of the destination node without any
neighbors) and to increase the throughput. Local repair time-out is the amount of time
that an intermediate node waits for other nodes to move into the vicinity of the great
circle arc before forwarding the packet. The time-out value is independently calculated
by each node depending upon mobility and density within the transmission range, as
described in the literature [Brown and Mohan 1997; Xu et al. 2005].
In the reported work, we performed simulations in J-SIM
1
to compare the perfor-
mance of the proposed routing protocol with two well known routing protocols, AODV
[Perkins and Royer 1999] and MFR [Takagi and Kleinrock 1984]. While AODV is a
reactive topology-based routing algorithm, MFR is a location-based routing algorithm.
Simulation results show that the proposed protocol, BFIRP, has high scalability, de-
livery ratio, and low hop-count compared to AODV and MFR. The proposed protocol
is also simulated in various scenarios. The results show that the average energy con-
sumed by BFIRP is relatively less compared to that by AODV, which proves that it is
energy-efcient.
The rest of the article is organized as follows. Section 2 introduces the existing liter-
ature in the eld of routing in ad hoc networks. In Section 3 we present an overview of
great circle navigation by birds. In Section 4, we introduce our algorithm. In Section 5,
we theoretically prove the correctness of our algorithm. In Section 6, we perform a
complexity analysis of our algorithm. Section 7 delineates the performance of the pro-
posed scheme relative to AODV and MFR routing protocols. Finally, in Section 8 we
present concluding remarks.
2. RELATED WORK
Over the past decade, the topic of routing in ad hoc networks has been explored in
detail and a considerable amount of literature has been published. Before we elaborate
our work, we introduce readers to some of the popular works on routing protocols in
ad hoc networks.
The proposed algorithms can be classied into two categories based on their
approaches. One of them is proactive, and the other is reactive, routing. Proactive
1
http://j-sim.uiuc.edu
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
Bird Flight-Inspired Routing Protocol for Mobile Ad Hoc Networks 25:5
routing protocols explore and maintain paths to all nodes in the network constantly
even if some of them are not useful. Whenever there is a change in topology, each
node noties its neighbors by sharing its routing table. This process of maintaining
paths results in a large amount of routing overhead, especially in large sized networks.
The excessive routing overhead required to maintain paths limits the scalability and
applicability of these protocols to small and medium size ad hoc networks. A few ex-
amples of proactive routing protocols include DSDV [Perkins and Bhagwat 1994] and
OLSR [Clausen and Jacquet 2003].
A reactive routing protocol nds a path to the destination node only when the node
has a packet to send. This reduces the communication overhead at the cost of incur-
ring additional delay. Reactive routing protocols use different methods such as blind
ooding, directional ooding to nd paths between the source and the destination. Ex-
amples of reactive routing protocols include AODV [Perkins and Bhagwat 1999], DSR
[Johnson and Maltz 1996], and ABR [Toh 1997]. For example, in AODV, the source
nds a path to the destination by broadcasting a route request packet (RREQ) to its
neighbors, then they forward the packet to their neighbors, and this process continues
until the RREQ packet reaches the destination or any intermediate node that has a
fresh route to the destination. During the process of forwarding the RREQ, each inter-
mediate node maintains a reverse path to the source in their routing tables by storing
the identier (ID) of the neighbor from which they have received the RREQ. If the
packet reaches the destination or an intermediate node with a fresh route to the des-
tination, the destination/intermediate node responds to it by sending an RREP packet
to the source. For each entry in the routing table a timer is set in order to purge the
routes whenever the timer expires. When an intermediate node in a route moves out of
the transmission range of the predecessor node, the predecessor node notices the link
breakage and broadcasts the RERR packets to notify its neighbors to reconstruct that
part of the route involving it. AODV also uses periodic broadcast of hello messages
to maintain local connectivity of the nodes. The broadcasting of RREQ packets to nd
the routing paths and maintenance of the routing table limits the scalability of the
protocol. ZRP [Haas et al. 2002], TORA [Park and Corson 2001] are some of the pro-
tocols that have been proposed to improve the scalability. These protocols combine the
features of proactive and reactive routing to reduce communication overhead. ZRP re-
duces the communication overhead by using proactive routing locally within the zone
and reactive routing for remote nodes that are outside the zone. But these protocols
have low scalability because of the overlapping of different zones, as in the case of ZRP.
These are topology-based protocols, which do not take the node location into consid-
eration. To reduce the communication overhead incurred by topology-based reactive
routing protocols, position-based routing protocols have been proposed. They use the
location information of the nodes to construct paths. Some of the protocols that belong
to this class of routing protocols include MFR [Takagi and Kleinrock 1984], DREAM
[Basagni et al. 1998], GPSR [Karp and Kung 2000], GFG [Bose et al. 1999], LAR [Ko
and Vaidya 2000], GPSAL [Camara and Loureiro 2000], GEDIR [Lin and Stojmenovic
1998], and TBF [Niculescu and Nath 2003]. These protocols can be classied into two
categories based on the location service requirements. The GPSR and DREAM proto-
cols are designed based on the assumption that the location service maps the location
of nodes to their IDs, and provides the location of the destination when queried using
same routing scheme that is used for data packets. LAR and GPSAL determine the
location of the destination and construct paths reactively. LAR reduces the commu-
nication cost incurred by restricting the ooding of packets to the region where the
destination is expected to be located based on the previous known location. GPSAL is
a proactive position-based routing where the articial agents traverse the entire net-
work and update the location of nodes in the routing table. In this protocol, the packet
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
25:6 S. Misra and G. Rajesh
Fig. 1. Example of MFR.
forwarding route is chosen by applying the shortest path algorithm. In GEDIR, pack-
ets are forwarded to the node closest to the destination. In MFR, a node projects the
distance to its immediate neighbor onto the straight line that connects itself and the
destination. The neighbor that has the largest projection is selected as the next hop.
For example, in Figure 1 if MFR is deployed then node S chooses node A as the next
hop to route the packet to node D because it has the largest projection. GPSR and GFG
use greedy forwarding techniques, in which packets are forwarded to the next hop node
which has the maximum progress towards the destination. In GFG, when the packet
ends up in a dead end position, the protocol switches to the FACE algorithm, which
uses perimeter-forwarding based on planar graph-traversal to route the packets. The
GPSR protocol improves the FACE algorithm proposed in GFG. TBF is one of the most
recently proposed protocols, in which the source piggybacks a trajectory to the pack-
ets and the intermediate nodes forward packets to their immediate neighbors that are
closer to the trajectory embedded in those packets. TBF introduces an extra burden
on the network with the increase of packet size. Although these protocols have a high
degree of scalability, most of them are not readily applicable to real-world applications,
because they are based on a two-dimensional view of the network geographic region.
A three-dimensional routing protocol named 3D-GPR [Day et al. 2008] has recently
been proposed, in which nodes are assumed to be located on a three-dimensional grid
of cells and routing decisions are taken based on the criterion that packets traverse
the least number of cells before reaching the destination. Routes to the destination
are maintained in a routing table, which reduces scalability due to the broadcasting of
NHREQ packets. Also, 3D-GPR has no recovery routing technique, which reduces the
packet delivery ratio.
3. OVERVIEW OF GREAT CIRCLE NAVIGATION BY BIRDS
In this section, we present an overviewof the great circle navigation by birds to readers
so that they can appreciate the notations and calculations involving the great circle.
In the past few years, extensive research has been done on how birds navigate thou-
sands of miles and reach the same destination year after year. Recently, scientists
have proposed a theory [Qiu 2005] proving that birds use a number of cues such as
the magnetic eld of the earth, the position of the stars during the night, and the po-
sition of the sun during the day to navigate long distances. Birds may use these cues
in combination for better accuracy.
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
Bird Flight-Inspired Routing Protocol for Mobile Ad Hoc Networks 25:7
Fig. 2. The actual path followed by birds compared to the great circle path over the globe.
Birds using different combinations of cues end up traveling in different trajectories.
Scientists have recorded the trajectories of birds [Qiu 2005] using radar technology and
compared them with the trajectories estimated using different combinations of cues.
The comparison revealed the important cues used by the bird ocks for navigation over
long distances. One of the important and most-used cues is the position of the sun.
After further study, scientists realized that predicting bird ight using the position of
the sun is complicated because of dependence of the direction on the birds sense of
time.
Alerstam [1993] pointed out that most of the birds navigate using the position of
the sun without compensating for the constant jet-lag and end up travelling along
trajectories that approximated the great circle route on the globe (Figure 2). This
is how birds save their energy while navigating thousands of miles. This particular
fact motivated us in designing the dynamic and efcient routing protocol discussed in
Section 1. Unless we do not have any obstacles along the trajectory, following the great
circle does not ensure that we reach the destination in the minimum amount of time.
The shortest arc connecting two points on the surface of a sphere is part of a great
circle, which represents the line of intersection of the circumference of the sphere and
an innite plane passing through the center of the sphere. The length of the shortest
arc is known as the great circle distance between two points. As shown in Figure 3,
apart from the great circles, there are small circles that represent the line of intersec-
tion of the circumference of the sphere and an innite plane not passing through the
center of the sphere. To summarize, great circle distance is the shortest path between
any two points on the surface of a sphere.
There is one and only one unique great circle connecting two points that are not
opposite each other. The points on the surface of a sphere located exactly opposite
each other are called antipodal points. There are an innite numbers of great circles
passing through these points, but they are of the same length.
3.1. Calculation of Great Circle Distance Between Two Points
The distance between two points on the surface of a sphere is not the same as the
magnitude of the vector joining these two points or the length of the straight line
joining these points as in the case of Euclidean geometry.
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
25:8 S. Misra and G. Rajesh
Fig. 3. Great circles and small circles on the surface of a sphere and the difference between the great circle
distance and Euclidean distance between two points A, B.
The shortest distance between two points with coordinates (x
1
, y
1
, z
1
), (x
2
, y
2
, z
2
) in
Euclidean geometry is given by:
Euclidean distance =
_
(x
1
x
2
)
2
+ (y
1
y
2
)
2
+ (z
1
z
2
)
2
. (1)
In Equation (1), x
1
, x
2
are the X coordinates, y
1
, y
2
are the Y coordinates and z
1
, z
2
are the Z coordinates of those two points.
In spherical geometry, straight lines are replaced by the geodesics, which are the
great circle arcs joining two points on the surface of a sphere. The great circle dis-
tance is the shortest distance between any two points located on a sphere, measured
along the surface. For example, if anyone tries to y along the straight line connect-
ing two points on the surface of a sphere, then he/she has to dig a tunnel through the
sphere. This makes it clear that the great circle lies vertically above the straight line
connecting these two points.
The orthodromic distance, or the great circle distance, between two points having
coordinates (x
1
, y
1
, z
1
), (x
2
, y
2
, z
2
) is calculated using the following procedure.
First, it is required to convert the coordinates from the Cartesian coordinate system
to the spherical coordinate system. We assume that both the systems have the same
origin.
Spherical coordinates (r, , ) of a point having Cartesian coordinates (x, y, z) are
given by:
r =
_
x
2
+ y
2
+ z
2
, (2)
= atan2(y, x), (3)
= acos
_
z
_
x
2
+ y
2
+ z
2
_
. (4)
In Equation (4), r is the radial distance of the point fromthe origin, is the elevation
angle and is the azimuth angle, atan2(y, x) is the angle in radians between the
positive x-axis of a plane and the point with Cartesian coordinates (x, y) on it.
Using the preceding formulas, we convert the Cartesian coordinates of the two
points (x
1
, y
1
, z
1
) and (x
2
, y
2
, z
2
) to spherical coordinates (r,
1
,
1
) and (r,
2
,
2
)
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
Bird Flight-Inspired Routing Protocol for Mobile Ad Hoc Networks 25:9
Fig. 4. Figure illustrating the topology of an ad hoc network on the surface of a sphere.
respectively. To simplify further calculations we make sure that all the spherical coor-
dinates are measured in radians.
There are two methods to calculate the great circle distance depending on the esti-
mate of the distance between two points to reduce the round-off errors.
(1) To measure longer distances we use the following equation.
= cos
1
(sin(
1
) sin(
2
) + cos(
1
) cos(
2
) cos(
1

2
)). (5)
(2) To measure shorter distances we use the following equation.
= 2 sin
1
_
_
(sin((
1

2
)/2))
2
+ cos(
1
) cos(
2
) (sin((
1

2
)/2))
2
_
. (6)
In Equations (5) and (6), is the angle subtended by the great circle arc at the
center of the circle with radius equal to radial distance r of the points.
The distance in units of radial distance is given by:
d = x r. (7)
4. THE BFIRP ALGORITHM
In this Section, we present our proposed position-based algorithm. We assume that
the nodes move on the surface of a sphere of radius R as shown in Figure 4. The ref-
erence planes for the x, y, z coordinates are three planes perpendicular to each other
passing through the origin, which is the center of the sphere. Each node obtains its
location coordinates (X, Y, Z), and speed information from the GPS. The packet source
obtains the destinations location coordinates (X, Y, Z) by querying a distributed local-
ized service using the same routing scheme that is used for data packets, as described
in the literature [Li et al. 2000]. We limit the scope of this article to only routing using
geographic location of the destination.
Our proposed routing algorithm is composed of three sections. (1) Neighbor-location
updates, (2) packet-forwarding procedure, and (3) recovery-routing procedure.
4.1. Neighbor Location Updates
Here we elaborate on the process of neighbor-location updates in our algorithm. Each
node maintains a neighbor table with entries consisting of ID, location coordinates,
speed (m/s), and energy (J) of the neighbor.
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
25:10 S. Misra and G. Rajesh
Fig. 5. Figure showing the neighbor table maintained by each node in the network.
Fig. 6. Figure showing the data structure carried by beacon packets.
Figure 5 gives an idea of the neighbor table used by the nodes in our algorithm. Each
node periodically broadcasts beacons to nodes within its radio transmission range. The
data structure for the beacons consists of the nodes current location coordinates, speed
and energy. A beacon data structure is composed of an eight-bytes long node id and
four-bytes oat for each of the coordinates, speed and energy. Figure 6 gives a view
of the beacon data structure used by the proposed algorithm. Broadcasting of the
beacons by each node only to its immediate neighbors accounts for the scalability of
the proposed algorithm. This is because the communication overhead incurred due to
beacons is very low. Broadcast of beacons by nodes in the network is an independent
and unsynchronized process, but there is a possibility of synchronization among the
nodes present within their transmission ranges. To avoid this problem, as observed by
Sally Floyd and Van Jacobson [Floyd and Jacobson 1994], in our algorithm, we assume
the fundamental beacon interval to be equal to T
p
. After each broadcast of beacons by
a node, the beacon interval is reset to a new value that is randomly chosen from the
uniform distribution [ 0.5 x T
p
, 1.5 x T
p
], where the mean of the distribution is equal
to the fundamental beacon interval (T
p
).
We introduce dynamic nature into the proposed routing protocol by varying the fun-
damental beacon broadcast interval according to the mobility of the nodes in the net-
work. The beacon interval is varied according to the following equations.
V
avg
=
V
avg
+ V
2
(8)
T
p
= T
p

_
1 +
(V
avg
V)
V
max
_
. (9)
In these equations, V
max
is the maximum speed, V
avg
is the average speed, and V
is the current speed of the node. T
p
is the fundamental beacon broadcast interval,
is the density factor, whose value depends on the density of nodes in the transmission
region of a node. It lies in the interval [0, 1] to reduce the congestion in the network
with increase in density. Each node changes its fundamental beacon broadcast interval
whenever its speed changes, without altering the value of the present set timer for the
broadcast of the beacon. Whenever a node receives a beacon fromits neighbor, it checks
for an existing entry in its neighbor table. If an entry already exists, then it updates
the corresponding neighbors position, speed, and energy. If an entry does not exist,
then it creates a new entry in the neighbor table with the entry consisting of the nodes
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
Bird Flight-Inspired Routing Protocol for Mobile Ad Hoc Networks 25:11
Fig. 7. The structure of the data packet sent by the nodes in the proposed algorithm.
ID, position, velocity vector, and energy. The timer for receiving the next beacon is set
to three times the fundamental beacon broadcast interval to account for the loss of
packets due to collisions and congestion. According to the experiments carried out in
the literature [Ebert et al. 2002], if the timeout value is greater than twice the beacon
interval, it results in an increase of neighbor table inaccuracy beyond 10%, which in
turn reduces the performance of the protocol. Also, if the timeout value is smaller,
it may lead to inefcient routing of packets along suboptimal paths. If a node does
not receive a beacon from a neighbor before the timer for its entry expires, then the
respective neighbors entry in the neighbor table is purged, assuming that the neighbor
is no longer within its transmission range.
4.2. Packet-Forwarding Procedure
We now describe in detail the procedure followed by the source and intermediate nodes
to forward data packets to their immediate neighbors. As stated earlier, each node
maintains a neighbor table with entries consisting of the neighbors position, speed,
and energy. Whenever a node needs to send a data packet to a particular node, it
sets the source and the destination elds of the data packet with its own and the
destination nodes IDs. It then queries the localized distributed location service for the
current location of destination node. After receiving the location information of the
destination node, the source node piggybacks it to the data packet, which is then used
by the intermediate nodes in forwarding the data packet. In the proposed algorithm,
we have also included an extra eld bid in the data structure of the data packet to
route the data packets away from the void, which is the region in the neighborhood
of the node in the direction of the destination node without any neighbors. Figure 7
provides a clear view of the data structure used by the algorithm. After all the elds
of the data packet are lled, the source node uses the priority-forwarding technique,
which is described later in the article, to forward the packet.
When an intermediate node receives a data packet, rst it veries whether the
packet is destined to itself or not. If it is, then it forwards the packet to the upper
layers for further processing. If the node is not the destination, then it forwards the
packet to its immediate neighbor using the location information of the destination node
contained in the data packet. The procedure for routing a data packet is shown in the
ow chart in Figure 8.
After comparing its own ID with the destination ID, each intermediate node does a
lookup into its neighbor table for the destination nodes entry. If the destination is its
immediate neighbor, then it forwards the packet without any further delay. If no match
is found for the destination in the neighbor table, then it executes the priority-based
forwarding procedure.
In the priority-based forwarding scheme, each intermediate node calculates the de-
gree of closeness of its neighbors to the great circle path between itself and the desti-
nation, using the following equation.
= cos
1
_
cos(d
nbrdst
) cos(d
nbrnode
) cos(d
nodedst
)
sin(d
nbrnode
) sin(d
nodedst
)
_
. (10)
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
25:12 S. Misra and G. Rajesh
Fig. 8. Illustration of the steps in the data packet forwarding algorithm.
Fig. 9. Figure illustrating the calculation of the degree of closeness of a node to the great circle path.
Figure 9 illustrates the calculation of the degree of closeness of a node to the great
circle path using Equation (7). In Equation (7), d
nodedst
denotes the great circle dis-
tance between the node, and destination, d
nbrnode
denotes the great circle distance
between the node and its immediate neighbor, and d
nbrdst
denotes the great circle
distance between the immediate neighbor and the destination.
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
Bird Flight-Inspired Routing Protocol for Mobile Ad Hoc Networks 25:13
Priority-based forwarding procedure. After calculating the degree of closeness as de-
scribed in the preceding, each intermediate node consults its neighbor table to cal-
culate, and assign priority to, each of its immediate neighbors and then forwards the
packet to the neighbor with highest priority. The priority of a neighbor is calculated
by using the following equation.
P
i
=
_
E
i
E
max
_
+
_
1
d
nd
d
id
_
+
_
1

_
. (11)
In Equation (11), P
i
is the priority of the neighbor node in the ith entry of the neigh-
bor table, E
t
is the energy of the neighbor, E
max
is the maximum energy of a node at
any instant, which is equal to its initial energy, d
nd
is the great circle distance between
the neighbor and destination, d
id
is the great circle distance between the current node
and the destination, is the degree of closeness of the neighbor to the great circle path
between the current node and destination ranging from [0, ]. , , , are the applica-
tion constants, which depend upon the application for which the ad hoc networks are
deployed. Application constants satisfy the following equation.
+ + = 1. (12)
These constants are varied either to increase or decrease the contribution of the
three terms in Equation (11). For example, when the density of nodes is high, the
value of can be set to a higher value to select the neighbor having the maximum
energy as the next hop. If the density is less, then can be set to a lower value to
select a neighbor closest to the trajectory of the great circle path between the current
and destination nodes.
In the proposed algorithm, to prevent loops in the path, each intermediate node cal-
culates and assigns priority for each neighbor that satises the following inequalities.
d
nd
< d
id
, d
ni
< RR, & 75

. (13)
In Equation (13), d
nd
represents the great circle distance between the neighbor and
the destination, d
id
is the great circle distance between the current and destination
nodes, d
ni
is the great circle distance between the current node and its immediate
neighbor, and R denotes the minimum required distance of the neighbor from bound-
ary of the transmission range of the current node. R is used to choose an immediate
neighbor, so that no broken link event occurs. In the worst case scenario, where an
intermediate node does not have a neighbor with an appropriate priority to forward
the packet, it uses the recovery routing algorithm, which is described in detail in the
next section. Figure 10 claries the forwarding procedure in the proposed algorithm.
4.3. Recovery Routing Procedure
As shown in Figure 8, when an intermediate node has no immediate neighbor with
appropriate priority to route data packets, it executes the recovery routing procedure,
which has been previously proposed [Xu et al. 2005] to escape from dead-end situa-
tions. The recovery routing procedure consists of two parts.
Local repair timeout. We assume uniform distribution of nodes in the network. The
density of nodes in the area of operation is denoted by . The rate of crossing of the
boundary of a void in the direction of the destination node according to Equation (23) is
r =
v
avg
L

. (14)
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
25:14 S. Misra and G. Rajesh
Fig. 10. Illustration of the routing of packets using the priority-based forwarding procedure.
In Equation (14), V
avg
denotes the average speed of the node and L is the length
of the boundary of the void in the direction of the destination node. The local repair
timeout value is determined from the following equation.
T =
1
r
=

v
avg
L
. (15)
The length of the boundary of the void is calculated using the following equation.
L = 2 R+ R S. (16)
In Equation (16), R is the maximum range of transmission of a node and the value
of Sis given by
S= cos
1
(cos(R) cos(R) + sin(R) sin(R) cos(2
max
)). (17)

max
is the maximum angle of deviation of the path between the current node and its
neighbor from the great circle path between itself and the destination. After the local
repair timeout, the node once again calculates and assigns priorities to its neighbors.
If any of one the neighbors has appropriate priority, then it forwards the packet to that
neighbor. If no neighbor has appropriate priority, then the node moves into the second
part of the recovery-routing procedure.
Sending FERROR. After the local repair timeout, the node once again carries out the
priority calculations and assigns respective priorities to its neighbors, as described in
the previous section. If no neighbor has appropriate priority, then the node looks into
its neighbor table for the neighbor that has forwarded the data packet to it. If an entry
is found, then the node sends a FERROR packet to it; otherwise it drops the packet.
The recipient of the FERROR packet once again carries out the priority forwarding
procedure excluding the neighbor that sent the FERROR packet to it, and forwards
the packet to the neighbor with the highest priority. This improves the throughput of
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
Bird Flight-Inspired Routing Protocol for Mobile Ad Hoc Networks 25:15
Fig. 11. The situation of a node encountering a void region.
the protocol at the cost of latency, which is acceptable in a vast majority of applications
of Mobile ad hoc networks.
Let us consider the example shown in Figure 11, in which Node S forwards a packet
to Node 2, which has no neighbor with appropriate priority to forward the packet. So,
it waits until the local repair timeout. Even after the timeout, Node 2 has no neighbors
with appropriate priority to forward the packet, so it sends a FERROR packet to its
predecessor Node S as shown in Figure 12, which eventually routes the packet through
Node 6. Thus, this procedure increases the throughput.
5. PROOF OF CORRECTNESS
In this Section, we theoretically prove the correctness of the proposed solution using
lemmas and theorems along with their proofs.
LEMMA 5.1. Each network has unique application constants , , .
PROOF. In wireless ad hoc network applications, ad hoc networks are deployed to
fulll different objectives, which in turn need ne tuning of different parameters. Let
us consider the case of an application requiring high throughput at the cost of a high
energy drain. On the other hand, there are applications in which the lifetime of the
network has to be maximized at the cost of throughput of the network.
So, every network deployed to fulll an objective needs to tune its parameters to
those values by which it fullls its objectives. In our proposed model, application con-
stants , , , vary the network performance in terms of different metrics such as
throughput, lifetime of the network, and latency. If values of , , are greater than
, the throughput is increased because packets are forwarded to the neighboring node
that has maximum energy and that is close to the great circle path between the send-
ing node and the destination. In the same way, by varying the three application con-
stants, different objectives can be fullled by the networks deployed. Hence, all three
application constants can be varied and chosen uniquely in a specic application.
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
25:16 S. Misra and G. Rajesh
Fig. 12. The case when an intermediate node sends a FERROR packet to its predecessor, which then routes
the packet through another node.
LEMMA 5.2. The assumption of nodes equipped with GPS and a distributed local-
ized location service providing the exact location of the destination node is contextual
in implementing this scheme in any ad hoc network.
PROOF. In the proposed scheme, we have assumed that each node is equipped with
GPS, which provides its location information and also a distributed localized location
service which provides location information of the destination node when queried using
the same routing scheme that is used for data packets. This reduces the communica-
tion overhead, latency, and congestion in the network, since source nodes do not ood
the networks with packets to nd the location of destination node. This assumption
is not out of context as there are various proposed location service schemes available
such as GPS and GLS [Li et al. 2000]. This assumption is not a drawback because
the proposed scheme works equally well without the positioning system providing the
location of destination node, but with an additional increase in latency, communication
overhead, and congestion in the network.
LEMMA 5.3. A nodes distance from the destination, the degree of closeness to the
great circle path, and its energy are key factors for increasing efcacy and lifetime of
the network.
PROOF. In the proposed scheme, as stated in earlier sections, each intermediate
node calculates the priority for each neighboring node based on the energy, the dis-
tance from the destination node, and the degree of closeness of the path between the
neighbor and the destination to the great circle path between the current node and
the destination. After sorting the priorities of neighboring nodes, it forwards pack-
ets to the neighboring node with maximum priority. The distance of a node from the
destination is taken into consideration in calculating the priority, because the smaller
the distance of the node from the destination node, the lower the number of inter-
mediate nodes along the path between the source and destination, which reduces the
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
Bird Flight-Inspired Routing Protocol for Mobile Ad Hoc Networks 25:17
congestion, energy consumption, and interference in the network. The degree of close-
ness of the neighbor to the great circle trajectory between source and destination nodes
is also taken into consideration, because the great circle path is the shortest route be-
tween two points on the surface of a sphere; so the packet has to be steered along the
great circle path between the source and destination nodes, which reduces the latency.
Last but not least, energy is an important factor that is required for reception or trans-
mission of a packet and processing before making routing decisions. If a node drains
its energy below a threshold, it will not be supportive for forwarding and receiving
packets. Thus, to increase the lifetime of the network, energy is considered as a key
factor in calculating the priority of nodes through which nodes with low energy are
assigned low priority and are not taken into consideration while forwarding packets.
Hence, the theorem is proved.
THEOREM 5.1. The proposed protocol is highly scalable, provides a realistic view of
the ad hoc networks geographical region, and it is highly adaptive to rapid changes in
the network.
PROOF. The proposed protocol is highly scalable due to the fact that it makes use of
the geographic location of the nodes provided by the positioning system while making
routing decisions, in contrast to the topology-based routing protocols, in which each
node periodically broadcasts its routing table to its neighbors to have updated infor-
mation of the changes in the topology of the network. Broadcasting of its routing table
by each node, increases the congestion in the network due to the large size of the rout-
ing table. Also, energy is squandered for the transmission and reception of large-sized
routing tables. This holds equally well even in the absence of the positioning system,
by virtue of Lemma 5.2. This scheme provides a realistic view of the networks geo-
graphical region because the nodes are assumed to be deployed over the surface of a
sphere, which generally has a similar shape as that of the globe over which the ad hoc
networks are deployed for various applications. Also, the proposed protocol is highly
adaptive to rapid changes in the network because the proposed scheme changes the
fundamental beacon broadcast interval with changes in the mobility of the node, so
that its neighboring nodes are regularly made aware of changes in its position. The
proposed scheme ensures that the fundamental beacon interval is not too low, which
results in high trafc in the network when the speed of the node is high, by calculating
the beacon interval using Equation (9), which involves the difference between the cur-
rent speed and the average speed over the time, with respect to the maximum speed of
the node, to make sure that fundamental beacon interval does not change rapidly. The
density factor is used to reduce congestion in the network when the density of nodes
is high. Hence, theorem is proved.
THEOREM 5.2. The priority of neighboring nodes is calculated as the weighted sum
of
E
t
E
max
,
_
1
d
nd
d
id
_
,
_
1

_
PROOF.
_
1
d
nd
d
id
_
gives the measure of how close the neighboring node is to the des-
tination with respect to the current nodes distance from the destination node, which
is a key factor in decreasing the trafc (by Lemma 5.3) by assigning lower priority
to the neighboring node that is less close to the destination than the node that is
more close. Hence, it is one of the key factors to be taken into consideration in the
calculation of priority.
E
t
E
max
indicates the physical capability of a node to efciently
forward the packet (by Lemma 5.3). So, it is also one of the key factors to be taken
into consideration in the calculation of priority.
_
1

_
reects the closeness of the
neighboring node to the great circle route between the source and destination nodes
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
25:18 S. Misra and G. Rajesh
(by Lemma 5.3). Therefore, it is an important factor to decrease the latency. The
primary objective of our scheme is to increase the throughput and lifetime, and to de-
crease latency. By Lemma 5.1, every network has its application constants, which are
ne tuned according to the objective to be fullled. Thus, this feature of our scheme
ensures that it is applicable to a large number of applications. Hence the theorem
holds.
THEOREM 5.3. The throughput of the network is increased by using the priority-
based forwarding procedure and the recovery routing procedure.
PROOF. One of the priorities of the proposed scheme is to improve the efciency. To
prove this theorem, we consider a network of size N in two different scenarios. In case
(1), the nodes use the proposed protocol as the routing protocol, and in case (2) the
nodes use AODV [Perkins and Royer 1999] as the routing protocol. We assume ideal
conditions for transmission and reception of packets in the network with no packet
loss and that the size of the header of all the packets is equal to four bytes.
We consider a path consisting of n-1 hops between the source and destination and a
packet of size 1 byte to be transferred from source to destination.
Case 1. Let a
i
be the number of immediate neighbors of the i
th
hop in the route.
Since the calculation of priority includes distance, energy, and the closeness of the
neighbor to the great circle path between the source and destination as given by Equa-
tions (7), (10), and (11), the probability that a node successfully forwards a packet is
proportional to the priority of the node. Therefore, we assume P
i
as the probability of
a node successfully forwarding a packet, which is given by.
P
i
= priority
i
, (18)
where priority
i
is the priority of the i
th
hop in the routing path. The probability that a
packet is successfully transferred from source to destination is given by.
P
suc
=
a
s

i=1

P
i
T
1
i

i1

j=1
(1 P
j
)

, (19)
where T
1
i
is calculated using the following equation.
T
k
no
=

a
no

i=1
(P
i
T
k+1
i

i1

j=1
(1 P
j
)) if k <= n1&k >= 1
1 otherwise
(20)
The probability that a packet is dropped is given by
P
drop
= 1 P
suc
. (21)
Case 2. Let P be the probability that transmission of a packet to a node results in a
broken link event, T
p
be the delay in completely transmitting a packet of size one byte
from a node to its neighbor, T
proc
be the processing delay, P
0
be the probability that the
source has a recently used route to the destination, T be the maximum amount of time
the packet is stored in the buffer by the source before it is dropped.
In this scenario, a packet is dropped in the following possible cases, (1) destination
has no neighbors (2) no route has been found within the timeout of the buffer in which
the packets are stored, and (3) movement of a node involved in the path resulting in a
broken link event and reinitiation of the route discovery process, which introduces the
possibility of case (2).
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
Bird Flight-Inspired Routing Protocol for Mobile Ad Hoc Networks 25:19
The timeout period for receiving RREP is given by
T
out
= (T
proc
+ (8 T
p
)) (n1). (22)
The probability that the packet is dropped due to the timeout of the buffer during
the process of nding a route to the destination is given by
P
1
= (1 P
0
) min

max retrans

i=1
_
T
out
i
_
1
_
n

k=1
(1 P
k)
__
k
_
T
, 1

, (23)
where max retrans is the number of times the source tries to nd a route to the desti-
nation. The probability that a packet is dropped when the source has a recently used
path to the destination is given by
P
2
= P
0
min

max retrans

i=1
_
_
8 T
p
+ T
proc
_
(n1) i
_
1
_
n

k=1
(1 P
k
)
__
k
_
+ g
i
T
, 1

,
(24)
where g
i
is given by the following equation.
g
i
=
_
10 (T
p
+ T
proc
)
n1

i=1
__
i1

k=1
(1 P
k)
__
P
i
_
. (25)
The probability that a packet is dropped is given by (cf. Equations (23) and (24)):
P
drop
=
P
1
+ P
2
2
. (26)
FromEquations (21) and (26), we conclude that the P
drop
in case (2) is high compared
to that in case (1), because the probability that a node successfully transmits a packet
to the next hop in the path is low due the fact that in case (2) the scheme used does
not take into consideration the energy of the node and the distance between the nodes
while forwarding the packet. Hence, the theorem is proved.
THEOREM 5.4. The proposed protocol improves the lifetime of the network.
PROOF. The proposed scheme identies the energy of the node as one of primary fac-
tors in calculating its priority. Hence, a node that has more energy has higher priority
compared to the nodes with lower energy. The lifetime of the network automatically
improves, as the nodes priority is calculated every time a packet is to be forwarded.
Since the node that has previously forwarded packets has lower priority due to the
energy drain, a new node is selected to forward packets. This procedure leads to im-
provement in the longevity of the network. This can be proved mathematically as
follows.
Let us consider a basic network architecture consisting of ve nodes as shown in
Figure 13. In (1), nodes are deployed with the proposed scheme as the routing proto-
col to forward data packets from source to destination, and in (2), the nodes are de-
ployed with a topology-based routing protocol to forward data packets from source to
destination.
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
25:20 S. Misra and G. Rajesh
Fig. 13. Illustration of the ad hoc network consisting of ve nodes.
In this proof, we assume that the lifetime of the network is the amount of time be-
fore which a node drains its battery below a certain threshold E
thr
. Source s transmits
packets for a time t at periodic intervals of one second in both cases, to the destination
node d with nodes 1, 2, 3 acting as the forwarding nodes. Each node in the network
has a maximum energy E
max
to start with, E
thr
being the threshold value of energy
below which a node does not forward any packet, E
Tr
denotes the energy required to
completely transmit a packet, E
R
denotes the energy required to completely receive a
packet, E
t
denotes the energy of the node at time t. P
i
denotes the probability that a
node with ID i is dead. P
r&t
denotes the probability that a node successfully receives
and transmits a packet to its neighbor. To ease complexity of the proof, we assume
that the mobility of the nodes is negligible and that conditions for the transmission
and reception of the packet are ideal.
Case 1. In this case, the routing protocol forwards the packets based on the priority
of the node. The probability of a node completely receiving and forwarding a packet is
directly proportional to the priority of the node, which in turn depends on the energy
of the node:
P
r&t
P
r
& P
r&t
E
t
. (27)
In Equation (27), P
r
denotes the priority of the node at time t. Here we assume that
of the three application constants , , , the value of is very high compared to the
other two. So, the priority of the node at time t is approximated as
P
r

=
E
t
E
max
. (28)
The probability of a node completely receiving and forwarding a packet is given by
P
r&t
= k P
r
= k
E
t
E
max
, (29)
where k is the proportionality constant, which varies between 0 and 1 (inclusive). The
probabilities of each individual node in the network being dead at time t are given by
P
i
=
(E
Tr
+ E
R) k
t

1
E
i
E
max
E
max
, i = 1, 2, 3. (30)
Case 2. In this scenario, the routing protocol used by the nodes is topology-based,
where packets are routed based on routes stored in the routing table. Since nodes are
almost stationary, the path through which packets are routed is the same throughout
the period of observation, t. So, P
r&t
in this case is either zero or one.
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
Bird Flight-Inspired Routing Protocol for Mobile Ad Hoc Networks 25:21
P
r&t
for node with ID 1 is unity and the rest of the intermediate nodes have P
r&t
equal to zero, because the route involving the node with ID 1 has a smaller number of
hops compared to the others.
The probabilities of each individual node in the network being dead at time t are
given by
P
i
=
(E
Tr
+ E
R) t P
r&t
E
max
, i = 1, 2, 3. (31)
From Equations (30) and (31), it is clear that the value of P
0
in the rst case is less
than the value in the second case because P
r&t
in the second case is unity, whereas in
the rst case it is less than one. This implies that network lifetime is greater in the
rst case compared to the second case. Hence, the theorem is proved.
THEOREM 5.5. The proposed protocol decreases the latency in the network.
PROOF. One of the priorities of the proposed scheme is to decrease the latency, or
end-to-end delay. To prove this theorem we consider a network with n
2
nodes in two
different scenarios. In Case (1), nodes use the proposed protocol, and in Case (2) nodes
use a topology-based routing protocol. We assume ideal conditions for transmission
and reception of packets in the network with no packet loss, and that the size of the
header of all packets is equal to four bytes.
Let T
p
be the delay in completely transmitting a packet of size of one byte from a
node to its neighbor, and T
proc
is the processing delay. We consider a path consisting of
n-1 hops between the source and the destination.
Case 1. Let T
rand
be the delay to obtain the location of the destination node. Let us
consider a packet of size four bytes to be transferred fromthe source to the destination.
The total end-to-end delay is given by
T = (n1)
_
5 T
p
+ T
proc
+ T
rand
_
. (32)
Case 2. Let P
i
denote the probability that a node with ID i has a route to the destina-
tion in its routing table. Let P
0
denote the probability that the source has a route to the
destination in its routing table. T
route
is the delay to nd the route to the destination,
which is given by
T
route
= (1P
0
)
_
8 T
p
+ T
proc
_

n1

i=1

i1

j=1
_
1 P
j
_

P
i
(1 P
i1)

, i >= 1. (33)
From Equation (33), total end-to-end delay in this scenario is given by
T =
_
(n1)
_
5 T
p
+ T
proc
__
+ T
route
. (34)
It is clear from Equations (32) and (34) that the end-to-end delay in Case (1) is less
as compared to Case (2). Hence, the theorem is proved.
6. COMPLEXITY ANALYSIS
In this Section, we compute the complexity of the proposed scheme in terms of message
overhead and time delay. As described in Section 4, the proposed scheme is composed
of three phases: (1) neighbor location updates, (2) priority-based forwarding procedure,
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
25:22 S. Misra and G. Rajesh
and (3) recovery routing procedure. We carry out the analysis under the following as-
sumptions: (1) the positioning system does not behave unexpectedly and works ideally,
and (2) ideal transmission and reception of packets.
As described in Section 4, the neighbor location update requires each node to broad-
cast a beacon of size 36 bytes, as shown in Figure 6. Let N be the total number of
nodes in the network, and n
0
be the number of times location update messages are
broadcast by each node in the network during the time of observation. The location
update message complexity is given by
Location Update
msg
= 36 Nn
0
. (35)
According to Equation (35), the location update message complexity increases with
the increase in size of the network. The time complexity of location update is given by
T
LU delay
= 36 T
transm
+ T
rand
, (36)
where T
transm
is the delay for broadcasting a packet of size one byte and T
rand
is the
delay incurred to get the present location of the node from the positioning system.
Let us now determine the time complexity for the priority-based forwarding proce-
dure. Let us consider a node with n number of neighbor entries in its neighbor table,
which has to forward a packet to one of them. The time delay to calculate priority of a
neighbor is given by
T
priority
= T
calc distance
+ T
degree o f closeness
. (37)
In Equation (37), T
calc distance
is the time delay incurred in calculating the great circle
distance using Equation (7), and T
degree of closeness
is the time delay to calculate the degree
of closeness of a neighbor to the great circle path using Equation (7). The total time
complexity of the forwarding procedure is given by
T
f wd proc
= n T
priority
+ T
rand
. (38)
The total delay to forward a packet of size a is given by
T
f wd delay
= T
f wd proc
+ a T
transm
. (39)
Finally, we evaluate the message overhead and time complexity of two parts of the
recovery routing procedure. Let T
timeout
be the timeout value, T
calctimeout
be the time
delay to calculate the timer value. The time complexity of the rst part of recovery
routing is given by
T
recovery1
= T
timeout
+ n T
priority
+ T
rand
+ a T
transm
. (40)
In Equation (40), a is the size of the packet in bytes, n is the number of nodes, and
T
priority
is the time delay to calculate priority. The time complexity of the second part
of the recovery routing procedure is given by
T
recovery2
= T
search
+ a T
transm
. (41)
In Equation (41), T
search
is the time delay to nd out whether a node has an entry in
the neighbor table or not. The second term indicates the time required to transmit a
FERROR packet. The message overhead of the second stage is equal to the size of the
data packet, which in turn depends on the payload.
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
Bird Flight-Inspired Routing Protocol for Mobile Ad Hoc Networks 25:23
Fig. 14. Example of MFR in a three dimensional geographical region.
7. PERFORMANCE EVALUATION
In this Section, we report the evaluation results of the performance of the proposed
scheme in various simulation scenarios. Simulations were performed in J-SIM, which
provides support for a number of routing protocols for wireless networks. We compared
the performance of the proposed protocol with well known protocols AODV and MFR.
We have modied MFR to make it viable in the three dimensional coordinate system,
by projecting the distance onto the geodesic joining the source and destination nodes.
The projection of the distance of the neighbor from the sender onto the great circle
arc joining the source and destination is carried out as shown in Figure 14 using the
spherical law of cosines.
In the rest of this Section, we describe the simulation conguration, trafc model,
and performance metrics used to evaluate the performance of selected protocols, and
nally we present the observed results.
7.1. Simulation Environment
We evaluated the performance of the proposed scheme in different simulation scenar-
ios by varying the following specic parameters. (1) The number of nodes, and (2) the
mobility of nodes in the network. In the two simulation scenarios, while varying any
one of these parameters, specications of other parameters were kept constant and
equal to the values specied in Table I.
The mobility model used for simulation is the Random Way Point Model (RWP), in
which each node selects a destination point randomly and moves towards it. The speed
with which each node moves towards the destination point is uniformly distributed be-
tween 0 m/s and the maximum speed limit. After the node arrives at the destination
point, it remains stationary for the specied time period and reiterates the described
procedure. In our simulations, we had set the pause time equal to zero in order to
test the dynamic nature of our proposed protocol. In order to emulate real-world ap-
plication of the protocol, we used power drain values of commercially available NICs,
namely Lucents 915 MHz WLAN card which consumes 1.6W while transmitting, 1.2W
while receiving and 1.15W when in idle state [Carvalho et al. 2004; Ebert et al. 2002].
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
25:24 S. Misra and G. Rajesh
Table I. Simulation Parameters
Simulation Parameters
Simulation time 600 seconds
Number of Nodes (default) 600
Field Dimensions Sphere of radius equal to 500m
MAC protocol IEEE 802.11
PHY-MODEL PHY802.11b
Propagation path loss Two Ray Ground Model
Placement of nodes Random
Mobility model Random wave point mobility model
Transmission Range 260m
Seed for random number generator 7777
Power Drain for Transmission 1.6 W
Power Drain for Reception 1.2 W
Power Drain for Idle state 1.15W
Fundamental Beacon Interval 0.5 seconds
Values of , , 1/3,1/3,1/3
In order to evaluate the protocol in a general scenario, application constants , , ,
are each given a value of 1/3. Also to assess the scalability of the protocol, we have
chosen number of nodes to be 600 as default and varied from 500 to 700.
7.2. Trafc Model
Data trafc was generated by three randomly selected constant bit rate (CBR) trafc
pairs with sending rates equal to 80 bytes per second. The payload size equals 512
Bytes.
7.3. Performance Metrics
We compared the performance of the proposed protocol against the benchmark routing
protocols by considering following metrics.
(1) Data Delivery Ratio. It is the ratio between the total number of packets received
by all the destination nodes to the total number of packets sent by all the sources.
It is an indicator of the quality of service provided by the network.
(2) Average Residual Energy of a Node. It is the ratio of sum of the residual energies
of all the nodes in the network at the end of simulation time to the total number of
nodes in the network. This is considered as one of the performance metrics because
energy determines the computational power and the lifetime of the node.
(3) Routing Overhead. It is the total number of packets other than data packets trans-
mitted by all the nodes in the network. This is taken into account because it re-
ects the scalability of a protocol. A protocol with a large routing overhead has
high probability of packet collisions, resulting in low throughput.
(4) End-to-End Delay. It is the amount of time elapsed between transmission of pack-
ets from source to destination, which includes the time taken for processing at
intermediate nodes, the delays due to the transmission of packets between inter-
mediate nodes, the buffering of packets at intermediate nodes, and the delays due
to retransmission of packets at the MAC layer.
Apart from these metrics, we also considered the maximum number of hops a data
packet has traversed during the entire simulation period as a performance metric.
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
Bird Flight-Inspired Routing Protocol for Mobile Ad Hoc Networks 25:25
Fig. 15. Variation in the delivery ratio with the variation in number of nodes in the network.
7.4. Performance Results
We evaluated the performance of the proposed scheme using two simulation scenarios:
by varying, (1) the number of nodes, and (2) the mobility of nodes in the network.
7.4.1. Variation in the Number of Nodes in the Network. In this simulation scenario, we
compared the performance of the proposed scheme with that of the benchmarks by
varying the number of nodes in the network from 500 to 700; the maximum speed
with which the node moves towards the destination point is set to 15 m/s.
The results of data packet delivery ratio shown in Figure 15 prove that the proposed
scheme outperforms AODV and MFR in terms of delivery ratio. This complements our
theoretical proofs in Section 5. AODV performs poorly because of the high mobility of
the nodesits throughput decreases with the increase in number of nodes due to long
delays encountered in transmission of RREQ, RREP, and RERR packets. The proposed
scheme has a higher delivery ratio compared to MFR because priority-based forward-
ing is used, which ensures that packets are routed along the great circle path between
the source and destination, while recovery routing is used to route packets around the
voids. Whereas MFR uses only the forwarding technique and has no recovery routing
technique to route the packets around the voids encountered along the route.
Figure 16 shows that the proposed scheme consumes relatively lower energy com-
pared to AODV, but relatively higher energy compared to MFR. The energy consumed
by all three protocols increases with an increase in the number of nodes, due to an
increase in the routing overhead. AODV performs poorly compared to other two due
to the rapid broken link events, which increases the ooding of RREQ, RREP, and
RERR packets in the network. BFIRP performs better than AODV because packets
are routed along the great circle path between the source and destination nodes. The
energy consumption of BFIRP is relatively higher compared to MFR because of the
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
25:26 S. Misra and G. Rajesh
Fig. 16. Variation in the average residual energy with the variation in number of nodes.
increase in the routing overhead due to the ooding of FERR packets and additional
processing at intermediate nodes in case of the presence of voids along the great circle
route. However, the difference in energy consumption between BFIRP and MFR is
low because MFR routes packets along detouring paths, whereas BFIRP routes data
packets along the great circle path, which is the shortest path.
Figure 17 shows that the proposed routing protocol has low routing overhead com-
pared to that of AODV but relatively higher overhead compared to that of MFR. The
routing overhead increases with the increase in size of the network, because of the
additional ooding of packets by the extra nodes added to the network.
AODV has high routing overhead compared to other two because of the increase
in ooding of RREQ, RREP, and RERR packets due to rapid broken link events in
the network. BFIRP has low routing overhead compared to AODV because the pro-
posed routing scheme uses location information of the destination node to route the
packet, whereas AODV uses a topology-based routing technique and has no local route
repair scheme to deal with broken link events. BFIRP has slightly higher routing over-
head compared to MFR, although both protocols periodically broadcast beacon packets,
because BFIRP changes its fundamental beacon interval with the increase in nodes
speed. The proposed scheme also transmits FERR packets in the case of a recovery
routing scheme to route packets away from voids, whereas in the case of MFR packets
are dropped whenever a void is encountered.
Figure 18 shows that BFIRP has very low average end-to-end delay compared to
that of AODV, but has relatively higher average end-to-end delay compared to MFR.
Average end-to-end delay increases with the increase in number of nodes in all three
protocols due to additional transmission and processing delays at intermediate nodes.
AODV performs poorly compared to the other two because of the large delays
encountered due to the buffering of packets, the transmission of RREQ packets,
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
Bird Flight-Inspired Routing Protocol for Mobile Ad Hoc Networks 25:27
Fig. 17. Variation in routing overhead with the variation in the number of nodes.
Fig. 18. Variation in the average end-to-end delay with variation in the number of nodes.
reception of the RREP packets, and the reconstruction of routes due to broken link
events. BFIRP has low average end-to-end delay compared to AODV because packets
are routed along the great circle path between the source and destination, which is the
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
25:28 S. Misra and G. Rajesh
Fig. 19. Variation in the maximumnumber of hops traversed by a data packet with variation in the number
of nodes.
shortest path. However, BFIRP has a relatively higher average end-to-end delay com-
pared to MFR because of the additional processing and transmission delays associated
with recovery routing scheme whenever voids are encountered.
Figure 19 shows that the maximum number of hops traversed by a data packet is
less in the case of BFIRP compared to that in AODV and MFR. The number of hops
decreases with an increase in the number of nodes, as shown in Figure 19, because
the number of neighbor nodes increases with the increased size of the network. The
maximumhop-count of BFIRP is less compared to that of AODVand MFR, because the
data packets are routed along the great circle path, which is the shortest path between
two points on a sphere.
7.4.2. Performance Results with the Variation in Speed. In this simulation scenario, we
compared the performance of BFIRP with that of AODV and MFR, by varying the
maximum speed of the nodes from 2 m/s to 15 m/s, and the number of nodes in the
network is set to 600.
Figure 20 shows that the proposed routing scheme has a higher delivery ratio com-
pared to that of AODV and MFR when the mobility of nodes is high, which ensures its
dynamic response to changes in the network topology. The data packet delivery ratio
decreases with an increase in the mobility of nodes because of an increase in broken
link events and the probability of collision resulting in dropping of packets. The data
packet delivery ratio of BFIRP is high compared to both AODV and MFR, because
routing in BFIRP uses priority forwarding and recovery routing techniques. AODV
performs poorly compared to other two because it has no local route recovery scheme
in case of the occurrence of broken link events at intermediate nodes, and source nodes
have to reinitiate the route discovery process to route packets, which increases the de-
lay in highly mobile networks, resulting in low throughput.
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
Bird Flight-Inspired Routing Protocol for Mobile Ad Hoc Networks 25:29
Fig. 20. Variation in the delivery ratio with the variation in the mobility of nodes.
Fig. 21. Variation in the average residual energy with the variation in the mobility of the nodes.
Figure 21 shows that the average energy consumed by BFIRP is relatively less com-
pared to that of AODV, but it is relatively higher compared to that of MFR. The average
residual energy decreases with the increase in the mobility of the nodes in the network
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
25:30 S. Misra and G. Rajesh
Fig. 22. Variation in routing overhead with the variation in the mobility of the nodes.
because of multiple retransmissions of data packets due to the occurrence of broken
link events. The average energy consumed by BFIRP to route data packets is less com-
pared to AODV because the routing overhead required in case of a broken link event is
less. This is because in AODV the source reinitiates the process of route discovery in
such events. The average energy consumed by BFIRP is relatively high compared to
that of MFR because MFR does not use any recovery routing procedure, which reduces
the routing overhead to some extent. The difference in energy consumption between
BFIRP and MFR is low because MFR routes packets in long detouring paths, whereas
BFIRP routes packets along the great circle path, which is the shortest path.
Figure 22 shows that the routing overhead in BFIRP is less compared to that in
AODV, but relatively high compared to that in MFR. The routing overhead increases
with an increase in the mobility of the nodes, due to an increase in the number of
broken link events. The routing overhead in AODV is high compared to that in BFIRP,
because of the increase in the number of transmitted RREQ, RREP, and RERR packets,
due to link breakages at intermediate nodes of a path. Routing overhead of BFIRP is
slightly high compared to MFR and it increases with an increase in the mobility of
the nodes because of the increase in the number of FERR packets required to route
packets whenever voids are encountered, as shown in Figure 11.
Figure 23 shows that the average end-to-end delay experienced by data packets in
BFIRP is less compared to AODV, but slightly higher compared to MFR. The end-to-
end delay increases with an increase in the mobility of the nodes because of multiple
retransmissions of data packets and additional processing and transmission delays at
intermediate nodes due to broken link events. AODV has high average end-to-end
delay compared to BFIRP because of large delays encountered due to the buffering
of packets, the transmission of RREQ, the reception of the RREP packets, and the
reconstruction of routes to destination nodes by source nodes, as it has no local route
recovery procedure.
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
Bird Flight-Inspired Routing Protocol for Mobile Ad Hoc Networks 25:31
Fig. 23. Variation in the average end-to-end delay with the variation in mobility of the nodes.
BFIRP has a low average end-to-end delay compared to that of AODV because pack-
ets are routed along the great circle path between the source and destination, which is
the shortest path. However, BFIRP has a relatively higher average end-to-end delay
compared to that of MFR because of the additional processing and transmission delays
associated with the recovery routing scheme whenever voids are encountered wherein
MFR packets are dropped.
Figure 24 shows that the maximum number of hops traversed by a data packet is
less in BFIRP compared to AODV and MFR. The maximum hop count increases with
an increase in the mobility of the nodes because packets are routed along detouring
long paths.
The maximum hop count of BFIRP is less compared to AODV and MFR, because
data packets are routed along the great circle path, which is the shortest path between
two points on a sphere. It also uses the recovery routing procedure to ensure that
packets are routed along the great circle path in case voids are encountered.
7.4.3. Performance Results of BFIR with Different Application Constants. In this section, we
evaluate the performance of the proposed BFIRP routing scheme by varying the ap-
plication constants , , . The values of the simulation parameters are the same as
specied in Table I. The maximum speed of the nodes is set to 4 m/s.
Figure 25 shows that the delivery ratio is high when all three terms in Equation (11)
are given equal importance in the calculation of priority, by assigning a value equal to
1/3. When the energy term in Equation (11) is given more importance compared to
the other two terms, BFIRP performs poorly because data packets are routed along
detouring paths, leading to the dropping of packets; but performance increases with
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
25:32 S. Misra and G. Rajesh
Fig. 24. Variation in the maximum number of hops traversed by a data packet with the variation in the
mobility of the nodes.
Fig. 25. Variation in the delivery ratio with the variation in number of nodes in the network.
an increase in the number of nodes. This is because the number of nodes within the
transmission range of a node increases. The same argument can be stated for the case
where the distance term is given more importance than the other two. When the
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
Bird Flight-Inspired Routing Protocol for Mobile Ad Hoc Networks 25:33
Fig. 26. Variation in the average residual energy with the variation in the number of nodes.
degree of closeness term is given more importance, the performance of BFIRP is almost
the same as that when all the three terms are given equal importance. This is because
the packets are routed along the great circle path, which is the shortest path.
The results shown in Figure 26 prove that the average energy consumed by BFIRP
is less when all the three parameters, , , , are assigned a value equal to 1/3. When
the energy term in Equation (11) is given more importance compared to the other
two terms, BFIRP performs poorly. This is because the data packets are routed along
detouring paths, leading to high hop count. The same argument can be stated for the
case in which the distance term is given more importance than the other two. The
performance of BFIRP, when the degree of closeness term is given more importance,
is almost the same as that when all three terms are given equal importance. This is
because the packets are routed along the great circle path (the shortest path), thus
reducing the hop count and the energy consumption for the routing process.
The results shown in Figure 27 prove that the routing overhead is less when
all three, , , , are assigned a value equal to 1/3. When the energy term in
Equation (11) is given more importance compared to other two terms, BFIRP has rela-
tively higher routing overhead, because packets are forwarded to nodes with maximum
energy, which may or may not be close to the great circle path between the source and
destination, resulting in detouring paths. The detouring paths lead to transmission of
additional FERR packets by intermediate nodes. The same argument can be stated for
the case where the distance term is given more importance than the other two. The
routing overhead of BFIRP, when the degree of closeness term is given more impor-
tance, is almost the same as that when all three terms are given equal importance.
This is because the packets are routed along paths close to the great circle, resulting
in nondetouring paths.
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
25:34 S. Misra and G. Rajesh
Fig. 27. Variation in the routing overhead with the variation in the number of nodes.
Fig. 28. Variation in the average end-to-end delay with the variation in the number of nodes.
The results shown in Figure 28 prove that the average end-to-end delay is less when
all three parameters, , , , are assigned a value equal to 1/3. When the energy term
in Equation (11) is given more importance compared to other two terms, the data
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
Bird Flight-Inspired Routing Protocol for Mobile Ad Hoc Networks 25:35
packets experience high end-to-end delay. This is because packets are forwarded to
nodes with low degree of closeness to the great circle path between the source and
destination nodes, resulting in additional transmission and processing delays at the
intermediate nodes.
8. CONCLUSION
In this article, we have proposed a new geographic routing protocol inspired by the
navigation of migratory birds over long distances, following the great circle arc, which
is the shortest path connecting two points on the surface of a sphere. To the best of
our knowledge, the proposed protocol is the rst effort in designing a routing protocol
based on the three dimensional view of the ad hoc network geographical region. In
the proposed protocol, the packets are forwarded taking into consideration the energy
of the node, distance of the node from the destination, and the degree of closeness
of the node to the great circle path connecting the intermediate and the destination
nodes. The proposed protocol is dynamic, highly scalable, energy efcient, and readily
applicable to real-world applications. The proposed solution is relatively better than
that of AODV and MFR in terms of the delivery ratio, the end-to-end delay, hop count,
and routing overhead. We have presented the results of simulations in various sce-
narios comparing the performance of BFIRP, AODV, and MFR. Results show that the
delivery ratio of BFIRP is better than that of AODV and MFR. The routing overhead
almost remains constant and is relatively less compared to AODV with an increase in
the number of nodes and the mobility of nodes, which makes it highly scalable. The
end-to-end delay is also less compared to that of AODV. The energy consumed with
the increase in number of nodes and the mobility of nodes is less compared to that of
AODV, thereby making it energy efcient. Overall, the performance of the proposed
protocol is observed to be better than that of AODV and MFR.
BFIRP is an algorithm that takes into account a realistic view of the ad hoc net-
work geographical region, which makes it suitable for employment in future ad hoc
networks. In the future, before employing BFIRP in mobile ad hoc networks, we in-
tend to carry out a thorough evaluation of its performance by deploying it on a test bed.
Our main focus will be to study bandwidth consumption issues, especially for different
degrees of node mobility. The future work will also include integration of the proposed
protocol with other works to make it secure. We also plan to increase the robustness
of the protocol by supporting node disjoint multipath routing.
REFERENCES
ALERSTAM, T. 1993. Bird Migration. Cambridge University Press.
BASAGNI, S., CHLAMTAC, I., SYROTIUK, V. R., AND WOODWARD, B. A. 1998. A distance routing effect
algorithm for mobility (DREAM). In Proceedings of the 4th Annual ACM/IEEE International Conference
on Mobile Computing and Networking. 7684.
BOSE, P., MORIN, P., STOJMENOVIC, I., AND URRUTIA, J. 1999. Routing with guaranteed delivery in ad
hoc wireless networks. In Proceedings of the 3rd International Workshop on Discrete Algorithms and
Methods for Mobile Computing and Communications. 4855.
BROWN, T. AND MOHAN, S. 1997. Mobility management for personal communication systems. IEEE Trans.
Veh. Technol. 46, 2, 269278.
CAMARA, D. AND LOUREIRO, A. A. F. 2000. GPS/ant-like routing in ad hoc networks. In Proceedings of the
Wireless Communications and Networking Conference. 12321236.
CAO, Y. AND XIE, S. 2003. A position based beaconless routing algorithm for mobile ad hoc networks. In
Proceedings of the IEEE International Conference on Communications, Circuits and Systems. 303307.
CARVALHO, M. M., MARGI, C. B., OBRACZKA, K., AND GARCIA-LUNA-ACEVES, J. J. 2004. Modeling energy
consumption in single-hop IEEE 802.11 ad hoc networks. In Proceedings of the International Conference
on Computer Communications and Networks (ICCCN). 367372.
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
25:36 S. Misra and G. Rajesh
CHANG, J.-H. AND TASSIULAS, L. 2000. Energy conserving routing in wireless ad-hoc networks. In Pro-
ceedings of the Conference on Computer Communications. 2231.
CHEN, C.-L. AND HSU, T.-P. 2004. A novel approach to great circle sailings: The great circle equation. J.
Navigation. 57, 311325.
CLAUSEN, T. AND JACQUET, P. 2003. Optimized link state routing protocol (OLSR). RFC Editor.
DAY, K., ARAFEH, B., TOUZENE, A., AND AL-KINDI, A. 2008. A 3D grid position-based routing protocol for
mobile ad hoc networks. In Proceedings of the International Conference on Computer and Communica-
tion Engineering. 151156.
EBERT, J., AIER, S., KOFAHL, G., BECKER, A., BURNS, B., AND WOLISZ, A. 2002. Measurement and
simulation of the energy consumption of a WLAN interface. Tech. rep. TKN-02-010, Telecommunication
Networks Group, Technical University Berlin.
FLOYD, S. AND JACOBSON, V. 1994. The synchronization of periodic routing messages. IEEE/ACM Trans.
Netw. 2, 122136.
HAAS, Z. J., PERLMAN, M. R., AND SAMAR, P. 2002. The zone routing protocol (ZRP) for ad hoc networks.
http://www.ietf.org/proceedings/02nov/I-D/draft-ietf-manet-zone-zrp-04.txt.
HEISSENBTTEL, M., BRAUN, T., BERNOULLI, T., AND WAELCHLI, M. 2004. BLR: Beacon-less routing algo-
rithm for mobile ad hoc networks. Elsevier Comput. Comm. J. 27, 11, 10761086.
HEISSENB UTTEL, M. AND BRAUN, T. 2005. Optimizing neighbor table accuracy of position-based routing
algorithms. In Proceedings of the Annual Joint Conference of the IEEE Computer and Communication
Societies (INFOCOM).
JAIN, R., PURI, A., AND SENGUPTA, R. 2001. Geographical routing using partial information for wireless
ad hoc networks. IEEE Pers. Commun. 8, 1, 1628.
JOHNSON, D. B. AND MALTZ, D. A. 1996. Dynamic source routing in ad hoc wireless networks. In Mobile
Computing. T. Imielinski, H. Korth, Eds., Kluwer Academic Publishers, 153181.
KARP, B. AND KUNG, H. T. 2000. GPSR: Greedy perimeter stateless routing for wireless networks. In Pro-
ceedings of the 6th Annual International Conference on Mobile Computing and Networking. 243254.
KO, Y.-B. AND VAIDYA, N. H. 2000. Location-aided routing in mobile ad hoc networks. Wirel. Netw. 6, 4,
307321.
KUHN, F., WATTENHOFER, R., AND ZOLLINGER, A. 2003. Worst-case optimal and average-case efcient
geometric mobile ad hoc routing. In Proceedings of 4th International Symposium on Mobile Ad Hoc
Networking and Computing (Mobihoc). 267278.
LI, J., JANNOTTI, J., DE COUTO, D. S. J., KARGER, D. R., AND MORRIS, R. 2000. A scalable location
service for geographic ad hoc routing. In Proceedings of the ACM/IEEE International Conference on
Mobile Computing and Networking. 120130.
LIAO, W.-H., SHEU, J.-P., AND TSENG, Y.-C. 2001. GRID: A fully location-aware routing protocol for mobile
ad hoc networks. Telecomm. Syst. J. 18, 3760.
LIN, X. AND STOJMENOVIC, I. 1998. Geographic distance routing in ad hoc wireless networks. Tech. rep.,
SITE, University of Ottawa.
MAUVE, M., WIDMER, J., AND HATTENSTEIN, H. 2001. A survey on position-based routing in mobile ad-hoc
networks. IEEE Network 15, 6, 3039.
MURTHY, S. AND GARCIA-LUNA-ACEVES, J. J. 1996. An efcient routing protocol for wireless networks.
Mob. Netw. Appl. 1, 2, 183197.
NA, J. AND KIM, C.-K. 2006. A novel geographic routing scheme for large wireless ad hoc networks. Comp.
Netw., Int. J. Comput. Telecomm. Netw. 50, 17, 34343448.
NAYAK, A., STOJMENOVIC, I., AND KURUVILA, J. 2005. Design guidelines for routing protocols in ad hoc
and sensor networks with a realistic physical layer. IEEE Comm. Mag. 46, 2, 101106.
NICULESCU, D. AND NATH, B. 2003. Trajectory based forwarding and its applications. In Proceedings of the
9th Annual International Conference on Mobile Computing and Networking. ACM Press, 260272.
PARK, V. D. AND CORSON, M. S. 1997. A highly adaptive distributed routing algorithm for mobile wireless
networks. In Proceedings of the Annual Joint Conference of the IEEE Computer and Communication
Societies (INFOCOM). 711.
PARK, V. AND CORSON, S. 2001. Temporally-ordered routing algorithm (TORA). Functional Specication,
IETF draft.
PEI, G., GERLA, M., AND CHEN, T.-W. 2000. Fisheye state routing: A routing scheme for ad hoc wireless
networks. In Proceedings of the IEEE International Conference on Communications (ICC). 7074.
PERKINS, C. E. AND BHAGWAT, P. 1994. Highly dynamic destination-sequenced distance-vector routing
(DSDV) for mobile computers. ACM SIGCOMM Comput. Comm. Rev. 24, 4, 234244.
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.
Bird Flight-Inspired Routing Protocol for Mobile Ad Hoc Networks 25:37
PERKINS, C. E. AND ROYER, E. M. 1999. Ad hoc on-demand distance vector routing. In Proceedings of the
2nd IEEE Workshop on Mobile Computer Systems and Applications. IEEE Computer Society, 90100.
QABAJEH, L. K., KIAH, L. M., AND QABAJEH, M. M. 2009. A qualitative comparison of position-based
routing protocols for ad hoc networks. IJCSNS Int. J. Comput. Sci. Netw. Secur. 9, 2, 131140.
QIU, J. 2005. Ornithology: Flight of the navigators. Nature Int. Weekly J. Sci. 437, 804806.
STOJMENOVIC, I. 2002. Position-based routing in ad hoc networks. IEEE Comm. Mag. 40, 7, 128134.
TAKAGI, H. AND KLEINROCK, L. 1984. Optimal transmission ranges for randomly distributed packet radio
terminals. IEEE Trans. Comm. 32, 3, 246257.
TOH, C. K. 1997. Associativity-based routing for ad-hoc mobile networks. Wirel. Pers. Comm. 4, 2, 103139.
UC AN, F. AND ALTILAR, D. T. 2008, Navigation and guidance planning for air vehicles. In Proceedings of
the 20th IEEE International Conference on Tools with Articial Intelligence. 534538.
XU, Y., LEE, W.-C., XU, J., AND MITCHELL, G. 2005. PSGR: Priority-based stateless geo-routing in highly
dynamic sensor networks. In Proceedings of the IEEE International Conference on Mobile Ad hoc and
Sensor Systems Conference. 673680.
Received January 2010; accepted August 2010
ACM Transactions on Autonomous and Adaptive Systems, Vol. 6, No. 4, Article 25, Publication date: October 2011.

You might also like