You are on page 1of 6

GC'12 Workshop: The 3rd International Workshop on Wireless Networking & Control for Unmanned Autonomous Vehicles: 1

Architectures, Protocols and Applications

A Geographic Mobility Prediction Routing


Protocol for Ad Hoc UAV Network 1

Lin Lin, Qibo Sun, Shangguang Wang, and Fangchun Yang

Abstract—Unmanned Aerial Vehicles (UAVs) play more and characteristic of AN [1]. And it is the main reason of causing
more important roles in modern warfare. However, the data intermittent and episodic connections which brings challenges
routing for communication among UAVs faces several challenges, such as frequent change in topology. So, it is hard to maintain
such as packet loss or routing path failure etc. The main problem
of UAVs data routing is caused by the high mobility of UAVs. In routing path and high routing performance in AN, especially
this paper, an effective geographic mobility prediction routing for topology based routing protocols which depend on stable
protocol is proposed to improve the performance of routing routing path. In contrast, the representative geographic protocol,
among UAVs. First, a Gaussian distribution of UAVs movement Greedy Perimeter Stateless Routing (GPSR) [3] makes routing
probability density function is deduced to reduce the impact of decision according to next hop without stable routing path. So,
high mobility. Then, two-hop perimeter forwarding is proposed to GPSR has the potential to get better performance in AN.
reduce the impact of routing void. The experiment results show
that the proposed approach can provide effective and reliable However, there are still some problems have to be addressed.
data routing with acceptable communication overhead in the The first problem is how to reduce the impact of high
highly dynamic environment of Ad Hoc UAV Network. mobility. The data routing of GPSR relies on the position of its
neighbors and the destination. However, it is difficult for UAVs
Keywords—Multi-UAV, Ad Hoc UAV Networks, Geographic updating information in battlefield. So, the routing depends on
Based Routing, Mobility Prediction the record obtained a period of time ago. Considering the
velocity of UAVs, the real position of each UAV may be
different from the record. Thus, the data routing may fail.
I. INTRODUCTION

T he use of unmanned aerial vehicles (UAVs) for military


missions receives growing attention in the recent years,
such as surveillance, search and attack missions. Considering
the situation of battlefield, these missions require real-time and
accurate data routing among UAVs for prosecuting tasks
cooperatively. To satisfy the requirements of self-organizing
and infrastructure-less among UAVs, an effective approach is
to use Ad Hoc UAV network which is an instance of Airborne
Network (AN) [1] as shown in Figure 1.
Figure 2 Example of Routing Void

The other problem is as shown in Figure 2. The node S has


packets to send to O. But the distance between each neighbor
and O is longer than S. This situation is termed as S meeting a
void. To make a distinction from normal meaning of void, it is
named as routing void in this paper. Then, S will switch on
perimeter mode. And S may choose B as next hop. Although
Figure 1 Example of Airborne Network there is a shorter path (SĺLĺMĺNĺPĺO), S may choose
Previous studies have shown the differences between AN the path (SĺBĺCĺDĺEĺFĺGĺHĺIĺO). At least, it
and Internet [2]: intermittent link, low bandwidth connection will result in extra hops and longer delay.
and high latency. The high mobility of UAVs is the main To address the routing problems of Ad Hoc UAV network, a
geographic based routing protocol Mobility Prediction based
Geographic Routing (MPGR) is proposed in this paper. MPGR
Manuscript submitted by June 12, 2012. This work is supported by Research can reduce the impact of high mobility and routing void while
Fund for the Doctoral Program of Higher Education of China
(20110005130001). the communication overhead is low to avoid being detected.
L. Lin, Q. Sun, S. Wang and F. Yang are with the State Key Laboratory of The contributions of this paper are summarized as follows:
Networking and Switching Technology, Beijing University of Posts and 1) A Gaussian distribution of UAVs movement probability
Telecommunications, Beijing, 100876, People’s Public of China.
Corresponding author: L. Lin, e-mail: bupt.linlin@gmail.com. density function is deduced to predict the mobility of

978-1-4673-4941-3/12/$31.00 ©2012 IEEE 1597


Authorized licensed use limited to: ULAKBIM UASL ISTANBUL TEKNIK UNIV. Downloaded on November 22,2023 at 12:51:50 UTC from IEEE Xplore. Restrictions apply.
2

UAVs and reduce the impact of high mobility while the satellites for information update. So, the data routing would
communication overhead is acceptable. depend on the record obtained a period of time ago. As shown
2) By analyzing routing void, two-hop perimeter forwarding in Figure 3, the relay node D will move away from the
is proposed to reduce the impact of routing void. communication range of the sender S. And the destination O
3) A simulation environment based on the Gauss mobility will move away from the original position before packets
model is designed to verify the performance of MPGR. arrival. Thus, the real position of each UAV is different from
The rest of this paper is organized as follows: In Section II, the record one. As a result, the packet may be lost. To reduce
the related work is presented. The Gaussian distribution of the impact of high velocity, it is necessary to predict the
UAVs movement probability density function is deduced in mobility of UAVs. Then using the predicted mobility for data
Section III. Section IV presents the analysis of routing void and routing can make the routing more accurately and improve the
two-hop neighbor selection method. The proposed protocol is routing performance.
detailed in Section V. In Section VI, the simulation results are
B. UAVs Mobility Prediction
described. Finally, the conclusions and future works are shown
in Section VII. The goal of mobility prediction is to predict the position
distribution of UAVs at tn according to the mobility feature and
II. RELATED WORK position at tn-1. A typical scenario of AN can be seen as an
Euclidean graph G = (U, E) with following properties:
After proposed, GPSR has inspired extensions in WSN and
y Q = [0, s] x [0, s] is the UAVs movement field.
vehicular Ad Hoc networks (VANET). Recent years, some
y U = {u0, u1, … , un} is the set of UAVs nodes.
works [4-9] attempted to exploit geographic information to
y Edge (i, j) ‫ א‬E iff the distance between ui and uj is not more
improve the routing performance of AN. The works of [4, 5]
than R which is the communication range of UAVs.
improved the protocol by predicting the movement of aircrafts
y The UAV ui is with capabilities of: high velocity over 300
while they assumed the destination was static. Daniel Medina
km/h; localization with GPS.
and Felix Hoffmann et al. [6] exploited the multiplicity of
y ‫ ׊‬ui ‫ א‬U, (xi, yi) ‫ א‬Q is the immediate position of ui. For
source-destination paths in moderate or high density AN while
simplicity, the movement of UAVs is assumed in 2-D.
they assumed the topology was quasi static. The works of [7]
y vi is the velocity of ui while și is the angle in degrees
used time varying topology for data routing while the UAVs’
between the direction of ui and positive x-axis.
orbits are assumed to be predefined. Actually, the trajectory of
Let ǻt be the time step between tn-1 and tn. So, ǻT = tn-tn-1 =
UAVs may change according to the situation of battlefield.
mǻt where m is the number of time step. The velocity of UAVs
Moreover, UAVs cannot connect to ground stations in enemy’s
is random at a given time tn. So, the velocity can be regarded as
region while they cannot communicate with each other
uniform motion while the acceleration is a series of White
frequently to avoid being detected. So, the methods of [8, 9]
Gaussian Noise (WGN) with constant variance İ during ǻt.
based on the assistance of ground stations and frequent
Theorem 1: For UAV ui at kth step between tn-1 and tn, x(k) is
interaction among UAVs cannot work well in battlefield. The
the position component on x-axis. v(xk ) is the velocity
problem of routing void has been discussed in [10] etc.
However, their scenario was in VANET with high vehicle component on x-axis while ε (xk ) is the acceleration component
density and moderate mobility while they solved the problem
on x-axis. Thus, x(k) and v(xk ) are expressed as follows:
with the characteristics of geography, e.g. the crossroads. As
k
discussed, high mobility of UAV is important for routing
v(xk ) = v(0)
x
+ ¦ ε (xi ) Δt (1)
performance while routing void will lead to forwarding failure. i =1
Both problems need to be improved in further research, which 1 k
are also fully discussed in rest of this paper.
x
¦ (2k − 2i + 1)ε (xi ) Δt 2
x( k ) = x(0) + kv(0)
2 i =1
Δt + (2)

Proof: (1) and (2) can be proved by Mathematical Induction.


III. MOBILITY PREDICTION
Deduction 2: According to the position at tn-1, the arguments
A. Problem Description of Gauss distribution at tn are expressed as follows (The
expressions of y-axis μ y and σ y2 are same with x-axis):
μ x = E ( xn ) = xn −1 + vnx−1mΔt (3)
m
σ x2 = D ( xn ) = E (¦ (2i − 1)2 (ε kx )2 )Δt 4 / 4 (4)
i =1

Deduction 3: According to the position (xn-1, yn-1) at tn-1, after


ǻT = tn-tn-1 = mǻt, the probability to arrive at (xn, yn) is
expressed as follows:
1 −(( xn − μx )2 + ( yn − μy )2 )
p{(xn , yn ) | ( xn−1, yn−1)} = exp{ } (5)
Figure 3. Potential neighbor moving towards destination 2πσ 2 2σ 2
With the mobility feature and position of UAV at tn-1, it is
In enemy’s region, it is difficult for UAVs connecting to possible to predict the position at tn according to (5) in the form
1598
Authorized licensed use limited to: ULAKBIM UASL ISTANBUL TEKNIK UNIV. Downloaded on November 22,2023 at 12:51:50 UTC from IEEE Xplore. Restrictions apply.
3

as follows: which are related to routing void. Then, the method of reducing
­° xn = predict _ pos _ x ( xn −1 , v , ε , ΔT )
x x impact of routing void could be deduced.
n −1 n −1
® (6)
B. Routing Void Analysis
°̄ yn = predict _ pos _ y ( yn −1 , vn −1 , ε n −1 , Δ T )
y y

Before analyzing routing void, there are some assumptions


C. Neighbor Connection Persistence as follows:
With the prediction of mobility, it is possible to analyze the y There are m neighbors on average in the communication
connection persistence which is useful to select next hop more range of each UAV.
accurately. With (6), the Euclidean Distance between each y The distance between sender S and destination O is D
neighbor and sender can be predicted. Given the sender uo with which is longer than communication range R.
coordinates (xo, yo) and a neighbor ur with (xr, yr) at time T, the y The position distribution of UAVs obeys the distribution
distance between ur and ud is: of random walker.
Theorem 4: The probability of UAV S with m neighbors
Δd = ( xr − xo ) 2 + ( yr − yo ) 2 (7)
meeting routing void is:
After a period of time ǻT, the distance should be: 1 1 R m
p ≈ ( − arcsin ) (12)
2
Δd ' = ( x 'r − x 'o ) + ( y 'r − y 'o ) 2
(8) 2 π 2D
Δd ' means the connection relationship between uo and ur
after ǻT which indicates connection persistence of two nodes.
A longer ǻT means more reliable connection between a node
and its neighbor. To eliminate the impact of high mobility, the
connection persistence should be taken into account.
D. Next Hop Selection
Optimal Objective
Let IA be the neighbor set of UAV uA. There are two aspects
needed to be taken into account with next hop selection: Figure 4 Example of UAV S Meeting Routing Void
1) Strong neighbor connection persistence: it will guarantee
the communication process of selected next hop more reliable. Proof: As shown in Figure 4, if S meets routing void, all of
The ǻT is maximum persistence time of sender’s neighbors: its neighbors should be in SACB. The area of SACB is:
ΔT = max j∈I ΔT j (9) π R2 R R 1
A Q= − R2 arcsin + 2D2 arcsin − R 2D2 − R2 (13)
2) Short distance from destination: it will induce a faster 2 2D 2D 4
forwarding and fewer hops. The distance D between neighbor Then the probability of neighbor uj in SACB is:
uj and destination ud is expressed as follows: Q 1 1 R
pj = ≈ − arcsin (14)
2 2 πR 2
2 π 2D
D = min j∈I A D j = ( x j − xd ) + ( y j − yd ) (10)
So, the probability of all neighbors in SACB is:
Joint Metric 1 1 R m
To combine the two objectives, a joint metric called Reliable p ≈ ( − arcsin ) (15)
2 π 2D
Next Hop (RNH) is proposed as the next hop selection metric.
Deduction 5: If the average number of neighbor is great
The neighbor with the minimum RNH will be selected:
enough, the probability of meeting routing void is almost zero.
­ η
° D j + ΔT , ( D j < Di ) ∧ ( ΔT j ≥ 1) (11)
However, in real environment, the density of UAVs cannot
RNH = min j∈I A RNH j = ® j be high. So, it is necessary to address the trouble of routing void
° MAX _ VALUE , otherwise without increasing the density of UAVs.
¯
Where Ș is the weight used to adjust the affect of neighbor Theorem 6: The probability of a random routing path
connection persistence. The difference in distance between without routing void is:
each neighbor and the destination is not longer than the proute = (1 − p) hop (16)
communication range R. So, Ș is set with the range of (1, R).
(2 m + 3) ³ Df ( D ) dD is the average hops on each
hop =
IV. ROUTING VOID AVOIDANCE 2 mR
path. f ( D ) is the distribution function of distance travelled by
A. Problem Description
Although the mobility can be predicted more accurately with a random walker [11]. ³ Df ( D)dD is the distance between
the approach above, the routing void would still happen. It will each neighbor and destination.
induce extra hops or routing failure. Especially, considering the Proof: As shown in Figure 5, the probability pdr that the
characteristic of AN, such as low node density and frequent distance between the farthest neighbor and S equals to r can be
topology changes, it is necessary to analyze the scenario factors expressed as follows:

1599
Authorized licensed use limited to: ULAKBIM UASL ISTANBUL TEKNIK UNIV. Downloaded on November 22,2023 at 12:51:50 UTC from IEEE Xplore. Restrictions apply.
4

m V. PROPOSED ALGORITHM
§ π r2 · 2π rdr
pd r ≈ ¨ 2 ¸
m (17) In this section, the routing protocol MPGR is presented,
©πR ¹ π R2
which is based on GPSR. The details are listed as follows.
The average distance of the farthest neighbor is as follows:
R 2m A. On Demand Neighbor Discovery
d = ³ rpd r dr = R (18) Considering the battlefield environment, it is difficult for
0 2m + 3
UAVs to connect to satellite all the time. An effective way is to
Then the average number of hops on the path is as follows:
get information of one-hop neighbors periodically to finish
(2m + 3) ³ Df ( D)dD global routing. However, it is impossible to interact frequently
hop = (19)
2mR in battlefield because it will increase the risk of being detected
by enemy. In this algorithm, when there are no packets to
Thus, the probability of data routing through a random path delivery, UAVs will travel silently. The velocity and position of
without routing void is as shown in Theorem 6. destination are assumed be obtained by reaction method which
is beyond the scope of this paper.
If one UAV has to send message, it will broadcast Neighbor
Discovery packet (ND), as shown in Table I. The ND mainly
provides information of destination. The neighbors will reply
with the information required by sender, as shown in Table II.
All of the information will be recorded in the neighbor table of
sender. Then the sender will keep silent to calculate and select
the next hop. The neighbors also keep silent to reduce the risk
of being detected. That is different from GPSR which requires
Figure 5 The Farthest Neighbor of UAV UAVs to broadcast periodically.
Deduction 7: With the increase in communication range of TABLE I
MAIN FIELDS OF NEIGHBOR DISCOVERY PACKET
UAVs, the probability of meeting routing void will decrease. Field Function
According to (19), when the communication range increases, P position of destination
the average hops on a path will decrease. So, the probability of M delivery mode (greedy or perimeter)
UAV without meeting routing void will also increase according Lp position of entering perimeter forwarding
D distance from destination
to (16). Hence, the packet delivery ratio will increase while the TABLE II
end-to-end delay will decrease correspondingly. MAIN FIELDS OF NEIGHBOR REPLY PACKET
Field Function
C. Two-Hop Neighbor Selection P position of neighbor
However, the communication range of UAVs is limited. In v velocity of neighbor
İ acceleration of neighbor
order to increase the range, the two-hop information is used for RNHnext RNH of two-hop neighbor
neighbor selection. Two-hop information means the position D distance between two-hop neighbor and destination
and velocity information of sender’s two hop neighbors. That
enlarges the communication range of sender equivalently. B. Mobility Prediction Greedy Forwarding
There may be no immediate information for sender to find
next hop or locate destination accurately. So, the sender can use
the prediction method of (6) to predict the position and mobility
of neighbors and destination. The packets delivery will become
more accurately than without taking mobility into account. Let
Ai be the UAV with packets to send by greedy forwarding. The
process of greedy forwarding is detailed as follows:
1) After receiving a packet, Ai judges whether the destination
is itself at first. If it is, the packet will be passed to the
application layer. Go to 5)
Figure 6 Example of Two-Hop Avoid Routing Void
2) Ai broadcasts ND packet and waits for neighbors’ reply. If
As shown in Figure 6, S meets routing void. Then, S checks there is no reply before timeout, go to 6).
its neighbors’ two-hop information. S finds that L has neighbor 3) After receiving replies, Ai calculates the position of
M whose distance from O is shorter than others. Thus, L is destination and neighbors according to mobility prediction
selected as the next hop. Then the greedy forwarding is valid function (6). Then Ai updates the mobility information of
again. It will quickly find the path (SĺLĺMĺNĺPĺO). The destination as well as the neighbor table.
extra hops caused by routing void are avoided. In some case, 4) Ai calculates the distance between destination and each
the routing failure caused by routing void can also be avoided. neighbor. Then it calculates the connection persistence of each
neighbor. According to RNH of (11), Ai selects neighbor as
next hop to send the packet. If there is no candidate, it means

1600
Authorized licensed use limited to: ULAKBIM UASL ISTANBUL TEKNIK UNIV. Downloaded on November 22,2023 at 12:51:50 UTC from IEEE Xplore. Restrictions apply.
5

that Ai meets routing void. So, MPGR will switch on two-hop B. Simulation Scenario
perimeter forwarding which is detailed in next section. The simulation runs on PC (Intel Core2 2.1GHz CPU, 2GB
5) The delivery is completed. RAM, Windows XP SP3) for 1000 times.
6) The delivery is failed. UAVs move at the altitude of 8km with constant velocity. To
C. Two-Hop Perimeter Forwarding reflect the impact of mobility, the velocity of UAVs is set from
1 m/s to 150 m/s. The default beacon mode of GPSR is set to 5
Once meeting routing void, the greedy forwarding is failed.
seconds while the compared mode is 3 seconds. The application
And MPGR will switch on perimeter forwarding. As analyzed
layer of source node uses CBR protocol to simulate the traffic
in Section IV, the probability of meeting routing void can be
which is generated based on Poisson traffic model. To test the
reduced by increasing communication range. It means that
performance of MPGR with different density, the number of
longer range is helpful for finding appropriate next hop and
UAVs id set from 50 to 100 with the velocity varying. Other
resuming greedy forwarding. According to the analysis of
details of parameters are shown in Table III.
Section IV.C, the two-hop perimeter forwarding is proposed.
There are three different metrics to measure the performance
For simplicity, the Euclidean distance is used to make
of each routing algorithm: 1) Packet Delivery Ratio (PDR) 2)
decision. Let Ai be the UAV who will send packet by perimeter
End to End Delay and 3) Packet Overhead.
forwarding. The details are listed as follows: TABLE III
1) After receiving a packet, Ai judges whether the destination SIMULATION PARAMETERS
is itself at first. If it is, the packet will be passed to the Variable Values
application layer. Go to 6). Simulation area 10 km x 10 km
Node number 50 - 100
2) Ai broadcasts ND packet and waits for neighbors’ reply. If UDP packet size 512 Bytes
there is no reply before timeout, go to 8). Propagation loss model Friis
3) After receiving neighbors’ replies, Ai calculates the Transmission range 1 km
Transport Protocol UDP
distance between each neighbor and destination. If there are Each simulation runs time 100s
neighbors whose distance from destination is shorter than Ai, Simulation runs 1000
the perimeter forwarding is complete. Go to 7).
4) Ai gets the two-hop information of each neighbor. If there C. Simulation Result on Packet Delivery Ratio
is no two-hop neighbor information, go to 8). The performance of PDR is shown in Figure 7. In general,
5) Ai calculates the distance between each two-hop neighbor PDR of each protocol increases with the growth in number of
and destination. The neighbor Aj who has the shortest distance UAVs while the increasing velocity induces the reduction of
between two-hop neighbor and destination will be the next hop. PDR. MPGR gets good performance of PDR with both low and
Ai sends the packet to Aj. high velocity. The reason details as follows.
6) The delivery is completed.
7) The protocol switches on greedy forwarding. Then Ai
sends packet to the neighbor whose distance from destination is
the shortest one. Go to 6).
8) The delivery is failed.

VI. SIMULATION RESULTS


In this section, the results of simulations conducted with the
NS-2 simulator [12] are presented to compare the performance Figure 7 Average packet delivery ratio
of MPGR with GPSR and Ad hoc On-Demand Distance Vector
As shown in Figure 7, AODV achieves good performance of
Routing (AODV).
PDR at low velocity. However, the high velocity affects the
A. Simulation Model of UAVs Mobility stability of topology which is needed by AODV to form routing
The movement of UAVs follows the rule of aerodynamics. path. So, the performance of AODV decreases rapidly with the
With the knowledge of current velocity, the movement of increase in velocity. On another aspect, GPSR selects next hop
UAVs can be simulated with Gauss-Markov mobility model by Euclidean distance. So, the routing path selection becomes
[13]. In this simulation, the velocity of UAVs is modulated as inaccurate when the velocity increases. That results in the
the following model: reduction of PDR for GPSR. However, the short beacon period
can improve PDR of GPSR because a shorter period would
­ v = β v + (1 − β ) v + (1 − β 2 ) v
° n n −1 xn −1
(20) update neighbor information more timely. Compared with the
®
°̄ d n = β d n −1 + (1 − β ) d + (1 − β ) d xn−1
2 formers, PDR of MPGR is always at high level. Because the
effect of mobility prediction strategy and two-hop perimeter
ȕ is the tuning parameter. v and d are the mean velocity and forwarding reduces the impact of both high mobility and
direction parameters. vxn−1 and d xn−1 are random variables routing void.
from a Gaussian distribution, which give some randomness to D. Simulation Result on End to End Delay
the new velocity and direction parameters. As shown in Figure 8, the end to end delay of each protocol is
1601
Authorized licensed use limited to: ULAKBIM UASL ISTANBUL TEKNIK UNIV. Downloaded on November 22,2023 at 12:51:50 UTC from IEEE Xplore. Restrictions apply.
6

increased with the increasing in velocity while the growth in Hoc UAVs network. To verify the effectiveness, a simulation
number of UAVs reduces the delay of geographic protocols in based on Gauss mobility model is designed to compare the
general. At low velocity, AODV gets better performance while performance of MPGR with other protocols. The simulation
the routing void happened to GPSR may induce extra hops or results show that MPGR outperforms AODR and GPSR in
routing failure. At high velocity, it is hard for both AODV and terms of PDR and end to end delay with minimum overhead.
GPSR forwarding data with high performance. Thus, MPGR can provide effective and reliable data routing for
communication among UAVs. In future work, the load
balancing for heavy traffic application of UAVs should be
taken into account with protocol design.

REFERENCES
[1] Airborne Network Architecture System Communication Description &
Technical Architecture Profile Version 1.1, USAF Airborne Network
Special Interest Group, 7 October 2004.
[2] Bosner, Wayne, Stranc, Ken, LaBarre, Lee, and Eyestone, David,
Figure 8 Average end to end delay “Internet Versus Airborne Network”, USAF Special Interest Group,
February 2008.
Compared with GPSR, MPGR has much better performance [3] Karp B and Kung H T. “GPSR: Greedy Perimeter Stateless Routing for
wireless networks”. In Proceedings of the Sixth ACM Annual
on end to end delay. The reason is that two-hop perimeter International Conference on Computing and Networking, Boston, MA,
forwarding effectively reduces the probability of meeting 2000: 243- 254.
routing void as well as the impact of routing void. That [4] J. P. Rohrer, A. Jabar, E. K. Cetinkaya and J. P. Sterbenz, “Cross-Layer
Architectural Framework for Highly-Mobile Multihop Airborne
obviously decreases the end to end delay. Telemetry Networks”, IEEE MILCOM 2008, pp.1-9, Nov. 2008.
[5] Peters K., Jabbar, A., Cetinkaya, E.K., Sterbenz, J.P.G., "A geographical
E. Simulation Result on Packet Overhead
routing protocol for highly-dynamic aeronautical networks" , Wireless
The packet overhead of each protocol is shown in Figure 9. Communications and Networking Conference (WCNC), 2011 IEEE , vol.,
In general, the packet overhead increases with the growth in no., pp.492-497, 28-31 March 2011
[6] Medina, D.,Hoffmann, F., Rossetto, F., and Rokitansky, C.-H., "Routing
number of UAVs. AODV is better than GPSR while MPGR in the Airborne Internet", Integrated Communications Navigation and
gets an acceptable performance almost all the time. Surveillance Conference (ICNS), 2010 , vol., no., pp.A7-1-A7-10, 11-13
May 2010
[7] Kothari, A.; Bao-Hong Shen; Tiwari, A.; Ganguli, A.; Shuli Xu;
Krzysiak, D.; , "Performance characterization of ad hoc routing protocols
with mobility awareness", Military Communications Conference, 2010.
MILCOM 2010. IEEE , vol., no., pp.80-85, Oct. 31 2010-Nov. 3 2010
[8] Dong-Woo Seo; Sok-Hyong Kim; Young-Joo Suh; , "System integration
of GPSR and ADS-B for aeronautical ad hoc networks", Military
Communications Conference, 2008. MILCOM 2008. IEEE , vol., no.,
pp.1-6, 16-19 Nov. 2008
[9] SeUng Hyeon; Ki-Il Kim; SangWoo Yang; , "A new geographic routing
protocol for aircraft ad hoc networks," Digital Avionics Systems
Conference (DASC), 2010 IEEE/AIAA 29th , vol., no.,
Figure 9 Average packet overhead
pp.2.E.2-1-2.E.2-8, 3-7 Oct. 2010
At first, consider the curves of GPSR. The packet overhead [10] Xiaoli Ma; Min-Te Sun; Gang Zhao; Xiangqian Liu; , "An Efficient Path
Pruning Algorithm for Geographical Routing in Wireless Networks”,
of default beacon mode is lower than the shorter one. Because a Vehicular Technology, IEEE Transactions on, vol.57, no.4,
shorter beacon period induces the increase of packet overhead. pp.2474-2488, July 2008
As shown in Figure 9(a) and (b), the packet overhead of AODV [11] Bettstetter C, Hartenstein H, and Perez-Costa X. “Stochastic properties of
the random waypoint mobility model: epoch length, direction
is grown when the velocity increases. Because frequent routing distribution, and cell change Rate”, In Proceedings of the Fifth ACM
failure caused by topology change induces much more route International Symposium on Modeling, Analysis and Simulation of
discovery of AODV. Taking into account with packet overhead, Wireless and Mobile Systems, Atlanta, GA, 2002: 7- 14.
[12] “Wireless and Mobility Extensions to the NS-2 Network Simulator.”
MPGR uses on demand neighbor discovery which obviously CMU Monarch Project, http:/monarch.cs.cmu.edu/cmu-ns.html.
reduces the packet overhead as shown in Figure 9(a) and (b). [13] Dan Broyles, et al. “Design and analysis of a 3–D gauss-markov mobility
Moreover, MPGR runs with mobility prediction strategy that model for highly-dynamic airborne networks”, International
Telemetering Conference (ITC) 2010, San Diego, CA October 2010.
high velocity impacts the packet overhead of MPGR slightly.
According to these simulation results shown above, MPGR
can provide more effective and accurate routing for highly
dynamic Ad Hoc UAVs network.

VII. CONCLUSION AND FUTURE WORK


The communication among UAVs requires specific routing
method to cope with the challenges emerged in the highly
dynamic environment. In this paper, a mobility prediction
based geographic routing protocol, MPGR, is proposed for Ad

1602
Authorized licensed use limited to: ULAKBIM UASL ISTANBUL TEKNIK UNIV. Downloaded on November 22,2023 at 12:51:50 UTC from IEEE Xplore. Restrictions apply.

You might also like