Content uploaded by Anas Alkhatieb
Author content
All content in this area was uploaded by Anas Alkhatieb on Aug 22, 2022
Content may be subject to copyright.
XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE
Performance Evaluation
of
Ad-Hoc
Routing
Protocols
in
(F
ANETs)
Anas
AlKhatieb
College of Computer & Information Systems
Umm al Qura University
Makkah, Saudi Arabia
s43880520@st.uqu.edu.sa
Emad Felemban
College of Computer & Information Systems
Umm al Qura University
Makkah, Saudi Arabia
eafelemban@uqu.edu.sa
Atif Naseer
Science and Technology Unit
Umm al Qura University
Makkah, Saudi Arabia
anahmed@uqu.edu.sa
Abstract—The utilization of Unmanned Aerial Vehicles
(UAVs) as aerial relays for the Internet of Drones (IoD)
network has several advantages such as civilian and military
applications. A Flying Ad-Hoc Network (FANETs) is a group
of Unmanned Aerial Vehicles (UAVs) which can complete their
function without human intervention. FANET is considered as
a subset of MANET, however, due to high mobility and rapid
topology changes in FANET applying routing protocols in
FANET is a big challenge. In this paper, we have extensively
evaluated existing Ad-Hoc routing protocols such as OLSR,
AODV, DSR, TORA & GRP for FANET environment. The
performance of those protocols was evaluated using an OPNET
17.5 network simulator. We have compared the protocols using
packet dropped ratio, end to end delay, number of hops and
throughput in different moving speeds and mobility models
such as Random Waypoint Mobility (RWPM), Manhattan
Grid Mobility Model (MGM), Semi-Random Circular
Movement (SCRM) and Pursue Mobility Model (PRS). For all
evaluation scenarios, the results indicate that OLSR and GRP
perform better than AODV, DSR, and TORA on average. This
paper shows that the variation of the network topology caused
by the relative speed of nodes is the main reason for the
fluctuation of network performance. Also, we found that the
(MGM) greatly affects the packet dropped ratio for all
protocols. As we increase mobility speed, we found that End-
to-End delay decreases in MGM, PRS and RWPM, while it is
high in SCRM.
Keywords— FANETs, Mobility Models, Ad-hoc Routing
Protocols, OPNET, UAVs.
I. INTRODUCTION
A Flying Ad-Hoc Network (FANET) is a group of
Unmanned Aerial Vehicles that form an ad-hoc network to
achieve a high-level goal with or without human interactions
[1]. FANET is a special class of Mobile Ad-Hoc Network
(MANET) similar to Vehicular Ad-Hoc Network (VANET),
Table 1 summarizes the differences between the three
different Ad-Hoc Network categories. FANET has become a
very interesting research topic recently due to its wide
application spectrum like disaster check [2], search and
rescue functions [3] , border inspection, fire discovery [4],
relaying networks [5][6], wind valuation [7], civil safety [8],
agricultural function [9], and traffic management [10].
Although FANET shares many similarities with VANET and
MANET, maintaining connectivity and routing in FANET is
more challenging due to its rapid and fast topology dynamics
[11]. Fig. 1 shows the Flying Ad-Hoc Network.
The main motivation of this research is to evaluate
classical Ad-Hoc Networks routing protocols under the
different mobility models of FANET. Simulate the routing
protocols with different mobility models like Random
Waypoint Mobility (RWPM), Manhattan Grid Mobility
Model (MGM), Semi Circular Random Movement (SCRM)
and Pursue Mobility Model (PRS).
Table 1: FANETs vs VANETs and MANETs
Types
Parameters
MANET
VANET
FANET
Node
Speed
Low
Compactness
Medium
Compactness
High
Compactness
Motion
Model
Random
Regular
Regular for
predetermined
paths
Density
of
Node
Low
Very High
Very low
Change of
Topology
Slow
Rapid
Rapid
Radio
Propagation
Model
Close to
ground
Close to
ground
High above
the ground
Energy
Consumption
Energy
effective
Protocols
Not needed
needed for
mini UAVs
Position
based
protocols
GPS
GPS, AGPS,
DGPS
GPS, AGPS,
IMU
Figure 1: Flying Ad-Hoc Network (FANET)
We run extensive simulation experiments to evaluate
FANET performance with multiple mobility models and
different routing protocols including Dynamic Source
Routing (DSR), Temporally Ordered Routing Algorithm
(TORA), Ad-Hoc On-Demand Vector (AODV), Geographic
Routing Protocol (GRP) and Optimized Link State Routing
(OLSR). We analyze the results to find the best protocol for
FANET.
The paper is organized as follows. Section 2 lists the
related work about FANET. In Section 3, a brief description
of the classical Ad-Hoc Networks routing protocols that we
used in the evaluation process is provided. Section 4
contains the extermination and results from the performance
analysis. Section 5 discusses the results. Finally, Section 6
concludes the paper.
II. RELATED WORK
There are few contrast analyses of routing protocols for
high-dynamic scenarios of FANET, and limited different
jobs where MANET routing protocols are applied for
FANET. Here, the performance analysis of routing
protocols is built with the aid of simulation. Actual tasks are
consisted to review characteristics of various types of
protocols, and they are compared that are taken into
consideration. In [12] OLSR routing protocol provided
better packet delivery ratio than other routing protocols in a
high mobility situation. In [13] AODV routing protocol was
higher of packet delivery fraction and the end-to-end delay
than with DSR routing protocol under higher mobility state.
Detailed comparisons were a made in [14] of packet
delivery ratio of AODV routing protocol by varying
mobility models such as RW, RWP, PURSUE and RDIR
mobility model and the packet delivery ratio declined more
sharply in RWP model after that in pursue model.
III. PROTOCOL DESCRIPTIONS
Although many routing protocols have been suggested
for MANET and VANET in the literature, but they cannot
be adapted easily to work on FANET due to dynamic
topology changes and 3D nature of communication. In
[1][15][16], the authors have classified the routing protocols
that can work with FANET into some big categories as
shown in Fig. 2. In this paper, we have worked with
Topology-based routing protocols which are also further
classified into Proactive, Reactive and Hybrid Protocols.
Specifically, we have selected five protocols, namely, DSR,
TORA, AODV, GRP, OLSR to evaluate on FANET
environment. The remaining of this section will give a brief
overview about each protocol and typical mobility models
for networks.
A. Routing Protocols
1) AODV (Ad-Hoc On-Demand Vector)
This protocol [17][16] was born as an evolution of
the DSDV, but in this, it seeks to maintain sequence
numbers and routing tables and add the concept of
routing on demand since as explained in the concept,
they only need to save information from the nodes that
intervene in the data transmission. In comparison with
its previous design, a decrease of the processing time,
decrease of the memory expense and reduction of the
control traffic by the network was achieved; In addition,
this protocol is very careful with the routes, keeping
them in cache while they are necessary and disabling
them when their information is not useful. For a node
that has a fairly recent address path, it is understood that
the node knows a path associated with a destination
number that is large, at least, as that contained in the
RREQ message. In addition, the nodes of the network
that are part of active routes can periodically transmit
special RREP messages, called “Hello” messages, to
their closest nodes. The lack of “Hello” messages from
neighboring nodes is interpreted as a loss of connection
with that node and the node correct its routing table and
eliminate the path.
2) DSR (Dynamic Source Routing)
The DSR protocol [18] is based on the concept of
routing at source, in which the nodes maintain caches,
whose entries contain the destination and the list of
nodes to reach it. It should be noted that updates are
given as new routes are learned and consists of two
main mechanisms that are discovery and route
maintenance.
3) Tora (Temporally Ordered Routing Algorithm)
TORA protocol [12][19] is of type Link Reversal
Routing and aims to maintain a directed graph that does
not contain cycles to reach a destination. This in order
to minimize the load on the network, but with the
impossibility of having to constantly estimate the
distance to the destination or always keep the shortest
route; however, it has the advantage that it is an
efficient protocol since it does not excessively saturate
the network with traffic. If a node needs to know a path
to a destination, it would broadcast a QRY packet that
propagates, until it reaches the recipient node or a node
that has a valid path to the destination. The responding
node will, in turn, be served by a UPD package that will
also add its weight. The UPD packets will be sent in
broadcast so that they allow all the intermediate nodes
to modify their weight conveniently. It is derived,
therefore, that the nodes that want to reach distant
destinations or directly unreachable, increase their local
weight to the maximum value allowed, while the node
that finds a nearby node with a weight that tends to
infinity, will change the path. The CLR (Clear) type
packet intervenes in some cases to reset all address
states of a network portion when the destination is
completely unreachable.
4) GRP (Geographic Routing Protocol)
GRP [16] uses the Global Positioning System to
detect the location of nodes to gather network
information at a source node with a little number of
control overheads. The Source node discovers routes
and constantly send data although the current route is
disconnected. This technique is generally known as a
hybrid routing protocol. Every node determines its own
position and for determining the position of the network
Fig 2. F
ANET Protocol Classification
node the different positioning schemes are used such as
GPS, GPRS etc.
5) OLSR (Optimized Link State Routing Protocol)
The OLSR protocol [16] maintains tables that store
the routing information and periodically, or before any
change in the topology of the network, trigger an update
propagation mechanism through the network, in order
to maintain a real idea of the state of the network. This
can cause a significant amount of signaling packets
(overhead) that affect bandwidth utilization, flow
(throughput), as well as energy consumption. The
advantage is that routes to each destination are always
available without the increment of signaling packets
caused by a route discovery mechanism, but protocol
faces issues like, when the network presents a high
mobility rate or when there are a large number of nodes
in the network.
B. Typical Scenarios and Mobile Models
Mobility is a fundamental feature of FANET. To test the
performance of FANET network, a set of realistic and
application-oriented mobility models should be selected
very carefully. In the literature, different mobility models
that simulate real life scenarios have been suggested. We
have selected four different models, namely, Random
Waypoint (RWPM) [20] [21], Manhattan Grid (MGM) [22],
Semi-Random Circular Mobility Model (SCRM) [23], and
Pursue Mobility Model (PRS) [24].
1) Manhattan Grid Mobility Model
The Manhattan Grid (MGM) mobility model uses a
grid road topology. In this type of mobility model, the
mobile nodes transfer in horizontal or vertical angles on
an urban map. The MG model highlights a probable
approach which a vehicle chooses to keep moving in the
same direction or to turn. This means that they can go
straight and turn left or right.
2) Random Waypoint Mobility Model
This model includes the pause times before changes
the speed and the direction of the nodes. A mobile node
begins transmission by waiting in a location for some
period of time. Once this period finishes, the mobile
nodes select a random position in the defined simulation
area and a speed from a certain range of minimum speed
uniformly and maximum speed [21]. Then mobile nodes
move in the direction of the newly selected location in
the defined area at the selected speed. This procedure is
repeated another time, but before that node takes a break
for a short time period.
3) Pursue Mobility Model
In this kind of mobility model, the mobile nodes
track a target. Which can be calculated newer location by
using the following equation.
Newlocation = oldplace + randomvector
+ acceleration(targetoldplace)
A random vector is an offset for individually mobile
node and acceleration says how mobile nodes are
pursuing towards the target. The degree of randomness
of each node is limited to maintain tracking.
4) Semi-Random Circular Movement Model
Semi-Random Circular Mobility Model is
applicable for collecting some information by turning
around a specific position for simulating UAV’s and this
model is formed for the curved movement scenarios. In
this model, the route is hexagon rather than Random
Waypoint Mobility Model where the plan of flight is not
predetermined. In this model, aircraft are being placed in
different locations wherein a square area chooses the
desired object. The application scenarios are brief to
several types based on nodes movement models, as
shown in TABLE2.
Table 2: MOBILITY MODELS FEASIBILITY FOR FANET
APPLICATION SCENARIOS
Application
Class
Mobility
Model
Scenario Description
Search and
rescue
MGM
SRCM
RWPM
Random exploration of a
definite target zone.
Scanning in a circular area
Each UAV chooses the scan
pattern in a random location
Traffic and
urban
monitoring
MGM
SRCM
Surveillance of city roads
Patrolling of a crash event
before the rescue team reaches
Survey &
patrolling
SRCM
Surveillance of a target
Target tracking
Pursue
Crime tracking
Pursuing of a critical moving
target
SRCM: Semi-Random Circular Movement, PRS: Pursue model;
MGM: Manhattan Grid Mobility, RWPM: Random Waypoint
model
IV. PERFORMANCE EVALUATION
In this paper, our main objective is to evaluate the
performance of five classical Ad-Hoc routing protocols in
FANET scenarios. To do that, we have used OPNET v 17.5
Network Modeler [25] as a simulation tool. The validity and
performance of OPNET have been proven and ratified by
similar research activities. In this study, the same type of
traffic, a constant number of flying nodes and different
speed levels have been used to evaluate the five routing
protocols performance. Table 3. shows the simulation
environment parameters used in OPNET.
Table 3: SIMULATION ENVIRONMENT PARAMETERS
PARAMETER
VALUE
Size
1500x2000 meters
Number of Nodes
15
Address Mode
IPv4
Protocols
DSR, TORA, AODV, GRP and
OLSR
Traffic Type
CBR
Mobility Models
RWPM, MGM, SCRM and
Pursue
Nodes Speed
5,10,20,30,40,50 m/s
Simulation Time
10 Minutes
The simulation is used to evaluate the performance
against Throughput, End-to-End Delay, Packet Hop Count
and Packet Dropped Ratio parameters. Following are the
evaluation of these parameters for different protocols.
A. Maintaining the Integrity of the Specifications
Our simulated results, shown in Fig. 3 (a), (b), (c) and
(d), indicates that by increasing the speed the end-to-end
delay is decreased for MGM, RWMP and Pursue. Regarding
performance under group mobility models, we find that GRP
and OLSR have comparatively lower delays in RWPM,
Manhattan and Pursue. We observed in our analysis that
TORA has a higher delay in SCRM and Pursue as shown in
Fig 3. (a)(d). Finally, it has been found that GRP and OLSR
are better during the delay as compared to AODV then
DSR. When comparing the End-to-End Delay for every of
these routing protocols, DSR had the highest End-to-End
Delay in wholly scenarios. AODV similarly exhibited a
higher delay due to its reactive protocol, where routes were
allotted on-demand. OLSR had good performance in terms
of End-to-End Delay in all scenarios, due to its proactive
characteristics. GRP had the lowest End-to-End Delay
because of the use of the position technique.
(a) PRS
(b) MGM
(a) RWPM
(a) SCRM
Figure 3: End to End Delay
B. THROUGHPUT
Fig. 4 (a), (b), (c) and (d) shows the throughput analysis
of all protocols under different mobility models. The results
indicate that TORA and DSR have comparatively lowest
Throughput in MGM, SCRM and Pursue. Finally, it has
been found that AODV has better throughput as compared
to others.
C. DATA DROP RATIO
It has been observed that the data dropped ratio for
OLSR is higher in various Mobility models. On the other
hand, the data drop is better for DSR in all mobility models.
While it is adequate for TORA and AODN in RWPM,
SCRM and Pursue as shown in Fig. 5 (b), (c) and (d).
(a) PRS
(b) MGM
(a) RWPM (a) SCRM
Figure 4: Throughput Analysis
(a)
PRS
(b)
MGM
(c)
RWPM
(d)
SCRM
Figure 5: Data drop Ratio
D. NUMBER OF HOPS
Fig. 6 (a), (b), (c), (d) shows the analysis after increasing
the no of hops. Results shows that OLSR is lowest in all
mobility models. We have also observed that the Number of
Hops in Pursue model is lowest than other mobility models.
The Number of Hops in AODV is higher than other routing
protocols.
V. DISCUSSION
When analysing the delay variations of the packets
against different mobility and speed, the results obtained for
the OLSR protocol are of particular importance since they
establish fast connections between nodes without significant
delays. On the other hand, the delays experienced in
FANET based on the TORA and DSR protocols are much
greater. Unlike other routing protocols, OLSR does not
spend much time on the route discovery mechanism since
routes are available in advance when data transmission is
needed, resulting in the least delay. Fig. 7, 8, & 9 shows the
comparison of end-to end delay, throughput, and data drop
ratio of all protocols in different mobility models.
(a) PRS
(b) MGM
(a) RWPM
(a) SCRM
Figure 6: No of Hops
Figure 7: End-to-end Delay in Different Mobility Models
Even with a higher speed of the nodes, the performance is
not degraded, and a lower constant delay is observed for the
OLSR protocol. This is because it has the advantage of
using the MPR nodes to enable the forwarding of control
messages to other nodes. Therefore, it eventually helps to
minimize network overload. The proactive routing protocol
(OLSR) and the Reactive protocol (AODV) experiences the
utmost stable performance with all mobility models than
GRP and DSR. TORA shows a high delay in whole
scenarios. This is due to the mechanism of TORA algorithm
which uses the concept of direction of the next destination to
forward the packets. Therefore, the source node continues
one or two downstream paths to the destination node
through multiple middle neighboring nodes. It will be
wasted if the source does not require the route prior to its
invalidation due to topological changes.
Figure 8: Throughput in Different Mobility Models
Figure 9: Data drop ratio in Different Mobility Models
VI. CONCLUSION
The objective of this study is to evaluate the performance
of five classical Ad-Hoc routing protocols in FANET
scenarios using different mobility models. The simulation
results shows that the effect of node mobility on
performance is higher than the effect of node speed on
performance. The Proactive protocol OLSR and the hybrid
protocol GRP are stable, the hybrid protocol TORA is
vulnerable, and the reactive protocol AODV and the
reactive protocol DSR are moderate. Among them OLSR
performs better and is more suitable for highly dynamic
scenarios. Node speed is not a major factor affecting
performance.
Rapid changes in topology caused by high-speed relative
motion of nodes are the main reason for network
performance changes. Topology changes in different ways
in each model. Among the four mobility behaviors, SCRAM
is the most demanding environment in high speed
conditions. In MGM and PRS mobility model, OLSR
outperforms the other two mobility model. These routing
algorithms are suitable to the environment needs to refer to
the application scenario such as (Crime tracking, Pursuing
of a critical moving target and Surveillance of city roads).
When designing routing protocols, it is wise to select
appropriate routing protocols or make appropriate
improvements on existing ones based on specific
requirements.
REFERENCES
[1] M. A. Khan, A. Safi, I. M. Qureshi and I. U. Khan, ”Flying ad-hoc
networks (FANETs):A review of communication architectures, and routing
protocols,” in IEEE First International Conference on Latest trends in
Electrical Engineering and Computing Technology (INTELLECT),
Karachi, Pakistan, 2017.
[2] M. Erdelj, E. Natalizio and K. a. A. I. Chowdhury, “Help from the
sky: Leveraging UAVs for disaster management,” IEEE Pervasive
Comput., vol. 16, p. 2432, 2017.
[3] George, J.; Sujit, P.B.; Sousa, J, “Search strategies for multiple UAV
search and destroy missions,” J. Intell. Robot. Syst., vol. 61, p. 355367,
2011.
[4] Hong, Jie, Dehai Zhang, and Xiaona Niu. , “Impact Analysis of Node
Motion on the Performance of FANET Routing Protocols,” in DEStech
Transactions on Computer Science and Engineering, wicom, 2018.
[5] De Freitas, E.P.; Heimfarth, T.; Netto, I.F.; Lino, C.E.; Pereira, C.E.;
Ferreira, A.M.; Wagner, F.R.; Larsson, T., “UAV relay network to support
WSN connectivity,” In Proceedings of the International Congress on Ultra
Modern Telecommunications and Control Systems, p. 309314, 2010.
[6] Jiang, F.; Swindlehurst, A.L, “Jiang, F.; Swindlehurst, A.L. Dynamic
UAV relay positioning for the ground-to-air uplink,” in In Proceedings of
the IEEE Globecom Workshops, Miami, 2010.
[7] Cho, A.; Kim, J.; Lee, S.; Kee, C., “Wind estimation and airspeed
calibration using a UAV with a single antenna GPS receiver and pitot
tube,” IEEE Trans. Aerosp. Electron. Syst, vol. 47, p. 109117, 2011.
[8] Maza, I.; Caballero, F.; Capitan, J.; Martinez-De-Dios, J.R.; Ollero,
”Experimental results in multi-UAV coordination for disaster management
and civil security applications,” J. Intell. Robot. Syst, no. 61, p. 563585,
2011.
[9] Xiang, H.; Tian, L., “Development of a low-cost agricultural remote
sensing system based on an autonomous unmanned aerial vehicle,” Biosyst.
Eng, no. 2, p. 174190., 2011.
[10] E. Semsch, M. Jakob, D. Pavlicek and M. Pechoucek, “Autonomous
UAV Surveillance in Complex Urban Environments.,” in In Proceedings of
the IEEE/WIC/ACM International Joint Conference on Web Intelligence
and Intelligent Agent Technology, Milan, Italy, 2010.
[11] O. K. S. a. S. lkerBekmezcia, “Flying Ad-Hoc Networks (FANETs):
A survey,” Science Direct, vol. 11, no. 3, pp. 1254-1270, 2013.
[12] S.Mohapatra and P.Kanungo, ”Comparative Performance Analysis of
MANET Routing Protocols Using NS2 Simulator,” in Computational
Intelligence and Information Technology: First International Conference
CIIT, Pune,India, 2011.
[13] A. Tuteja, R. Gujral and S. Thalia, “Comparative Performance
Analysis of DSDV, AODV and DSR Routing Protocols in MANET Using
NS2,” in International Conference on Advances in Computer Engineering,
2010.
[14] S. a. P.Kanung, ”Comparative Performance Analysis of MANET
Routing Protocols Using NS2 Simulator,” in Computational Intelligence
and Information Technology: First International Conference, Pune, India,
2011.
[15] Adnan Nadeem, Turki Alghamdi, Muhammad Shoaib Siddiqui, ”A
Review and Classification of Flying Ad-Hoc Network (FANET) Routing
Strategies,” International Journal of Sciences: Basic and Applied Research
(IJSBAR) , vol. 3, pp. 1-8, 2018.
[16] K.Yadav, A. Nain, S. Maakar and B.Bhiwani, ”Routing Protocols in
FANET: Survey,” in National Conference on Innovative Trends in
Computer Science Engineering, 2015.
[17] Maistrenko, Vasily A., Leonov V. Alexey, and Volkov A. Danil,
“Experimental estimate of using the ant colony optimization algorithm to
solve the routing problem in FANET,” in In Control and Communications
(SIBCON), 2016 International Siberian Conference on, 2016.
[18] Kaur, Parampreet, and A.G. Singh, “Chaining Mobility Models for
AODV and DSR protocols in FANETs,” in PhD diss, 2017.
[19] Park, V., Corson, S, “Temporarily-Ordered Routing Algorithm
(TORA) Version,” Internet draft, IETFMANET working group, 2013.
[Online]. Available: http://tools.ietf.org/html/. [Accessed 04. Accessed 3
Aug 2018].
[20] R. R. Roy, Handbook of Mobile Ad Hoc Networks for Mobility
Models, Springer New York Dordrecht Heidelberg London, 2011.
[21] F. Bai and A. Helmy, “A Survey of Mobility Models in Wireless
Adhoc Network,” in University of Southern California , USA, 2004.
[22] Kaur, Kulwinder, and B. Singh, “Survey analysis of routing protocols
and mobility models in manets,” International Journal of Advanced Science
and Technology, vol. 85 , pp. 55-64, 2015.
[23] K Kumari, S Maakar, B sah, “A Survey: Different Mobility Model for
FANET,” International Journal of Advanced Research in Computer Science
and Software Engineering, 2015.
[24] Vijayavani, G. R., and G. Prema, “Performance comparison of
MANET routing protocols with mobility model derived based on realistic
mobility pattern of mobile nodes,” in 2012 IEEE International Conference
on Advanced Communication Control and Computing Technologies
(ICACCCT), IEEE, 2012.
[25] O. Simulator. [Online]. Available: https://www.riverbed.com/sg
/products/steelcentral/opnet.html.
[26] S.Mohapatra and P.Kanungo, ”Comparative Performance Analysis of
MANET Routing Protocols Using NS2 Simulator,” in Computational
Intelligence and Information Technology: First International Conference,
Pune, India, 2011.
[27] Banoj Kumarpanda, Bibhudatta Dash, Rupanita DasandAjitSarangi,
“Mobility and its impact on Performance of AODV and DSR in Mobile Ad
hoc Network,” in Third Asian Himalayas International Conference on
Internet, 2012.
[28] S H Manjula, C N Abhilash, Shaila K, K R Venugopal and L M
Patnaik, “Performance of AODV Routing Protocol Using Group and Entity
Mobility Models in Wireless Sensor Networks,” in Proceedings of the
International MultiConference of Engineers and Computer Scientists, 2008.
[29] De Freitas, E.P.; Heimfarth, T.; Netto, I.F.; Lino, C.E.; Pereira, C.E.;
Ferreira, A.M.; Wagner, F.R.; Larsson, T., ” UAV relay network to support
WSN connectivity,” In Proceedings of the International Congress on Ultra
Modern Telecommunications and Control S.