Towards Enhancement of Network Communication

Architectures and Routing Protocols for FANETs: A Survey

Sara Al-Emadi, Aisha Al-Mohannadi
Department of Computer Science and Engineering
Qatar University
Doha, Qatar
Email: saraalemadi@[Link], aa1401818@[Link]
Abstract—The popularity of Unmanned Aerial Vehicles (UAVs), also reliable end-to-end transmission of the data remains an active area of
known as drones, has been growing rapidly in the recent years. Their research [5].
applications extend to using a group of UAVs, forming what is known Currently, researchers are investigating the effectiveness of the
as the Flying Ad hoc Network (FANET), to perform a certain mission
in various applications such as search and rescue and various military routing protocols used in the Mobile Ad-hoc Networks (MANETs)
applications and many others. In order for the UAVs to communicate, a and those implemented in the Vehicle Ad-hoc Networks (VANETs).
network communication architecture with the implementation of suitable With the born of UAVs’ communication, MANETs have a sub-field
routing protocols needs to be established while taking into consideration focused on FANETs network communication[6]. Hence, FANETs
the limitation of this system. Therefore, in this paper, we survey in
details the latest state-of-the-art network communication architectures
share some of the MANETs features such as operating in an ad-
and various ad-hoc routing protocols, mainly those falling under the hoc manner, the lack a central control and having a self-organising
proactive, reactive and hybrid routing protocols and their enhanced topology[5]. These features contribute towards extending the commu-
versions which are used currently for FANETs. Furthermore, we propose nication range and expanding the connectivity in an infrastructure-
a unique network communication architecture based on the concept of
less area [5]. Furthermore, FANETs also share features of VANETs,
clustering with the aim of overcoming the limitation in the state-of-the-
art. Moreover, we put forward a new idea of a routing mechanism to aid which they are a subset of, such as the mobility aspect and the
in implementing an efficient and effective routing system for FANETs. dynamic topology change. This relation between MANET, VANET
Index Terms—Ad hoc networks, Flying ad hoc network, FANETs, and FANET is illustrated Figure 1.
Drones, Routing protocols, Network architecture, Reactive routing,
Proactive routing, Dynamic topology, Scalability, Unmanned Aerial Ve-
hicles, Drone ad hoc network, Rapid mobility, Clustering.

I. I NTRODUCTION
Flying Ad-hoc Networks (FANETs) have rapidly spread, where
they are used in many applications such as military sectors[1] and
civil security[2][3]. FANETs can be given a task such that they can be
in the field to operate along with people to perform critical tasks and
make operations easier to achieve. The network consists of multiple
Unmanned Aerial Vehicle (UAV) nodes that communicate using
wireless communication. Using a single UAV instead of multiple ones
can limit the number of applications. While it might be feasible and Fig. 1. FANET vs. VANET vs. MANET
has less complexity in network connectivity, it does not have a large However, unlike MANETs and VANETs, the UAVs in FANETs
coverage compared to the multiple UAVs system. Moreover, single move significantly faster not only in two dimensional but in three
unconnected UAVs take more time in order to complete given task, dimensional space which introduces a great challenge. Furthermore,
since they operate individually [4]. Therefore, FANETs can be very the rapid movement of the UAVs in a FANET’s architecture imposes
helpful in time-critical situations where they operate faster and take the issue of performance degradation in the transmission quality.
less time to finish a given task. In addition, they are compatible with Moreover, a great risk in implementing a system with UAVs is the
wide range of applications. However, they also might have a complex fact that they can crash, hence, maintaining an active link in a FANET
network topology as well as power consumption limitations [4]. is challenging [7].
One of the most common applications in FANETs are used for Although FANETs are a subset of VANETs due to their mobility
are military and civil applications. They are widely used in searching characteristics, it is worth mentioning that the node mobility in a
and rescuing operations, as they can operate faster since they have FANET is much higher than in VANETs causing an increase of
large coverage area, which makes the task of finding potential targets difficulty and complexity in the design of the network communication
easier[1]. Furthermore, civil applications also include surveillance for FANETs [8].
and civil security. These applications depend on extreme security, In a similar vein, computational power plays a crucial role in
where any interruption might cause major damage to the public [2]. designing the communication system for ad-hoc networks. Another
FANETs are known for their dynamic nature in which drones are difference between FANETs and other ad-hoc networks is the fact
flying with different speeds and in some cases in different directions. that it has very limited power [6]. Unlike VANETs, where power
In these types of topologies, it is very challenging to implement the consumption for communication is not a concern since the power
conventional routing protocols. required is very minor in comparison to the overall vehicle’s battery,
The routing protocols are necessary in FANETs in order to estab- the power consumption is a significant concern in UAVs [9]. This is
lish the UAV-to-UAV communication and finding the ultimate routing due to the fact that the batteries used to power UAV are considerably
protocol which has the lowest routing overhead while providing a small. They typically last up to approximately 28 minutes of flight

978-1-7281-8704-4/20/$31.00 ©2020 IEEE

Authorized licensed use limited to: UNIVERSITY OF ROCHESTER. Downloaded on September 22,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
for commercial drones such as DJI Phantom 4 [10]. Therefore, power
consumption is a great factor to be taken into consideration when
designing a routing protocol for FANETs.
Hence, the main aim of this study is to trace the latest development
in the network communication architectures and routing protocols that
are designed in the literature for FANETs. Being inspired by the state
of the art, we propose a new network communication architecture and
routing protocol, which will overcome a number of shortcomings
faced with the exists techniques.
The main contributions of this work can be summarised as follows: Backbone UAV

• Provide a comprehensive literature survey on the latest and the

state of the art FANETs’ network architectures
• Discuss and compare the different state of the art routing
protocols used in FANETs
• Propose a new network communication architecture for FANETs
• Provide a recommendation for the most effective routing proto-
cols which can be used in FANETs. Base Station
The rest of this paper is distributed as follows: in Section II,
we explore various FANETs network architectures. Moving forward, Fig. 2. UAV Ad hoc Network
Section III provides a thorough discussion on the current state of the
art routing protocols used in FANETs. In Section IV, we discuss useful in application where large number of UAVs with different flight
the challenges of FANETs. Whereas, in Section V, we discuss paths and communication characteristics are required in a mission.
our proposed solution which consist of both, the optimal routing Although the authors have discussed the advantages of implement-
mechanism along with the design of a new network architecture for ing these methods, we believe that it can have a large overhead due to
FANETs. Finally, we conclude this paper in Section VI. the fact that if a communication between the two groups is required, it
would have to be redirected through the base station. Furthermore, as
II. C OMMUNICATION A RCHITECTURES with the first topology, there is still a single point of failure between
the gateway drones in each group and the base station.
As introduced previously, the dynamic nature of FANETs imposes
great challenges while finding the ultimate network architecture. The
design of the communication architecture varies depending on factors
such as the number of drones required to achieve certain mission,
the distance between drones, the total area of coverage and the delay
tolerance of the application.
In [5], the authors have discussed three interesting topologies that Backbone UAV Backbone UAV

are being used in FANETs, those three topologies were adapted from
another study [11]. Starting with the first topology, referred to as UAV
Ad-hoc network, illustrated in Figure 2. This topology is the most
common topology proposed in various of other studies due to its Base Station

simple design. The UAV Ad-hoc network is made up of a number of

drones with an ongoing internal connection and have a connection Fig. 3. Multi-Group UAV Ad hoc Network
with a backbone drone. Hence, in order to communicate the data The third communication architecture proposed in [11] is the
outside the UAV network, the backbone drone would act as a gateway Multi-Layer UAV Ad hoc Network. This topology is made up of
between the other UAVs and the base station. Hence, the gateway multiple layers; the first layer is the communication between the
drone will forward the data from the designated drone to the base UAVs within an individual group, the second layer is made up of
station. The advantage of this topology is the fact that it can expand the communication between the backbone drones and the last layer
the communication range significantly. However, for this architecture is between one of the backbone drones and the base station. Basically,
to be feasible, the mobility patterns such as, the speed and direction, each group of heterogeneous drones has a backbone drone. All
need to be the same for all connected UAVs in a FANET. Therefore, backbone drones are connected together. Finally, a single backbone
it is considered a suitable option for small sized FANETs. drone reports to the base station. This topology design reduces the
From this topology, one can observe that having a single drone communication load and computation on the ground station given
to act as a gateway leads to having a single point of failure in that the ground station only processes the information it is destined
the network. Additionally, the battery of the gateway drone will be to, which is received through only one backbone drone. The authors
drained very quickly. in [11] and [5] mentioned that this architecture design eliminates
Another topology discussed in [5] is shown in Figure 3. This the single point of failure that is present in Multi-Group UAV Ad-
topology is referred to as, Multi-Group UAV Ad hoc Network, where hoc Network and UAV Ad-hoc network, hence, it is considered a
it uses both techniques; centralized and ad-hoc. As it can be observed robust design. However, it can be clearly observed from Figure 4
in the figure, two drones, each with a connection to a different drone that the single backbone drone communicating with the base station
group, are connected to a base station separately with the intra- acts as a single point of failure in the network. That is, if the link
connection being independent of the base station. Whereas, the inter- between the communicating drone and the base station goes down,
connection is with base station. This communication architecture is the whole network will not be able to communicate with the base

Authorized licensed use limited to: UNIVERSITY OF ROCHESTER. Downloaded on September 22,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
station. Hence, the same phenomenon of single point of failure would (OLSR) and another type for distance vector that is Destination-
exist. Therefore, we propose in this paper, an new architecture which Sequenced Distance Vector (DSDV). It is based on Bellman-Ford
eliminates the single point of failure by assigning the gateway task algorithm and both protocols are investigated from the literature to
to multiple backup drones. Hence, the communication with the base- be able to provide choices for FANET networks.
station will be shifted during different time intervals. This solution 1) Optimized Link State Routing (OLSR): OLSR is mainly about
is further discussed in Section V. optimizing the traditional link state protocol (LS), where it uses
a unique feature that introduces Multi-point Relay (MPR), where
MPR nodes are responsible for forwarding routing information.
Furthermore, it makes use of two types of control messages in its
routing[15]:
1) Hello messages: these messages are responsible for completing
Backbone UAV Backbone UAV
link detection, as well as neighbor node detection and MPR
selection[15].
2) Topology Control (TC): it holds the topology information of
the network[15].
Moreover, the selected nodes as MPR, by neighbor nodes broadcast
Base Station
this information periodically in their control messages. Thus, when
data packets are needed to be transmitted, it is found directly in the
Fig. 4. Multi-Layer UAV Ad hoc Network
table. This leads to reduced data transmission delay, since there is no
III. ROUTING P ROTOCOLS route discovery[16].
A study in [17] showed a simulation for the traditional OLSR
In this paper, we address multiple applications in the literature in
routing protocol, where they focused on different metrics such as,
order to analyze and find an optimal routing protocol to be used in
UAV density, transmission range, and node mobility. During the
FANETs. Figure 5 shows the types of routing protocols discussed
investigation, authors claimed that tracking the dropped packets and
in this section particularly they fall under the proactive, reactive,
packet delivery ratio would help in obtaining the efficiency of the
and hybrid routing types. Furthermore, we surveyed many different
three mentioned metrics. Thus, a simulation was done in NS2 simu-
literature work that performs an evaluation analysis as shown in Table
lator with varying number of UAVs and a range of distances where
I
packet delivery ratio was examined. Authors used mobility scenarios
Each routing protocol has different characteristics with numerous
existed in NS2, where they defined the scenario for transmission
metrics that can be used to evaluate its performance, and in this
range as 10, 30, 50, 80, and 100 number of UAVs moving around
paper we will be examining the routing protocols using the following
an area of 3600x1200m2 for 300 sec time and a varying speed.
metrics:
Finally, the performance was evaluated where they got the packet
• Throughput: this metric is evaluated by averaging the number delivery ratio to be bigger than 95% in OLSR when UAV number
of bits arrived per second at the destination node [12]. is 50 and transmission range is 600m. In addition, it was proved
• End-to-End Delay: measures all delays that occurred in the that with higher denisty OLSR can have higher packet delivery ratio,
network [12]. since it has the advantage of handling higher density networks better
• Packet Delivery Ratio (PDR): measures how successful the comparing to other routing protocols.
protocol was in performing the delivery from source to destina- Another study was made by authors in [13] regarding traffic
tion[12]. monitoring using two routing protocols namely, OLSR and AODV.
• Routing Overhead (RO): measures the ratio of the control The study was done on NS2 simulator in a scenario of traffic
messages over time [13]. monitoring, where UAVs are supposed to be connected while tracking
A. Proactive Routing Protocols MANETs on the ground. The simulation for OLSR routing protocol
was done in similar environment of that in paper [17], specifically,
Proactive routing protocols also are called table-driven, it focuses it was done in an area of 3600x1200m2 for 1800s time and varying
on managing the routing information at each node, where all in- speed of 0-60m/s with different metrics such as, packet delivery ratio,
formation is updated using broadcast messages to check for any throughput, end-to-end delay, and routing overhead. Evaluation for
changes in the topology[14] and the routing protocol consists of this study showed that PDR decreases significantly in OLSR when
one or multiple tables. Thus, this routing protocol has the benefit of increasing the speed. However, there is slight increase of end-to-end
having routes directly accessible when needed. On the other hand, it delay in OLSR when raising speed comparing to the other protocol.
has one limitation where it suffers from additional overhead cost due The paper also showed that OLSR goes through a high load due to
to maintaining up-to-date information, as a result, throughput may TC and Hello messages generation.
be affected since there will be redundant messages sent out even In order to tackle these limitations in OLSR routing protocol,
when there is no data traffic [5]. Hence, it can be slow to reacting another communication technique was implemented by authors in
on failures and may not be suitable for large network topologies. [16] focusing on the enhancement of the OLSR routing protocol
This routing protocol includes two types of protocols: specifically on node link expiration and residual energy. It was
• Link state: it depends on the knowledge of the whole topology developed to accommodate topology changes and avoid interruption
and computing the shortest path. of communication. The authors also pointed out that the main idea
• Distance vector: it focuses on the knowledge of the cost to every was about making use of GPS information in order to calculate
destination. node link expiration time along with taking into account the residual
Two types of PRP will be discussed in this paper, one type is for energy. Therefore, they proposed a new solution for ocean FANETS
the link-state routing protocol which is Optimized Link State Routing with OLSR protocol using the improved multi metric Expected

Authorized licensed use limited to: UNIVERSITY OF ROCHESTER. Downloaded on September 22,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
 Fig. 5. Routing protocols diagram

Transmission Count (ETX). They managed to apply it by increasing disadvantages and limitations, however, these limitations were tackled
packet delivery ratio as well as reduce end-to-end delay and routing and appeared to be improvable in many metrics.
overhead, to ensure adaption to changes of network topology more 2) Destination-Sequenced Distance Vector (DSDV): DSDV rout-
efficiently. The authors mentioned as well that they aimed for ing protocol contains a table that consists of the next hop neighbor as
improving the routing algorithm to choose a stable Multi-Point Relay well as the distance to the destination specifying its cost. Typically,
(MPR) with a new ETX metric and residual energy nodes. They the tables are updated by nodes broadcasting their tables to their
simulated the solution using NS3 to verify the efficiency of the neighbors. One advantage of this protocol is that it assures loop free
improved protocol, where they found out that end-to-end delay in routes going to each destination, as well as it uses the settling delay
the proposed method has become less than the traditional protocol that is the delay before advertising the route in order to find the
and that is because of the selection of MPR node that is more stable. optimal path. [12].
Moreover, in case of packet delivery ratio, it was lower than the one One study was made by authors in [12] to simulate and evaluate the
in the traditional protocol, and this is due to taking into account the performance of DSDV and AODV. This study was simulated using
energy which is not the case in the traditional protocol. Furthermore, NS2 simulator and evaluated using different metrics mentioned in
they discussed their results in routing overhead which was less in section III. Authors showed that DSDV has good PDR ranging from
the improved method than the one in the traditional protocol, this is 97.7% to 95.7%. In addition, DSDV outperformed the other protocol
also due to the above mentioned unique features that the improved in end-to-end delay with slight increase due to its fast features in
solution has. small networks specifically. The DSDV routing protocol also showed
Another possible enhanced solution for the OLSR routing proto- an energy consumption of 14.2J.
col’s limitations was done by authors in [15]. This study focused Another approach was implemented by authors in [19], where
on cross-layer design, where the proposed solution counts the link they simulated and analyzed different routing protocols including
quality information directly using the physical layer, while the DSDV, AODV, DSR, OLSR, and TORA. They stated that one of
network layer calls the information to help in optimizing the routing. the limitations for DSDV is that it can not be used in multi-path as
This showed reduction of load on the protocol, where simulation of well as it faces some complications in large networks where it cannot
the results was implemented in OMNeT++ platform. The enhanced maintain the route table for it. The simulation was done on NS2 and
paper which uses C-OLSR showed a higher throughput comparing to the evaluation showed that in throughput DSDV works better than
original routing protocol, as well as the end-to-end delay was reduced most protocols, while in packet delivery ratio DSDV works better
more than half comparing to the one in the literature. than all other protocols.
Authors in [18] examined a new technique for enhancing OLSR There is a lack of enhancement studies for this particular routing
routing protocol by using directional OLSR that concentrates on protocol and this is due to the fact that DSDV meets the requirements
minimizing the MPRs by selecting number of MPRs, which would in most applications in FANETs and does not need to be enhanced
improve the performance of OLSR in terms of end-to-end delay and in performance.
routing overhead. They also compared the enhanced solution with
the original one along with other routing protocols such as, DSR B. Reactive Routing Protocols
and AODV. Their simulation was done on OPNET tool and they Unlike the proactive routing protocols, the Reactive Routing Proto-
tackled a network with 25 nodes. According to their results, DOLSR cols (RRP) only discover and maintain the routing tables of the node
outperformed the original protocol in end-to-end delay significantly. to which data needs to be sent. Therefore, since not all the nodes
This indicates that flood of control messages have significant impact in a network will have all of the information of every other node
on the packet delivery. Moreover, DOLSR outperformed all other in the network, this family of routing protocols are considered more
routing protocols in terms of end-to-end delay. efficient than the proactive routing protocol as it has less overhead,
Given the above improvements, it showed that OLSR has its hence, they are bandwidth efficient [5].

Authorized licensed use limited to: UNIVERSITY OF ROCHESTER. Downloaded on September 22,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
 TABLE I
S UMMARY OF ROUTING P ROTOCOLS A PPLICATION PAPER FOR FANET S

Ref. Protocols Throughput E2ED Overhead (PDR) Size

[9] AODV, DSDV, DSR, OLSR, AOMDV, HWMP X X - X 20 nodes
[19] AODV, DSDR, DSR, TORA, OLSR X X - X 6 nodes
[20] OLSR, AODV, TORA - X X X -
[17] AODV, OLSR - - - X 10, 30, 50, 80, 100 nodes
[13] AODV, OLSR X X X X 50 nodes
[12] AODV, DSDV X X - X 4 nodes
[21] AODV, ZRP, AODVDR - X - X 50 nodes
[16] OLSR-ETX, Enhanced OLSR ETX - X X X 30 nodes
[15] OLSR-ETX, C-OLSR X X - - 20 nodes
[18] OLSR, DOLSR, AODV, DSR - X X - 25 nodes
[22] AODV, ZRP, NA-ZRP, ZRP-B, DSDV - X X X 200 nodes

A limitation of the reactive routing protocol is that it consumes In [18], the authors have performed an experimental comparison
time to find and establish the optimal route between the sending on the implementation of different routing protocols for FANETs.
node to the receiving node which contributes to high latency [23][5]. It was concluded from their analysis, that DSR suffers from a larger
The following are two main techniques used in transmitting packets end-to-end delay in comparison to OLSR and their proposed method.
using RRP [23]: Furthermore, they claim that such behavior is due to the fact that DSR
1) Source Routing: in which the source to destination address is is a reactive routing protocol which construct its routing table on-
included in the packet header. Once the packet travels across demand. Moreover, the end-to-end delay could have been the result of
the intermediate nodes, the intermediate nodes will forward the the high flooding of the traffic control packets, which are preformed
packet to the next appropriate node based on this information. by DSR that also could contribute in the end-to-end delay and the
2) Hop-to-Hop routing: in this routing mechanism, the packet decrease in the amount of packets received over the same time interval
holds the information of the destination as well as the next in comparison to OLSR and their proposed method.
hop. Hence, to send the packet to the destination, each node In another study [25], the authors have conducted an experimental
will maintain a routing table. study between DSR, OLSR, AODV and ZRP to find the best routing
Furthermore, a node that uses the RRP protocol initialises the protocol for FANETs. In their simulation, they have considered
communication and sends the packet that uses two types of messages: scalability of the FANET, speed of UAVs movement and the overall
• Route Request Message (RREQ): used when a node want to area which the drones occupy. They have concluded that DSR had the
send a packet to the destination. The RREQ message includes highest throughput with the UAV network had 30 nodes regardless
the initiator, destination and request ids and the route record of the change in the speed of the moving UAVs. Hence, the authors
which are identified initially by the source[24]. have concluded that DSR is the most suitable routing protocol among
• Route Reply Message: the destination node sends a route reply the others in a highly mobile and dynamic network. However, it
using the reverse of the route used in receiving the route is important to note that DSR yielded the highest network load
request[24]. when the speed of the UAVs was high. The authors argued that this
In this paper, we will discuss the most popular types of the RRP phenomenon was due to the reactive nature of the routing protocol.
used for the communication between the UAVs in a FANET namely On a similar note, it was deduced that as the number of nodes increase
the Dynamic Source Routing (DSR) and the Ad-hoc On-Demand to 100 nodes, DSR did not perform as well and was outperformed
Distance Vector (AODV). Moreover, we will dive into the latest by AODV. Furthermore, DSR had the highest delay when the area
work in the literature that enhances those two routing protocols to occupied by the drones was larger than 1km x 1km. Therefore,
specifically suit the requirements of a FANET network architecture. the authors suggest that in highly dynamic network where the path
1) Dynamic Source Routing (DSR): DSR is used heavily in multi- discovery would occur repeatedly, it is not recommended to use
hop mesh ad-hoc network applications[23]. This routing protocol DSR. Hence, one can deduce that DSR is efficient and is considered
supports the feature of self-organising and self-discovering and the optimal routing protocol out of the other tested in the study of
configuring which is made possible given the fact that it does not [25] when FANET is relatively small, of around 30 drones, and the
retain the routing tables and establishes the route between the source coverage area is less than 1km x 1km. Furthermore, DSR is not
and the destination using the two types of messages discussed above. a recommended routing protocol since the topology changes very
However, since nodes in DSR based networks use the caches to store frequently and if any drone in a FANET fails, it will cause a path
the single route between the source node and the destination, an distribution [9].
extra storage space should be considered when using this network. 2) Ad Hoc On-Demand Distance Vector (AODV): AODV routing
Furthermore, having outdated route information in the cache might protocol was designed in an attempt of improving on DSR. Similar
result in the source node having to attempt sending multiple times to DSR, AODV is a multi-hop reactive routing protocol with an on-
before finding an alive route. demand route discovery characteristic. The main difference between

Authorized licensed use limited to: UNIVERSITY OF ROCHESTER. Downloaded on September 22,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
the two reactive routing protocols is that DSR embeds the source used in FANETs.
routes in the packet headers, as discussed in Section III-B1 above,
C. Hybrid Routing Protocols
which leads to having a large packet’s header causing a degradation
in performance. To overcome this limitation, in a network where Hybrid Routing Protocol (HRP) is mostly a combination of
AODV is being implemented, intermediate nodes would maintain proactive and reactive routing protocols. This technique takes the
routing tables such that the data packets transmitted would not include best characteristics of each type and combine it to overcome the
any of the source route information. In addition, the intermediate limitations. The reactive routing is known for the extra delayed time
nodes’ function would extend to not only forwards the RREQ but they to discover routes, while proactive routing faces a large overhead
would have also to keep their routing table updated with the latest of control messages. These limitations are tackled by implementing
source information [23]. Therefore, this protocol uses the advantages both proactive and reactive routing. Furthermore, HRP work better
of the reactive routing protocol and the proactive routing protocols. in large topology as well as it is divided into two zones[5]:
The advantages of using AODV in FANETs is that it is capable of 1) Intra-zone: applied using proactive routing
providing high mobility rates with various data traffic levels, yet, it is 2) Inter-zone: applied using reactive routing
considered energy efficient and has a low data traffic overhead [12]. In this study, two routing protocols from HRP are investigated,
However, due to the dynamic nature of FANETs, AODV suffers with which are Zone Routing Protocol (ZRP) and Temporarily Ordered
high network congestion [25]. Routing Algorithm (TORA).
In [25], the authors deduced that when it comes to the highest 1) Zone Routing Protocols (ZRP): ZRP is known for using differ-
throughput rate while scaling up the network from 30 UAVs to 100 ent techniques in variety of zones. The protocol divides the network
UAVs with speeds like 30m/s and 60m/s, enhanced AODV, TSODR, into number of zones, where each zone has multiple nodes. Moreover,
had an significantly outstanding performance in comparison to DSR, it uses proactive routing protocols in single clusters, whereas it uses
OLSR and ZRP. The authors have conluded that such performance reactive routing protocols between different clusters [29].
was related to the fact that TSODR have time-slotted for each packet A study in [21] was made to implement and simulate ZRP and
transmission to avoid network congestion. However, it suffered from compare it along with other routing protocols. The authors used
high delay rates. Furthermore, it had shown a consistent network load two different metrics to evaluate the protocols which are the end-
regardless of the flying speed and the area occupied by the drones. to-end delay as well as the packet delivery ratio. The simulation was
To observe the performance of AODV on MANETs, the authors done on NS2, where they explained the environment of the MANET
in [19] have performed a performance analysis of AODV, DSDR, implementation was an area of 1000m x 1000m and speed ranging
DSR, OLSR and TORA in a simulation environment. In a network from 0-20m/s with number of nodes as 50. Evaluation showed that
which is made up of 6 nodes, they have deduced that AODV had the ZRP outperformed other routing protocols in terms of end-to-end
best throughput rate among all the other routing protocols used in delay, where it succeeded in having the lowest delay time when the
this study. Although this finding does not align with the findings of number of connection nodes in the network is less than or equal to
[25], in which they have deduced that in smaller FANET network 10. Also, ZRP showed the same results in terms of packet delivery
DSR would outperform AODV and AODV would have the best ratio, this indicates that ZRP works better when there is less nodes
performance when the network scales significantly, the difference in connections in the network comparing to other routing protocols.
the results could be attributed to the fact that FANETs are much more By addressing these limitations, an enhanced solution was de-
dynamic and the mobility is much higher than in MANETs. veloped by authors in [29] to overcome some of ZRP limitations.
Authors in the ad-hoc network field have been working towards Authors stated that their proposed solution focused on the drawback
enhancing the conventional AODV such that it would overcome the of ZRP feature that loses messages and wastes time by sending
shortcoming of the existing AODV to better suit the needs of different unwanted messages due to the fact that these limitations come from
types of ad-hoc networks. Two interesting enhancement on AODV proactive and reactive routing. Thus, the solution consists of an added
were discussed in [26], the first being the time slotted ADOV, or flag in the protocol that checks the importance of the incoming
Time-Slotted On-Demand Routing Protocol (TSODR) which was messages and filter according to the value. The simulation was done
proposed in [27] explicitly for FANETs and Unmanned vehicle using OPNET tool and it was evaluated using three metrics namely,
systems within MANETs. In TSODR, a single node can send a data throughput, end-to-end delay, and bandwidth utilization. In addition,
packet at a given time slot to overcome the packet collision and results showed rapid increase in throughput for the enhanced design
increase the packet delivery ratio. However, the limitation of using of the ZRP protocol comparing to the original. Also, an enhancement
this protocol is that it increases the bandwidth utilisation. The second was shown in terms of delay where it was mostly stable in the original
routing protocol discussed which enhances AODV was Reactive- protocol and in the improved one, it had lower delay and it was
Greedy-Reactive (RGR) which was proposed and studied in [28]. The decreasing after specific time. Finally, it was accomplished that the
uniqueness of this protocol is that it combines between the Greedy bandwidth utilization was better in the enhanced solution than in the
Geographic Forwarding and the reactive routing features of AODV. original one.
RGR transmits the data packet which also includes the location Another improvement was implemented by authors in[22], where
information of the UAV through the network through the same they concentrated on the network overhead limitation. The authors
technique as in AODV. However, if for any reason the reactive route proposed a solution that includes a multi-center intra-domain routing
broke, then RGR will continue the transmission through switching to area that uses the neighbor connection degree. As for the inter-domain
the greedy geographic section of the protocol. Hence, the authors in technique, it uses an adaptive routing that is based on link quality.
[28] have concluded that RGR had better performance in comparison The paper simulated the improved solution using NS2, where it used
to the AODV in terms of packet delivery ratio, delay while the a scene of 2000x2000m2 and a number of nodes equal to 200 and
overhead is close to the conventional AODV. a speed ranging from 5-25m/s. The performance of the enhanced
These enhanced versions of AODV gives hope to the researcher in solution was evaluated regarding the large network and the high speed
the field to find the optimal reactive routing protocol which can be movement, and it was illustrated that the enhanced solution showed

Authorized licensed use limited to: UNIVERSITY OF ROCHESTER. Downloaded on September 22,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
 Cluster 1 Cluster 2

Alternative Cluster Alternative Cluster

Head (ACH) Head (ACH)

Cluster Head (CH) Cluster Head (CH)

Base Station

Fig. 6. Proposed Communication Architecture. The drone circled in red represents the current nominated CH. Whereas, the drone circled in orange represent
the drone elected as the ACH.

better performance than other protocols in terms of dealing with high two routing protocols. Furthermore, the packet delivery ratio was
speed in delay as well as delivery rate. Also, the improvement in ZRP the lowest and the routing overhead was average. However, in a low
showed that it can adapt well in large-scale high-speed scenarios, mobility scenario, the end-to-end delay was low with high PDR while
where there is rapid change in topology and and a large number of the routing overhead remained average. The authors have concluded
service transmission that can overcome this routing overhead. that TORA has a good performance in a dense network. The findings
Two enhancement scenarios were shown in ZRP protocol that of this study suggest that TORA might not be the ultimate routing
can approve that using hybrid routing protocol is more adaptive to protocol for implementation in FANETs.
large networks and can have better performance after applying some
improvements on the limitations.
IV. C HALLENGES
2) Temporarily Ordered Routing Algorithm (TORA): The Tem-
porarily Order Routing Algorithm (TORA), also known as link This paper discussed multiple routing protocols, as it showed vari-
reversal protocol, is designed based on the features of both the ety of applications and how the literature tackled their shortcomings.
reactive and the proactive routing protocols [26]. It is a multi-hop With that being said, there are still some limitations and challenges
protocol where nodes maintain only information about the adjacent that needs to be addressed in future studies to be able to achieve the
nodes [23]. The main technique used in this algorithm is that it optimal routing protocol for FANETs. Some challenges are discussed
constructs and maintains a directed acyclic graph (DAG) between below,
the source and destination nodes where the path preferences does
not depend on the number of hops. This is due to the fact that • Scalability: one of the main challenges faced when designing
sometimes longer paths are preferred to overcome the issue associated an optimal routing protocol is the adaption of scalability, where
with large network overhead [26]. Furthermore, each node might have changing number of nodes has a strong effect on adaptation to
multiple outgoing edges and hence it might have multiple routes to the topology change and load coming from packets.
destination. This redundancy is considered an advantage of using this • Speed: all routing protocols have their unique features and most
routing protocol since in case of a failure in one of the active links, of them suffer when there is a rapid speed change from UAVs
another active link will instantly take the lead in transmitting the within a FANET.
packet [20]. One of the main characteristics of this routing protocol • Altitude: altitude of UAVs needs to be taken into account
is that it is loop-free and it uses the on-demand feature as in reactive since most routing protocols did not tackle this feature in their
routing protocol. That is, the node will only send a query request once simulation, as the height of the UAVs in a given environment
it has a packet to be sent to the destination node. Therefore, reducing might differ in some scenarios. Thus, it would have changes in
the bandwidth utilization and overhead[20]. However, a limitation of performance.
this routing protocol is that it requires clock synchronisation between • Mobility: FANETs’ routing protocols depend on the mobility
all nodes in the network and in case of the dynamic nature of of UAVs as well, obtaining a high mobility feature in routing
FANETs, this could be challenging to achieve. protocols is a challenge, since there is constant movement in a
network with varying in speed.
To better understand the capability of TORA on MANETs, the
authors in [20] have performed a study on the performance of of There are some challenges as well in applying such conventional
AODV, OLSR and TORA. In a network which is made up of 6 nodes, protocols in FANETs which are not faced in VANETs, given that
they have deduced that in a high mobility networks the End-to-End VANETs operate only in 2D and FANETs work in 3D. These
delay is very high when using TORA in comparison to the other movements make changes in altitude which effects the performance.

Authorized licensed use limited to: UNIVERSITY OF ROCHESTER. Downloaded on September 22,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
 V. P ROPOSED S OLUTION
Flying mission begins

A. Proposed Communication Architecture

Intra-connection establishment
As observed in Section II, many of the currently used topologies between drones within a cluster
suffer from having a single point of failure in the network as well
as not considering the power efficiency and the limited battery life
Assignment of each CH in each
available in the UAV. These are constraint restrict the design of the cluster based on its current
communication architectures for FANETs. Therefore, we propose capabilities
in this paper, a new network communication architecture which
eliminate the issue associated with a single point of failure by using
ACH is selected based on the
cluster heads (CH) and alternative cluster heads (ACH) as illustrated same parameters
in Figure 6. Initially, the intra-connection will be established between
the drones in each cluster. Then, each cluster will assign a CH
while taking into consideration its current capabilities, such as, the Each CH maintains an on-going
connection with the BS
size of the battery, the signal strength and the available battery life.
Moreover, an ACH will be selected based on similar parameters. It is
important to note that each CH will have an on-going communication An active inter-connection
with the BS, however, after a specific time slot or when the battery No
between the different clusters is
established when required
level falls under a certain threshold, the current CH will shift its task
as a gateway between the cluster and the BS to the ACH, making
the ACH a CH. The next ACH will be nominated based on the Is
the battery level
parameters mentioned previously. Furthermore, there will be an active
of CH under the predefined
inter-connection between the different clusters. Figure 7 illustrates the threshold value?
breakdown, in steps, of this process. No
This approach will ensure the following: Yes

1) Eliminate a single point of failure: if a drone assigned as CH Current CH shifts its task as a
gateway to the ACH, making the
fails, shut down or the link between the CH and the BS breaks, current ACH a CH
there is always an alternative drone in the cluster that will take
the role of CH immediately, hence, ensuring redundancy.
2) Ensures that the responsibility is shared among all drones in A new ACH is selected based on
the chosen parameter
a cluster. That is, there will not be a burden and extra battery
usage on a specific drone. Therefore, the drones in the cluster
will hopefully live longer without being depleted.
Is mission complete?
3) Having a connection between the two clusters which does
not have to go through the BS means that there is an extra
redundancy in the network. For example, in case Cluster 1 is Yes

not able to establish the connection to the BS, it can re-route the End
traffic through Cluster 2. This is very efficient in case Cluster 1
was performing an operation away from the BS where Cluster 2
will act as intermediate median for the communication between Fig. 7. Proposed Communication Architecture Breakdown
Cluster 1 and the BS. B. Proposed Routing Mechanism
4) This FANET architecture is expected to perform outstandingly
in a larger network with many clusters as it will increase In Section V-A, we proposed a new network communication archi-
redundancy and the inter-connection link between the clusters. tecture which is, theoretically, able to overcome the limitation of the
conventional network architecture for FANETs discussed in Section
With having the advantage of extra redundancy comes the costs II. In this section, we will propose the routing protocol solution which
of expect network overhead, extra computation, and storage require- will ultimately provide the most efficient communication median
ments. This is due to the fact that in order to have an active ACH for the routers in FANETs without sacrificing in the quality nor
that is ready to take over the role of the CH at any instance, the jeopardize the bandwidth for transmitting the control packets.
ACH would have to have updated information of the network which Figure 8 illustrates the extended network architecture which has
are identical to those in the CH. However, storage is not an issue in been discussed in the previous section, each UAV in each cluster
drones as the recent development enabled them to be equipped with has an inter-connection to the adjacent UAVs in the same cluster. In
extra storage capacity when needed. Whereas, the network overhead Figure 8, these links are represented in blue. Based on the findings
remains a challenge. This is due to the fact that for the drones of the literature review conducted in this paper, we propose the
to re-elect the next ACH, information as those mentioned above usage of ZRP protocol [30] for this connection. ZRP will provide
would need to be exchanged between the drones to nominate the the functionality of both the proactive and reactive routing protocols
next ACH. Furthermore, the new CH would need to establish the [31]. The proactive nature will occur with the drones within the zone
communications with both the BS and the other CHs which increases of the drone, that is, within the cluster in this scenario. Whereas, if
the network overhead. the drone is outside the cluster, it will use the reactive functionality

Authorized licensed use limited to: UNIVERSITY OF ROCHESTER. Downloaded on September 22,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
 Cluster 1 Cluster 2 Cluster 3

Alternative Cluster
Head (ACH)

Alternative Cluster Alternative Cluster

Head (ACH) Head (ACH)

Cluster Head (CH)

Cluster Head (CH) Cluster Head (CH)

Base Station

Fig. 8. Proposed Routing Mechanism. The blue lines represent the inter-connection links between a single cluster. Where the yellow lines illustrates the links
between each CH and the adjacent CH and the green lines resembles the communication lines between the CH and the BS.

to find the designated drone which it needs to communicate with. R EFERENCES

As for the routing protocol used for the communication be- [1] S. Chowdhuri, S. Banerjee, and S. S. Chaudhuri, “Implemen-
tween cluster head, the enhanced version of AODV, TSODR is tation of a novel Minimum Energy Consume Secure Routing
recommended. The selection is based on the fact that this routing Protocol for reliable and energy efficient rescue operation of
protocol have proven its efficiency and effectiveness in very dynamic FANET,” in 2018 IEEE Applied Signal Processing Conference
environment. Additionally, since there does not need to be an ongoing (ASPCON), 2018, pp. 129–133.
connection between all the CHs in the proposed FANET network, this [2] F. A. Fayez, M. Hammoudeh, B. Adebisi, and K. N. A.
routing protocol will avoid the limitation of the proactive routing Sattar, “Assessing the effectiveness of flying ad hoc networks
protocols. Furthermore, if the entire network is subject to scalability, for international border surveillance,” International Journal of
where new clusters will join or leave, there will not be a great Distributed Sensor Networks, vol. 15, no. 7, 2019. DOI: 10.
distribution on the current links between the active clusters. Hence, 1177/1550147719860406. [Online]. Available: [Link]
ensuring the optimal performance of the network in terms of delay, 10.1177/1550147719860406.
throughput, network overhead and packet delivery ratio. [3] Y. Liu, Z. Liu, J. Shi, G. Wu, and C. Chen, “Optimization
VI. C ONCLUSION AND F UTURE W ORK of Base Location and Patrol Routes for Unmanned Aerial
In this paper, the latest literature work on different routing pro- Vehicles in Border Intelligence, Surveillance and Reconnais-
tocols namely OLSR, DSDV, DSR, TORA, ZRP and a number sance,” Journal of Advanced Transportation, Jan. 2019. DOI:
of enhanced routing protocols for FANETs were surveyed in de- 10.20944/preprints201809.0487.v1.
tails. This survey included a discussion on the advantages of each [4] R. Shakeri, M. A. Al-Garadi, A. Badawy, A. Mohamed, T.
routing protocol and shortcoming for FANETs. Furthermore, we Khattab, A. Al-Ali, K. A. Harras, and M. Guizani, “Design
have provided a review on the most used network communication Challenges of Multi-UAV Systems in Cyber-Physical Applica-
architecture used in FANETs. From the knowledge gathered through tions: A Comprehensive Survey, and Future Directions,” 2018.
exhaustive survey carried out throughout this paper, we propose a arXiv: 1810.09729 [[Link]].
new network communication architecture designed specifically for [5] M. A. Khan, A. Safi, I. M. Qureshi, and I. U. Khan, “Flying
FANETs which is based on the concept of clusters and cluster ad-hoc networks (FANETs): A review of communication ar-
heads. Theoretically, the proposed architecture will overcome the chitectures, and routing protocols,” in 2017 First International
limitation of the conventional network communication architecture by Conference on Latest trends in Electrical Engineering and
eliminating single point of failure and ensure redundancy. Moreover, Computing Technologies (INTELLECT), Karachi, Pakistan,
we put forward an idea of a routing mechanism which utilises two 2017, pp. 1–9.
of the optimal routing protocols, ZRP and TSODR, to be used in the [6] N. Lin, F. Gao, L. Zhao, A. Al-Dubai, and Z. Tan, “A 3D
architecture proposed. Smooth Random Walk Mobility Model for FANETs,” in 2019
In the future work, we aim to implement the proposed solution IEEE 21st International Conference on High Performance
using a simulator to examine its effectiveness. Furthermore, we plan Computing and Communications, 2019, pp. 460–467.
to investigate the proposed solution in comparison to the routing [7] I. Sumra, P. Sellappan, A. Abdullah, and A. Ahmad, “Security
protocols studies in this paper in terms of scalability, mobility and issues and Challenges in MANET-VANET-FANET: A Survey,”
rapid topology change. EAI Endorsed Transactions on Energy Web, vol. 5, no. 17,
2018.
ACKNOWLEDGEMENT
[8] A. A. Munshi, S. Sharma, and S. S. Kang, “A Review on
We thank Dr. Elias Yaacoub, Associate Professor in Computer Routing Protocols for Flying Ad-Hoc Networks,” in 2018
Engineering at Qatar University, who provided his insights and International Conference on Inventive Research in Computing
support throughout this research.

Authorized licensed use limited to: UNIVERSITY OF ROCHESTER. Downloaded on September 22,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
 Applications (ICIRCA), Coimbatore, India, 2018, pp. 1270– [21] A. K. Sharma and M. C. Trivedi, “Performance Comparison
1274. of AODV, ZRP and AODVDR Routing Protocols in MANET,”
[9] A. Nayyar, “Flying Adhoc Network (FANETs): Simulation in 2016 Second International Conference on Computational
Based Performance Comparison of Routing Protocols: AODV, Intelligence Communication Technology (CICT), Ghaziabad,
DSDV, DSR, OLSR, AOMDV and HWMP,” in 2018 Inter- India, 2016, pp. 231–236.
national Conference on Advances in Big Data, Computing [22] K. Yang and W. Huang, “Performance Research of an Im-
and Data Communication Systems (icABCD), Durban, South proved Hybrid Routing Protocol,” in 2019 IEEE 19th Inter-
Africa, 2018, pp. 1–9. DOI: 10.1109/ICABCD.2018.8465130. national Conference on Communication Technology (ICCT),
[10] Phantom 4 - DJI’s smartest flying camera ever. [Online]. Xi’an, China, 2019, pp. 1242–1247.
Available: https : / / www . dji . com / phantom - 4 (visited on [23] M. Y. Arafat and S. Moh, “Routing protocols for unmanned
04/30/2020). aerial vehicle networks: A survey,” IEEE Access, vol. 7,
[11] J. Li, Y. Zhou, and L. Lamont, “Communication architectures pp. 99 694–99 720, Jul. 2019.
and protocols for networking unmanned aerial vehicles,” in [24] R. Ghosh, Wireless Networking and Mobile Data Management.
2013 IEEE Globecom Workshops (GC Wkshps), Atlanta, GA, Springer Singapore, 2017, ISBN: 9789811039416.
USA, 2013, pp. 1415–1420. [25] M. A. Khan, I. U. Khan, A. Safi, and I. M. Quershi, “Dynamic
[12] A. Garcia-Santiago, J. Castaneda-Camacho, J. F. Guerrero- routing in flying ad-hoc networks using topology-based routing
Castellanos, and G. Mino-Aguilar, “Evaluation of AODV and protocols,” Drones, vol. 2, no. 3, p. 27, 2018.
DSDV routing protocols for a FANET: Further results towards [26] A. Chriki, H. Touati, H. Snoussi, and F. Kamoun, “FANET:
robotic vehicle networks,” in 2018 IEEE 9th Latin American Communication, mobility models and security issues,” Com-
Symposium on Circuits Systems (LASCAS), Puerto Vallarta, puter Networks, vol. 163, pp. 106–877, Nov. 2019.
Mexico, 2018, pp. 1–4. [27] J. H. Forsmann, R. E. Hiromoto, and J. Svoboda, “A time-
[13] A. V. Leonov and G. A. Litvinov, “Considering AODV and slotted on-demand routing protocol for mobile ad hoc un-
OLSR Routing Protocols to Traffic Monitoring Scenario in manned vehicle systems,” in Unmanned Systems Technology
FANET Formed by Mini-UAVs,” in 2018 XIV International IX, G. R. Gerhart, D. W. Gage, and C. M. Shoemaker, Eds.,
Scientific-Technical Conference on Actual Problems of Elec- International Society for Optics and Photonics, vol. 6561,
tronics Instrument Engineering (APEIE), Novosibirsk, Russia, SPIE, 2007, pp. 530–540. DOI: 10.1117/12.719355.
Oct. 2018, pp. 229–237. DOI: 10.1109/APEIE.2018.8545667. [28] R. Shirani, “Reactive-greedy-reactive in unmanned aeronau-
[14] N. Sharma and S. Ali, “Study of Routing Protocols in tical ad-hoc networks: A combinational routing mechanism,”
MANET-A Review,” in 2019 6th International Conference on PhD thesis, Carleton University, 2011.
Computing for Sustainable Global Development (INDIACom), [29] A. Sengupta, D. Sengupta, and A. Das, “Designing an en-
New Delhi, India, 2019, pp. 1245–1249. hanced ZRP algorithm for MANET and simulation using
[15] X. Fan, J. Lin, C. Zhang, and S. Zhang, “A cross-layer anti- OPNET,” in 2017 Third International Conference on Research
jamming routing protocol for FANETS,” in 2018 IEEE 4th in Computational Intelligence and Communication Networks
International Conference on Computer and Communications (ICRCICN), Kolkata, India, 2017, pp. 153–156.
(ICCC), Chengdu, China, 2018, pp. 301–305. [30] Z. Haas, M. Pearlman, and P. Samar, The Zone Routing
[16] P. Xie, “An enhanced olsr routing protocol based on node Protocol (ZRP) for Ad Hoc Networks. [Online]. Available:
link expiration time and residual energy in ocean FANETS,” https : / / tools . ietf . org / html / draft - ietf - manet - zone - zrp - 04
in 2018 24th Asia-Pacific Conference on Communications (visited on 04/30/2020).
(APCC), Ningbo, China, 2018, pp. 598–603. [31] Zone Routing Protocol, Feb. 2020. [Online]. Available: https://
[17] A. V. Leonov, G. A. Litvinov, and D. A. Korneev, “Simula- [Link]/w/[Link]?title=Zone Routing Protocol&
tion and Analysis of Transmission Range Effect on AODV oldid=943004272 (visited on 04/30/2020).
and OLSR Routing Protocols in Flying Ad Hoc Networks
(FANETs) formed by Mini-UAVs with Different Node Den-
sity,” in Systems of Signal Synchronization, Generating and
Processing in Telecommunications (SYNCHROINFO), Minsk,
Belarus, 2018, pp. 1–7.
[18] A. I. Alshabtat, L. Dong, J. Li, and F. Yang, “Low latency rout-
ing algorithm for unmanned aerial vehicles ad-hoc networks,”
International Journal of Electrical and Computer Engineering,
vol. 6, no. 1, pp. 48–54, 2010.
[19] A. Sharma and R. Kumar, “Performance comparison and
detailed study of AODV, DSDV, DSR, TORA and OLSR
routing protocols in ad hoc networks,” in 2016 Fourth Interna-
tional Conference on Parallel, Distributed and Grid Computing
(PDGC), Waknaghat, India, 2016, pp. 732–736.
[20] P. Kuppusamy, K. Thirunavukkarasu, and B. Kalaavathi, “A
study and comparison of OLSR, AODV and TORA routing
protocols in ad hoc networks,” in 2011 3rd International Con-
ference on Electronics Computer Technology, vol. 5, Kanyaku-
mari, India, 2011, pp. 143–147.

Authorized licensed use limited to: UNIVERSITY OF ROCHESTER. Downloaded on September 22,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.