Digital Communications and Networks ∎ (∎∎∎∎) ∎∎∎–∎∎∎

HOSTED BY Contents lists available at ScienceDirect

Digital Communications and Networks

journal homepage: www.elsevier.com/locate/dcan

A reliable, delay bounded and less complex communication protocol

for multicluster FANETs
Wajiya Zafar n,1, Bilal Muhammad Khan 1
National University of Sciences and Technology (NUST), Pakistan

art ic l e i nf o a b s t r a c t

Article history: Recently, Flying Ad-hoc Networks (FANETs), enabling ad-hoc networking between Unmanned Aerial
Received 5 August 2015 Vehicles (UAVs) is gaining importance in several military and civilian applications. The sensitivity of the
Received in revised form applications requires adaptive; efficient; delay bounded and scalable communication network among
4 May 2016
UAVs for data transmission. Due to communication protocol complexity; rigidity; cost of commercial-off-
Accepted 13 June 2016
the-shelf (COT) components; limited radio bandwidth; high mobility and computational resources;
maintaining the desired level of Quality of Service (QoS) becomes a daunting task. For the first time in
Keywords: this research we propose multicluster FANETs for efficient network management; the proposed scheme
FANETs considerably reduces communication cost and optimizes network performance as well as exploit low
IEEE 802.15.4
power; less complex and low cost IEEE 802.15.4 (MAC) protocol for intercluster and intracluster com-
Reliability
munication. In this research both beacon enabled mode and beaconless modes have been investigated
Latency
Multicluster with Guaranteed Time Slots (GTS) and virtual Time Division Multiple Access (TDMA) respectively. The
methodology plays a key role towards reserving bandwidth for latency critical applications; eliminate
collisions and medium access delays. Moreover analysis ad-hoc routing protocols including two proactive
(OLSR, DSDV) and one reactive (AODV) is also presented. The results shows that the proposed scheme
guarantees high packet delivery ratios while maintaining acceptable levels of latency requirements
comparable with more complex and dedicatedly designed protocols in literature.
& 2016 Chongqing University of Posts and Telecommunications. Production and Hosting by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction for efficient network management, reducing communication cost

and optimizing network performance. The communication be-
Airborne nodes in FANETs move at a high speed of 30–460 km/ tween cluster heads not only shares redundant information en-
hr [1], resulting in frequent topology changes which in turn cause hancing robustness and reliability but also adds to scalability and
link fluctuations and breakages [2]. FANETs are used for highly coverage area of the network.
sensitive applications such as traffic monitoring, remote sensing, This article introduces a novel approach of employing IEEE
disaster monitoring, search operations, border surveillance and 802.15.4 MAC layer protocol in both beacon enabled and beacon-
relaying networks [3–6], these applicaions require precise and less modes for UAV-to-UAV communication in multicluster FANET
prompt data delivery. Thus, the most important challenges that scenarios. Both single cluster and multicluster setups have been
investigated. Although IEEE 802.15.4 is considered a low data rate
need to be addressed by FANET MAC and routing layers are of high
protocol i.e. up to 250 kbps at 2.4 GHz band, we propose the use of
reliability and delay bounded data delivery [7,8].
802.15.4 MAC protocol for intercluster and intracluster commu-
Single cluster networks can only cover small mission areas [9],
nication in multicluster FANETs that can be deployed for less
hence, we propose multicluster FANETs, in which cluster head in
bandwidth hungry applications such as monitoring and control,
one cluster communicates with the cluster head in another cluster
wildfire management, remote sensing and relaying networks. Such
applications do not necessarily require a high data rate protocol
n
Corresponding author. such as 802.11, which is being used in recent FANET studies [10,11],
E-mail addresses: [email protected] (W. Zafar), as the nature of the in-network messages exchanged is primarily
[email protected] (B.M. Khan). of time-sensitive control data, UAV location information and tasks
1
Pakistan Navy Engineering College (PNEC), PNS JAUHAR, Habib Ibrahim Re-
hmatullah Road, Karachi, Pakistan.
assignment updates [12]. Thus, UAV-to-UAV communication can
Peer review under responsibility of Chongqing University of Posts and be realized efficiently by deploying a low data rate protocol such
Telecommunications. as IEEE 802.15.4 which has not been exploited so far for FANETs.

http://dx.doi.org/10.1016/j.dcan.2016.06.001
2352-8648/& 2016 Chongqing University of Posts and Telecommunications. Production and Hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Please cite this article as: W. Zafar, B.M. Khan, A reliable, delay bounded and less complex communication protocol for multicluster
FANETs, Digital Communications and Networks (2016), http://dx.doi.org/10.1016/j.dcan.2016.06.001i
2 W. Zafar, B.M. Khan / Digital Communications and Networks ∎ (∎∎∎∎) ∎∎∎–∎∎∎

For both single cluster and multicluster scenarios, 802.15.4 can be required for the modifications is appended in HELLO and Topology
used within the cluster so as to achieve better results in terms of Control (TC) messages of OLSR. The results presented lack the
reduced complexity and bandwidth minimization. For a swarm of discussion on routing overhead which is expected to be high.
small UAVs collecting data from remote ground sensor networks, Communications bandwidth is a meager resource in FANETs and
its short range and low data rate can achieve the required con- routing protocols with excessive overhead are considered
nectivity offering a line-of-site (LOS) range of more than 75 m. For inefficient.
applications requiring exchange of heavy traffic, 802.15.4 can offer Another proactive protocol, Directional Optimized Link State
dedicated and guaranteed cyclic time slots (GTS) as used in our Routing Protocol (DOLSR) has been designed for FANETs in [21] on
scheme. the assumption that the network uses directional antennas and
The rest of this paper is outlined as follows: Section 2 gives an reflects dependency on COT component to get desired outputs. In
overview of the related works on FANET communication protocols [22], Reactive-Greedy-Reactive (RGR) protocol is proposed which
in the literature, Section 3 describes the proposed methodology adopts reactive routing based on AODV or geographic routing
modeled, Section 4 discusses the simulation setup and perfor- depending on validity of the links. A modification of this protocol
mance metrics, Section 5 reports graphical results and analysis [23] addresses the issue of high overhead by broadcasting route
followed by conclusion and future work. request (RREQ) packets to the neighbors instead of flooding them
in the network but computational complexity and average packet
delays are increased.
2. Related work Considering the above mentioned issues of scalability, robust-
ness, protocol complexity, rigidity and cost; in this paper we have
Not a lot of work is being done over the higher layers of FA- presented the concept of multicluster FANETs deploying a low cost
NETs. In this section we have summarized the recent work that has and simple IEEE 802.15.4 MAC layer protocol for UAV-to-UAV
been done at MAC and routing layer for UAV systems. communication, which, to the best of our knowledge, has not been
proposed in the literature. Comparing with the existing studies,
2.1. MAC protocols our work has three noteworthy novelties. Firstly, 802.15.4 can be
used within the cluster so as to achieve bandwidth savings for less
Till date many studies have investigated UAV-to-Ground com- bandwidth hungry applications. Secondly, the concept of multi-
munication aspect of FANETs. In [13], the authors have used UAV as cluster FANETs facilitates network scalability and robustness.
a relaying network between ground relaying node and sink to Thirdly, less complex and existing MAC and routing protocols used
limit the transmission range of the relaying nodes (RN). The MAC in multicluster scenario have achieved the QoS gains similar to
protocol used for RN-UAV and UAV-sink is IEEE 802.11 (Wi-Fi). modified protocols used in single cluster networks [19–22].
Similarly, in [12] a wireless communication infrastructure has
been proposed for short range UAV using single 802.15.4 trans-
ceiver. In order to enhance the limited offerings of 802.15.4, a 3. Proposed methodology
computationally complex adaptation layer is implemented be-
tween IPv6 stack and 802.15.4 link-layer. The work in [14] suggests In this section we describe the network model, including
the use of 802.11 MAC protocol for UAV swarm in relaying net- multicluster formation, cluster head selection, propagation model,
works by assigning dedicated tasks of collection, facilitation and mobility model, and IEEE 802.15.4 protocol as it is used in our
delivery to each UAV. In all these network models only one UAV model.
has been used to relay the information raising issues of robustness At the start a swarm of UAVs is deployed in the mission area
and system collapse in case of UAV failure. which is divided into zones. During the mission, formation of
In [15], an adaptive MAC protocol has been designed by adding clusters takes place adaptively which are then intact till the net-
directional network allocation vector (DNAV) to the existing 802.11 work operation completes. The movement of clusters occurs in
MAC protocol for intra-UAV communication. The proposed MAC accordance with Reference Point Group Mobility model which is
protocol employs four antennas on each UAV increasing com- discussed in Section 3.4.
plexity and cost of the network design. In [16] Location Oriented A multicluster FANET is shown in Fig. 1. In the network there
Directional MAC (LODMAC) protocol has been proposed for FA- are N clusters depending on the mission area and application re-
NETs operating in conjunction with High Altitude Platforms quirement. Each cluster has Y UAVs among which one is serving as
(HAPs). The work addresses the problems that arise due to the use
the head UAV called as cluster head in our scheme. In multicluster
of directional antennas in FANETs. The proposed protocol meets
performance gains but with the help of COT hardware equipment.

2.2. Routing protocols

Some studies have evaluated existing routing protocols used for

MANETs for performance analysis in UAV networks. In [17] a
comparison of AODV, OLSR and OSPF-MDR has been performed for
airborne tactical networks. Similarly, [18] evaluates mesh routing
protocols open80211s, BATMAN, BATMAN Advanced and OLSR for
UAV swarming applications. The simulations are performed for
very low speeds of the nodes which suppresses the dynamism and
high mobility characteristic of FANETs and fails to generalize large
scale realistic FANET environments.
Some studies have modified the existing MANET protocols ac-
cording to the requirements of UAV networks. In [19,20] OLSR has
been modified to Predictive-OLSR (P-OLSR) and Mobility and Load
aware OLSR (ML-OLSR) respectively. The additional information Fig. 1. Multicluster FANETs.

Please cite this article as: W. Zafar, B.M. Khan, A reliable, delay bounded and less complex communication protocol for multicluster
FANETs, Digital Communications and Networks (2016), http://dx.doi.org/10.1016/j.dcan.2016.06.001i
 W. Zafar, B.M. Khan / Digital Communications and Networks ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 3

3.2. Cluster head selection

Fig. 2. Node_info message structure.

All the UAVs in the cluster maintain a table of link quality. The
UAV with the best link quality broadcasts “cluster head declara-
FANETs, cluster head of each cluster plays an important role in tion” message to all member UAVs of the cluster which reply with
regulating the coordination and collaboration among the UAVs. “Follow” message. In case of link failure or poor channel condition
We assume that when multicluster FANET starts, each UAV is well experienced by the cluster head, the node with the second best
aware of its neighbors, location, zone ID, SNR value and speed. The link quality becomes the cluster head and broadcasts cluster head
work is based on fixed number of UAVs per cluster. declaration message. Algorithm 2 describes the cluster head se-
lection mechanism. Cluster head selection on the basis of link
quality, energy and number of neighbors is a well-established
3.1. Multicluster formation approach in Wireless Sensor Networks (WSN) [24]. However, in
FANETs there are no energy or battery constraints hence the key
At the start of the operation, each UAV in the formation shares factor for cluster head selection is link quality to ensure reliable
its known information with its neighbors by exchanging “no- communication in inter-cluster as well as intra-cluster.
de_info” messages. The message structure is shown in Fig. 2.
The message fields are described below: Algorithm 2. Cluster head selection algorithm for each cluster in
multicluster FANETs
Broadcast ID: 1 if the message is a broadcast message for i¼1 to Y do
0 if the message is a unicast message if (UAVi has SNR of Y nodes)
Neighbor list: Contains the pre-learned one hop neighbor IDs Sort SNR against UAV IDs
Location Contains the location coordinates obtained via end if
coords: GPS if (UAVi has the highest SNR)
Node SNR: Contains the node SNR broadcast “Cluster head declaration”
Zone ID: Contains the pre-assigned zone ID message
Speed: Contains average speed of the node else
Assigned task: Nat (Not any task) at the start of the network Unicast “follow” message to the UAV with
1 if selected as cluster head the highest SNR
0 if selected as member end if
end for
UAVs with similar zone ID group with the help of location co-
ordinates. Once the clusters are formed and cluster heads are se-
lected, each cluster starts following the pre-assigned path ac- 3.3. Propagation model
cording to the group mobility model used in our study. Whenever,
during the mission the cluster heads come into radio range of each The propagation model used in our study is freespace propa-
other, control packets are exchanged between them directing the gation model. The reason for selecting this model over shadowing
decision of whether to collaborate with the neighboring UAVs or model is that in FANETs there is a clear LoS between the trans-
not. mitter and receiver UAVs. The synchronization of the nodes is
Each cluster head monitors the wireless link quality during the performed using GPS. This leads to valid connection establishment
exchange of information. If the link quality metric alpha drops between any two UAVs in the operative circular communication
below a certain threshold, the cluster head of that cluster sends a range. Friis equation for freespace propagation model is given as:
collaboration request to the cluster head of any other cluster
Pt Gt G r (λ )2
within the communication range. Alpha is obtained by monitoring P (d) =
4πd2L (1)
Signal to Noise Ratio (SNR) values. Algorithm 1 describes the col-
laboration phase of multicluster scheme. Hence, collaboration will where, P(d) is the received signal power at distance d, Pt is the
be done on the basis of the reliability so as to have redundant transmission power, Gt and Gr are transmit and receive antenna
information of both clusters at the ground station. Similarly, in gain respectively, λ is the wavelength, L is the system loss and d is
order to mitigate the problems of latency as well, collaboration the distance between transmitter and receiver.
between cluster heads can take place. If one cluster head experi-
ences a poor link quality and retransmissions, it may route its 3.4. Mobility model
information via cluster head of the neighboring cluster.
Group mobility models can be used to simulate group of UAVs
Algorithm 1. Collaboration amongst cluster heads in performing autonomous military operations without cen-
for (i ¼ 1 to Y) tralized control. Based on the previous study [25] by the authors
if (alpha o thresh hold for satisfactory link quality) which investigated the performance of different mobility models
if (UAVi is cluster head) for FANETs, in this work, Reference Point Group Mobility (RPGM)
sense for other network has been used [26]. Every group or cluster has a reference point
forward collaboration request upon which the movement, direction, speed and acceleration of all
else the group members depend. The cluster head is positioned at the
notify cluster head of link quality reference point during the entire simulation with member nodes
end else if following it. Thus, the UAV network within the cluster is in a fully
end if connected state most of the time. The nodes fly in a formation and
end for the cluster remains intact with no new UAVs joining or leaving it.

Please cite this article as: W. Zafar, B.M. Khan, A reliable, delay bounded and less complex communication protocol for multicluster
FANETs, Digital Communications and Networks (2016), http://dx.doi.org/10.1016/j.dcan.2016.06.001i
4 W. Zafar, B.M. Khan / Digital Communications and Networks ∎ (∎∎∎∎) ∎∎∎–∎∎∎

3.5. IEEE 802.15.4 for multicluster FANETs capacity which means that no bandwidth wastage occurs and the
UAVs have data to transmit all the time. The chosen traffic rate is
In this paper we propose the use of IEEE 802.15.4 [27] for car- 50 kbps such that one slot is of 19.2 ms and can transmit 120 bytes
rying reliable and timely intercluster and intracluster commu- of data.
nication in FANETs. We have used both the models provided by FANETs do not encounter energy resource constraints; hence,
this MAC standard: beacon enabled mode and non-beacon en- 100% duty cycle operation is carried in our simulations. The duty
abled mode. cycle is calculated by the ratio of active period (SD) of the super-
frame to beacon interval (BI) as given in Eq. (4). We have selected
3.5.1. Beacon enabled mode SO¼ BO¼2 so as to eliminate the inactive period.
In beacon enabled mode the cluster head periodically sends
beacon frames to synchronize member UAVs. The duration be- ⎛ SD ⎞
Duty cycle = ⎜ ⎟ *100% = 2(BO − SO) *100% (BO = SO = 2)
tween these beacon frames is divided into 16 slots and is termed ⎝ BI ⎠ (4)
as Beacon Interval (BI). The metric used to define beacon interval is
The low value of SO and BO is selected so as to achieve max-
known as macBeaconOrder (BO). The relationship is expressed in
imum bandwidth savings as higher values of SO increases the slot
Eq. (2).
size.
Beacon Interval (BI ) = aBaseSuperFrameDuration*2BO (2)
3.5.2. Non-beacon enabled mode
Where aBaseSuperFrameDuration¼Number of slots * Slot duration,
Since energy efficiency is not the primary issue in FANETs,
for which default values of 16 and 60, respectively, are used.
therefore we have also used non-beacon enabled mode of IEEE
BI is further divided into two regions: active period defined by
802.15.4. In beaconless mode, there is no transmission of periodic
macSuperframeOrder (SO) and the inactive period during which the
beacons from the cluster head and UAVs use unslotted CSMA/CA
node sleeps. The active period is denoted by Superframe Duration
for accessing the channel. SO ¼BO ¼15 has been used to enable
(SD) according to Eq. (3).
this mode. In this mode the UAVs sense the channel and transmit
Superframe Duration (SD) = aBaseSuperFrameDuration*2SO (3) data if found to be idle. If the channel is found to be busy, the
algorithm backs off for a random period and sense the channel
SD is divided into Contention Access Period (CAP) during which
again. This approach eliminates beaconing overheads. In order to
all devices that wish to communicate, compete using slotted
CSMA/CA mechanism to win the channel and Contention Free have a fair comparison, we have implemented TDMA approach in
Period (CFP) during which dedicated Guaranteed Time Slots (GTS) beaconless mode also so that the UAVs have periodic transmis-
are reserved by the devices to utilize the channel alone. The su- sions and only one UAV is allowed to transmit in a virtually as-
perframe structure of standard IEEE 802.15.4 is shown in Fig. 3. signed time slot.
In our study we have utilized the CFP portion of the superframe
since multicluster FANETs can be used in critical and time-sensi- 3.6. Routing protocols
tive missions. Hence, contention-based transmission has not been
considered. The cluster head allocates GTS to the member UAVs Following routing protocols have been used for evaluation in
using First Come First Serve (FCFS) algorithm. According to the this study:
standard, seven UAVs can hold the GTS slots. In our simulations,
there are 3 nodes per cluster. So, the default GTS slotting can serve 3.6.1. AODV
the purpose. As soon as the network starts, the member UAVs The Ad-hoc On-demand Distance Vector (AODV) routing pro-
consume GTS slots and hold them forever or until the need of tocol is a reactive protocol which determines route to destination
explicit deallocation arises which means that the UAV itself re-
only when source initiates it and keeps it as long as the source
quests the cluster head to deallocate the slot. By exploiting GTS in
desires. A route request (RREQ) packet is broadcasted by the
our scheme, we are basically implementing TDMA in the entire
source node to discover route to the destination. The intermediate
active period. As the packets are generated, nodes in their re-
nodes not only forward RREQ but also update themselves with the
spective reserved allocated time slots start transmitting without
source information as contained in RREQ thus setting up a reverse
competing for the channel. Two critical issues of FANETs are re-
route entry to the source. Any intermediate node that has route to
liability and latency. The use of GTS mechanism satisfies both as
the destination replies the source with route reply (RREP) packet
the UAVs can transmit time-sensitive data reliably to the cluster
head in a collision-free manner avoiding delays due to re- containing the number of hops required to reach the destination.
transmissions. Also, the GTS slots are utilized to their maximum In case of an invalid route, a route error packet (RERR) is generated
to inform the source about link failure so that it can re-start the
route discovery process. In terms of overhead AODV dominates its
proactive counterparts as routes are discovered only on demand.
However, it suffers from latency issues since a packet has to wait
till the route to the new destination is found [28].

3.6.2. DSDV
The Destination Sequenced Distance Vector (DSDV) is a
proactive table-driven protocol. All nodes in the network have
routes stored in their tables to all other nodes. Looping is avoided
by updating routes periodically and removing stale routes using
sequence numbers. The route with the highest sequence number
is considered as the most recent route. The updates may contain
whole routing table or only the routes that have been changed
Fig. 3. Superframe structure. [29].

Please cite this article as: W. Zafar, B.M. Khan, A reliable, delay bounded and less complex communication protocol for multicluster
FANETs, Digital Communications and Networks (2016), http://dx.doi.org/10.1016/j.dcan.2016.06.001i
 W. Zafar, B.M. Khan / Digital Communications and Networks ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 5

4. Simulation setup and performance metrics

In this section we evaluate the performance of the proposed

scheme by employing IEEE 802.15.4 protocol in multicluster FA-
NETs via computer simulations.
The mobility scenarios have been generated using bonnmotion
utility which generates node movement patterns according to the
mobility models specified by taking certain parameters as input
such as maximum minimum speed, number of nodes, speed
standard deviations, group change probabilities, pause time and
simulation area and duration [31].
The simulations were performed on ns2.35 which is a discrete
network simulator. In the physical layer we assume that UAVs
operate in 2.4 GHz ISM frequency band with radio bandwidth of
250 Kbps. The network topology at initial time is shown in Fig. 5 in
which two clusters of 3 nodes each are deployed in an area of
200 m  200 m. The velocity of the UAVs is varied from 10 m/s to
500 m/s [1] and they move in a multicluster fashion in accordance
with RPGM model. The high mobility values have been selected to
create more dynamism in the network. Our approach is to evaluate
the performance of IEEE 802.15.4 MAC standard in both beacon
enabled and beaconless modes as described in Section 3 for UAV-
to-UAV communication in combination with existing proactive
(DSDV, OLSR) and reactive (AODV) routing protocols. Other details
of the simulation parameters are given in Table 1.
During the simulation duration of 200 s, the cluster heads
communicate with each other twice. As per the simulation sce-
nario, the intra cluster communication continues until the link
quality metric alpha drops below the desired threshold at 80ths.
Fig. 4. Flow chart representation of proposed multicluster scheme. During the interval between (80–100) s the cluster head of first
cluster collaborates with the cluster head of the second cluster and
data transmission occurs till the threshold limit is retrieved. The
3.6.3. OLSR same event occurs during the end of the simulation that is (180–
Optimized Link State Routing (OLSR) protocol maintains a 200) s when information from the cluster head of the second
routing table at each network node by gathering topology in- cluster is transferred to the cluster head of the first cluster.
formation through Topology Control (TC) messages. HELLO mes- The evaluation is done on the following performance metrics:
sages are also used to find one hop and two hop neighbors. An
important feature for reducing control messages in OLSR is multi 4.1. Packet Delivery Ratio (PDR)
point relaying. Multi Point Relay (MPR) nodes are a subset of
It is the ratio of total data packets delivered to the destination
nodes responsible for forwarding link state updates. This optimi-
node to the data packets generated by the source node as given by
zation to pure link state routing protocol proves beneficial in
Eq. (5). Higher value of PDR corresponds to low packet loss rate
highly dense environments where MPR mechanism is efficiently and thus reliable communication. It is an important parameter to
utilized [30]. evaluate the behavior of protocols in our proposed multicluster
approach because of the sensitivity of the information exchanged
in intercluster and intracluster communication. It also indicates
3.7. Protocol operation the importance of collaboration between cluster heads as sug-
gested in Algorithm 1.
Fig. 4 shows flowchart representation of our proposed multi-
N
cluster scheme for FANETs. ∑1 Dpackets
PDR = N
The steps of our proposed scheme are listed below: ∑1 Spackets (5)

1. All UAVs deployed randomly in the entire coverage area and where,
divided into zones.
2. UAVs with similar zone IDs group together and follow RPGM Dpackets ¼ Total data packet delivered at the destination.
model for path planning. Spackets ¼ Total data packets generated by the source.
3. Every cluster selects its cluster head according to Algorithm 2.
4. In beacon enabled mode all UAVs in the cluster request GTS slot
at the start of the network.
5. All UAVs transmit in reserved slots in case of beacon enabled
network and accesses channel through unslotted CSMA/CA in
beaconless mode.
6. If any other network is sensed by the cluster head, it checks the
link quality metric and sends collaboration request to the
cluster head of that network.
Fig. 5. Network topology.

Please cite this article as: W. Zafar, B.M. Khan, A reliable, delay bounded and less complex communication protocol for multicluster
FANETs, Digital Communications and Networks (2016), http://dx.doi.org/10.1016/j.dcan.2016.06.001i
6 W. Zafar, B.M. Khan / Digital Communications and Networks ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Table 1
Simulation parameters.

Parameter Value

Simulation platform NS-2.35

Simulation duration 200 s
Simulation area 200 m  200 m
Propagation model Freespace
MAC protocol 802.15.4, 802.11
Mobility model Reference Point Group Mobility Model
Routing protocols DSDV, AODV, OLSR
Number of clusters 2
Number of nodes per cluster 3
Packet size 100 bytes
Traffic type CBR
Data rate 50 kbps
Transmission range 85 m Fig. 6. PDR vs. speed for beacon enabled 802.15.4.

drop below 80% because within the cluster UAVs are moving in a
N ¼ Total number of packets. swarm and are in the radio range of each other. The effect of high
mobility in intra cluster communication nullifies due to RPGM
4.2. Average End-to-End (E2E) delay model whereas random movement of clusters causes link
breakages in inter cluster communication.
The average time it takes for the successful transmission of a In our scenario of less node density, OLSR gives better and
data packet from source node to destination node as given by Eq. stable performance results for varying node mobility values. The
(6). This metric investigates the efficiency of our proposed scheme proactive nature of OLSR maintains routing table so that the routes
to meet delay bounds of critical command and control data. This are immediately available when the nodes have data packets to
parameter includes all delays: transmit. The highly dynamic scenario of proposed multicluster
N FANETs cause frequent topology change which is also addressed by
E2Edelay = ∑ (Tt + Rt + Bt + Prt ) OLSR which keeps the routes up-to-date by generating periodic
1 (6) Topology Control (TC) messages. DSDV is also table driven protocol
where, and gives delivery ratios comparable to OLSR in our topology.
However, with the increase in mobility, PDR for DSDV drops due to
Tt ¼Transmission time. unsuccessful collaboration amongst clusters. AODV exhibits on-
Rt ¼Retransmission time. demand behavior which means that routes are requested only
Bt ¼Buffering time. when needed and there is no maintenance of network topology. In
Prt ¼Processing time. our case the network density is low and there are frequent route
breakages so the reactive nature of AODV tends to initiate a route
4.3. Normalized Routing Load (NRL) discovery process, making it impossible for immediate route
calculations.
NRL is the ratio between routing packets generated to the data Fig. 7 shows that the approach of assigning dedicated slot to
packets successfully delivered at the destination as given by Eq. each UAV in beacon enabled mode in our scheme not only ensures
(7). It determines the efficiency of the routing protocol. In order to reliable but also well-timed data delivery. The omission of con-
achieve bandwidth savings, control overhead should be minimum tention based approach suppresses packet collisions and channel
in multicluster FANETs. access delays. Also, 100% duty cycle ensures and exhibits minimal
N average end-to-end delay of data packets.
∑1 Dpackets DSDV and OLSR exhibit lowest average end-to-end delay in our
NRL = N
∑1 Rpackets (7) network. Being table driven protocols and only 3 nodes per cluster,
these protocols do not encounter long route setup time as updated
routes are present in the routing tables. AODV on the other hand
does not reuse routing information and has to initiate route dis-
Rpackets ¼No. of routing packets transmitted. covery process again and again when a node wishes to transmit.
Dpackets ¼No. of data packets received.

5. Results and analysis

In this section we describe performance comparison AODV,

DSDV and OLSR routing protocols for IEEE 802.15.4 beacon enabled
and beaconless modes in multicluster FANETs.

5.1. IEEE 802.15.4 beacon enabled mode

Fig. 6 shows that using GTS mechanism in beacon enabled

mode of 802.15.4 results in higher PDR values (above 85%) for both
reactive and proactive routing protocols. The transmissions are
collision free which results in higher average packets received by
the destination nodes. Also, as the speed increases, PDR does not Fig. 7. E2E delay vs. speed for beacon enabled 802.15.4.

Please cite this article as: W. Zafar, B.M. Khan, A reliable, delay bounded and less complex communication protocol for multicluster
FANETs, Digital Communications and Networks (2016), http://dx.doi.org/10.1016/j.dcan.2016.06.001i
 W. Zafar, B.M. Khan / Digital Communications and Networks ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 7

Fig. 8. NRL vs. speed for beacon enabled 802.15.4.

Fig. 9. PDR vs. speed for non-beacon enabled 802.15.4.

By the time route reply message reaches the source in response to

route request query, the destination has moved away due to high
mobility. Hence, in our time-sensitive network topology AODV
shows poor delay bounds as compared to proactive protocols:
DSDV and OLSR.
Fig. 8 shows Normalized Routing Load (NRL) for AODV, DSDV
and OLSR in our network. OLSR exhibits high routing load because
of massive generation of Topology Control (TC) and HELLO packets
to maintain routing table. OLSR computes Multi-Point Relays
(MPRs) through these TC and HELLO messages. MPRs are selected
to access two hop neighbors within the network and reduce
broadcast overhead but due to less node density and random
movement of nodes within the cluster in our network, MPR me-
chanism is inefficiently used. Although AODV does not generate
Fig. 10. E2E delay vs. speed for non- beacon enabled 802.15.4.
periodic topology update messages but the dynamic topological
behavior of our network causes AODV to broadcast frequent route
updated routes in the table by transmitting periodic HELLO mes-
request messages which results in high routing overhead. DSDV
sages. OLSR also shows average delay performance comparable to
transmits periodic HELLO messages only to update routing tables
DSDV due to its table driven nature but the maintenance of MPRs
and shows least routing overhead.
adds to average delays. Due to our network topology of scarce
node density, MPRs are not effectively used. AODV shows higher
5.2. IEEE 802.15.4 beaconless mode
packet delays in beaconless mode than in beacon enabled network
because of absence of synchronization mechanism which adds to
Due to the mobility model used, all UAV nodes within the
frequent route discoveries that arise as a result of high mobility
cluster are in the radio range of each other so hidden node pro-
induced route breakages.
blem does not exist in beaconless mode used in our simulations.
Fig. 11 shows normalized routing load for beaconless mode.
To have fair comparison of beacon enabled and non-beacon en-
OLSR shows similar behavior to beacon enabled mode due to ex-
abled modes, virtual timeslots have been assigned to each UAV.
cessive TC and HELLO messages generation. Although these con-
This type of network scenario is applicable to applications that
trol messages help to maintain more knowledge of the network
require reliable data transfer and all UAVs in the cluster monitor-
ing homogeneous data. Hence, Fig. 10 shows that high delivery than AODV and DSDV but at the expense of increased routing
ratios are obtained in beaconless mode also. overhead. AODV exhibits more routing overhead in non-beacon
Proactive protocols OLSR and DSDV outperform AODV in terms mode than in beacon enabled mode because route request mes-
of PDR because updated routes are present in their routing tables. sages may suffer failed response from the destination which has
Although this incurs high overhead and bandwidth usage for both not completed previous transmission. Hence, more route request
table driven protocols but ensures that only fresh and valid routes messages are generated which boosts up the routing load offered
are used to deliver the packets to the highly mobile destination.
AODV on the other hand performs worse than beacon enabled
mode as shown in Fig. 9. In beaconless mode, since there is no
synchronization mechanism, hence, the destination UAV refuses to
send route reply message if route request message arrives before
the destination has send ACK for the previous transmission. This is
due to the unavoidable turn-around time in 802.15.4 during which
the node changes from transmit to receive mode.
Fig. 10 shows minimal delay values are obtained in case of
beaconless mode in our network. Any UAV that wishes to transmit
data to the cluster head does not have to wait to receive beaconing
messages from the cluster head before sending data. All trans-
missions are performed using unslotted CSMA/CA which reduces
the beaconing overhead and in turn latency of the network.
DSDV offers lowest average delays as it maintains most Fig. 11. NRL vs. speed for non-beacon enabled 802.15.4.

Please cite this article as: W. Zafar, B.M. Khan, A reliable, delay bounded and less complex communication protocol for multicluster
FANETs, Digital Communications and Networks (2016), http://dx.doi.org/10.1016/j.dcan.2016.06.001i
8 W. Zafar, B.M. Khan / Digital Communications and Networks ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Fig. 12. PDR vs. speed for IEEE 802.11. Fig. 14. NRL vs. speed for IEEE 802.11.

by the protocol in beaconless mode.

Table 2
Feasibility of multicluster FANETs for various applications.
5.3. Ieee 802.11
Potential Multicluster Performance Suitable
Fig. 12 shows that due to the transmission of Request To Send Applications FANETs metric routing
(RTS) and Clear To Send (CTS) packets in IEEE 802.11, both reactive protocol

and proactive protocols exhibit high packet deliveray rates. How- Traffic monitoring 802.15.4 PDR (reliability) OLSR
ever, with the increase in speed of the UAVs, more dynamism is Agricultural Beacon enabled E2E delay (latency) OLSR, DSDV
incorporated in the network and the collision avoidance me- management NRL (bandwidth DSDV
chanism of IEEE 802.11 does not work efficiently as it can be ob- efficiency)
Relaying networks 802.15.4 PDR (reliability) OLSR
served that less number of packets reach their destination Non-beacon E2E delay (latency) OLSR, DSDV
successfully. enabled NRL (bandwidth DSDV
Both reactive and proactive routing protcols exhibit similar efficiency)
behavior with the increase in speed. However, the PDR sig- Disaster management 802.11 PDR (reliability) OLSR, AODV
Reconnaissance E2E delay (latency) DSDV
nificanltly decreases at higher speeds for DSDV and OLSR. Due to missions NRL (bandwidth DSDV
frequent topology change at high speeds, the number of stale efficiency)
routes increases in the routing tables of these proactive protocols
and the rediscovery of the broken link takes time. The dips ob-
served at certain speeds are due to the random scenario genera-
tion of bonnmotion utility that has been used, however, the higher than IEEE 802.15.4.
average results show decreasing trend. The usage of FANETs for several military, civilian and com-
Fig. 13 shows that average end-to-end delay of 802.11 based mercial applications is expected to deliver favorable results in
network is comparable to 802.15.4 beaconless mode based net- terms of reliable and delay bounded data delivery. Table 2 de-
work. However, the slightly better results are due to the fact that scribes the feasibility of our proposed scheme for different
802.11 supports higher data rate and requires larger bandwidth as applications.
compared to 802.15.4 which is suitable for low data rate
applications.
Considering the routing protocols DSDV shows better results 6. Conclusion
but at the cost of low packet delivery rates. It must be noted that
In this paper the concept of multicluster FANETs employing
the delays are effective delays of data packets that successfully
IEEE 802.15.4 MAC layer protocol for UAV-to-UAV communication
arrive at the destination.
is presented which is to the best of author's knowledge the first of
Normalized Routing Load (NRL) for IEEE 802.11 based network
its type proposal. The proposed scheme allows collision free, re-
follows the same trend as that of IEEE 802.15.4 as seen in Fig. 14.
liable and timely data transmission by employing GTS and virtual
However, the average values for all three routing protocols are
TDMA approaches in beacon enabled and beaconless modes of
802.15.4, respectively. In this context the proposed scheme is in-
vestigated using ad-hoc routing protocols: OLSR, DSDV and AODV.
The results clearly reveal that this novel approach meets the QoS
gains comparable to existing studies which are performed for
single cluster networks as well as employ more complex routing
protocols. 802.15.4 has proved to be a potential candidate showing
80–98% packet delivery rates and comparable network delays to
IEEE 802.11 which involves complexity and high bandwidth usage.
Hence, IEEE 80.15.4 can be a suitable choice for applications that
are not bandwidth exhaustive and require lesser data rate for
communication.
This work has been conducted assuming a network of fixed
number of UAVs. For the future work, we aim to refine our pro-
posal for the case when new UAVs join the existing clusters during
Fig. 13. E2E delay vs. speed for IEEE 802.11. the mission.

Please cite this article as: W. Zafar, B.M. Khan, A reliable, delay bounded and less complex communication protocol for multicluster
FANETs, Digital Communications and Networks (2016), http://dx.doi.org/10.1016/j.dcan.2016.06.001i
 W. Zafar, B.M. Khan / Digital Communications and Networks ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 9

References Control, IEEE, 2010: pp. 598–603. http://dx.doi.org/10.1109/ICNSC.2010.

5461589.
[16] S. Temel, I. Bekmezci, On the performance of Flying Ad Hoc Networks (FANETs)
[1] İ. Bekmezci, O.K. Sahingoz, Ş. Temel, Flying Ad-Hoc Networks (FANETs), Surv.
utilizing near space high altitude platforms (HAPs), in: 2013 6th Int. Conf.
Ad Hoc Netw. 11 (2013) 1254–1270, http://dx.doi.org/10.1016/j.
Recent Adv. Sp. Technol., IEEE, 2013: pp. 461–465. http://dx.doi.org/10.1109/
adhoc.2012.12.004.
RAST.2013.6581252.
[2] M. Zhao, W. Wang, The Impacts of Radio Channels and Node Mobility on Link
[17] B.-N. Cheng, S. Moore, A comparison of MANET routing protocols on airborne
Statistics in Mobile Ad Hoc Networks, in: IEEE GLOBECOM 2007-2007 IEEE
tactical networks, in: MILCOM 2012  2012 IEEE Mil. Commun. Conf., IEEE,
Glob. Telecommun. Conf., IEEE, 2007: pp. 1206–1210. http://dx.doi.org/10.1109/
2012: pp. 1–6. http://dx.doi.org/10.1109/MILCOM.2012.6415798.
GLOCOM.2007.232.
[18] J. Pojda, A. Wolff, M. Sbeiti, C. Wietfeld, Performance analysis of mesh routing
[3] L. Geng, Y.F. Zhang, J.J. Wang, J.Y.H. Fuh, S.H. Teo, Mission planning of auton-
omous UAVs for urban surveillance with evolutionary algorithms, in: 2013 protocols for UAV swarming applications, in: 2011 8th Int. Symp. Wirel.
10th IEEE Int. Conf. Control Autom., IEEE, 2013: pp. 828–833. http://dx.doi.org/ Commun. Syst., IEEE, 2011: pp. 317–321. http://dx.doi.org/10.1109/ISWCS.2011.
10.1109/ICCA.2013.6564992. 6125375.
[4] B. Coifman, M. McCord, R.G. Mishalani, M. Iswalt, Y. Ji, Roadway traffic mon- [19] S. Rosati, K. Kruzelecki, G. Heitz, D. Floreano, B. Rimoldi, Dynamic Routing for
itoring from an unmanned aerial vehicle, Intell. Transp. Syst. IEE Proc. 153 Flying Ad Hoc Networks, IEEE Trans. Veh. Technol. (2015), http://dx.doi.org/
(2006) 11–20. 〈http://ieeexplore.ieee.org/articleDetails.jsp? 10.1109/TVT.2015.2414819.
arnumber ¼ 1618950〉 (accessed November 8, 2015). [20] Y. Zheng, Y. Wang, Z. Li, L. Dong, Y. Jiang, H. Zhang, A Mobility and Load Aware
[5] I. Jawhar, N. Mohamed, J. Al-Jaroodi, S. Zhang, Data communication in linear Olsr Routing Protocol for Uav Mobile Ad-Hoc Networks, in: Int. Conf. Inf.
wireless sensor networks using unmanned aerial vehicles, in: 2013 Int. Conf. Commun. Technol., pp. 1–7.
Unmanned Aircr. Syst., IEEE, 2013: pp. 492–499. http://dx.doi.org/10.1109/ [21] A. Alshabtat, L. Dong, J. Li, F. Yang, Low latency routing algorithm for un-
ICUAS.2013.6564725. manned aerial vehicles ad-hoc networks, Int. J. Electr. Comput. Eng. 6 (2010)
[6] C. Barrado, R. Messeguer, J. Lopez, E. Pastor, E. Santamaria, P. Royo, Wildfire 48–54 〈http://waset.org/journals/waset/v56/v56-137.pdf〉.
monitoring using a mixed air-ground mobile network, IEEE Pervasive Comput. [22] Y. Li, R. Shirani, M. St-Hilaire, T. Kunz, Improving routing in networks of un-
9 (2010) 24–32, http://dx.doi.org/10.1109/MPRV.2010.54. manned aerial vehicles: reactive-greedy-reactive, Wirel. Commun. Mob.
[7] Y. Cai, F.R. Yu, J. Li, Y. Zhou, L. Lamont, MAC performance improvement in UAV Comput. 12 (2012) 1608–1619, http://dx.doi.org/10.1002/wcm.2333.
ad-hoc networks with full-duplex radios and multi-packet reception cap- [23] Y. Li, M. St-Hilaire, T. Kunz, Enhancements to Reduce the Overhead of the
ability, in: 2012 IEEE Int. Conf. Commun., IEEE, 2012: pp. 523–527. http://dx. Reactive-Greedy-Reactive Routing Protocol for Unmanned Aeronautical Ad-
doi.org/10.1109/ICC.2012.6364116. Hoc Networks, in: 2012 8th Int. Conf. Wirel. Commun. Netw. Mob. Comput.,
[8] R.L. Lidowski, B.E. Mullins, R.O. Baldwin, A novel communications protocol IEEE, 2012: pp. 1–4. http://dx.doi.org/10.1109/WiCOM.2012.6478515.
using geographic routing for swarming UAVs performing a Search Mission, [24] A.E. Tumer, M. Gunduz, An improved leach protocol for indoor wireless sensor
2009 IEEE Int. Conf. Pervasive Comput. Commun. (2009) 1–7. http://dx.doi.org/ networks, in: 2014 Int. Conf. Signal Process. Integr. Networks, IEEE, 2014: pp.
10.1109/PERCOM.2009.4912764. 432–437. http://dx.doi.org/10.1109/SPIN.2014.6776992.
[9] J. Li, Y. Zhou, L. Lamont, M. Déziel, A token circulation scheme for code as- [25] W. Zafar, B.M. Khan, Performance analysis of mobility models for Flying Ad-
signment and cooperative transmission scheduling in CDMA-based UAV ad hoc Networks (FANETs), in: 30th IEEEP Multi-Topic Int. Symp., 2015.
hoc networks, Wirel. Netw. 19 (2013) 1469–1484, http://dx.doi.org/10.1007/ [26] R.R. Roy, Handbook of Mobile Ad Hoc Networks for Mobility Models, Springer,
s11276-013-0545-5. US, Boston, MA 2011, http://dx.doi.org/10.1007/978-1-4419-6050-4.
[10] E.W. Frew, T.X. Brown, Airborne communication networks for small un- [27] IEEE Standard for Information technology– Local and metropolitan area net-
manned aircraft systems, Proc. IEEE 96 (2008) 2008–2027, http://dx.doi.org/
works– Specific requirements– Part 15.4: Wireless Medium Access Control
10.1109/JPROC.2008.2006127.
(MAC) and Physical Layer (PHY) Specifications for Low Rate Wireless Personal
[11] E. Yanmaz, S. Hayat, J. Scherer, C. Bettstetter, Experimental performance ana-
Area Networks (WPANs), (2006) 1–320. http://dx.doi.org/10.1109/IEEESTD.
lysis of two-hop aerial 802.11 networks, in: 2014 IEEE Wirel. Commun. Netw.
2006.232110.
Conf., IEEE, 2014: pp. 3118–3123. http://dx.doi.org/10.1109/WCNC.2014.
[28] C.E. Perkins, E.M. Royer, Ad-hoc on-demand distance vector routing, in: Proc.
6953010.
[12] J.P. Bodanese, G.M. de Araujo, C. Steup, G.V. Raffo, L.B. Becker, Wireless Com- WMCSA’99. Second IEEE Work. Mob. Comput. Syst. Appl., IEEE, 1999: pp. 90–
munication Infrastructure for a Short-Range Unmanned Aerial, in: 2014 28th 100. http://dx.doi.org/10.1109/MCSA.1999.749281.
Int. Conf. Adv. Inf. Netw. Appl. Work., IEEE, 2014: pp. 492–497. http://dx.doi. [29] C.E. Perkins, P. Bhagwat, Highly dynamic Destination-Sequenced Distance-
org/10.1109/WAINA.2014.154. Vector routing (DSDV) for mobile computers, ACM SIGCOMM Comput. Com-
[13] I. Jawhar, N. Mohamed, J. Al-Jaroodi, S. Zhang, A framework for using un- mun. Rev. 24 (1994) 234–244, http://dx.doi.org/10.1145/190809.190336.
manned aerial vehicles for data collection in linear wireless sensor networks, J. [30] P. Jacquet, P. Muhlethaler, T. Clausen, A. Laouiti, A. Qayyum, L. Viennot, Opti-
Intell. Robot. Syst. 74 (2013) 437–453, http://dx.doi.org/10.1007/ mized link state routing protocol for ad hoc networks, in: Proceedings. IEEE
s10846-013-9965-9. Int. Multi Top. Conf. 2001. IEEE INMIC 2001. Technol. 21st Century., IEEE, 2001:
[14] O. Cetin, I. Zagli, Continuous airborne communication relay approach using pp. 62–68.0020. http://dx.doi.org/10.1109/INMIC.2001.995315.
unmanned aerial vehicles, J. Intell. Robot. Syst. 65 (2011) 549–562, http://dx. [31] N. Aschenbruck, R. Ernst, E. Gerhards-Padilla, M. Schwamborn, BonnMotion: a
doi.org/10.1007/s10846-011-9556-6. mobility scenario generation and analysis tool, in: Proc. 3rd Int. ICST Conf.
[15] A.I. Alshbatat, L. Dong, Adaptive MAC protocol for UAV communication net- Simul. Tools Tech., ICST, 2010: p. 51. http://dx.doi.org/10.4108/ICST.SIMU
works using directional antennas, in: 2010 Int. Conf. Networking, Sens. TOOLS2010.8684.

Please cite this article as: W. Zafar, B.M. Khan, A reliable, delay bounded and less complex communication protocol for multicluster
FANETs, Digital Communications and Networks (2016), http://dx.doi.org/10.1016/j.dcan.2016.06.001i