UAVs Traffic Control Based on Multi-Access Edge Computing

Full text

UAVs Traffic Control based on Multi-Access Edge
Computing
Oussama Bekkouche, Tarik Taleb and Miloud Bagaa
Communications and Networking Department, Aalto University, Finland.
Email: firstname.lastname@aalto.fi
Abstract—Given the continuously increasing use of Unmanned
Aerial Vehicles (UAVs) in different domains, their management
in the uncontrolled airspace has become a necessity. This has
given rise to new systems called UAVs Traffic Management
(UTM) systems. Nevertheless, currently, there is a lack of com-
munication infrastructures that can support the requirements
of UTM systems. Luckily, the envisioned 5G mobile network
has introduced the concept of Multi-access Edge Computing
(MEC) in its architecture to support mission-critical applications
by decreasing the end-to-end latency and the unreliability of
communication. In this paper, we evaluate the impact of the
network latency and reliability on the control of UAVs’ flights.
The obtained results show that a UAV can deviate from its
intended path with more than 5mif the network latency exceeds
400ms and with more than 2mif the packet loss probability
exceeds 0.2. To overcome these limitations, we have leveraged
MEC to provide a new UTM framework that enables an efficient
traffic management. Moreover, due to MEC resource-limited
nature and in order to give an insight about the resource
provisioning, we have evaluated the scalability of the proposed
solution in terms of the number of UAVs that can be handled
without affecting the efficiency of the proposed UTM framework.
I. INTRODUCTION
Unmanned Aerial Vehicles (UAVs), commonly known as
drones, are one of the emerging technologies that attract a lot
of industrial and academic interests. The UAVs market has
been valued at $18 billion in 2017 and is expected to reach
$52 billion by 2025 [1] given their endless applications in
numerous domains, such as military reconnaissance, inspec-
tion of infrastructures, smart agriculture, traffic management,
border surveillance and cargo delivery [2], [3].
Despite the several benefits and practical applications of
UAVs, several challenges are yet to be addressed before the
final integration of these UAVs in our everyday lives. Indeed,
today’s commercialized Unmanned Aerial Systems (UASs)
are qualified to be semi-autonomous as the intervention of a
human pilot equipped with a Ground Control Station (GCS)
within a Visual-Line-of-Sight (VLOS) to the UAVs is required
for either programming a predefined mission or controlling
the UAVs in a real-time fashion, which hinders the full
exploitation of UAVs’ potential.
From another side, human-controlled UAVs may cause
several security and privacy issues. According to the Federal
Aviation Administration (FAA) [4], more than 100 reports
are received each month signaling the sighting of UAVs by
pilots of commercial aircrafts. This is mainly due to the fact
that each UAV is operated manually and individually without
taking into consideration the state of air traffic and the geo-
fences. Furthermore, keeping a UAV within a VLOS to the
human-pilot, during missions in challenging environments,
presents high risks in the context of safety. For these rea-
sons, the introduction of a system that can ensure efficient,
intelligent, secure and reliable control of UAVs Beyond the
Visual Line-of-sight (BVLOS) is mandatory. In this context,
the existing Air Traffic Management (ATM) systems rely on
voice communication between air traffic controllers and pilots,
and not designed to handle the expected high density of UAVs
traffic [5]. This has given rise to new systems called UAV
Traffic Management (UTM) systems. Several UTM projects
are being developed [6], but the most advanced and mature
one is the UTM system proposed by the National Aeronautics
and Space Administration (NASA) in collaboration with the
FAA [7].
The main functionalities that a standard UTM must offer are
the identification, the localization and the steering of UAVs,
which means that direct communication between the UTM
and the UAVs is required. This communication is qualified
to be an Ultra-Reliable and Low-Latency Communication
(URLLC), as the loss or the lateness of a packet may cause
several damages. Cellular mobile networks are well placed to
support the communication between the UTM services and
UAVs. Indeed, mobile networks can ensure a scalable, secure
and ready-for-use connectivity to UAVs using a licensed
spectrum. Moreover, cellular-connected UAVs are considered
as an Aerial User Equipment (AUE) that can benefit from
all services provided by the mobile operators to conventional
User Equipment (UE), such as localization, identification, and
secure communication. These services can be harnessed by the
UTM.
In spite of the advantages that current mobile networks
offer, their performances dramatically drop when used to serve
AUE with URLLC. For instance, the work in [8] presents
the challenges that face the integration of UAVs in the LTE
mobile network. Fortunately, the envisioned 5G system had
introduced a set of new paradigms that would be of benefit
to cellular-connected UAVs.
Multi-access Edge Computing (MEC) is one of the most
exciting paradigms in the era of 5G and Internet-of-Things
(IoT). Leveraging MEC technology, cloud-like services are
pushed to the network edges (e.g., base stations) [9], [10].
This will reduce the latency and increase the reliability of
the communication between the applications running on the
978-1-5386-4727-1/18/$31.00 ©2018 IEEE

connected devices (e.g., UAVs and IoT devices) and the
remote services hosted at the edge network. As a result, MEC
is well placed to host the UTM services.
The primary contributions of this paper are two-fold. First,
we investigate the effects of the latency and the reliability of
the communication between the UTM services and the UAVs
on the management of the air traffic. Second, we propose a
high-level architecture for a MEC-based UTM inspired by the
NASA’s UTM architecture, followed by its evaluation using
a real MEC server in order to give an insight about resources
provisioning for such services.
The remainder of this paper is organized as follows. Section
II discusses some related research work. Section III investi-
gates the communication requirements in terms of latency and
packet loss for an efficient UAVs’ air traffic management. An
overview of the proposed architecture, along with its main
components, is presented in Section IV. Section V illustrates
the experimental setup and discusses the obtained results.
Finally, Section VI concludes the paper.
II. RE LATE D WO RK
A lot of research work has been conducted to investigate
how the current and the next generations of mobile networks
can be harnessed to provide an efficient backend for the
autonomous systems in general, and for UAVs in particular.
Sasaki et al. [11] have proposed a vehicle control system
coordinated between cloud and MEC. The presented work
investigates the effect of the latency on the stability of the
driving trajectory by applying two control methods. In the
first method, the vehicle controller is hosted in the cloud far
away from the vehicle, whereas in the second method the
vehicle controller is hosted in a MEC node at the edge of
the network. The obtained results show that the vehicle has
higher stability when controlled from a MEC node.
The work presented in [12] discussed another layered
vehicle control system, coordinated between multiple edge
servers to overcome the limitations induced by the resource-
restricted nature of MEC nodes. In [13], a cloud-based MEC
offloading framework in vehicular networks is proposed to
minimize the total cost of the offloading process. Aissioui
et al. [14] discuss the use of the Follow Me edge-Cloud
(FMeC) concept to enable the migration of vehicular services
among the edge nodes, in a way that the services follow the
mobile vehicles. The idea of offloading heavy computation
from UAVs to a MEC node, discussed in [15], shows the
benefits of MEC in terms of energy saving and response
time improvement. Furthermore, a 5G-based framework for
preventive maintenance of critical infrastructure using UAVs
is proposed in [16]. This framework takes advantages of the
envisioned 5G network architecture to introduce an extended
version of MEC that can support the real-time control of
UAVs, as well as the processing and the analysis of their
collected data. The research work, presented in [17] and
[18], addressed the optimization of UAVs’ flight trajectory
for an efficient computation offloading at the edge of multiple
cells. Moreover, UAVs’ control and tasks offloading require
a reliable connection to the mobile network. For this matter,
Motlagh et al. [19] propose an efficient connection steering
mechanism between mobile networks for reliable UAV-based
communication. Work in [20] discusses the mitigation of the
interferences created by cellular connected UAVs by adjusting
the transmission power. 3GPP in [8] presents an advanced
study of the performance of Release-14 LTE networks when
used to serve UAVs, and identify the possible performance-
enhancing solutions.
III. UAV TRAFFI C MANAGE ME NT: COMMUNICATION
PERSPECTIVE
The management of UAVs’ traffic is mainly based on the
dynamic definition of aerial geo-fences to specify the allowed
and the restricted flight zones. For this matter, the UTM
controls the flight plan of each UAV to ensure that it is
flying inside an allowed geo-fence. However, the restriction
of the airspace may influence the services provided by the
UAV. e.g., a UAV that is charged to collect data from a
specific region cannot perform its mission if this region is
situated inside a restricted geo-fence. Therefore, the UTM
must capitalize on the allowed airspace; so it must be able
to efficiently control the UAVs even in the borders of the
allowed geo-fence and without crossing the borders. In order
to achieve such level of control, an acceptable Quality-of-
Service (QoS) in terms of latency and packet loss is required
for the communications between the UTM and the UAVs. In
this section, we investigate the impact of the latency and the
packet loss on the remote control of the UAVs’ flights.
A. Methodology of experimentation
Fig. 1: Experimentation methodology.
In order to investigate the effect of latency and packet
loss on the efficiency of UAVs’ control, we consider the
most challenging case of control. As illustrated in Fig. 1,
a UAV must fly on the border of an allowed geo-fence,
then the stability of the flight is evaluated by measuring two
parameters:
•The deviation ratio Dr: This parameter represents the
percentage of the distance traveled outside the allowed
geo-fence, which can be calculated using the following
equation:
Dr=OD×100
TD

where ODis the distance traveled outside the allowed
geo-fence and TDis the total traveled distance.
•The average of crossing distances Cavg: This parameter
represents the average of the distances by which the
UAV has crossed the allowed geo-fence. To calculate this
parameter, periodical measurements of the distance that
separates the UAV from the borders of the allowed geo-
fence are performed. Then Cavg is calculated using the
following equation:
Cavg =
Nbm
P
i=1
Mi
Nbm
where Miare the measured values and Nbmis the
number of the performed measurements.
B. Experimentation setup
To evaluate the previous parameters, we emulate the remote
steering of UAVs using the Software-In-the-Loop (SITL)
emulator that allows running the open-source UAVs’ autopilot
ArduCopter in a normal PC without any dedicated hardware
and with the same behavior of a real UAV. The experimen-
tation setup consists of a UAV Controller (UC) developed
using the Python drone-kit API which is an implementation
of the MAVLink protocol [21], used for the communication
with drones, and a SITL-emulated UAV. The communication
traffic between the UC and the emulated UAV pass through
a network emulator to simulate the communication latency
and packet loss. we have used three virtual machines (VMs)
deployed on top of our cloud network that uses Openstack as
virtual infrastructure manager (VIM). While the first VM is
used as a UAV controller, the second VM plays the role of
the network emulator, whereby we can enforce some rules to
change the network delay and reliability. Meanwhile, the third
VM runs SITL emulator for mimicking the UAV’s behavior.
The UC defines a circular geo-fence with a radius of
200mand drives the simulated UAV through its borders at
a speed of 10m/s during a period of 300s. The developed
UC functions with a frequency of 100Hz, which means that
it checks the state of the UAV (i.e., position and velocity)
each 10ms and sends commands accordingly to correct any
possible deviation.
The measurements required to calculate the Cavg parameter
are performed by the UC itself, whereas the post-processing of
the UAVs’ data-flash logs, using the Mission Planner software,
allows us to reproduce the flight trajectory of the UAV in order
to compare it with the borders of the geo-fence and calculate
the evaluation parameter Drusing Google Earth software.
C. Results discussion
1) Effect of latency: Fig. 2a illustrates the trajectory trav-
eled by the UAV (the yellow circle) versus the desired
trajectory (the red circle) in the presence of high latency.
The first observation that can be drawn from the figure is
that the latency has a negative impact on the control of the
UAV flight. Indeed, the UAV has crossed the borders of the
allowed geo-fence in almost all the flight trajectory. Fig. 2c
shows that the deviation ratio Drincreases considerably from
50% to 89% when we vary the latency from 0ms to 400ms.
Moreover, as depicted in Fig. 2b, the average of crossing
distances Cavg increases proportionally with the increase of
the latency, which means that a high latency will cause the
crossing of borders with higher distances.
2) Effect of packet loss: Fig. 3c and Fig. 3b, respectively,
show that the deviation ratio Drand the average of crossing
distances Cavg increase gradually when we vary the packet
loss percentage from 0% to 20%. It should be noted that it
is not possible to establish any connection between the UC
and the UAV when the packet loss ratio is higher than 20%.
From the obtained results of Fig. 3, we have noticed that the
impact of packet loss on the flight trajectory is less than the
impact of the latency. This is due to the fact that the UC
functions with a relatively high frequency (i.e., 100Hz), so it
is able to correct any deviation caused by the loss of a control
packet within a time interval of 10ms. Nevertheless, the UC
is ineffective when facing consecutive losses of packet. As a
result, peaks of deviation are observed on the flight trajectory
of the UAV (points A and B in Fig. 3a).
In these experiments, we measured the exact averages
of crossing distances and the deviation ratios for different
values of network latency and packet loss. This would help to
choose the placement of the remote UAV controller according
to the tolerated deviation.
IV. MEC-BA SE D UAVS’TRAFFIC MANAGEMENT
FRAMEWORK
Based on the aforementioned results, the latency and packet
loss have a dramatical impact on the control of the UAVs’
flights. For this reason, in this section, we propose a new
framework that leverages MEC for ensuring that the UAVs
stay within their defined geo-fences as much as possible. In
fact, the use of MEC for the management of UAVs’ traffic
would alleviate the problems caused by the poor quality
of communication. Indeed, many issues known in legacy
networks, such as congestion and packet loss, are eliminated
when the UAVs are controlled from edge-hosted services.
As shown in Fig. 4 the architecture of the proposed frame-
work can be divided into three parts.
A. Cloud domain
The cloud domain hosts all the management services. Here,
three principal services can be identified:
•UAVs Traffic Management Service (UTMS): The
UTMS has all the information regarding the air traffic,
namely registered UAVs, UAVs’ locations, UAVs’ flight
plan and airspace restrictions. Also, this service is re-
sponsible for defining static and dynamic geo-fences.
•Supplementary Data Provider Service (SDPS): This
service provides additional information that may be
useful for planning the UAVs’ flights, such as weather
forecasts and locations of interest (e.g., accidents and
disasters locations, service requester’ s location).

(a) Desired (red) vs Traveled (yellow)
trajectory in presence of high latency.
1
1.5
2
2.5
3
3.5
4
4.5
5
5.5
0 50 100 150 200 250 300 350 400
Average of crossing distances (m)
Latency (ms)
Average of crossing distances
(b) Average of crossing distances.
50
55
60
65
70
75
80
85
90
0 50 100 150 200 250 300 350 400
Deviation ratio (%)
Latency (ms)
Deviation ratio
(c) Deviation ratio.
Fig. 2: Effect of latency.
(a) Desired (red) vs Traveled (yellow)
trajectory in presence of Packet loss.
1.3
1.4
1.5
1.6
1.7
1.8
1.9
0 5 10 15 20
Average of crossing distances(m)
Packet loss (%)
Average of crossing distances
(b) Average of crossing distances.
50
55
60
65
70
75
80
0 5 10 15 20
Deviation ratio (%)
Packet loss (%)
Deviation ratio
(c) Deviation ratio.
Fig. 3: Effect of Packet loss.
•Operator Command and Control Service (OCCS):
The OCCS provides an interface for the remote operators
to monitor and interact with UAVs in real time when
needed.
Fig. 4: Architecture of the proposed framework.
B. Core and transport network
The core and transport networks ensure the transmission of
communication traffic between MEC-hosted services and the
previous Cloud-hosted services. Given that this communica-
tion belongs to URLLC, the generated traffic must be treated
with a high QoS. In this context, Software-defined Networking
(SDN) is one of the emerging technologies introduced in the
architecture of the next generation mobile networks [22] to
ensure a flexible QoS provisioning [23], [24]. In our proposed
framework, the core and transport networks are considered to
be SDN-enabled networks.
C. Mobile Edge Computing
In our proposed framework, MEC nodes are used to host the
UAVs Flight Controller service (UFC). The UFC service is
responsible for the monitoring and the control of the on-going
flights of the set of UAVs connected to the access point associ-
ated with the MEC node. Indeed, the UFC collects information
(i.e., geo-fences, weather forecasts, and locations of interest)
and commands from cloud-hosted services, and accordingly
adapts UAVs’ flights. For example, when receiving a new geo-

fence event from the UTMS, the UFC will enforce all the
connected UAVs to respect this new geo-fence. Furthermore,
the UFC service forwards the telemetry data streamed from
the UAVs to the UTMS service. Thanks to the deployment of
UFC at the edge of the network, the communication latency
should be close to zero, which is the best possible case.
V. EVAL UATION
Despite the advantages of using MEC for the next gener-
ation of mobile applications, the resource-limited nature of
the edge servers imposes several constraints. Indeed, it is not
possible to install powerful servers and data-centers on all
network edge. Therefore, an optimal resource provisioning is
required before the final exploitation of the edge services. In
this section, we investigate the computational and network
resources required by the framework proposed in Section IV.
For this matter, we have considered the same experimental
setup and scenario discussed in Section III, and we improve
the UAV controller to meet the specification of the UFC
service, so it would be able to control multiple UAVs at the
same time by the creation of multiple control agents (CAs).
Furthermore, the UFC service is running in a real MEC server
(i.e., Intel Fog Reference Design as depicted in Fig. 5) 1,
instead of running in a virtual server hosted in the cloud.
Also, as this experiment mimics the real case of MEC-
controlled UAV, the communication latency and the packet
loss are considered to be nil. Table. I summarizes the different
parameters of our experiment testbed.
Fig. 5: The Edge server used in our testbed.
TABLE I: Experimentation parameters.
Parameter Description Values
CPU
cores
The number of physical CPU
cores in the edge server.
4
RAM Available RAM in the edge
server.
32 GB
Throughput Available throughput for the
communication between the
edge server and the UAVs.
100 Mbps
UAVs
number
The number of UAVs connected
to edge.
20, 40, 60, 80, 100
Speed The speed of the UAVs 10 m/s
Duration The duration of the flights and
the simulations.
300 s
Radius The radius of the allowed geo-
fence.
200 m
1The Fog Reference Design is not a product sold by Intel, and is instead
a reference design offered to select industry leaders to enable quick Edge
product development.
Fig. 6: CPU usage VS average of crossing distances.
Fig. 6 shows the CPU percentage allocated to a control
agent (CA) versus the average of the Cavg values of the set
of UAVs connected to the edge server. The first observation
that can be drawn from this figure is that the allocated CPU
percentage for each CA decreases from 7.7% to 2.9% when
we vary the number of UAVs from 20 to 100. This is mainly
due to the fact that the CPU time is fairly shared among the
increased number of CAs. The decrease of CPU percentage
allocated to the CAs is accompanied by an increase of the
average of crossing distances from 1.36mto 2.24m, which
means that the flight stability is highly impacted by the
computational resources allocated to the CA. It should be
noted that a CA requires at most 8% of CPU usage. As a
result, the resource provisioning at the edge must aim to keep
this value constant regardless the number of connected UAVs.
0
1e+08
2e+08
3e+08
4e+08
5e+08
6e+08
7e+08
8e+08
20 30 40 50 60 70 80 90 100
Number of bytes
Number of UAVs.
Edge received bytes
Edge sent bytes
Fig. 7: Exchanged traffic.
Fig. 7 illustrates the evolution of the traffic exchanged be-
tween the edge server and the connected UAVs as a function of
the number of UAVs. We notice that the amounts of received
and sent bytes increase linearly along with the increase of
the number of connected UAVs. The extracted slopes are
αr= 3028437 and αs= 7152453 for received and sent bytes,
respectively. As a result, each UAV has generated around
3028Kb of traffic during the experiment time (300s), which
corresponds to a rate of 10Kbps, and has received around

7152Kb of traffic, which corresponds to a rate of 24Kbps.
In light of these results, the network provisioning must aim
to ensure at least the previous communication throughputs in
both the wireless and wired networks connecting to the MEC
nodes.
VI. CONCLUSION
Along with the ever-increasing number of UAVs, the on-
site control of UAVs, by pilots within VLoS, is becoming
all but impossible. In this vein, UAVs Traffic Management
(UTM) systems are gaining lots of momentum. However,
UTM systems can be affected by different parameters, such
as the network state and weather conditions. In this paper, we
have studied the impact of the network parameters, in terms
of communication latency and reliability, on the efficiency of
the UTM systems. We have shown that the control of UAVs is
negatively impacted by the increase of the latency and packet
loss rate. To overcome these limitations, we have proposed a
new UTM framework that enhances the existing approaches
by leveraging MEC technology. In fact, in the proposed frame-
work, the UAVs Flight Controller (UFC) service has been
placed at the edge near to the UAVs to ensure highly reliable
and low latency communication. Moreover, in this paper, we
have evaluated the scalability of the proposed framework by
varying the number of UAVs connected to the MEC and
measuring the computational and network resources consumed
during the management process. The obtained results give an
insight about the required resource provisioning at the edge
of the network in order to ensure an efficient UAVs’ traffic
management.
Whilst the performance evaluation was conducted using one
single UAV and one single Edge node, as future research
work, we intend using more realistic scenarios involving
multiple UAVs flying over a wide region served by multiple
edge nodes and that is leveraging our Mobile Service Usage
Cartography tool. We also plan devising and evaluating dif-
ferent algorithms for the placement of UFC services across
these edge nodes as per the mobility features of UAVs [25].
ACKNOWLEDGMENT
This work is partially supported by the European Unions
Horizon 2020 research and innovation programme under the
5G!Pagoda project with grant agreement No. 723172. It is also
supported in part by the Aalto 5G meets Industrial Internet
(5G@II) project.
REFERENCES
[1] (2018, February) Unmanned aerial vehicle (uav) market by system uav
platforms, uav payloads - 2025 - marketsandmarkets. Market and Mar-
ket. [Online]. Available: https://www.marketsandmarkets.com/Market-
Reports/unmanned-aerial-vehicles-uav-market-662.html
[2] N. H. Motlagh, T. Taleb, and O. Arouk, “Low-altitude unmanned aerial
vehicles-based internet of things services: Comprehensive survey and
future perspectives,” IEEE Internet of Things Journal, vol. 3, no. 6, pp.
899–922, Dec 2016.
[3] S. Ouahouah, T. Taleb, J. Song, and C. Benzaid, “Efficient offloading
mechanism for uavs-based value added services,” in 2017 IEEE Inter-
national Conference on Communications (ICC), May 2017, pp. 1–6.
[4] (2017, December) Uas sightings report. Fed-
eral Aviation Administration. [Online]. Available:
https://www.faa.gov/uas/resources/uas sightings report/
[5] (2018, February) Mobile-enabled unmanned aircraft. GSMA. [On-
line]. Available: https://www.gsma.com/iot/mobile-enabled-unmanned-
aircraft/
[6] H. Hellaoui, O. Bekkouche, M. Bagaa, and T. Taleb, “Aerial control
system for spectrum efficiency in uav to cellular communications,” To
appear in IEEE Communications Magazine.
[7] (2018, April) Nasa utm :documents. National Aero-
nautics and Space Administration. [Online]. Available:
https://utm.arc.nasa.gov/documents.shtml
[8] Study on Enhanced LTE Support for Aerial Vehicles, 3GPP TR 36.777
V1.0.0, 2017.
[9] T. Taleb, K. Samdanis, B. Mada, H. Flinck, S. Dutta, and D. Sabella,
“On multi-access edge computing: A survey of the emerging 5g net-
work edge cloud architecture and orchestration,” IEEE Communications
Surveys Tutorials, vol. 19, no. 3, pp. 1657–1681, thirdquarter 2017.
[10] P. Porambage, J. Okwuibe, M. Ylianttila, and T. Taleb, “Survey on
multi-access edge computing for internet of things realization,” To
appear in IEEE Communications Surveys Tutorials.
[11] K. Sasaki, N. Suzuki, S. Makido, and A. Nakao, “Vehicle control system
coordinated between cloud and mobile edge computing,” in 2016 55th
Annual Conference of the Society of Instrument and Control Engineers
of Japan (SICE), Sept 2016, pp. 1122–1127.
[12] ——, “Layered vehicle control system coordinated between multiple
edge servers,” in 2017 IEEE Conference on Network Softwarization
(NetSoft), July 2017, pp. 1–5.
[13] K. Zhang, Y. Mao, S. Leng, Y. He, and Y. ZHANG, “Mobile-edge
computing for vehicular networks: A promising network paradigm with
predictive off-loading,” IEEE Vehicular Technology Magazine, vol. 12,
no. 2, pp. 36–44, June 2017.
[14] A. Aissioui, A. Ksentini, A. Gueroui, and T. Taleb, “On enabling
5g automotive systems using follow me edge-cloud concept,” IEEE
Transactions on Vehicular Technology, vol. PP, no. 99, pp. 1–1, 2018.
[15] N. H. Motlagh, M. Bagaa, and T. Taleb, “Uav-based iot platform: A
crowd surveillance use case,” IEEE Communications Magazine, vol. 55,
no. 2, pp. 128–134, February 2017.
[16] T. Zahariadis, A. Voulkidis, P. Karkazis, and P. Trakadas, “Preventive
maintenance of critical infrastructures using 5g networks drones,” in
2017 14th IEEE International Conference on Advanced Video and
Signal Based Surveillance (AVSS), Aug 2017, pp. 1–4.
[17] X. Cao, J. Xu, and R. Zhang, “Mobile Edge Computing for Cellular-
Connected UAV: Computation Offloading and Trajectory Optimization,”
ArXiv e-prints, Mar. 2018.
[18] F. Cheng, S. Zhang, Z. Li, Y. Chen, N. Zhao, R. Yu, and V. C. M.
Leung, “Uav trajectory optimization for data offloading at the edge of
multiple cells,” IEEE Transactions on Vehicular Technology, vol. PP,
no. 99, pp. 1–1, 2018.
[19] N. H. Motlagh, M. Bagaa, T. Taleb, and J. Song, “Connection steering
mechanism between mobile networks for reliable uav’s iot platform,” in
2017 IEEE International Conference on Communications (ICC), May
2017, pp. 1–6.
[20] H. Hellaoui, A. Chelli, M. Bagaa, and T. Taleb, “Towards mitigating
the impact of uavs on cellular communications,” in GLOBECOM 2018
- 2018 IEEE Global Communications Conference, Dec 2018, pp. 1–6.
[21] (2018, March) Mavlink micro air vehicle communication protocol.
Dronecode. [Online]. Available: https://mavlink.io/en/
[22] P. K. Agyapong, M. Iwamura, D. Staehle, W. Kiess, and A. Benjebbour,
“Design considerations for a 5g network architecture,” IEEE Commu-
nications Magazine, vol. 52, no. 11, pp. 65–75, Nov 2014.
[23] D. L. C. Dutra, M. Bagaa, T. Taleb, and K. Samdanis, “Ensuring
end-to-end qos based on multi-paths routing using sdn technology,” in
GLOBECOM 2017 - 2017 IEEE Global Communications Conference,
Dec 2017, pp. 1–6.
[24] I. Farris, T. Taleb, Y. Khettab, and J. S. Song, “A survey on emerging
sdn and nfv security mechanisms for iot systems,” To appear in IEEE
Communications Surveys and Tutorials.
[25] A. Laghrissi, T. Taleb, M. Bagaa, and H. Flinck, “Towards edge
slicing: Vnf placement algorithms for a dynamic amp;amp; realistic
edge cloud environment,” in GLOBECOM 2017 - 2017 IEEE Global
Communications Conference, Dec 2017, pp. 1–6.