Received April 9, 2019, accepted May 10, 2019, date of publication May 15, 2019, date of current version

May 30, 2019.

Digital Object Identifier 10.1109/ACCESS.2019.2917070

Performance Evaluation of Multi-UAV System

in Post-Disaster Application: Validated
by HITL Simulator
MAHER ALJEHANI , (Student Member, IEEE), AND MASAHIRO INOUE, (Senior Member, IEEE)
Shibaura Institute of Technology, Saitama 337-8570, Japan
Corresponding author: Maher Aljehani (nb16507@shibaura-it.ac.jp)
This work was supported by the JSPS KAKENHI under Grant JP15k00929 and Grant JP19K0315.

ABSTRACT This paper proposes an evaluation of unmanned aerial vehicles (UAVs) performance in the
mapping of disaster-struck areas. Sendai city in Japan, which was struck by the Tohoku earthquake/tsunami
disaster in 2011, was mapped using multi-heterogeneous UAV. Normal mapping and searching missions are
challenging as human resources are limited, and rescue teams are always needed to participate in disaster
response mission. Mapping data and UAV performance evaluation will help rescuers to access and commence
rescue operations in disaster-affected areas more effectively. Herein, flight plan designs are based on the
information recorded after the disaster and on the mapping capabilities of the UAVs. The numerical and
statistical results of the mapping missions were validated by executing the missions on real-time flight
experiments in a simulator and analyzing the flight logs of the UAVs. After considering many factors and
elements that affect the outcomes of the mapping mission, the authors provide a significant amount of useful
data relevant to real UAV modules in the market. All flight plans were verified both manually and in a
hardware-in-the-loop simulator developed by the authors. Most of the existing simulators support only a
single UAV feature and have limited functionalities such as the ability to run different models on multiple
UAVs. The simulator demonstrated the mapping and fine-tuned flight plans on an imported map of the
disaster. As revealed in the experiments, the presented results and performance evaluations can effectively
distribute different UAV models in post-disaster mapping missions.

INDEX TERMS Path planning evaluation, UAVs, mapping missions, disaster response, searching and rescue
mission, flight plan design, hardware in the loop simulator.

I. INTRODUCTION delivery services. Furthermore, UAV can be merged with

Unmanned aerial vehicles (UAVs) are a futuristic technology other technologies such as the internet of things (IoT), M2M
and a potential solution to many humanitarian problems. communication systems, networking, functional control sys-
By virtue of their high-quality and low-cost sensors, UAVs tems, system models, and satellite systems. Also, UAVs can
have become a low-cost and accessible tool in academic and contribute to mission-response assessment after a disaster.
industrial research. Since their first deployment as air-force Specifically, impacted residential areas can be evaluated after
weapon systems in the early 1960s [1], UAVs have been con- a mapping mission. Damage evaluation relies on aerial and
tinually developed for military operations. The technology terrestrial images of the disaster-struck areas [2]. From these
was initially privatized by governments due to its expense imagery data, an updated map (safe map) can be built and
and significant security threat. After 50 years of develop- new rescue routes can be produced [3]. There are various
ment, UAVs have entered the public domain, where they UAV designs, each with a unique flight behavior [4]. Also,
have been widely developed by universities, companies, and flight plan designs should consider UAV’s mapping capabil-
research centers. UAV deployment has now been extended ities and area’s features. This paper introduces system model
to hobbies such as photography and racing, and to business and evaluations of flight performance of UAVs to map a
applications such as agriculture, mapping, surveying, and disaster area in which based in two different types of UAVs
on the market (i.e., fixed wing and multi-rotor designs) map
The associate editor coordinating the review of this manuscript and the impacted areas in a disaster scenario in Japan [5]. The
approving it for publication was Zhen Li. framework system model considers factors that influence the
2169-3536
2019 IEEE. Translations and content mining are permitted for academic research only.
64386 Personal use is also permitted, but republication/redistribution requires IEEE permission. VOLUME 7, 2019
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
M. Aljehani, M. Inoue: Performance Evaluation of Multi-UAV System in Post-Disaster Application

FIGURE 1. Scenario of a post-disaster mapping mission undertaken by fixed-winged and drone UAV models.

flight plan design. The experimental area had been stricken obstacle environments and solves the optimization problem
by the Tohoku earthquake and tsunami in 2011. Parts of the using stochastic methods. Energy consumption and the effec-
damaged areas are mapped by our system module, which was tiveness of path planning have also been researched. The
applied on our developed simulator. Figure 1 illustrates the algorithm in [14], minimizes the energy consumption of
surveying of the disaster area by the multi-UAV system. UAVs deployed in missions. This algorithm considers the
The remainder of this paper is organized as follows, velocity and acceleration parameters of the mission and plans
Section One continues below with an overview of related the path that best minimizes the energy consumption without
works and a formal problem statement. Section Two describes affecting the missions’ outcomes.
the system of the developed simulator, which is used in In autonomous missions, a UAV mainly uses its GPS and
the results and experimental validation of the UAV system compasses to navigate the given waypoints. Real-time kine-
and Section Three presents the fine-tuning process UAV and matic (RTK) GPS receivers can enhance the precision of GPS
camera specification. Section Four presents the experimental systems from meters (in conventional GPS) to centimeters.
area and flight plan design. The experiments and their results Other sensors that detect optical flow, ultrasonic, and dis-
are discussed in Section Five. tance (LIDARs) ensure a stable flight experience. In a related
research project, Razi et al. [15] employed a UAV in a 3D
A. RELATED WORK photogrammetry technique that validates the methodology
The deployment of multi-UAV systems in disaster response of monitoring an area. After mapping the land deforma-
missions is a relatively innovative solution faced with many tion and conducting a terrestrial survey, they reported more
practical challenges, such as the effectiveness and stability of precise monitoring results by 3D photogrammetry than by
the operations during the mission, the network and commu- Advanced Land Observation Satellite Phased Array L-band
nication system, the flight plan design, energy requirements, Synthetic Aperture Radar (ALSO PALSAR) [15]. Another
and mission management [6]. When integrating a multi-UAV study confirmed the excellent performance of UAVs in dis-
system into a post-disaster application, the area of interest aster recovery networks, regardless of whether the disaster is
must be entirely covered. Coverage problems in static wire- natural or man-made [16]. In that study, the multiple UAVs
less sensor networks and the coverage enhancement of mobile distributed in the disaster area acted as a relay between the
nodes have been well researched in several studies [7]–[10]. surviving mobile-base stations. In [16] a stochastic geometric
Communication and networking (particularly the commu- framework based on clustered deployment of drone small
nication barriers) in UAV systems are discussed in depth cells around the site of a destroyed base station is considered.
in the next subsection. When planning the flight path of a The UAVs significantly enhanced the quality of information
UAV system with new modules, the designer must consider delivered to the ground users and the surviving base stations
the high mobility and flight time while minding the limited in post-disaster and other unforeseen events. UAVs are suit-
battery life [11] The authors of [11] proposed two methods able not only for mapping and surveying missions, but also for
for tackling the dynamic coverage problem of multi-UAVs: tracking humans after a disaster [3], [17]. In [17], a group of
dividing the area into subareas and planning the flight path of UAVs semi-autonomously traced the paths of human refugees
each UAV serving a designated sub-area in a two-dimensional and generated escape routes based on their movements, while
scenario. Occlusion-awareness of a multi-UAV system has other groups of UAVs surveyed the area. Human planners
been addressed in [12]. However, the system in [12] covers then examined these routes along with aerial image data, and
a specific area and the waypoints are covered several times created a safe map. That safe map consists of new routes
without considering the coverage quality. In another study, based on evaluation from aerial images. These images and
a flight plan for two-dimensional mapping was designed by a highlighted routes can be uploaded to a social network service
genetic algorithm [13]. This planning considers undiscovered for easy access by endangered residents.

VOLUME 7, 2019 64387

 M. Aljehani, M. Inoue: Performance Evaluation of Multi-UAV System in Post-Disaster Application

B. COMMUNICATION SYSTEM RELATED WORKS between UAVs and the GCS in the impacted area, and pro-
Several investigations have focused on the communication poses a system model for UAV-based disaster missions.
systems of multi-UAV systems in post-disaster operations.
Many considerations and scenarios must be addressed when C. POST-DISASTER SIMULATIONS
developing communication systems. Fortunately, this prob- This subsection addresses the importance of simulations in
lem has been well studied, and several frameworks, solutions, disaster applications. A simulator helps to verify the system
and system models have been proposed to overcome the models in the post-disaster scenario. The authors of [24]
communication delays in multi-UAV systems. The authors deployed UAVs as a wireless sensor network in a post-disaster
of [18] introduced a system model and flight mechanisms situation. Simulations are useful not only in communication
for wireless-enabled UAVs. They designed a communication scenarios but also for navigation and performance evaluations
system that collects information from the ground station of disaster applications. For instance, the authors of [25]
under different considerations, such as power constraints and evaluated the efficiency and navigation performance of UAVs
collision avoidance. Their iterative algorithm jointly opti- in a robotic simulator. By running artificial intelligence algo-
mizes the communication schedule, power allocation and rithms in the same simulator, they also established the UAV
flight plan designs to improve the performance between control settings that maximize the communication service
ground terminals. However, the UAVs altitude was fixed to coverage in a disaster scenario. Christy et al. [26] presented
the values that ensured collision avoidance of the UAVs. UAVs as a flying mobile-base tower station for device-to-
Delay between the ground control station (GCS) and online device finding in a disaster area. In simulations, they verified
UAVs due to low bandwidth during the mission is another that the flight plans were enhanced and that the coverage of
problem related to the communication system in various the impacted area exceeded 80% with relatively low power
applications. The authors of [19] and [20] proposed an consumption. As demonstrated in these examples, simula-
event-triggered heterogeneous nonlinear filter and a particle tions are widely used in UAV-based post-disaster response,
filter, respectively, for real-time dynamic estimation in wide- mapping, and communication. However, finding a suitable
area measurement. With its nonlinear master-slave struc- UAV (i.e., multi-UAV, customized camera, and UAV param-
ture, the filter in [20] considers the communication and eters) for a specific usage in disaster applications is a chal-
computation power generated from the node. This design lenging task. Moreover, most of the existing simulators have
delivers high accuracy while relieving the communication limited functionalities for mapping-mission support. There-
burden. Meanwhile, the event-triggered particle filter in [20] fore, their simulator was developed to verify the evaluation
obtains the real-time state with suitable estimation accuracy, results of the UAV performance in the present study.
reduces the communication burden, and overcomes the lim-
ited transmission capability of UAVs in their network envi- D. PROBLEM STATEMENT
ronment. Tuna et al. [21] proposed a UAV-aided emergency Although many researches and studies have tested the prac-
communication system and end-to-end communications in ticality of deploying UAV systems in post-disaster response
post-disaster scenarios. In experiments and simulations, the missions, flight planning considering capabilities of UAV’s
UAV-aided communication system proved its feasibility as parameters needs further evaluation. Essential challenges of
a post-disaster emergency communications solution. UAVs UAV’s parameters need to discuss the efficiency of the flight
can also be integrated with wireless sensor networks for performance and the number of UAV based on a scenario of
natural disaster management. The main contributions of UAV, disaster area. UAV’s coverage capabilities must be analyzed
along with their unsolved challenges such as coverage, dis- statistically before implementation in the mission and the
connectivity, security and privacy, and quality of service, limit performance must be revealed to effectively map the
are discussed in [22]. In a previous work, the authors com- area. Such system models and performance evaluations are
bined three communications techniques into a hybrid com- missing and have not been systematically presented for het-
munication system [17]. A default communication module erogeneous UAV system. In an efficient flight plan, the UAVs
(i.e., a cellular network) was configured with two communi- should navigate only the critical areas, particularly in areas
cation modules (a wireless and a radio control module), which with a complex geometry accommodating various popula-
can be activated when the cellular network is unavailable. tion distributions and natural areas. This study presents a
Additionally, if the communication has been intermittent system model and performance evaluations of two different
throughout the mission, the system attempts to reconnect types of UAVs on the market. An effective post-disaster
to the GCS via one of the communication methods. The flight plans have been implied considering communication
communication system was manually selected by the user range, GCS location, and disaster area data. Because pre-
while considering the statuses of the available communica- processing stages and simulators can help in verifying the
tion network(s). In such a case, the UAV executes the return system models, the authors evaluated the system model in a
to launch (RTL) point autonomously. The RTL can be con- self-developed simulator operated in three–dimensional (3D)
figured in the flight controller used in the tested UAVs [23]. mode with different flight heights. Furthermore, numerical
Similarly, the present paper considers the communication and statistical expressions were generated by implementing

64388 VOLUME 7, 2019

M. Aljehani, M. Inoue: Performance Evaluation of Multi-UAV System in Post-Disaster Application

FIGURE 2. Block diagram showing the overall system-level design of the UAVs, HTL simulator and ground control station.

and fine-tuning the UAV parameters and selecting the camera

specifications for mapping an area of interest. These param-
eters and specifications are based on real products on the
market.

II. SYSTEM DESIGN

A. REQUIREMENTS
Before designing a system model, the system requirements
must be determined in a preconditioning process [27]. This
section outlines the system objectives from a development
point of view, with reference to previous UAV implemen-
tations in disaster response planning. The purposes of the
present study are listed below:
• Developing a cooperative multi-UAV system for post-
FIGURE 3. Schematic of a user and ground control station use case.
disaster mapping missions.
• Selection different UAV models.
• Fine-tuning to minimize the flying time.
receives feedbacks from the UAVs and workstation (which
• Evaluating the UAV performance in the simulator.
can be a processing unit such as a computer or mobile device).
The mission plan comprises flight strips, height, footprints,
B. HARDWARE IN THE LOOP SIMULATORS and the vehicle velocity. The controller on the UAV board
Many hardware in the loop (HITL) simulators are avail- is connected to multiple sensors inside the HITL simulator.
able for testing UAV performances. However, most of these The sensor data and disaster environment have been totally
simulators do not provide multi-UAV options, and almost simulated in HITL, providing UAV feedback on a given map
every simulator only tests the performance of a specific rather than sensor data in real flight. After designing the
UAV model, which is impractical for evaluating different missions, the waypoints are uploaded and sent to the UAV
UAV models. To overcome this difficulty, the authors of [23] controller through the network. The mission planning begins
developed a HITL simulator based on a standard protocol with the mission design and terminates after syncing the flight
(i.e., the MAVlink protocol). This simulator can simulta- plans to the flight controller. The GCS stores the UAV flight
neously run two or more UAVs in one GCS workstation. data logs in the workstation and the flight controller storage.
Moreover, the default map can be updated by importing These data are sensor data provided by the HITL simulator
photographs, highlighted items, and map information into (or by sensors in real flight experiments). The user (usually
the simulator. The authors imported new map parameters called the ground user) refers to the person or the pilot who
from A geographic information system (GIS) like authors controls the UAV, and designs the mission plans. The ground
in their work [28], which provides frameworks of gathering user has four main assignments: flight plan design, sending
and analyzing geography data. The simulator was developed commands, managing the communications, and network and
in the C# programming language. Before demonstrating the monitoring (see Figure 3).
multi-UAV system design in the HITL simulator, we must
systematically define the UAV control and GCS. Figure 2 is a C. GROUND CONTROL STATION
block definition diagram of the overall system design. The The main tasks of the GCS are controlling the UAVs and
GCS application (GCSApp) transmits the commands and running the GCSApp to design the flight plans. The GCSApp

VOLUME 7, 2019 64389

 M. Aljehani, M. Inoue: Performance Evaluation of Multi-UAV System in Post-Disaster Application

tool communicates with the UAV via a human–machine inter-

face. Through the GCS unit, the ground user can control
the UAVs, assign missions, and receive the telemetry data of
the UAVs. The GCS unit performs multiple tasks as listed
below:
• Managing the communications between the UAVs and
GCSApp.
• Interfacing the control tools and sensors of the UAVs.
• Setting up the UAV parameters.
• Designing the UAV flight plan and manages missions. FIGURE 5. Model-Based Design (MBD).
• Sending commands and receiving information from
a UAV. As a system-level testing technique, the HITL simulator is
• Monitoring the UAVs in real time. applicable to realistic electronic components and embedded
• Analyzing and storing the videos and photographs which systems [29]. HITL simulation is one stage of model–based
captured from the UAV. design; specifically, it validates the system requirements and
• Processing the received data and analysis. tests the system design (see Figure 5). The HITL simulator
• Simulating the mission and presents its results before runs commands into a gateway connected to the vehicle con-
execution. troller in real time, and displays the outputs. The technique
• Recording the flight logs of the mission and stores them can be used in systems of various complexity, from simple
in the workstation. to very complicated. In the present study, the HITL simulator
was developed for testing the multi-UAVs and flying designs.
As HITL runs its commands on real hardware, it can evaluate
a proposed system that cannot be tested easily, and verify the
system outputs in realistic scenarios. When the vehicle flight
controller connected to the GCSApp is provided with a vehi-
cle simulator of the selected vehicle type, it is ready to receive
commands from the GCSApp and execute commands. There-
after, the HITL simulates the flight logs and sends its data to
the controller. The simulator and GCSApp must be connected
to the same network. The vehicle simulator provides the
necessary feedbacks and displays them in the GCS interface.
The more accurate the data from the simulator, the better are
the validation results of the system.
In this study, all feedbacks were transmitted through User
Datagram Protocol (UDP) protocols. The output data can also
be configured to pass through the transmission control proto-
col [23]. The flight plan is transmitted to the vehicle controller
FIGURE 4. Developed HITL simulator for the multi-UAV system. through the network and GCS. The plan contains a series of
waypoints and mission parameters such as velocity, altitude,
Figure 4 shows the HITL simulator developed by the acceleration and control mode. In all missions, the control
authors. The UAVs are represented by 3D-dynamic robot mode was set to Autopilot. The UAV model must be known
simulators connected to one GCS. HITL simulator generates before sending the flight design to the vehicle controller.
sensor data based on the UAV model parameters. To start the In this system, the GCS autonomously recognizes the UAV
simulation, the HITL simulator must be simultaneously con- type at the instant of connecting the vehicle controller to
nected to the control unit and GCSApp. The three dynamic– the vehicle simulator. In real flight experiments, the GCS
robot simulators generate the sensor data in real time. Various recognizes the UAV type from the vehicle controller, which
algorithms, flight tests, and training of UAV systems can be has been configured to the UAV type before connection.
implemented in the robot simulators in realistic scenarios.
HITL simulators are connected to the GCSApp through serial III. FLIGHT PLAN DESIGN
ports. Although the HITL executes the mission on the con- Fine-tuning in the flight design is the very important task
troller unit, the feedbacks are simulated on the GCS or an before starting the mission. Generally, fine-tuning helps to
external system with the same protocol as the flight controller. complete the mission within the smallest time and require
After executing the mission and receiving the simulated flight the fewest flight paths. On a polygon, fine-tune is char-
logs, the controller reports these data as telemetry data to the acterized by two parameters, the number of turns w and
GCSApp, and displays the flight logs of the mission in real the number of flight strips l. Figures 6 and 7 illustrate
time a non-fine-tuned and fine-tuned, respectively, on the same

64390 VOLUME 7, 2019

M. Aljehani, M. Inoue: Performance Evaluation of Multi-UAV System in Post-Disaster Application

FIGURE 6. Example of polygon design without fine-tuning configuration

over the disaster area (shaded polygon).

FIGURE 8. Overlaps in the mapping mission of a rectangular polygon

using a vertical field of view.
FIGURE 7. Example of fine-tuning the configuration over the disaster
area (shaded polygon).
TABLE 1. Camera specifications in the multi-rotor and fixed-wing
frameworks.
given polygon. When the mission takes many turns, the flight
time is increased. For example, the suboptimal flight plan
design in Figure 6 requires w = 7 turns and l = 8 strips.
In Figure 7 reduces these values to w = 3 and l = 4. Note
that when the UAV flight strips are parallel to the longest
line of the given rectangular polygon, the number of turns,
and hence the number of flight strips, is reduced. When
mapping a complex area, the polygon must sometimes be
divided into multiple polygons and the flight must start from
different points to minimize the w and l [30]. Before starting
a flight plan design, the ground user requires the following of this study, the end and side laps are set to 70% and 50%,
specifications: respectively.
• The UAV models and their specifications (i.e., ground
speed, maximum height, and flight behavior). B. GROUND SAMPLING DISTANCE
• The camera specifications (focal length, field of The flight plan design in the proposed framework considers
view (FOV), and ground sampling distance (GSD). the digital camera types. The camera model plays the main
• The amount of overlap (end lap and side lap). role in the flight plan design process, because the FOVs
• Area size and status. and GSD (which determines the resolution of the image)
The data of aerial photogrammetry mappings are com- differ from one camera to another. A digital camera with a
monly collected from three FOVs: the true vertical FOV low FOV increases the w and time of the mission, whereas
(from 0◦ up to ±3◦ from the nadir of the UAV), a lowly a camera with a high FOV decreases the w because the
oblique FOV (titled photography; above ±3◦ but within ±30◦ distance between lines is increased. A low GSD acquires
of the nadir), and a highly oblique FOV ( between ±35◦ high-quality images, but requires a heavy, expensive camera.
and ±55◦ off-nadir). The present study adopts the vertical Table 1 shows the camera specifications used in the two UAV
mapping orientation in all flight plan designs. When the UAV models. The GSD is computed as follows:
is dispatched to fly forward in the real experiments, a gimbal h ∗ Sensor height
stabilizes to maintain a vertical camera view at all times. GSDh = (1)
FL ∗ Image height
h ∗ Sensor width
A. PHOTOGRAMMETRIC OVERLAPS GSDw = (2)
FL ∗ Image width
Overlaps in the photogrammetric process are of two main
types: end laps and side laps. The end lap represents the where h is the UAV height and FL represents the focal length
amount of data shared between images of photographic points (see Figure 9). In Table 1, the GSD of the digital camera was
along the flight strip, whereas the side lap describes the computed in both flight designs. For example, the GSD at
amount of data shared between images of adjacent flight h = 400 m is 11.81 cm/px for the multi-rotor camera and
strips. In typical cases, the end and side overlaps are 50–70% 24.72 cm/px for the fixed-wing camera. However, the fixed-
and 30–50%, respectively for fine-tuning [31]. Figure 8 wing camera offers a wider FOV than the multi-rotor camera,
demonstrates the end and side laps in the photogrammetric which reduces the number of flight strips in the mission.
process of a rectangular polygon. In the experimental section When computing the GSD, a designer must always consider

VOLUME 7, 2019 64391

 M. Aljehani, M. Inoue: Performance Evaluation of Multi-UAV System in Post-Disaster Application

FIGURE 9. Focal length and flight height in the GSD computation.

FIGURE 11. Operation of the fixed-wing framework in a mapping mission.

dynamically (see Figure 11). The advantages of the fixed

wing over the multi-rotor framework are high velocity and
rapid flight time. Therefore, fixed-wing aircraft are frequently
deployed over wide areas.

E. SEQUENCE DIAGRAM OF THE MISSION DESIGN

Figure 12 shows the sequence diagram of the overall sys-
tem, starting from the flight design and HITL execution,
and concluding at the receipt and display of the flight logs
of the mission. After establishing the connection between
FIGURE 10. Operation of the multi-rotor framework in a mapping
a UAV and GCSApp, Work_Station sends the command
mission. CMD_Plan_Design() to the ground user, granting permission
to start designing the flight plan in GCSAp. GCSApp then
the worst-case scenario. For instance, if the GSDh exceeds confirms the requests via ACK_Plan_Design massages. The
the GSDw , the GSDh should be considered in the flight plan design process is begun in Plan_Design. During the flight
design. plan design, GCSApp estimates the mission outcomes such
as the flight time, visual trajectories, and footprints. The
C. MULTI-ROTOR FRAMEWORK ground user uploads the completed flight plan design via
The multi-rotor framework embraces several designs, for the Upload_Flight_Design() command. GCSApp then pro-
example, the quad-copter with four rotors, the hexa-copter cesses the uploaded flight design and converts it into flight
with six motors, the octa-copter with eight motors, and commands and a series of waypoints based on the protocol
the tri-copter with three motors and one servo. The quad, adopted by the UAV. In the next step, the flight design is
hexa, and octa multi-rotor can be configured in cross (x) synced with the UAV controller via the Sync_Flight_Design()
or plus (+) forms. Among these designs, the quad-copter message. In the present study, the UAVs adopt a serially com-
with the x configuration was simulated in the HITL. Unlike municated UDP protocol. All flight plan designs must begin
fixed-wing and helicopter model, multi-rotor are controlled with a Take Off command and end with a return to launch
by changing the motors’ speeds to achieve 3D trajectory point (RTL). After the syncing process, the flight controller
movements i.e., pitch, yaw, and roll [32]. When the multi- reforms the mission to suit the framework of the registered
rotor copter reaches the end of a flight strip, it can yaw at vehicle, and executes the mission commands in HITL. The
an aggressive angle (see Figure 10). In contrast to its fixed- mission is processed inside the HITL simulator, which pro-
wing counterpart, the multi-rotor copter need not maintain its duces telemetry data for the GCSApp. These data generate
speed while executing roll, yaw or elevation maneuvers. The the flight logs of the vehicle’s mission in real-time on a
flexibility and maneuverability of multi-rotor frameworks are virtual map.
advantageous for mapping complex geometric areas.
F. SYSTEM MODEL
D. FIXED-WING FRAMEWORK This section discusses the problems of deploying UAVs in
Unlike the multi-rotor UAV, the fixed-wing vehicle can fly disaster applications. These problems are solved through
when air passes across its wings, creating an upward force on system modeling of a multi-UAV system. Let u = UAV
the wings that overcomes the downward pull of gravity. The (u > 1 for a multi-UAV system) and g = GCS (g = 1
three basic flying maneuvers, namely, roll, yaw, and pitch, for a single GCS, or g ≥ 1 for multiple GCSs). In this
are promoted by adjusting the wing angles [33]. However, study, the application (post-disaster deployment) required
the fixed-wing aircraft must maintain its speed while turning a mobile GCS. The location of the GCS was denoted as

64392 VOLUME 7, 2019

M. Aljehani, M. Inoue: Performance Evaluation of Multi-UAV System in Post-Disaster Application

FIGURE 12. Sequence diagram of autonomous mission design and execution on the unmanned vehicles’ flight controller and
HITL simulator.

TABLE 2. Specifications of multi-rotor and fixed-wing vehicles. Pythagoras’ theorem in Cartesian coordinates:
q
du,g [t] = (xu [t] − xg [t])2 + (yu [t] − yg [t])2 + (hu [t])2
(5)
The communication range of the GCS was denoted as βg ,
with βg ≤ βMax, and the area of interest was denoted by A.
The GCS can locate outside the impacted area, but the com-
munication must cover the area and collect the information
w = [xg (t), yg (t)] where is xg and yg are the GCS coordinates from the UAVs during the mission. In particular, the GCS
in the field at time instant t. Additionally, 0 ≤ t ≤ T(D) must locate within the UAV communication range. The GCS
where T(D) is the total period of the service. 3D modeling was can communicate with the UAVs by several methods. The
applied to accommodate the varying flight height. In Carte- GCS–UAV communication is generally expressed as
sian coordinates, the 3D location of the UAV was expressed / A ⇒ βg ≥ A
w∈ (6)
as (m ∈ R3 ) where is m represents the location of the UAV
during the mission, given by The distance between the UAV and GCS during the mission
must not exceed the maximum range of the communication
m = [xu (t), yu (t), hu (t)], hu (t) > 0, (3) coverage at any instant. This constraint is given by
where Hu is the altitude of the UAV, and x and z represent the ∀du,g (t) ∈ βMax (7)
UAV location from the FOV angle. The time t satisfies 0 ≤
The UAV capabilities and flight times of the UAVs (n)
t ≤ TD . TD , where TD is the overall mission time (including
might be insufficient for mapping a large area. The number of
the takeoff time T(TK ) , the service (i.e., mapping) time T(S) ,
UAVs that can be assigned to map A is calculated as follows:
and the time to back to the RTL point T(RTL) . As the UAV
will not take off unless the flight plan has been uploaded in A
knk = ( )+1 (8)
the flight controller, the out-of-service time was ignored, and C(u)MAX
all missions were assumed to be completed successfully. The where C(u)MAX represents the maximum coverage of the
mission time was simply computed as follows: single UAV in mapping mission. It can be acquired from
the flight logs results. Adding one more UAV is always
T(D) = T(TK ) + T(S) + T(RTL) (4)
appreciated in missions requiring fast completion. The times
where TD must be less than the maximum flight time (Tu ) TTK , TS , and TRTL differ among models with different UAV
of the UAV (i.e., TD < T(u)MAX ). Herein, the maximum parameters and flight mechanisms, as explained in the flight
flight time of the UAVs is given in Table 2. The distance du,g design section. Therefore, further evaluation of the UAV
between a UAV and the GCS at time t t was calculated by performance is needed.

VOLUME 7, 2019 64393

 M. Aljehani, M. Inoue: Performance Evaluation of Multi-UAV System in Post-Disaster Application

FIGURE 13. Area of Japan (Miyagi Prefecture) affected by the Tohoku earthquake and tsunami disaster of 2011:
(a) Areal population density (2010), (b) Areal population distribution (2010), (c) Earthquake sites (2011), and
(d) Areas flooded by the tsunami (2011).

FIGURE 14. GSI monitors the land conditions imported into GCS map.
FIGURE 15. Area polygons designed inside the GCS map after importing
GSI reports.
IV. IMPACTED AREA AND POLYGON DESIGNS
A. BACKGROUND OF THE EXPERIMENTAL AREA
TABLE 3. Information of the affected area after the disaster.
A great earthquake struck the Pacific coast of Tohoku,
Japan (38.1035◦ N, 142.861◦ E, M 9.0) at 14:46:18 JST.
The earthquake was followed by a huge and devastating
tsunami, causing 15,729 fatalities and 4,539 missing per-
sons in the Hokkaido, Tohoku, and Kanto regions [5] (see
Figures 13 and 14). Highways and railways were blocked
and some were totally destroyed, preventing evacuees from
leaving or reaching the nearest shelter. Nearly 5,200 spots
were surveyed in the disaster area, providing one of the largest
post-disaster datasets in surveying history [34]. The inun-
dation height was surveyed by advanced instruments such
as satellite systems, lasers, and GPSs. Based on the survey
dataset, the maximum run-up height of the tsunami exceeded B. DESIGN POLYGONS
10 m and was distributed along 500 km of coastline. The Table 2 lists the specifications of the UAVs deployed in
area affected by the 2011 event was the largest recorded in this work, which are based on real commercial products.
Tohoku’s disaster history. Search and rescue teams, volun- The UAV parameters were inserted in the HITL simulator.
teers, and firefighters attempted to find survivors and access The pre-information and post-information data were then
the impacted areas. However, as the routes were severely inserted into the map application of GCSApp. The tsunami,
damaged and many areas were flooded, the disaster recov- earthquake and population density data were imported from
ery required a prior mapping of the area and the identifica- the Geospatial Information Authority of Japan (GSI) [36].
tion of feasible entry routes. In such a scenario, a low-cost The GSI monitors the land conditions in Japan and pro-
multi-UAV system contribution is invaluable for mapping vides the latest results of land surveys and disaster impacts.
and providing terrestrial and aerial images [35]. Therefore, Figure 14 displays the post-disaster impact data imported
the authors applied the proposed system model in the Tohoku into the map application. The polygon design is merely a
disaster area and verified its performance in a simulator. rectangular design in GCS default map (See Figure 15).

64394 VOLUME 7, 2019

M. Aljehani, M. Inoue: Performance Evaluation of Multi-UAV System in Post-Disaster Application

FIGURE 16. Relations among altitude, number of photographs and flight FIGURE 18. Relations among area, number of photographs and altitude
strips for the multi-rotor UAV in 3D view and down-view graphs. for the multi-rotor UAV in 3D view and down-view graphs.

FIGURE 19. Relations among altitude and number of photographs for the
multi-rotor UAV model. Where f(x) represents the non-robust residuals
function and weighted by the area as shown in Table 4.

a convex grid composed of a number of cells. The cell size

depends on the coverage capabilities of the UAVs and the
area of interest. The exact method decomposes the region
into a set of polygonal subareas that exactly covers the
FIGURE 17. Relations among altitude, area and flight strips for the original area. This study applies the approximate method
multi-rotor UAV in 3D view and down-view graphs.
because the area of interest has a very complex infrastructure.
As the assigned area cannot be mapped by a single UAV
C. MISSION DESIGN AND EXECUTION with the specifications shown in Table 2, it was divided
Polygons can be mapped by exact or approximate meth- into sub-polygons that fit the UAVs capabilities. The poly-
ods. It has been well researched in many studies [37]–[39], gons were designed manually and decomposed into a set
and [40]. The approximate method covers the region with of square designs to show the UAV performance in various

VOLUME 7, 2019 64395

 M. Aljehani, M. Inoue: Performance Evaluation of Multi-UAV System in Post-Disaster Application

TABLE 4. Statistical evaluation of multi-rotor UAV performance based on 200 missions under different considerations of relations.

polygon design and different flight heights (See Figure 15) UAV deployment is suitable for midtown areas, areas
as described in the following process: near the sea, flooded areas, and remote areas far from
• Decompose the area into polygons that fit the UAV the GCS.
capabilities.
• Ignore the polygons covering natural areas (rivers, V. EXPERIMENTS
forests, mountains), and areas with no population. This section evaluates the performances of the UAVs in the
• Distribute the GCSs on the area, such that each GCS simulator. The experimental results will help rescue teams
serves the maximum number of polygons and is located and pilots to integrate the UAV(s) fulfilling the mapping
at the nearest start point of its mission. mission requirements. For this purpose, the evaluation inves-
• Assign multi-rotor vehicles in missions requiring high- tigates many relations between the UAV parameters, and
resolution images (complex geometries such as urban considers many factors: flight time, velocity, flight behavior
areas, small countryside areas, and areas needing (which influences the numbers of photographs and flight
many maneuvers). These data are available in the pre- strips), camera specifications, fine-tuning, flight height, and
information and post-information tables. If the area area coverage. The values of these factors are based on
coverage exceeds 5.l km2 (the capability limit of those of marketed products. As the technical evaluations of
mapping by multi-rotor vehicles), divide the polygon marketed products are either missing or determined from
into multiple polygons. few data, the performances of commercial UAVs are unclear.
• Assign fixed-wing UAVs in missions requiring a fast Herein, the relations between these factors were demon-
response over a wide area (≥17 km2 ). Fixed-wing strated statistically on graphs and based on multiple missions

64396 VOLUME 7, 2019

M. Aljehani, M. Inoue: Performance Evaluation of Multi-UAV System in Post-Disaster Application

FIGURE 20. Relations among altitude, number of photographs and flight

strips for the fixed-wing UAV in 3D view and down-view graphs.

FIGURE 22. Relations among area, number of photographs and altitude

for the fixed-wing UAV in 3D view and down-view graphs.

FIGURE 23. Relations among altitude and number of photographs for the
fixed-wing UAV model. Where f(x) represents the non-robust residuals
function and weighted by the area as shown in Table 5.

• Number of photographs, area, and flight height inves-

tigated by polynomial fitting with robust least absolute
FIGURE 21. Relations among altitude, flight strips and area for the
residuals (LAR).
fixed-wing UAV in 3D view and down-view graphs. • Number of photographs and flight height with area-
weighted, investigated by non-robust polynomial fitting.
in the simulator. The investigated factor relations are listed Functional expressions were generated for each rela-
below: tion. The authors executed 880 missions on the simula-
• Number of flight strips, area, and flight height, investi- tors (200 multi-rotor and 680 fixed-wing UAVs) with real
gated by linear interpolation. flight times. The mission numbers were based on the UAVs’
• Number of photographs, number of flight strips, and maximum mapping abilities over the given area (5 km and
area, investigated by linear interpolation. 17 km for the multi-rotor and fixed-wing UAVs, respectively).

VOLUME 7, 2019 64397

 M. Aljehani, M. Inoue: Performance Evaluation of Multi-UAV System in Post-Disaster Application

TABLE 5. Statistical evaluation of the fixed-wing UAV performance based on 680 missions under different considerations of relations.

Therefore, in the polygon representation, the impacted area contour plots. The black dots are the data sequences obtained
was expressed as A = {1, 2, . . . 5} for the multi-rotor UAVs from the HITL. The simulated velocity (set to the maxi-
and A = {1, 2, . . . 17} for the fixed-wing UAVs. The height h mum velocity of the multi-rotor UAV as mentioned above)
was varied as 100, 110, . . . 400. The velocity was set to its decreased during each turn of the UAV. The UAV perfor-
maximum value (22 m/s and 25 m/s for the multi-rotor and mance as a function of area size and flight height is pre-
fixed-wing UAVs, respectively). All missions in the simula- sented in Figure 17. Again, these graphs present the linear
tors were assumed to be successfully completed (i.e., with interpolation results and contour plots. The colors delineate
successful takeoff, mapping and landing), and the communi- the number of flight strips. Figure 18 relates the number
cation problem was within the range of βMax in the system of photographs, area, and altitude. Herein, the color range
model, as described in the previous subsection. graduates the number of photographs. The data were fitted by
a polynomial with robust LARs. Finally, Figure 19 plots the
A. MULTI-ROTOR PERFORMANCE number of photographs versus the flight altitude, weighted
This section evaluates the flight performances of the multi- by area. These data were fitted to a 9-degree polynomial
rotor UAV model for different polygon sizes and flight (blue solid curve in the figure). The functional expressions
heights over the impacted area. The polygon design methods of the multi-rotor relations in Figures 16–19, and their statis-
and the experimental area were introduced in Section Four. tics, are presented in Table 4. The error sum of squares
Figure 16 relates the number of flight strips, number of in Table 4 quantifies the divergence between the individual
photographs and flight heights for the multi-rotor UAV. observation and the group mean. The coefficient of deter-
The results are displayed as linear interpolations and mination root square (R-squared) defines the proportionate

64398 VOLUME 7, 2019

M. Aljehani, M. Inoue: Performance Evaluation of Multi-UAV System in Post-Disaster Application

deviation in the variable y explained by the independent vari- simulator was constructed from 3D dynamic robot simula-
ables x in the linear regression model. The R-squared value tors, which replace real flight experiments. The GCSApp was
is a property of the fitted model, and can be either unadjusted connected to all UAVs at different ports. The simulator was
or adjusted (if the adjustment statistically improves the model run on two common UAVs with different camera specifica-
performance) for the number of predictors in the model. The tions. Both UAV types (multi-rotor and fixed-wing) are based
root mean squared error (RMSE) quantifies the deviation of on currently commercial products. Based on the camera spec-
the estimated values from the flight results. The last column ifications, the ground surface distance was also determined
in Table 4 presents the standard deviations, and the coeffi- for each type. This study confirmed the importance of UAV
cients of the relations. performance data to design flight plans. After defining the
system objectives, the ground user can design polygons that
B. FIXED-WING PERFORMANCE meet the mapping capabilities of the UAVs considering com-
The flight times, flight behaviors (i.e., velocity, flight time, munication and flight behaviors of each model. Multi-rotor
and turn angles), and camera specifications all differ among UAVs are suitable for monitoring small urban areas, as their
fixed-wing and multi-rotor UAV models. These differences complexity and maneuverability enable obstacle avoidance
are mainly responsible for the disparate results of fixed- in these areas. In flooded areas, areas far from the GCS,
wing and multi-rotor UAVs in autonomous mapping mis- and remote areas, high maneuvers for avoiding obstacles are
sions [4]. The specifications of both types were introduced unnecessary. Such areas should be monitored by fixed-wing
and discussed in previous sections of this paper. Herein, UAVs, which have longer flight times than multi-rotor UAVs.
the fixed-wing model has a maximum flight time of 45 min, In this work, the flight plan design was based on UAV capabil-
a maximum velocity of 25 m/s, and an overall coverage ability ities and area features. After executing the mapping mission
of 17 km. All these factors were inserted manually to the sim- in the HITL simulator, the performances of the UAVs were
ulator in each mission. Similar to the multi-rotor performance demonstrated in the flight logs generated during the mission.
evaluations, the mapping results were displayed as con- All mapping mission flight times included the takeoff, time
tour plots, and the fixed-wing performances in the mapping of service and RTL time costs. The simulator can run multiple
missions were statistically expressed by linear interpolation instances at the same time, enabling simultaneous mission
or polynomial fitting. The number of flight strips, number executions, which is developed by the authors. The best-fit
of photographs and flight height are related in Figure 20. numerical expressions for both UAV types were derived by
The black dots represent the data from the HITL simulator. analyzing the flight logs of both UAV models while varying
Figure 21 coverage ability of 17 km. All these factors were the number of photographs, number of flight strips, areal
inserted manually to the simulator in each mission. Similar to size, and altitude within the UAV and camera specifications.
the multi-rotor performance evaluations, the mapping results The 880 experimental mapping missions revealed the real
were displayed as contour plots, and the fixed-wing perfor- performance of the UAVs. The empirical equations and sta-
mances in the mapping missions were statistically expressed tistical evaluations of the UAV performances were obtained
by linear interpolation or polynomial fitting. The number in a HITL simulator. The obtained graphs and statistical
of flight strips, number of photographs and flight height results will help rescue teams and pilots to effectively deploy
are related in Figure 20. The black dots represent the data UAVs in disaster applications. The numerical and statis-
from the HITL simulator. Figure 21 shows the fixed-wing tical performances of marketed UAVs are minimally pro-
performances over differently sized areas at different flight vided or missing, impeding the effective application of UAVs
heights. Both figures display the linear interpolation results in post-disaster mapping missions. Therefore, the marketed
and contour plots. The number of photographs, coverage area, UAVs must be comprehensively evaluated in several time-
and altitude are related in Figure 22, where the colors delin- consuming flight experiments. This study has highlighted
eate the number of photographs in the executed missions. The the benefits of the simulator in performance evaluations of
data were fitted to a polynomial with robust LARs. Figure 23 two different UAV models (based on their flight logs) under
plots the number of photographs versus the flight altitude, various considerations of mapping factors.
weighted by area. These data were fitted to a 9-degree polyno-
mial. The functional expressions of the fixed-wing UAV per- REFERENCES
formances in Figures 20–23, and their statistics, are presented [1] T. P. Ehrhard, ‘‘Air force UAV’s: The secret history,’’ Mitchell
in Table 5. Inst. Airpower Studies, Arlington County, VA, USA, Tech. Rep.,
Jul. 2010.
[2] S. M. Adams and C. J. Friedland, ‘‘A survey of unmanned aerial vehicle
VI. CONCLUSION (uav) usage for imagery collection in disaster research and management,’’
This study demonstrated the mapping performance of a multi- in Proc. 9th Int. Workshop Remote Sens. Disaster Response, vol. 8, 2011,
pp. 1–8.
UAV system in a disaster-struck area of Japan. Experiments [3] M. Aljehani and M. Inoue, ‘‘Multi-UAV tracking and scanning systems
were conducted on a self-designed HITL simulator, a very in M2M communication for disaster response,’’ in Proc. IEEE 5th Global
realistic testing simulator that estimated the outcomes of Conf. Consum. Electron., Oct. 2016, pp. 1–2.
[4] M. Aljehani and M. Inoue, ‘‘Comparison of UAV flight behaviors after
the mapping missions. 880 missions have been executed in autonomous mapping of an urban area,’’ in Proc. IEEE 7th Global Conf.
the simulator with real flight time consuming. The HITL Consum. Electron. (GCCE), Oct. 2018, pp. 1–2.

VOLUME 7, 2019 64399

 M. Aljehani, M. Inoue: Performance Evaluation of Multi-UAV System in Post-Disaster Application

[5] Y. Fujii, K. Satake, S. Sakai, M. Shinohara, and T. Kanazawa, ‘‘Tsunami [28] T. Nagata, Y. Kimura, and M. Ishii, ‘‘Use of a geographic information
source of the 2011 off the Pacific coast of Tohoku Earthquake,’’ Earth, system (GIS) in the medical response to the Fukushima nuclear disaster in
Planets Space, vol. 63, no. 7, p. 55, 2011. Japan,’’ Prehospital Disaster Med., vol. 27, no. 2, pp. 213–215, Apr. 2012.
[6] L. Gupta, R. Jain, and G. Vaszkun, ‘‘Survey of important issues in UAV [29] J. A. Ledin, ‘‘Hardware-in-the-loop simulation,’’ Embedded Syst. Pro-
communication networks,’’ IEEE Commun. Surveys Tuts., vol. 18, no. 2, gram., vol. 12, pp. 42–62, Feb. 1999.
pp. 1123–1152, 2nd Quart., 2016. [30] Y. Li, H. Chen, M. J. Er, and X. Wang, ‘‘Coverage path planning for UAVs
[7] G. Fan and S. Jin, ‘‘Coverage problem in wireless sensor network: A sur- based on enhanced exact cellular decomposition method,’’ Mechatronics,
vey,’’ J. Netw., vol. 5, no. 9, pp. 1033–1040, 2010. vol. 21, no. 5, pp. 876–885, Aug. 2011.
[8] R. Mulligan and H. M. Ammari, ‘‘Coverage in wireless sensor networks: [31] P. R. Wolf and B. A. Dewitt, Elements of Photogrammetry: With Applica-
A survey,’’ Netw. Protocols Algorithms, vol. 2, no. 2, pp. 27–53, 2010. tions in GIS, vol. 3. New York, NY, USA: McGraw-Hill, 2000.
[9] M. Vecchio, R. López-Valcarce, ‘‘Improving area coverage of wireless sen- [32] Y.-J. Tsai, C.-S. Lee, C.-L. Lin, and C.-H. Huang, ‘‘Development of flight
sor networks via controllable mobile nodes: A greedy approach,’’ J. Netw. path planning for multirotor aerial vehicles,’’ Aerospace, vol. 2, no. 2,
Comput. Appl., vol. 48, pp. 1–13, Feb. 2015. pp. 171–188, 2015.
[10] B. Liu, P. Brass, O. Dousse, P. Nain, and D. Towsley, ‘‘Mobility improves [33] M. Pepe, L. Fregonese, and M. Scaioni, ‘‘Planning airborne photogram-
coverage of sensor networks,’’ in Proc. 6th ACM Int. Symp. Mobile Ad Hoc metry and remote-sensing missions with modern platforms and sensors,’’
Netw. Comput., 2005, pp. pp. 300–308. Eur. J. Remote Sens., vol. 51, no. 1, pp. 412–436, 2018.
[11] Y. Chen, H. Zhang, and M. Xu, ‘‘The coverage problem in UAV network: [34] N. Mori, T. Takahashi, and Tohoku Earthquake Tsunami Joint Sur-
A survey,’’ in Proc. 5th Int. Conf. Comput., Commun. Netw. Technol. vey Group, ‘‘Nationwide post event survey and analysis of the 2011
(ICCCNT), Jul. 2014, pp. 1–5. Tohoku earthquake tsunami,’’ Coastal Eng. J., vol. 54, no. 1, 2012,
[12] M. Jakob, E. Semsch, D. Pavlicek, and M. Pechoucek, ‘‘Occlusion-aware Art. no. 1250001.
multi-uav surveillance of multiple urban areas,’’ in Proc. 6th Workshop [35] M. Aljehani and M. Inoue, ‘‘Communication and autonomous control of
Agents Traffic Transp. (ATT), 2010, pp. 59–66. multi-UAV system in disaster response tasks,’’ in Proc. KES Int. Symp.
[13] S.-Y. Fu, L.-W. Han, Y. Tian, and G.-S. Yang, ‘‘Path planning for unmanned Agent Multi-Agent Syst., Technol. Appl. Springer, 2017, pp. 123–132.
aerial vehicle based on genetic algorithm,’’ in Proc. IEEE 11th Int. Conf. [36] A. Zerger and D. I. Smith, ‘‘Impediments to using GIS for real-time
Cogn. Inform. Cogn. Comput., Aug. 2012, pp. 140–144. disaster decision support,’’ Comput., Environ. Urban Syst., vol. 27, no. 2,
[14] C. Di Franco and G. Buttazzo, ‘‘Energy-aware coverage path planning pp. 123–141, Mar. 2003.
of UAVs,’’ in Proc. IEEE Int. Conf. Auto. Robot Syst. Competitions, [37] B. J. Bulušek, ‘‘Coverage path planning in non-convex polygon areas for
Apr. 2015, pp. 111–117. orthophotomap creation using UAVs,’’ in Proc. Conf. IEEE, vol. 111, 2015,
[15] P. Razi, J. T. S. Sumantyo, D. Perissin, H. Kuze, M. Y. Chua, and p. 117.
G. F. Panggabean, ‘‘3D land mapping and land deformation monitoring [38] E. Santamaria, F. Segor, I. Tchouchenkov, and R. Schönbein, ‘‘Rapid
using persistent scatterer interferometry (PSI) ALOS PALSAR: Validated aerial mapping with multiple heterogeneous unmanned vehicles,’’ in Proc.
by geodetic GPS and UAV,’’ IEEE Access, vol. 6, pp. 12395–12404, ISCRAM, 2013, pp. 1–10.
2018. [39] L. H. Nam, L. Huang, X. Li, and J. F. Xu, ‘‘An approach for coverage
[16] A. M. Hayajneh, S. A. R. Zaidi, D. C. McLernon, M. Di Renzo, and path planning for UAVs,’’ in Proc. IEEE 14th Int. Workshop Adv. Motion
M. Ghogho, ‘‘Performance analysis of uav enabled disaster recovery net- Control (AMC), Apr. 2016, pp. 411–416.
works: A stochastic geometric framework based on cluster processes,’’ [40] N. Dadkhah and B. Mettler, ‘‘Survey of motion planning literature in
IEEE Access, vol. 6, pp. 26215–26230, 2018. the presence of uncertainty: Considerations for UAV guidance,’’ J. Intell.
[17] M. Aljehani and M. Inoue, ‘‘Safe map generation after a disaster, assisted Robot. Syst., vol. 65, no. 1, pp. 233–246, 2012.
by an unmanned aerial vehicle tracking system,’’ IEEJ Trans. Electr.
Electron. Eng., vol. 14, no. 2, pp. 271–282, 2019.
[18] Y. Xu, L. Xiao, D. Yang, Q. Wu, and L. Cuthbert, ‘‘Throughput maxi-
mization in multi-UAV enabled communication systems with difference
consideration,’’ IEEE Access, vol. 6, pp. 55291–55301, 2018.
[19] S. Li et al., ‘‘Event-trigger heterogeneous nonlinear filter for wide-area
measurement systems in power grid,’’ IEEE Trans. Smart Grid, vol. 10, MAHER ALJEHANI received the B.E. degree
no. 3, pp. 2752–2764, May 2018. in telecommunication engineering from the
[20] X. Liu et al., ‘‘Event-trigger particle filter for smart grids with limited Huazhong University of Science and Technology,
communication bandwidth infrastructure,’’ IEEE Trans. Smart Grid, vol. 9, the Associate degree in telecommunication from
no. 6, pp. 6918–6928, Nov. 2018. the Jeddah College of Telecom and Electronics,
[21] G. Tuna, B. Nefzi, and G. Conte, ‘‘Unmanned aerial vehicle-aided com- and the M.S. degree in systems engineering and
munications system for disaster recovery,’’ J. Netw. Comput. Appl., vol. 41, science from the Shibaura Institute of Technol-
pp. 27–36, May 2014.
ogy, in 2016, where he is currently pursuing the
[22] M. Erdelj, M. Król, and E. Natalizio, ‘‘Wireless sensor networks and multi-
UAV systems for natural disaster management,’’ Comput. Netw., vol. 124, Ph.D. degree with the Department of Communi-
pp. 72–86, Sep. 2017. cation Function Control Engineering. He has been
[23] L. Meier et al. (2013). MAVLink: Micro Air Vehicle Communication Proto- engaged in UAV Research Fields, since 2014 and a Research Visitor with
col. [Online]. Available: http://qgroundcontrol.org/mavlink/start.[Hämtad Valley Campus Japan Inc. He is a Student Member of IEEE and a Young
2014-05-22] Professional Member.
[24] G. Tuna, T. V. Mumcu, K. Gulez, V. C. Gungor, and H. Erturk,
‘‘Unmanned aerial vehicle-aided wireless sensor network deployment
system for post-disaster monitoring,’’ in Emerging Intelligent Comput-
ing Technology and Applications (Communications in Computer and
Information Science), vol. 304, D. S. Huang, P. Gupta, X. Zhang, and
P. Premaratne, Eds. Berlin, Germany: Springer, 2012. [Online]. Available: MASAHIRO INOUE received the B.S. and M.S.
https://link.springer.com/chapter/10.1007/978-3-642-31837-5_44 degrees in pure and applied physics from Waseda
[25] J. Sánchez-García, J. M. García-Campos, S. Toral, D. Reina, and University, and the Ph.D. degree in computer sci-
F. Barrero, ‘‘An intelligent strategy for tactical movements of UAVs in ence from Shizuoka University. He was engaged
disaster scenarios,’’ Int. J. Distrib. Sensor Netw. Arch., vol. 2016. New in research and development with Mitsubishi Elec-
York, NY, USA: Hindawi Publishing Corp., Mar. 2016, Art. no. 18.
[26] E. Christy, R. P. Astuti, B. Syihabuddin, B. Narottama, O. Rhesa, and tric Corporation. Since 2005, he has been a
F. Rachmawati, ‘‘Optimum UAV flying path for device-to-device com- Professor with the Department of Electronic Infor-
munications in disaster area,’’ in Proc. Int. Conf. Signals Syst. (ICSigSys), mation Systems, College of Systems Engineer-
May 2017, pp. 318–322. ing and Science, Shibaura Institute of Technology.
[27] B. Nuseibeh and S. Easterbrook, ‘‘Requirements engineering: Since 2017, he has been the Vice President of the
A roadmap,’’ in Proc. Conf. Future Softw. Eng., 2000, pp. 35–46. Shibaura Institute of Technology. He is a Senior Member of IEEE.

64400 VOLUME 7, 2019