drones

Review
Investigation of Autonomous Multi-UAV Systems for Target
Detection in Distributed Environment: Current Developments
and Open Challenges
Wilfried Yves Hamilton Adoni 1,2, *,† , Sandra Lorenz 2,† , Junaidh Shaik Fareedh 2,† , Richard Gloaguen 2,†
and Michael Bussmann 1,†

1 Helmholtz-Zentrum Dresden-Rossendorf, Center for Advanced Systems Understanding, Untermarkt 20,

02826 Görlitz, Germany
2 Helmholtz-Zentrum Dresden-Rossendorf, Helmholtz Institute Freiberg for Resource Technology,
Chemnitzer Str. 40, 09599 Freiberg, Germany; s.lorenz@hzdr.de (S.L.); j.shaik-fareedh@hzdr.de (J.S.F.)
* Correspondence: w.adoni@hzdr.de; Tel.: +49-351-2604487
† These authors contributed equally to this work.

Abstract: Uncrewed aerial vehicles (UAVs), also known as drones, are ubiquitous and their use
cases extend today from governmental applications to civil applications such as the agricultural,
medical, and transport sectors, etc. In accordance with the requirements in terms of demand,
it is possible to carry out various missions involving several types of UAVs as well as various
onboard sensors. According to the complexity of the mission, some configurations are required
both in terms of hardware and software. This task becomes even more complex when the system is
composed of autonomous UAVs that collaborate with each other without the assistance of an operator.
Several factors must be considered, such as the complexity of the mission, the types of UAVs, the
communication architecture, the routing protocol, the coordination of tasks, and many other factors
related to the environment. Unfortunately, although there are many research works that address the
use cases of multi-UAV systems, there is a gap in the literature regarding the difficulties involved
Citation: Adoni, W.Y.H.; Lorenz, S.; with the implementation of these systems from scratch. This review article seeks to examine and
Fareedh, J.S.; Gloaguen, R.; understand the communication issues related to the implementation from scratch of autonomous
Bussmann, M. Investigation of
multi-UAV systems for collaborative decisions. The manuscript will also provide a formal definition
Autonomous Multi-UAV Systems for
of the ecosystem of a multi-UAV system, as well as a comparative study of UAV types and related
Target Detection in Distributed
works that highlight the use cases of multi-UAV systems. In addition to the mathematical modeling
Environment: Current Developments
of the collaborative target detection problem in distributed environments, this article establishes a
and Open Challenges. Drones 2023, 7,
263. https://doi.org/10.3390/
comparative study of communication architectures and routing protocols in a UAV network. After
drones7040263 reading this review paper, readers will benefit from the multicriteria decision-making roadmaps
to choose the right architectures and routing protocols adapted for specific missions. The open
Academic Editor: Diego
challenges and future directions described in this manuscript can be used to understand the current
González-Aguilera
limitations and how to overcome them to effectively exploit autonomous swarms in future trends.
Received: 17 March 2023
Revised: 4 April 2023 Keywords: UAV; RPAS; UAS; uncrewed aerial vehicles; drones; multi-UAV systems; autonomous
Accepted: 9 April 2023 swarm; autonomous aerial vehicles; communication architectures; FANET; routing protocols;
Published: 12 April 2023 collaborative missions; distributed environment; distributed path planning

Copyright: © 2023 by the authors.

1. Introduction
Licensee MDPI, Basel, Switzerland.
This article is an open access article Uncrewed aerial vehicles (UAVs) or drones [1] belong to the large family of connected
distributed under the terms and objects. After being utilized extensively in military operations [2,3], drones have remained
conditions of the Creative Commons out of reach for civilians [4] due to their high price. With recent innovations in micro-
Attribution (CC BY) license (https:// controllers and sensors, drone prices have been reduced to become affordable. Currently,
creativecommons.org/licenses/by/ they are widely used for commercial purposes such as surveys, photography, and cine-
4.0/). matography. UAVs, in general, are outfitted with on-board sensors that allow them to

Drones 2023, 7, 263. https://doi.org/10.3390/drones7040263 https://www.mdpi.com/journal/drones

Drones 2023, 7, 263 2 of 37

collect geospatial information about their environment and are remotely controlled from a
ground control station [5]. From this station, the operator can plan and supervise the evolu-
tion of the mission. Their use in the civil and industrial sector [4] has allowed us to optimize
industrial processes or to carry out missions in hostile environments that are partially or
entirely inaccessible to humans [6–9]. For example, in the agricultural field, farmers are now
facing various problems that impact the quality of their crops. Drones are an efficient and
cheap means to collect information on ecosystems and their variations due to, e.g., climate
change, soil erosion, water availability, and meteorological extreme events. They are, for
example, also used in spraying plantations, which saves time and optimizes yields [10].
Goodrich Payton et al. [11] have demonstrated the efficiency of drones in precision agricul-
ture. From the collected data, they reconstructed a 3D cartographic representation of the
plots to better analyze the density of vegetation and soil heterogeneities.
Drones provide a broad variety of purposes in the health sector, including delivering
medical supplies to remote or hard-to-reach areas, e.g., transporting blood samples and lab
results [12]. In the field of transportation, drones can be used for package delivery [13–15],
traffic monitoring [16], and infrastructure inspections [17]. Drones have also been used to
map volcanoes’ terrain and to detect volcanic activity. Thiele et al. [5] used drones equipped
with thermal cameras, gas sensors, and other instruments to measure temperature, gas
concentrations, and other indicators of volcanic activity. This information can be used to
predict eruptions, to elaborate rescue operations, for photogrammetry and infrastructure
monitoring, or even for delivery services. UAVs can also be used to study the geology of
a volcano and its surrounding area, providing valuable information for volcano research.
The use of UAVs in this field can greatly improve the efficiency, accuracy, and safety of
operations, as well as decrease costs by reducing the need for human intervention in
dangerous areas [5]. However, it implies several challenges related to communication
services, such the range, security system, and communication architecture [18].
The communication architecture of drones relies on a Flying Ad-Hoc Network (FANET) [19],
without the requirement for a fixed infrastructure. FANET is a decentralized ad-hoc
network that enables communication between the ground station and flying vehicles, such
as drones and aircraft [20]. As shown in Figure 1, FANET inherits from both Vehicular
Ad-Hoc Network (VANET) [21] and Mobile Ad-Hoc Network (MANET) [22] networks. It
is a subclass of MANET that is an extension to highly mobile devices such as smartphones
and laptops. These three types of ad-hoc networks share the ability to form and maintain
a network connection dynamically [23]. Combined, they are an effective tool in creating
a wide-area network that links UAVs, vehicles, and communication devices. They can
also be used for applications such as collision avoidance [24], traffic jam prevention, and
intelligent transportation systems [25]. In this manuscript, we will focus on the FANET
protocols as well as the different paradigms of data exchange between UAVs, especially for
multi-UAV systems.

Figure 1. Different types of wireless Ad-Hoc Networks: FANET [19] vs. VANET [21] and MANET [22].

As the need for UAVs grows, there is an increasing requirement for systems that can
coordinate several UAVs in a dispersed environment [1,26]. This is where multi-UAV sys-
tems are relevant. Multi-UAV systems, commonly called swarms, are composed of multiple
Drones 2023, 7, 263 3 of 37

UAVs that are coordinated to collaborate in order to accomplish a shared purpose [27].
These systems have the potential to coordinate mission tasks across multiple UAVs in a
parallel manner. They have several possible applications, including rescue missions [7,28],
surveillance and reconnaissance [6,26,29,30], environmental monitoring [17], and even
payload carrying [13–15]. One of the primary benefits of swarms is their capacity to operate
cohesively, allowing them to cover large regions quickly and make choices collectively,
making them more robust to failure than individual UAVs [31]. Swarms operate according
to the design of a multi-UAV system architecture that is controlled by a central system [6]
or through decentralized algorithms [1,32,33].
To work in a coordinated manner, a multi-UAV system requires the following key
components. (1) The UAVs themselves: these are the drones that make up the swarm.
(2) A ground control station: this is the central point of control for the drone swarm. It
oversees the transmission of commands to the drones, as well as the retrieval of data from
them. It can be a single computer or a cluster of computing nodes. (3) A communication
system: it consists of devices and antennas that enable communication between the drones
and the ground station via a common protocol. The communication system can use a
variety of technologies, including MANET [22], FANET [19], and VANET [21]. (4) A
navigation system: this is the system that enables the UAVs to fly and locate themselves in
the environment. It includes sensors such as GPS and Inertial Measurement Units (IMUs).
(5) A control system: it allows the ground station to control the drones and to coordinate
their actions. It consists of software for mission planning, decision making, and swarm
behavior. (6) Finally, a data processing system: this is the system that allows the ground
station to process the data in real time and feed the data back to the control system. It
can include tools for image processing, data analysis, and machine learning. All these
components work together to allow the swarm to function as a cohesive unit, with the
ground station providing overall command and control, while the UAVs collaborate to
attain a common aim.
Depending upon the operations’ nature and mission requirements, the architecture
of a multi-UAV system might be centralized [6] or decentralized [32,33]. Because of its
complexity and the very dynamic environment of operation, there are several challenges
related to using multi-UAV for highly mobile networks operating in large-area missions:
• Communication: maintaining communication between the drones in a swarm can be
challenging, especially in highly mobile and wide areas. The drones must be able to
maintain communication even when they are moving at high speeds or when they are
far apart from each other.
• Coordination: coordinating the actions of the UAVs is complex, especially in elaborate
missions. To reach a shared purpose, the drones must be able to successfully collaborate.
• Autonomy: the drones must be able to operate autonomously without human in-
tervention. This requires collaborative actions for decision making, navigation, and
swarm behavior. Theses challenges include spatial awareness, maintaining a distance
from each other, and communicating potential threats to other drones, such as heavy
wind gusts, rain, and obstructions.
• Scalability: multi-UAV systems must be able to support a large number of drones,
they must be reliable, and they must be able to scale up or down depending on the
requirements of the mission.
• Reliability: the system must be reliable, even if one or more UAVs fail. This requires
robust algorithms for fault detection, diagnosis, and recovery.
• Security: a multi-UAV system must be encrypted, guarding against hacking, jamming,
and other types of interference to the drones and their communications [18,34,35].
• Interference: in highly dynamic surroundings, drones may encounter interference
from other wireless devices, which can affect communication or navigation [18].
• Energy consumption: to ensure that the drones can operate for long periods of time,
energy management is crucial. Drones must be able to manage their power consump-
tion and plan their routes to optimize battery life [35–42].
Drones 2023, 7, 263 4 of 37

• Interoperability: the system must guarantee the exchange of information between

different types of drones regardless of their communication protocols. Unfortunately,
there is still no common protocol for drones to communicate with each other.
One key challenge in the design and implementation of multi-UAV systems is ensuring
that the UAVs work together effectively in a distributed environment. This requires the
development of robust communication and coordination architectures, as well as the
integration of sensors and other equipment to allow the UAVs to sense and adapt to their
environment. One of the main problems with swarms is the coordination of multiple
UAVs carrying out a task in a distributed area. This problem can be addressed by using
distributed algorithms. Algorithms such as consensus algorithms [43], leader–follower
algorithms [44], and distributed optimization algorithms [45] can be used to coordinate
multiple drones in a distributed area. These algorithms allow the UAVs to communicate
and coordinate with each other so that they can complete the task in an efficient manner.
Furthermore, these algorithms can also be used to ensure safety and prevent collisions
between UAVs (see Section 10.1). Overall, these challenges require sophisticated algorithms
and technologies, and the development of new solutions to these challenges is an active
area of research in the field of drone swarm technology.
In this research, we explore the problem of target detection in a distributed environ-
ment. We focus on a review of communication architectures and routing protocols, as well
as the integration of routing protocols to enable effective coordination among the UAVs. In
the second part, we show how to elaborate a centralized and decentralized communica-
tion architecture to achieve a collaborative mission. With this work, we aim to contribute
to the progress of multi-UAV systems and their potential use in a wide range of applications.

Innovations By leveraging the latest advancements in the current state of the art, this
work fills the gaps in the integration from scratch of autonomous multi-UAV systems
for collaborative actions. More specifically, the main added value of this work lies in the
following points:
• The field of autonomous swarms is still in its infancy and, as a result, there is a lack
of research that consistently provides a formalism of the constraints related to the
collaborative actions of multi-UAV systems. In this context, our work has added sig-
nificant value by filling the gap concerning a well-defined mathematical formulation
of collaborative actions of multi-UAV systems in a large distributed environment.
• The study of an autonomous and cooperative drone swarm infrastructure is an incred-
ibly complex field, and there are still very few scientific studies that fully explore the
intricacies of how it operates. Due to this lack, we conduct an extensive investigation
of multi-UAV infrastructures, which results in a scientific work providing valuable
knowledge to answer the question of how to choose the right communication archi-
tecture as well as the most suitable routing protocol. This will save a lot of time and
avoid erroneous conceptual studies.
• Despite the importance of understanding the collaborative paradigm of autonomous
multi-UAV systems for target detection in distributed environments, there is a lack of
clear visibility of the challenges and future directions in developing effective collabo-
rative swarm. As added value, we have highlighted the most important challenges
and clarified the open perspectives for research in this area.
Objectives The goal of this work is to motivate the use of multi-UAV systems by a well-
informed study of the constraints related to their implementation. We offer useful sug-
gestions on how to choose the type of UAV, the communication architecture, as well as
the type of protocol. These suggestions are based on a multi-criteria survey that considers
the complexity of the mission, the environment, and the capabilities of the UAVs used.
This work can serve as a background for all types of projects involving the use of multi-
UAV systems. In this paper, we formalize the problem of collaborative target detection in
large-scale environments. We conduct a comparative analysis of existing communication
Drones 2023, 7, 263 5 of 37

architectures and routing protocols for the efficient deployment of a swarm, especially for
complex missions. The following are the key contributions.
1. Types of UAVs: We review existing drones. The UAVs are classified into four cate-
gories according to their shape and size.
2. Problem formulation: We start by identifying the key features that allow us to recog-
nize a multi-UAV ecosystem. We emphasize the complexities and major challenges
associated with the collaborative target detection problem. Then, we propose a math-
ematical formulation of the problems of k-partition of the environment and path
planning to reach the targets. The proposed models are based on multi-objective
functions that satisfy the constraints of the workload balancing of the swarm and the
optimization of the paths.
3. Architectures and protocols: We give cutting-edge knowledge on multi-UAV commu-
nication architectures as well as the existing routing protocols. Then, we evaluate
the performance and draw a comparison for each communication architecture and
routing protocol listed in this paper.
4. Suggestions: Based on the comparisons of the existing architectures and routing
protocols, we give some suggestions for choosing the right architectures and routing
protocols that are tailored to a well-defined mission.
5. Open challenges: We discuss the ongoing issues and potential future directions, such
as collaborative decision making, communication infrastructure, spatial organization,
and regulation for multi-UAV deployment.
Organization The rest of this article is structured as follows. Section 2 provides useful
information on multi-UAV ecosystems. Subsequently, in Section 3, we perform a compar-
ative study of the types of autonomous uncrewed aerial vehicles. Based on this study,
we highlight the advantages and drawbacks of different types of UAVs. In Section 4, we
present formal definitions and mathematical explanations about the problem of collabora-
tive target identification. Furthermore, in Section 5, we provide the state of the art of some
applications of multi-UAV systems in the civilian domain. Meanwhile, in Sections 6 and 7,
we discuss the existing communication architectures and provide a roadmap that helps
to choose the most adapted. Similarly, Sections 8 and 9 provide a comparison of routing
protocols, followed by a roadmap of how to choose the right protocols. Section 10 discusses
the open difficulties and potential directions. Finally, in Section 11, we conclude this work.

2. Background
We provide here a detailed description and useful definitions and terminology for the
ecosystem of a multi-UAV system.

Uncrewed Aerial Vehicle (UAV)

An uncrewed aerial vehicle or Remotely Piloted Aircraft (RPA), also referred to as a drone,
is an aircraft that does not require a human pilot onboard. It flies autonomously or remotely
through an onboard computer, remote control, or a combination of the two. It is a type
of Uncrewed Aerial System (UAS), also known as a Remotely Piloted Aircraft System
(RPAS) [8], which is defined as an aircraft system that operates without direct human
intervention in the aircraft [46]. Most of the time, the interaction with the aircraft is made
possible through a ground control station, a remote controller, a data transmission system,
and dedicated hardware and software support (see Figure 2).

Autonomous UAV
Autonomous UAVs are a specific case of aerial vehicles that can fly to a given location
without the assistance of a remote human. This type of vehicle is able to perceive, interact,
and act with respect to its environment in an autonomous way for decision making on a
mission [7].
Drones 2023, 7, 263 6 of 37

Remote Controller (RC)

A remote controller is a device that can be hand-held by a pilot to communicate with a
drone via a live stream of data signals.

Ground Control Station (GCS)

A ground control station is the command post that offers the necessary elements to the hu-
man for the remote control of UAVs. A GCS is designed for flight planning and monitoring,
as well as the real-time processing of UAV–environment interactions. A GCS is equipped
with a Human–Machine Interface that conducts the planning of the missions, display and
control consoles, a real-time video stream, an on-board computer, telemetry instrumenta-
tion, a data communication platform, and links with the UAVs via antennas [30].

Figure 2. Overview of RPAS components.

Multi-UAV (M-UAV)
Multi-UAV is a group or swarm of UAVs that operate simultaneously to achieve a mission.
In the architecture of M-UAV, we distinguish two types of interaction: cooperation and
coordination. Coordination [43–45] involves sharing the operational environment with
other drones safely. It therefore performs an important role in large space exploration
missions involving numerous UAVs. Meanwhile, in the logic of cooperation [47,48], drones
collaborate on different tasks to achieve a common goal.

Environment
An environment represents the exploration domain of the UAV. The domain could be either
aerial, terrestrial, or marine (oceanic or aquatic). It can be static or dynamic depending
on the natural factors that influence its complexity, such as climatic conditions, weather,
vegetation, as well as generic elements.

Perception
Perception is the input information that the UAV receives through embedded sensors from
its environment at any moment in order to make decisions about actions to take. The
sensors commonly used for UAV perception are proximity sensors, cameras for object
recognition, and laser sensors for range detection.

Target
A target is a physical entity or an object located in an exploration space. It is considered
static if it does not present any form of mobility in its environment or dynamic in the case
of spatio-temporal changes.

Mission
A mission is a task entrusted to a UAV. Drones are increasingly employed in a range of
operations, including the following: mapping, exploration, rescue, investigation, reconnais-
sance, traffic monitoring [16,48], weather monitoring, earth observation, firefighting [5],
Drones 2023, 7, 263 7 of 37

agriculture [11], photography, and videography. A mission is said to be collaborative [43–45,47–49]

when multiple UAVs work together to accomplish the mission.

Waypoints
Waypoints represent a set of intermediate positions through which a UAV must pass
to reach a well-known final position. Obviously, a deviation tolerance threshold must
be considered.
Figure 3 depicts an interactive view of the ecosystem of a multi-UAV system. It is
composed of a ground environment in which there are two categories of targets, i.e., mobile
(2 persons and 1 dog) and fixed (trees), and an aerial environment in which there are
non-autonomous and autonomous UAVs. The non-autonomous UAVs are guided by an
operator from the ground, while the autonomous UAVs carry out their mission without
human assistance. Each drone has the capacity for the perception of the environment. It
can collaborate with other UAVs (air-to-air). It can also collect data and communicate them
to the GCS from the ground (air-to-ground). The mission planning as well as the command
and control operations are centralized in the GCS, which stores the data received from
the UAVs.

Figure 3. Overview of a multi-UAV ecosystem.

3. Classification of UAVs
The growing interest in UAVs in recent years has led to the strong emergence of
various types of aircraft with varying configurations and components in terms of shape
and size. UAVs are divided into four types: single-rotor, multi-rotor, fixed-wing, and
hybrid [33], as shown in Figure 4.
1. Single-rotor [50] (or helicopter): this category of UAVs takes off and lands vertically.
Generally, they use a main rotor for attitude control (roll, pitch, and yaw) and a tail
rotor to control the direction. Its main advantage is the ability to carry heavy payloads
over a longer flight time. However, the complexity of their mechanical systems, as
well as the large size and high cost of the rotors constitute a danger for the uncrewed
versions. It is more unstable in bad weather.
2. Multi-rotor [51] (or multicopter): this is a UAV having more than two rotors. This
UAV category is further subdivided into five sub-categories, which are birotor, trirotor,
Drones 2023, 7, 263 8 of 37

quadrotor, hexarotor, and eight-rotor (octocopter) [13,15]. As with single-rotor UAVs,

multi-rotor UAVs also ensure a vertical takeoff and landing. They are fast and agile in
flight, allowing them to perform complex maneuvers and flights in confined spaces.
However, the short flight time is the main weakness of these types of aircraft.
3. Fixed-wing [52]: the navigation principle of this category of UAVs is based on a
simple structure of a fixed rigid wing. The classification of these drones is not only
based on the type of wing, but also on the body and the power system (Li-ion, Li-Po
batteries, or gas-powered). They are subdivided into four subcategories: normal,
swept back, swept forward, and delta. Furthermore, they can carry heavier payloads
than multi-rotors [53]. The disadvantage of these UAVs is that they have limited
agility in flight, which does not allow them to perform complex maneuvers and fly
over confined spaces, as well as the necessity for a runway for takeoff and landing.
4. Hybrid [54]: this last category is still under development. It is an improved version
that takes advantage of both multi-rotor and fixed-wing UAVs. They offer good
agility and velocity on long-distance flights. They can carry large payloads and do
not require a runway. The main disadvantages are the high price, the complicated
mechanics, and the lower performance in terms of flight stability and the restrictive
speed ranges.

Figure 4. Different categories of UAVs.

Table 1 presents a brief comparative study of the advantages and disadvantages of

each category of UAV. This comparison focuses on payload capacity, endurance, agility,
speed, and area of coverage.
Generally, the classification is made according to the weight, size, range, and en-
durance [32]. We will focus on the classification according to the size, which will be useful
for this discussion. As shown in Table 2, four classes of UAVs, organized as very small,
small, medium, and large, are usually defined.
Very small UAVs refer to micro- or nano-UAVs. These UAVs resemble insects or birds
with wings, and their dimensions generally vary between 5 and 50 cm. The components are
extremely small and lightweight. They can reach a top speed of >10 km/h and are limited
to a maximum altitude of ca. 120 m by law. Typical examples of very small UAVs include
the German TobyRich SmartPlane Pro (https://www.drohnen.de/tag/tobyrich/, accessed
on 15 March 2023) with a wing length of 30 cm; it is ultra-stable and equipped with a VGA
camera of 640 × 480 pixels. Moreover, the DJI Mavic 3 (https://store.dji.com/de/product/
dji-mavic-3, accessed on 15 March 2023), configured with dimensions of 28.3 × 10.7 cm2 , is
Drones 2023, 7, 263 9 of 37

another small model of UAV; it is equipped with a 4:3 CMOS camera and can fly for 46 min
with a maximum transmission range of ca. 15 km.

Table 1. Comparative study of UAVs: advantages vs. disadvantages.

Categories Strengths Weak Points

Heavy payload
Flight time Mechanical system
Single-rotor
Hovering flight Rotor size
VTOL
Hovering flight
VTOL
Multi-rotor Velocity Flight time
Agile maneuverability
Confined-space flying
Complex aerofoil
Flight time
Limited payload
Fixed-wing High Velocity
No hovering flight
Large-area coverage
Limited maneuverability
Flight time
Hovering flight
Very expensive
VTOL
Hybrid Unstable transition mechanism from
Agile maneuverability
horizontal to vertical flight
Velocity
Large-area coverage

Small UAVs, often known as mini-UAVs, refer to UAVs whose dimensions exceed at
least 50 cm and not more than 2 m. The design of these aircraft is essentially based on the
fixed-wing model, and the majority are launched by propelling them into the air by the
operator. They can carry a maximum payload of 9 kg and fly at a maximum speed of ca.
150 km/h at an altitude usually not exceeding 400 m. The German Wingcopter 198 (https:
//wingcopter.com/, accessed on 15 March 2023) is an electric VTOL drone used for delivery
services. It is designed with a configuration of 198 × 154 cm2 with a flight time of 90 min.
It is hybrid and can fly in both multicopter and fixed-wing modes. The second one is Astro
(https://freeflysystems.com/alta-x, accessed on 15 March 2023), a quadrotor designed
with a configuration of 141 × 51 cm2 with a flight time of 37 min. It incorporates a gimbaled
a7R IV mapping camera and the data transfer is based on the MAVLink communication
protocol [55]. The last one is Scorpion (https://www.quantum-systems.com/, accessed on
15 March 2023), a VTOL that has a wingspan of 1.37 m2 and a flight endurance of 35 min. It
provides data transfer up to a range of 25 km and it supports a maximum weight of >7 kg.
A UAV is considered “medium” if it is too heavy to be handled by one person but
smaller than an aeroplane, since they can only carry a payload of 200 kg. They are usually in
the category of fixed-wing UAVs and typically have a wingspan whose length ranges from
5 to 10 m. They can fly at a maximum velocity of 463 km/h without exceeding an altitude of
almost 1 km. There are numerous examples of UAVs in this scope of size, such as the Trinity
F90+ (https://www.quantum-systems.com/, accessed on 15 March 2023), Vector (https://
www.quantum-systems.com/, accessed on 15 March 2023), Alta X (https://freeflysystems.
com/alta-x, accessed on 15 March 2023), and, recently, the Yangda YD6-1600S (https:
//www.yangdaonline.com/, accessed on 15 March 2023). The German intelligent VTOL
Trinity F90+ and Vector have average fixed wings of 2.8 m and, respectively, a flight time of
90 min and 120 min. The Alta X is a VTOL quadrotor with a 50 min flight time and it carries
a maximum payload of 15 kg for 8 min. It has a peak speed of more than >95 km/h and
uses the MAVLink [55] protocol for data transmission. The hexacopter Yangda YD6-1600S
is designed with dimensions of 1.6 × 2.35 m2 . It can reach a cruise speed of >72 km/h and
can also carry a maximum payload of 5 kg for a flight time not exceeding 45 min.
Drones 2023, 7, 263 10 of 37

The last class of UAVs is mainly used for military purposes. The large UAVs have a
wide range and endurance. Furthermore, their designs are usually based on a fixed-wing
structure, which allows them to carry heavy payloads over long distances while reaching a
maximum altitude of 5.5 km. Some examples of these UAVs are the Czech Primoco UAV
One 150 (https://uav-stol.com/primoco-uav-one-150/, accessed on 15 March 2023) and
the Chinese Feng Ru 3-100 (https://ev.buaa.edu.cn/info/1133/3165.htm, accessed on 15
March 2023). Primoco is designed to fly for 15 h. It can carry a payload of 50 kg and has a
radio range of 200 km. Feng Ru is designed for couriers; it has a wingspan of 19.6 m and
can fly for 5 days.
Currently, there is no standard classification of UAVs. In the case of this study, we
have focused on size. However, other important parameters can be considered, such as
autonomy, speed, range, altitude, and payload capacity.

Table 2. UAV classification according to the size (https://www.e-education.psu.edu/geog892/node/5,

accessed on 15 March 2023).

Dimensions Payload Velocity Altitude

Size Example (Accessed on 15 March 2023)
(m2 ) (kg) (km/h) (km)
SmartPlane Pro (https://www.drohnen.de/tag/tobyrich/)
Very small 0.3–0.5 ≤10 <0.12
DJI Mavic 3 (https://store.dji.com/de/product/dji-mavic-3)
<9 Wingcopter 198 (https://wingcopter.com/)
Small 0.51–2 <185 <0.4 Astro (https://freeflysystems.com/alta-x)
Scorpion (https://www.quantum-systems.com/)
Trinity F90+ (https://www.quantum-systems.com/)
Vector (https://www.quantum-systems.com/)
Medium 5–10 <200 <1.1
Alta X (https://freeflysystems.com/alta-x)
<463 Yangda YD6-1600S (https://www.yangdaonline.com/)
Primoco UAV One 150 (https://uav-stol.com/primoco-uav-one-150/)
Large >10 <600 <5.5
Feng Ru 3-100 (https://ev.buaa.edu.cn/info/1133/3165.htm)

4. Problem Statement
This section introduces several notations and formal definitions relevant to the problem
statement. Following this, we will present a mathematical description of the problem of
finding targets in a distributed environment. Overall, the use of a multi-UAV system to
reach several targets in a large environment goes through two steps, as shown in Figure 5.
First, the pre-processing step, which is performed on the GCS, consists in analyzing the
properties of the environment in order to decide on the partitioning logic as well as the path
planning algorithm. Then, in the second phase, each UAV assigned to its sub-environment
is tasked with detecting, classifying, and finding the optimal path to reach a well-defined
target in a distributed manner.

Figure 5. Workflow of target detection with multi-UAV system.

4.1. Notations and Formalism

4.1.1. Environment E
An environment E = ( P, L, W ) is a three-dimensional finite space composed of the
following:
• A set of points P = { p1 , ..., pn } → R3 such as ∀i ∈ J1..nK; the positions pi =
( xi , yi .zi ) ∈ R3 of these waypoints are defined by 3D Cartesian coordinates.
Drones 2023, 7, 263 11 of 37

• A set of lines L = {( pi , p j )| pi , p j ∈ P × P} that connect the waypoints.

• A weighting function W : E → R associated with the lines of E, such as W =
{w( pi , p j )| pi , p j ∈ P2 and pi 6= p j }.
The number of finite elements of E is the cardinal | P| and the density (weight) of the
environment is | L|. We will denote the shape of the environment by s( E)

4.1.2. Sub-Environment Ei
Let E = ( P, L, W ) and Ei = ( Pi , Li , W ) be two environments. Ei is a sub-environment
of E if Pi is a subset (Pi ⊂ P) of P and Li is a subset (Li ⊂ L) of L.
Then, the shape and density of the sub-environment Ei are, respectively, s( Ei ) ≤ s( E)
and | Li | ≤ | L|.

4.1.3. Partition PEk

Let E be a non-empty environment and k ≥ 2 is the number of sub-environments in
the partition. We say that PEk = { E1 , ..., Ek } is a k-partition of E if
• ∀i ∈ J1..kK, Ei 6= ∅.
• ∀i, j ∈ J1..kK such as i 6= j; we obtain Ei ∩ Ej = ∅.
k
[
• Ei = E.
i =1
Thus, the elements of PEk (sub-environments of E) are not empty and they are disjoint
two by two. In other words, they have no element in common two by two.
The balance B( PEk ) indicates that the partitioning is spread equally. It is calculated
as follows:
max {| L1 |, · · · , | Lk | }
B( PEk ) = (1)
| L avg |
where | L avg | is the average weight of the partition.

k
∑ Li
i =1
| L avg | = (2)
k
If the criterion B( Pk ) < 1 + e is met, then the partition Pk is well balanced considering
an acceptance error e.

4.1.4. Target T
A collection of finite elements T = { T1 , ..., Tk } that lie in E are defined by a class C
of discrete objects such that E × C → T. Here, Ti is the set of targets associated with the
sub-environment Ei , such as the following:
• T = ∅, if E contains no target.
Sk
• i =1 Ti = T.
• ∀i, Ti 6= ∅, if and only if Ei contains at least one target.
• ∀i, j ∈ J1..kK such as i 6= j, we obtain | Ti | > | Tj | ⇒ Ei and it contains more targets
than Ej .
• ∀i, j ∈ J1..kK such as i 6= j, we obtain Ti ∩ Tj = ∅ ⇒ and they share common targets.

4.1.5. M-UAV
M-UAV is a swarm of k UAVs such as M-UAV = {UAV1 , ...,UAVk }. We assume that we
assign each drone or UAVi to its sub-environment Ei in such a way that the environment E
is covered by M-UAV.
Drones 2023, 7, 263 12 of 37

4.1.6. Perception P
Perception P = ( E, T, A) is the capability of the swarm M-UAV to take information
related to E as input and T to perform a set of actions A according to the function f defined
as f : E × T 7→ A.
∀i ∈ J1..kK, the perception Pi of a given drone UAVi from its observation on Ei and Ti
are defined as Pi = ( Ei , Ti , A).

4.2. Collaborative Target Detection Problem

The exploration of an environment E by M-UAV consists of detecting, analyzing, and
finding the path to reach these targets. For a large-scale environment, the complexity
becomes exponential with respect to the number of these elements. This becomes even
more problematic when using a UAV based on a centralized architecture (UAV-to-GCS)
because of the latency. In addition, the UAVs’ limited flying autonomy must be considered.
Therefore, to cope with these problems, a decentralized M-UAV system based on the divide-
and-conquer [56] technique should be used. It consists in dividing the exploration problem
into sub-problems according to the number k of drones used [12,40,49,57].
The division of the problem involves two challenges, which are (1) the partitioning of
the environment E, commonly called the k-partition problem, and (2) the use of M-UAV
such that each UAVi explores its sub-environment Ei and computes the shortest path
leading to a target of Ti ; it is known as the Single-Source Shortest Path problem [58].

4.2.1. k-Partition Problem

For a fixed k, the partitioning of a large environment E consists in generating a k-
partition PEk such that ∀i, j ∈ J1..kK, and we obtain Ei ∩ Ej = ∅ while minimizing the
number | Ti ∩ Tj | of shared targets between Ei and Ej . A path is optimal if it minimizes the
sum of the costs between two points. With the addition of obstacles and some constraints,
this problem can change from a polynomial complexity to an NP-Complete problem [58].
This problem can be formulated as a linear optimization problem.
The first challenge is the minimization of shared targets between Ei and Ej , as it
reduces the communication costs between drones UAVi and UAV j . The second challenge is
balancing the weights of the k sub-environments. For k = 2, it is a simple bi-partition of
the environment, whereas for k > 2, the problem is NP-Complete [58] because there is no
optimal solution algorithm in polynomial time.
Partitioning can be done in three ways, namely shape-based partition, density-based
partition, and target-based partition, as illustrated in Figure 6. Shape-based partition-
ing consists of partitioning the environment E into k sub-environments with identical
dimensions, such as ∀i, j ∈ J1..kK, and we have s( Ei ) ≈ s( Ej ). On the other hand, the
density-based partition consists in having a k-partition such as ∀i, j ∈ J1..kK, and we have
| Ei | ≈ | Ej |. Finally, the one based on targets allows us to obtain a partition of the environ-
ment by considering the uniform distribution of targets on each sub-environment, such as
∀i, j ∈ J1..kK, where we have | Ti | ≈ | Ej |.
We conclude that the mathematical model associated with the partitioning of the
environment E within the constraints is described as follows:

minimize
 | Ti ∩ Tj |

 i 6= j∈J1..kK
 k
subject to: B( PE ) ≤ 1 + e



Ei ∩ Ej = ∅ (3)


 Ei 6= Ej
∪ik=1 Ei = E




∪ik=1 Ti = T


Drones 2023, 7, 263 13 of 37

Figure 6. Illustration of a bi-partition (k = 2) of an environment for two UAVs with the three
partitioning approaches.

4.2.2. Single-Source Shortest Path Problem (SSSP)

After the partitioning of the environment E into PEk , the next step consists in finding
the paths that join a set of targets Ti detected in Ei in a distributed way. Each drone UAVi
calculates the shortest path to reach a target at the position ptarget from its initial position
pinit ∈ Vi , as shown in Figure 7. It is an algorithmic optimization problem referred to as
the Single-Source Shortest Path problem and is well known in operations research and
graph theory [59]. It consists in finding an optimal or near-optimal path between two nodes
of a directed graph. It is a multi-criteria optimization problem aiming at satisfying the
reachability of the final point ptarget while minimizing the costs (e.g., distance or time) of
the path.
There are three variants of the SSSP problem: (1) the dual-path problem, which consists
of finding a set of paths leading to a target from several sources; (2) the all-pair problem,
which consists in finding the optimal path between all pairs of points; and, finally, (3) the
single-source problem, which consists in finding an optimal path from a source that joins all
the targets Ti of a sub-environment Ei . In our study, we focus on the latter type of problem.
Suppose a drone UAVi assigned to the sub-environment Ei = ( Pi , Li , W ) for an explo-
ration mission that consists in finding a set of targets Ti such as the following:
• pinit ∈ Pi : the origin point or current position of the drone UAVi .
• ptarget ∈ Pi ∩ Ti : the goal point or position of the target ti ∈ Ti .
• pinit 6= ptarget .
• path: the path to reach ptarget from pinit .
We consider the linear program by introducing a binary decision variable xs,d ∈ {0, 1}.
This variable indicates whether the weight w( ps , pd ) associated with the line between ps
and pd minimizes the path cost. It is equal to 1 if the weight minimizes the cost w( path) of
the path and 0 in the opposite case:
Drones 2023, 7, 263 14 of 37

min w( path) = ∑ w( ps , pd ) × xs,d (4)

( ps ,pd )∈ Li

Thus, each UAVi will seek to determine the minimum weight lines that optimize and
favor the progression towards the target ptarget position from its initial position pinit . This
condition is defined as a flow conservation problem [60]; it stipulates that the sum of the
flows into a given point must be equal to the sum of the flows out of that point pi , apart
from the start pinit and end ptarget points.
The mathematical formulation of this condition is formulated as follows:

∀ pi ∈ Pi \{ pinit , ptarget }, ∑ xi,j = ∑ x j,i (5)

( pi ,p j )∈ Li ( p j ,pi )∈ Li

For each UAVi , the mathematical model of the problem of reaching the position ptarget
of a target t ∈ Ti in its environment Ei is described as follows:

∑


 minimize w( pi , p j ) × xi,j
 pi 6= p j ∈ Pi


 ( pi ,p j )∈ Li

∑ ∑

subject to: xi,j = x j,i (6)

 ( pi ,p j )∈ Li ( p j ,pi )∈ Li

xi,j ∈ {0, 1}




pi , p j ∈ Pi \{ pinit , ptarget }


If we set d( pi ) as the distance to reach the point pi from the initial point pinit and if we
set d( ptarget ) = 0, then, for each point pi to be explored by the UAVi in its environment Ei ,
the resulting mathematical model will be written as follows:

∑ (d( pinit ) − d( pi )) = d( ptarget ) − d( pinit )


minimize
pi 6= p j ∈ Pi pi ∈ Pi (7)
subject to: d ( p j ) − d ( pi ) > w ( pi , p j )


There are several shortest path search algorithms; the most popular are Bellman–Ford [61],
Dijkstra [62], Floyd–Warshall [63], and A* [64,65], which uses a heuristic function to predict
the path ahead and avoid obstacles.

Figure 7. Illustration of a decentralized path on UAV1 and UAV2 to reach targets, respectively, in
sub-environments E1 and E2 .

5. State of the Art

The concept of multi-UAV was first used in the military domain for recognition, before
being popularized in the industrial and civilian domains [66]. The fact of having several
Drones 2023, 7, 263 15 of 37

UAVs has been shown to minimize the duration of missions and increase the satisfaction
rate, especially for complex missions [27].

5.1. Military
The use of multi-UAV in the military field represents a real advantage, such as low-
ering the loss of both soldiers’ lives and the military arsenal [2]. In October 2016, the US
Department of Defense (https://www.defense.gov/News/Releases/Release/Article/1044
811/department-of-defense-announces-successful-micro-drone-demonstration/, accessed
on 15 March 2023) successfully tested one of the greatest deployments of 103 small UAVs
in California. The micro-UAVs flew together under an adaptive formation and each drone
collaborated with the others by using one distributed brain to achieve their mission.
In the military area, swarm UAVs are generally used for attack, defense [3], and
intelligence missions such as surveillance and reconnaissance [2]. The authors in [2,3] have
shown that the swarm UAV attack is effective for offensive missions because it allows
them to completely invade the enemy’s defenses with a high probability of hitting several
targets. This is the tactic of “quality by quantity”. It consists of attacking simultaneously
a target according to a defined formation. The most used formation is the simultaneous
attack in a dispersed way. For example, if a swarm composed of 5 UAVs attacks a target
simultaneously, the probability of shooting down one UAV is 15 , which is 20%. This
probability evolves according to the scalability of the UAVs in the swarm.
Although UAVs show a good ability in attack missions, they can also be used for
defensive or counter-offensive purposes against other enemy UAVs. Scharre [3] showed
that the use of numerous miniature decoy UAVs could disrupt the mission of the attacking
swarm by eliminating some enemy UAVs. This will have a considerable impact on the
cooperation and coordination mechanism of the attacking swarm. It should be noted that
the number of UAVs per swarm is very decisive regarding who would win.

5.2. Surveillance and Reconnaissance

Surveillance and reconnaissance are two of the most popular use cases in which
both the military and civilian domains cooperate. The rapid emergence of connected
objects and the standardization of those related to UAVs as well as artificial intelligence
through computer vision has created an ecosystem favorable to the integration of several
surveillance sensors, such as RGB and thermal cameras. Mohamed et al. [27] showed
that UAV surveillance is more advantageous than the conventional surveillance technique
based on fixed CCTV cameras during reconnaissance patrols. In addition to minimizing
the risk of casualties, it allows for the rapid and wide exploration of surveillance areas.
Multi-UAV systems are more suitable for surveillance and reconnaissance over a large
area because the scalability of the number of UAVs per swarm favors the coverage of large
areas and intelligence in less time. Ali et al. [26] presented a path planning system for
real-time surveillance employing cooperative multi-UAVs. Since the main challenge of this
work is similar to the problem of cooperative motion planning, they used a swarm of UAVs
that need to control, monitor, and periodically generate mosaic maps of their respective
mission areas.
Similarly, Matej et al. [29] presented a surveillance mission planning based on a
cooperative multi-UAV system for space coverage optimization to reach suitable sensing
locations. The proposed solution works in two phases. Firstly, they used the Rapid
Exploration Random Tree algorithm to quickly explore the space until the sensing locations
were found. Then, they used Particle Swarm Optimization to maximize the gathered
information about the reached sensing locations.
The fields of application of multi-UAV systems in surveillance are varied [30,67].
One of them [67] integrated the Oriented Line Segment Coverage Problem (OLSC) model
to improve intra-swarm cooperation. Each drone is fitted with an onboard camera that
provides large coverage of the space in order to detect and identify the size and direction
of targets. The architecture of the proposed model is centered around a GCS that plans
Drones 2023, 7, 263 16 of 37

the missions and a master UAV that controls and coordinates the movements of the slave
UAVs to cover all the targets. Following the same work, Daniel Perez et al. [30] have
developed a GCS for surveillance missions with multi-UAV. The proposed system is
decentralized, allowing the dispatching of tasks to the UAVs. Each UAV displays in real
time its visualization operation in the 3D environment. They demonstrated the reliability
of the system by using two quadrotors equipped with cameras.

5.3. Payload Carrying

The carriage of payloads such as packages is another area of growing interest in the
field of logistics and supply chain management. Several companies in the distribution
and delivery sector are conducting research on the automation of their delivery services
(healthcare delivery, food delivery, postal delivery) by multi-UAV systems [12,39]. Most of
the proposed approaches are adapted mainly for collaborative payload carrying with a set
of UAVs.
Daniel Mellinger et al. [13] presented cooperative grasping and transport using a
swarm of quadrotors to move a heavy payload of known structure and weight in a three-
dimensional space. Since each UAV knows its position, the task of transporting the payload
is decentralized in such a way that each UAV controls its trajectory with respect to the
payload stabilization laws. On the other hand, the velocity and the state estimation of the
positions of the swarm towards the goal are centralized on the GCS.
Following the same approach, the authors in [15] addressed the problem of carrying
heavy rigid payloads with a multi-UAV system composed of small quadrotors. Each UAV
is equipped with a camera and Inertial Measurement Unit sensor. The proposed approach
allows them to (1) independently control the motion of each UAV while maintaining the
stability of the system regarding the inertial and (2) each UAV benefits from a cooperative
scheme, allowing them to measure the positions of the other UAVs.
Robin and Raffaello proposed in [14] a method for carrying a flexible structure of
payload in three-dimensional space by using a swarm of quadrotors rigidly attached to
it. The proposed approach is comparable to previous work [13], except that the stabiliza-
tion and estimation of the payload deformation are centralized on the GCS. They use a
linear-quadratic regulator (LQR) function, which allows them to control and regulate the
parameters governing the payload shape.

5.4. Exploration
The ability of UAVs to reach inaccessible areas is a great advantage for exploratory
missions, especially in land and marine environments. Nowadays, UAVs are commonly
used for geological exploration and archaeological excavation missions. In [68], the authors
applied model-free deep reinforcement learning on a UAV swarm to accomplish exploration
missions. Each UAV is defined as an autonomous agent that interacts with its environment
through some actions. For each movement and action, the agent receives a reward (bonus
or malus). The objective is to maximize the reward of the agents’ actions with the aim of
better understanding the exploration policy of the environment.
Singh et al. [69] addressed the challenge of multi-UAV exploration by optimizing the
paths of UAVs investigating a large and dynamic environment. The optimal environment
exploration challenge is described as a pre-learned Gaussian process and is solved cen-
trally. Each UAV agent evaluates the quality of its exploration process by using mutual
information between both the explored and the rest of the environment.
Furthermore, decentralized techniques provide well-known exploration benefits.
In [70], the authors proposed a decentralized multi-agent version for environment ex-
ploration. They tackled the following three challenges: determining which information
should be shared to enable unsupervised multi-agent coordination; (2) avoiding collisions;
and (3) learning the environment’s hyper-parameters while exploring.
Following the previous work, a decentralized control approach for exploring unknown
environments with a multi-Turtlebot system composed of three robots was presented in [17].
Drones 2023, 7, 263 17 of 37

For controlling the coordination between the uncrewed robots, they implemented a greedy
decentralized method, allowing each robot to navigate and map an unknown environment
without exceeding its boundaries.
There are also other research projects [8,9] that expand the exploration to archaeological
excavation missions. For example, Efstathios and Fulvio [8] showed a state-of-the-art and
deep analysis of archaeological remote sensing based on multi-UAV systems. Most existing
methods are time-consuming, require a significant amount of hardware resources, and are
computationally expensive. Using a multi-UAV system not only decreases the mapping
time by distributing the coverage area, but it also enhances the map quality through
information sharing [49]. It is possible to create a fast 3D map of archaeological sites using a
collaborative mapping approach. This approach consists in partitioning the archaeological
space and assigning portions to each UAV. Thus, each UAV performs the mapping task of
its space and exchanges information on the boundaries of its space to the whole swarm for
the achievement of the mapping mission in a distributed way [9].

5.5. Other Applications

The application domains of multi-UAV are varied and almost ubiquitous. Apart from
the mentioned works, several use cases remain to be explored. Sharma et al. [48] present an
application of multi-UAV systems for road traffic monitoring. There, they present a swarm
coordinated with Vehicular Ad-Hoc Networks (VANETs) for vehicle tracking and driver
behavior analysis. This approach reduces the number of accidents due to unsafe drivers
and regulates traffic better. More applications and challenges of multi-UAV systems related
to intelligent transportation systems for smart cities are highlighted in [16,71].
Another research work [72] in the same scope presents a cooperative traffic monitoring
system based on a two-layer network (UAV aerial layer and vehicle ground layer). The
UAVs in the FANET layer work together to control vehicles from the ground in the VANET
layer. The link between UAVs is performed through air-to-air (A2A) communication,
while the one inside the ground vehicular sub-network relies on air-to-ground (A2G)
communication.
Another relevant use of multi-UAV is search and rescue [7,28]. The authors in [7]
implemented a rescue mission using a set of stream videos from a swarm to a distributed
ground control station. A similar work [28] introduces a machine learning approach to
predict the best UAV swarm formation, allowing them to find the optimal path loss for
the rescue.

6. Existing Multi-UAV Communication Architectures

The communication architecture of a multi-UAV system is a key factor in the structure
of the network infrastructure and the interoperability between UAVs. It has undergone
various improvements, moving from centralized to decentralized communications [1,19]
(see Figure 8). Centralized architectures are essentially based on the multi-directional
communication model between UAVs and GCS [6], while the decentralized architectures
are based on various models, allowing us to guarantee the scalability of UAVs, the interop-
erability between UAVs, and the coordination of several tasks on the UAVs to achieve a
common goal [45].
The multi-UAV communication architectures are classified into four types, centralized,
single-group, multi-group, and multi-level architectures [1], as shown in Figure 8. In this
section, we will discuss the different research works conducted toward the decentralization
of multi-UAV communication architectures.
Drones 2023, 7, 263 18 of 37

Figure 8. Types of multi-UAV communication architectures discussed in this paper.

6.1. Centralized Communication Architecture

A centralized multi-UAV architecture is a centralized data-oriented architecture that
promotes only communication between UAVs and the GCS (air-to-ground) [6]. As shown in
Figure 9, each UAV establishes directly two-way communication with the GCS. Moreover,
the control and command operations are centralized in the GCS. Generally, this architec-
ture supports mini-UAVs and implements simple routing protocols [73,74]. Its field of
application is limited to straightforward missions with limited coverage.
However, this type of architecture does not support intra-UAV communications (air-to-
air) [75]. Therefore, it is not possible to implement collaborative missions. Moreover, since
there is a single point of failure in this architecture, it is vulnerable and can lead to system
paralysis in case of failure [73]. In this context, the decentralization of the architecture will
allow us to have intra-communication between the UAVs and to eliminate the dependencies
on the GCS. The next section will cover the different types of decentralized architectures.

Figure 9. Multi-UAV communication architecture based on centralized framework.

Drones 2023, 7, 263 19 of 37

6.2. Single-Group Communication Architecture

A single-group architecture is commonly used for collaborative missions in small
coverage areas. It is a decentralized architecture that promotes intra-communication
between UAVs (air-to-air), thus eliminating the dependency on the GCS, as shown in
Figure 10. The communication between the GCS and the UAVs (air-to-ground) is achieved
through a particular UAV called a “Gateway UAV” [1]. It oversees the analysis of the
information from the GCS before sharing it with all the UAVs according to the ad hoc
network topology. In [1], the authors present three types of intra-UAV communication
topologies, ring, star, and meshed, as shown in Figure 11.
1. Ring: in an intra-UAV communication-based ring topology, the UAVs communicate
in a bidirectional loop in a closed network. To guarantee the high availability of the
information, any UAV can be used as a gateway to relay the information between
the GCS and the rest of the UAVs. On the other hand, it is difficult to maintain the
scalability of the UAVs because of the network topology.
2. Star: in a star topology, a single UAV relays information from the GCS infrastructure
and shares it with the others. Any communication between two UAVs must pass
through the gateway UAV. It is easy to maintain scalability. However, it does not
guarantee fault tolerance because, if the gateway UAV fails, the whole system stops.
3. Meshed: it takes advantage of the ring and star topologies. It allows us to guarantee
both scalability and fault tolerance.

Figure 10. Multi-UAV communication architecture based on single group.

The diversified applications of UAVs have led to the increasing use of a variety of
UAVs of different sizes and functions. Unfortunately, this architecture does not meet this
type of demand because it does not allow the incorporation of different types of UAVs [1].
Moreover, it is adapted for short-range intra-communication, but it may be the case that
some UAVs move away from the others during certain missions in large environments. To
overcome this problem, an alternative is to create several groups of UAVs. In the following,
we will discuss the two types of multi-group architectures.
Drones 2023, 7, 263 20 of 37

Figure 11. Types of intra-UAV communication: (a) ring topology, (b) star topology, (c) meshed topology.

6.3. Multi-Group Communication Architecture

A multi-group architecture takes advantage of both centralized and single-group
architectures. As shown in Figure 12, it extends the single-group architecture into several
groups of homogeneous UAVs that communicate with each other according to the intra-
UAV topology [19] seen in Section 6.2. Each group has a gateway UAV. In addition,
the communication between groups of UAVs (group-to-group) is done through gateway
UAVs that communicate bidirectionally with the GCS. The robustness of this architecture
favors collaborative missions over large coverage areas while maintaining control from the
GCS [28]. However, there are some limitations to be mentioned:
• Single point of failure: if the gateway UAV fails, then all other UAVs in its group
remain inactive. If the GCS fails, then the system will become totally paralyzed.
• Homogeneity of UAVs: in a group, the UAVs must be of the same type, and this is a
hindrance for some missions. Moreover, this implies serious interoperability problems
between groups of UAVs.
In the next section, we will discuss a type of architecture that overcomes the limitations
of the multi-group architecture.

Figure 12. Multi-UAV communication architecture based on multi-group.

6.4. Multi-Level Communication Architecture

A multi-layer communication architecture is an extension of the multi-group archi-
tecture to a three-tier architecture. It is adapted to support the scalability of UAVs [19].
Unlike the multi-group architecture, the communication between the groups of UAVs is
Drones 2023, 7, 263 21 of 37

achieved without using a GCS infrastructure [28]. It is composed of three layers, as shown
in Figure 13:
1. Layer 1 (A2A) is adapted for UAV-to-UAV communication. It allows the intra-
communication of a group of adjacent UAVs according to the three network topologies
seen in Section 6.2.
2. Layer 2 (G2G) is suitable for group-to-group communication. It is responsible for the
inter-communication of groups of UAVs. Each group of UAVs uses its gateway UAV
to communicate with the adjacent groups. Moreover, this communication is done
directly, without using the GCS network infrastructure.
3. Layer 3 (G2A) is suited to ground-to-air communication. It allows us to relay bidi-
rectional information between the GCS and the UAV groups. Generally, this is done
through the UAV gateway closest to the GCS infrastructure.
Compared to the three other architectures, the multi-level architecture is robust and
supports fault tolerance. It is used for complex missions that involve different types
of UAVs. In fact, the multi-layer structure of this architecture allows for the massive
deployment of UAVs to cover a large airspace. Unlike the multi-group architecture, UAVs
with different types of protocols can be deployed in the same group [19,28].

Figure 13. Multi-UAV communication architecture based on multi-level.

7. Comparison and Roadmap of Multi-UAV Architectures

The rapid emergence of UAVs and the diversity of their applications has led to a
variety of communication architectures. Each one has its advantages and disadvantages.
Table 3 summarizes the key performance of the four communication architectures seen
in this paper. Multi-level and multi-group architectures are the major players. These are
interoperable systems that support various types of UAVs. Their communication range
extends over long distances, which allows them to ensure permanent connectivity in the
network. They are also suitable for complex missions involving several types of UAVs that
collaborate in real time. The robustness of these two architectures guarantees the scalability,
fault tolerance, and stability of the system. They are currently implemented in the military
domain for coordinated attack and defense missions [2,3,66].
The single-group architecture is suitable for small and medium UAVs. It is not in-
teroperable and therefore does not allow the deployment of various types of UAVs. Its
communication range is relatively dependent on the distance between UAVs. The con-
nectivity becomes unstable when the UAVs move away from each other. It is appropriate
for simple collaborative missions in limited environments. Bidirectional communication
Drones 2023, 7, 263 22 of 37

(air-to-ground) between the GCS and the group of UAVs is achieved through the gateway
UAV. When the gateway fails, the system shuts down. Most of the projects [13–15,57]
related to the carrying of payloads implement this architecture since it offers peer-to-peer
communication.
Finally, the centralized architecture is less efficient than the other three. It also does
not support interoperability and its communication scheme is centralized on the GCS. This
leads to latency in air-to-ground communications. Intra-UAV communication is also not
supported in this type of architecture. Therefore, it is not suitable for collaborative mission
planning. The only point of failure is the GCS: if it fails, then the whole system is paralyzed.
Generally speaking, this architecture is used for very basic missions with relatively small
UAVs [26,27,48,57].

Table 3. Key features of multi-UAV communication architectures discussed in this paper.

Features Centralized Single-Group Multi-Group Multi-Level

Very small UAV    
Small UAV    
Medium UAV é   
Large UAV é é  
Gateway UAV é é  
Single-hop wireless    
Multi-hop wireless é   
Interoperability é é  
Stability é é  
Fault tolerance é é  
Cooperative mission é   
Streaming é   
Large aerial coverage é é  
Air-to-air é   
Air-to-ground    
Group-to-group é é  

The choice of an architecture is a delicate task during the conceptual phase of a

multi-UAV system because it depends on the type and size of the UAVs, the area of the
environment to explore, and the complexity of the mission to plan. If the choice is made out
of context, a robust architecture could be inefficient on some types of UAVs and vice versa.
Therefore, it is very important to define the type of mission and then fix the evaluation
parameters seen in Table 3 before selecting an appropriate architecture.
Figure 14 shows a roadmap for multi-criteria decision support. It provides information
on how to choose an architecture that is well adapted to the context. For simple missions
including very small and small UAVs, it is preferable to opt for classical architectures. Ide-
ally, if all operations are centralized on the GCS, then it is advisable to deploy a centralized
architecture. If the mission involves sharing resources or exchanging information between
UAVs, it is better to opt for a single-group architecture. It is important to ensure that each
UAV remains within range of adjacent UAVs to maintain connectivity in the network. For
complex missions, the ideal would be to use decentralized architectures. In this sense,
multi-group and multi-layer architectures remain the best choices. The robustness of the
multi-level architecture makes it an appropriate choice for mission planning involving
several different tasks [57].
Drones 2023, 7, 263 23 of 37

Figure 14. Roadmap of multi-UAV communication architectures.

8. Routing Protocols of Multi-UAV Systems

The increasing number and variety of mobile connected objects has led to numerous
studies [22,76] to develop routing protocols adapted to highly mobile environments. Each
protocol is adapted according to a particular routing standard, which is largely influenced
by the communication infrastructures. A protocol fits into the acceptance standards if it
allows us to convey, ensure the integrity of, and secure packets (messages) sent between
the nodes of its network. In the case of UAV networks, the communication protocols used
are adapted for Flying Ad-Hoc Network (FANET) routing protocols [19]. In this paper,
we investigate 17 FANET routing protocols. We distinguish three categories of routing
protocols: topology [23,75], geographic position [20,77,78], and hierarchical-based [79,80]
routing protocols; refer to Figure 15.

Figure 15. Classification of FANET routing protocols discussed in this paper.

8.1. Topology-Based Routing Protocols

Topology-based protocols [23,75] are classic routing protocols that employ a routing
table that contains all the data transfer rules on a network; each of them is identified by its IP
address (IPv4 or IPv6). For two UAVs in the network, the routing algorithm computes the
appropriate route to send and receive packets between them based on the rules defined in
the routing table. As shown in Figures 15 and 16, topology-based protocols are divided into
four sub-categories: static [73,81,82], proactive [83,84], reactive [23,85–87], and hybrid [88]
routing protocols.
Drones 2023, 7, 263 24 of 37

Figure 16. Paradigm of topology-based protocols: (a) static; (b) proactive; (c) reactive; (d) hybrid.

8.1.1. Static
Static-based routing protocols [73,81,82] are a subcategory of topology-based protocols
that use a preconfigured static routing table for data transfer between UAVs referenced by
their static IP addresses. To add a new UAV, the administrator must manually reconfigure
the rules in the routing table to fit the scalability of the multi-UAV system. Therefore, this
type of protocol is not suitable for a high-mobility network because of the lack of scalability
and fault tolerance. The three variants of this sub-category of protocols are Load Carry
And Deliver (LCAD) [73], Data-Centric Routing (DCR) [81], and Multi-Level Hierarchical
Routing (MLHR) [82].
The LCAD protocol [73] is based on the Store-Carry-Forward mechanism [74]. It is
adapted for a centralized architecture because it allows the transport of messages between
an infrastructure (e.g., ground control station) and all UAVs. The communication mecha-
nism of this protocol consists of three stages: (1) the load stage, in which the packets are
stored in the UAVs; (2) the carry stage, in which the UAVs transmit these packets via the
gateway UAVs; (3) finally, in the Deliver stage, the gateways deliver the packets to the
final UAVs.
The DCR protocol [81] is based on a broadcasting mechanism. It is suitable for a single-
group architecture. The transmission mechanism is based on the header attributes contained
in the data. The data are sent only to the UAVs that want them, thus optimizing the network
throughput. Then, the MLHR protocol [82], which is based on a node clustering paradigm,
allows us to address the scalability challenge. The UAVs are hierarchically divided into
clusters according to their types and roles. Each cluster contains a gateway UAV, which is
responsible for sharing data with other clusters. Therefore, this protocol is well suited for a
multi-group architecture.

8.1.2. Proactive
Proactive protocols [83,84] use dynamic routing tables. Each UAV has a routing table
that stores, periodically for each UAV, the routes to broadcast messages to the destination
UAVs. The delay between sending and receiving the message is very short because the
paths are already known beforehand. However, when the swarm reaches a certain level of
scalability, a network bottleneck ensues, as changes in the routing table are bandwidth- and
Drones 2023, 7, 263 25 of 37

energy-intensive. For this reason, this protocol is not used for applications requiring a large
number of highly mobile UAVs. The most well-known proactive protocols are Optimized
Link State Routing (OLSR) [83], Destination Sequence Distance Vector (DSDV) [84], and
Better Approach to Mobile Ad-Hoc Network (BATMAN) [89].
The OLSR protocol uses Dijkstra’s algorithm [62] to calculate the routes for sending
messages between the different UAVs, while the DSDV protocol uses the Bellman–Ford
algorithm [61]. On the other hand, the BATMAN protocol optimizes the calculation of
routes between UAVs. Unlike the OLSR and DSDV protocols, which calculate routes for all
possible sending combinations, the BATMAN protocol only calculates routes for UAVs that
are directly or indirectly (via the gateway UAVs) connected to each other.

8.1.3. Reactive
Reactive protocols [23,85–87] are optimized versions of proactive protocols that aim
to minimize the bandwidth consumption. This sub-category of protocols is based on on-
demand routing protocols, because routes are created only when necessary. If there are no
data to share between two UAVs, then there is no need to calculate the route. This approach
results in the efficient use of the bandwidth and reduced overhead. The most commonly
used reactive protocols are Dynamic Source Routing (DSR) [85], Ad Hod On-demand
Vector (AODV) [86], and Time-Slotted On-Demand (TSOD) [87].
DSR [85] is a dynamic source-routing protocol whose routing is initiated by the source
UAV that wants to send a message. The complete route to transmit a message is indicated
in the routing table of the source UAV. However, it is not possible to send messages if the
route to the destination UAV is not registered in the header of the source UAV. The AODV
protocol [86] leverages this limitation; it is the most used routing protocol in both VANET
and FANET. Before sending a message, the source UAV checks if the routing path already
exists. If so, it uses the existing route. Otherwise, it triggers a process to compute the route
to the destination UAV.
On the other hand, a risk of collision between packets can occur when there are enough
exchanges in the network. TSOD [86] allows us to avoid collisions; it is an extended version
of AODV that introduces a time slot routing protocol. Each source UAV has a time slot
allocated to it for sending messages, and they can only send messages during this slot. This
protocol reduces significantly the lost packets and minimizes network bottlenecks.

8.1.4. Hybrid
This sub-category of protocols [88,90] is both proactive and reactive and is suitable for
large mobile networks. It takes advantage of both to overcome the problems of bandwidth
saturation, long delays, and collisions. Its operating principle consists in dividing the
network of UAVs into sub-networks, commonly called zones. Intra-zone communications
are based on the proactive protocols, while inter-zone communications are handled by
the reactive protocols. The most popular hybrid protocols are the Zone Routing Protocol
(ZRP) [90] and Temporarily Ordered Routing Algorithm (TORA) [88].
ZRP [90] divides the network into several zones of relatively equal radius and uses a
dynamic routing table for the exchanges within the zones. Messages exchanged between
UAVs from different zones are based on on-demand routing protocols. UAVs should always
stay within the radius of their zones in order to maintain network stability. To address
this limitation, researchers [88] have proposed TORA. TORA is a distributed version of
a hybrid protocol that, instead of dividing the network, distributes the routing table in
such a way that each UAV contains only information about its neighbors, rather than the
entire network.

8.2. Geographic Position-Based Routing Protocols

Sometimes called georouting, these types of protocols rely on the geographical loca-
tion services of the UAVs [20]. They are more adapted for highly mobile UAVs in large
FANETs [19]. In this type of protocol, each UAV determines its GPS position and shares it
Drones 2023, 7, 263 26 of 37

with its neighbors. A sender UAV must know the position of the receiver UAV, as shown
in Figure 17. Based on this knowledge, messages can be routed without knowing the
network topology beforehand. The paradigm of these protocols is similar to reactive proto-
cols since the routes are used only when necessary. These protocols combine geoservice
techniques [20,77] with the Greedy Forwarding approach [78] and they consist of Greedy
Perimeter Stateless Routing (GPSR) [91], Geographic Load Share Routing (GLSR) [92], Geo-
graphic Position Mobility-Oriented Routing (GPMOR) [93], and Reactive Greedy Reactive
(RGR) [78].

Figure 17. Geographic position routing paradigm.

GPSR was originally proposed for MANET and allows us to calculate the shortest
route for sending messages. From the location information of each UAV, it uses the greedy
method [94] to find the nearest neighbors of the UAV that receives the message. Then, it
calculates the shortest route to send the message through one of the closest neighbors of
the receiver UAV. If no route is found, then the sending fails, without any possibility to
recover the message.
Cardellini et al. [92] proposed GLSR, an extended version of GPSR that uses a predic-
tion function to select the next promising gateway UAV to compute multiple routes from
the sender to the receiver UAV. In another work [93], the authors proposed GPMOR, an
adapted version of GLSR for a highly mobile flying ad-hoc network. The proposed protocol
uses Gauss–Markov stochastic processes to model and predict the movement of the UAVs.
This approach makes it possible to anticipate the most optimal routes for sending messages.
To increase the deliverability of the messages and minimize the bandwidth consump-
tion, Shirani et al. [78] introduced the RGR protocol. This protocol is based on both the
AODV routing protocol and greedy forwarding method. Since it is a reactive protocol, it
calculates “on-demand” only the routes for the UAVs waiting for the messages.

8.3. Hierarchical-Based Routing Protocols

In these types of protocols [79,80], the network is divided hierarchically into clusters.
Each cluster is used for a specific function and has a head that facilitates inter-cluster
communication through the other heads, see Figure 18. The communication process is
supported by proactive routing protocols, while changes in the network or failures are
managed by reactive protocol processes. The most popular of these routing protocols are
Mobility Prediction Clustering (MPC) [79] and the Clustering Algorithm for Networking
(CAN) [80].
Drones 2023, 7, 263 27 of 37

Figure 18. Hierarchical routing paradigm.

CNU is a clustering-based protocol that divides the network into clusters starting with
the hierarchical level closest to the GCS. Thus, in an iterative way, new clusters are formed
from the previous level’s clusters until they cover the entire network. For each cluster,
UAVs elect a cluster head based on their power law distributions in the graph network. The
cluster head will be responsible for calculating the route to effectively broadcast messages
inside the cluster. Thereafter, it shares them with other clusters through the cluster heads.
MPC is an adapted version of CAN for highly mobile FANETs. The MPC protocol
uses mainly the Doppler shift method to compute the velocity and movements of a couple
of UAVs. The clustering is done according to the state of each UAV: orphan (stateless),
member (cluster member), or head (cluster head). Initially, all UAVs are in the orphan
state. Afterwards, each UAV sends a “HELLO” message containing, in its header, the
information relative to its velocity, direction, and identifiers. Depending on the change in
wave frequency (obtained from Doppler shift) between the transmitter UAV that moves
compared to the receiver UAV, a state (member or head) is assigned to each UAV until the
clusters are completely formed.

9. Comparison and Roadmap of Routing Protocols

As we have seen in the previous sections, UAVs are varied from very small to large.
They are either adapted to be deployed with a traditional or a decentralized communi-
cation architecture. To this, we must add the interoperability between UAVs, which is
mainly based on routing protocols. Thus, each protocol has its benefits and drawbacks.
Table 4 summarizes the key performance of the FANET routing protocols reviewed in this
paper. We note that topology-based routing protocols are only suitable for centralized and
single-group architectures. They do not provide routing for highly mobile networks. It
should also be noted that only hybrid and BATMAN protocols are suitable for sending
messages without a risk of collision or packet loss. OLSR and DSDV protocols are much
more expensive in terms of bandwidth, even if they are fast and avoid redundancy dur-
ing message transmission. As for geographic position-based protocols, they operate on
centralized, single-group, and multi-group architectures. These types of protocols pro-
mote interoperability and ensure good message deliverability while reducing bandwidth
throughput. On the other hand, they do not guarantee fault tolerance in case of the failure
or inaccessibility (wrong GPS coordinates) of a UAV. Only the GPSR and GLSR protocols
are suitable for communication in a large-scale network. Finally, only the hierarchical-based
routing protocols are operational on the four types of architectures. They are suitable for
collaborative missions that involve frequent data exchange over several hierarchical levels
of the network. Although they are efficient, some limitations should be noted. (1) The MPC
protocol requires more time to send a message, which largely impacts the actions of other
UAVs if they depend on it. (2) In case of message loss, the CAN protocol does not ensure
its recovery, which can lead to the paralysis of the mission and cause a network bottleneck.
Drones 2023, 7, 263 28 of 37

Table 4. Key features of FANET routing protocols discussed in this paper.

Architectures Performance Metrics

Interoperability
High-Mobility
Deliverability
Single-Group

Multi-Group

Redundancy

Link Failure
Throughput
Multi-Level
Centralized

Scalability
Collision

Delay
Protocols
LCAD  é é é  é é é   é é é
DCR   é é  é é   é é é é
MLHR    é  é    é é é 
OLSR   é é é é é  é  é  é
DSDV   é é é é é  é  é é é
Topology BATMAN   é é       é  é
DSR     é é  é   é é 
AODV   é é  é é é   é  é
TSOD   é é   é é   é  é
ZRP    é       é  
TORA    é         
GPSR    é    é    é 
GLSR    é  é    é  é 
Geographic position
GPMOR    é  é é   é  é 
RGR    é  é é   é é é 
CAN            é 
Hierarchical
MPC        é     

The choice of routing protocol depends on the type of communication architecture,

the mission requirements, and the relevant features seen in Table 3. It is very important
to analyze this step thoroughly in order to avoid problems of incompatibility between
the protocols and the communication architecture. Figure 19 shows how to choose the
right protocol depending on the infrastructure and the communication architecture. For a
centralized architecture, it is better to opt for topology-based protocols. If the deliverability
of messages is an important factor, then it is better to use static protocols. In case interactions
with as few collisions as possible are required, it is more advantageous to use hybrid
protocols. For decentralized architectures involving highly mobile UAVs in a large-scale
environment, it is better to use geo-based protocols. In contrast, if it is a collaborative
mission, hierarchical-based routing protocols are ideal.

Figure 19. Roadmap of FANET routing protocols.

Drones 2023, 7, 263 29 of 37

10. Open Challenges and Future Directions

The deployment of a multi-UAV system involves several challenges to overcome,
ranging from the choice of infrastructure to the planning of the mission. Despite efforts
in the field, most research work is still in the prototyping or experimental phase for some.
Several challenges remain to be met to effectively exploit swarms in upcoming trends.
These topics include the collaborative decision making, the communication infrastructure,
the data streaming, and the regulation of the use of combined UAVs.

10.1. Collaborative Decision Making

The perception of the environment, the states, the allocation and coordination of tasks,
as well as the consensus are important factors that influence the behavior of UAVs in a
swarm. Each autonomous UAV contributes to the mission progress by performing actions
in a collaborative manner. In this context, the level of human control is reduced to a few
start-up operations or maintenance operations.
Several challenges remain to be addressed, including consensus building, task alloca-
tion, collective failure management, and perception.
• Definition of a consensus on which the UAVs base themselves to carry out individual
actions that contribute to a convergence towards a common goal.
• Task allocation, which consists in dynamically assigning tasks according to the state
and capacity of the UAVs. To allow a high level of parallelism, tasks can be distributed
fairly if the UAVs are homogeneous. Otherwise, it will be done according to the
workload and computing resources of each UAV. It should also be noted that the
assignment of tasks is based on a parallelism paradigm (synchronous parallelism,
asynchronous parallelism, or bulk synchronous parallelism) that ensures that the tasks
are executed concurrently.
• Collective failure detection in the swarm makes it possible to detect the UAVs that
have abnormal behavior, or that carry out actions that do not contribute to the real-
ization of the mission. For example, due to a hardware malfunction (e.g., low battery
or defective rotor), a UAV may not respond appropriately to some tasks assigned to
it. The collective management of the failures in the swarm must make it possible to
detect the UAVs whose behavior does not contribute to the progress of the mission. It
is also necessary to consider an activity continuity plan that allows the other UAVs to
continue the mission and to adapt to any type of change.
• Perception of the environment represents the information collected by the UAV. Each
UAV uses onboard sensors to build a local representation of the environment. The
combined set of perceptions allows the swarm to build a global representation and
make more predictive decisions. For example, to calculate the most optimal path that
reaches a target, we can choose the closest drone. See Section 4.2.2 for more details.
These challenges open future directions as follows:
• The deployment of a swarm using a multi-agent-based model for better coordi-
nating the actions and interactions of the drones to enhance the individual and/or
collective decision making.
• The implementation of a reinforcement learning algorithm on each drone to control
its behavior based on the reward of good actions.

10.2. Communication Infrastructure

The existence of several communication architectures, the variety of routing protocols,
and the variety of UAV types raise the question of the choice of an adapted communication
infrastructure. This issue higlights the challenges of interoperability in a distributed
manner [95]. Generally, in a decentralized architecture, UAVs are required to exchange
data or services based on nested protocols. Unfortunately, it is difficult to do so because
the protocols vary from one manufacturer to another. Thus, two UAVs having different
protocols will not be able to communicate. It becomes even more constraining when it is
Drones 2023, 7, 263 30 of 37

a collaborative mission in which UAVs operate synchronously or asynchronously. This

challenge opens the following research tracks:
• The development of a protocol converter that allows us to convert a given FANET
protocol into a desired protocol in order to ensure interoperability between non-
homogeneous UAVs.
• The standardization of UAV communication rules in order to evolve towards a
reference model used by manufacturers as well as end users.

10.3. Collective Navigation

Collective navigation is a commonly used practice for swarm mobility. It consists of
flying UAVs in a tight formation to maintain network connectivity. Difficulties arise with a
large and highly mobile swarm, where the positions and directions of the UAVs change
relative to each other. The most frequent issues to address are collision and crash avoidance,
obstacle detection, and the connectivity of the swarm [24]. A better way to control and
harmonize the movements of the swarm is to use the “divide-and-conquer” technique [56].
It consists of dividing the swarm into several sub-swarms. Each sub-swarm is assigned a
static formation with a well-defined polygon shape (regular, irregular, concave, convex, or
complex). The UAVs inside the sub-swarms move and maintain a distance from each other
without leaving the boundary of their formation, while the sub-swarms move in relation to
their delimitation while remaining two by two disjoint, as shown in Figure 20.

Figure 20. Example of swarm navigation based on divide-and-conquer.

10.4. Energy Consumption

To face the challenges of the maneuverability and portability of the UAVs, hardware
components are reduced to a miniaturized size. As a result, they cannot carry many
batteries or large-capacity batteries [41]. This constraint limits the energy consumption
and contributes to the degradation of the communication and mobility performance of
Drones 2023, 7, 263 31 of 37

UAVs [36]. The energy consumption related to the communication system is used to
ensure the signal processing, the packet transmission, the routing protocol, and the data
transmission of the onboard sensors. On the other hand, the power consumption of the
propulsion system is much higher than the energy consumption for communication. It
generates thrust and lift forces to ensure the mobility and movement of the UAVs in the
air [41]. This depends on the type of UAV and number of rotors (see Section 3), payload
weight, flying altitude, speed, and acceleration. In addition, there are environmental
factors such as gravity, air density, wind conditions, and weather that affect the drone’s
energy consumption.
In the case of a multi-UAV system, the problem becomes even more challenging de-
pending on the communication architecture, the mobility of the swarm, and the complexity
of the mission. Several challenges remain to be addressed, including the optimization of
the path planning [12,49,57], task optimization, and fast battery charging [40,42].
• The optimization of the coverage paths in order to reduce considerably the power
consumption of the UAVs. This problem consists in finding the optimal paths that max-
imize the number of covered waypoints in a network while minimizing both energy
costs and emissions [12,49]. A concrete application of this problem is the drone deliv-
ery operation, which consists in serving all the delivery waypoints without passing
twice by the same point [38]. It is a multi-criteria optimization problem [58] that con-
sists in determining the best energy consumption model that optimizes both the path
cost, the energy consumed, and the emissions associated with the delivery strategy.
• The efficient planning of the collaborative tasks of the swarm. In a multi-UAV sys-
tem, the tasks are running concurrently and in a parallel way to achieve a common goal.
The UAVs share a large amount of information synchronously or asynchronously in the
network for waypoint generation, map reconstruction, and collision avoidance [24,40].
These tasks are memory-intensive and also consume a significant amount of energy
resources related to the communication system [36].
• The maintenance of the battery is another problem that affects the flight times of
UAVs as the batteries used have a very limited lifespan [35,41]. It allows us to improve
the battery life cycle and to extend its duration of use. Furthermore, this is one of the
effective alternatives to have a battery considered safer and with a longer life cycle [35].
However, this task is complex because there is no exact method to determine the life
of a battery and each type of UAV has its own maintenance pattern [42]. There are
numerous factors affecting the life of the battery that must be considered, such as the
amperage, voltage, temperature, type of charger, and charge/discharge cycles [37].
These challenges open future directions as follows:
• The design of energy-efficient trajectory-based reinforcement learning, which al-
lows us to generate an adaptive path. Each UAV acts as an autonomous agent that
readjusts its trajectory, speed, acceleration, and altitude to maximize energy gains.
Moreover, it reinforces its adaptive capacity by using a multi-objective function that
rewards the actions that minimize the cost of its path, its energy consumption, and its
emissions. Some recent examples were presented in [36,40].
• The optimization of the UAV flight tour inspired by combinatorial optimization
and the integer programming problem, in which, starting from the launch position
of the UAV, we try to find an optimal set of paths that the swarm must traverse in order
to return to its initial position. This is a problem of NP-Hard complexity [12,49,59]
because we cannot pass twice through the same waypoint. Moreover, this problem
becomes even more complex when we add the constraints of energy consumption
and emissions. Typical examples were presented by Dorling et al. [39] and Febria
et al. [12] for drone delivery services.
• A novel energy harvester for powering UAVs that will allow the possible recharge
of an ultra-fast rechargeable battery onboard a UAV and an improvement in the
lifespan. A research work conducted in this direction has been proposed in [35] and the
analytical results show possible flight time improvements in the range of 3 to 16 min,
Drones 2023, 7, 263 32 of 37

depending on the flight conditions. In terms of energy usage, it appears that decreasing
the UAV size and payload weight may be more important than communication
methods for enhanced energy efficiency.

10.5. Regulation
The legislation on the use of UAVs is covered by rules and procedures. Before launch-
ing a UAV, some important points about the type of UAV, the operating license, the area
of exploration, and the mission must be checked. For details of the commission imple-
menting European Union regulation 2019/947 of 24 May 2019 on the rules and procedures
governing the operation of uncrewed aerial vehicles, see Table 5. UAV operating licenses
are categorized into open, special, and requiring approval. Depending on the mission, an
assessment of the operational requirements, risks, and preservation of confidentiality of
information must be approved. In the case of swarms, the first requirement is to comply
with the maximum number of UAVs allowed to fly simultaneously. Since a swarm can be
composed of various types of UAVs, it is important to ensure compliance with the opera-
tional requirements and risks (collisions and signal interference) associated with handling
these types of aerial vehicles. These non-technical challenges may provide an opportunity
for research into legislation on the use of homogeneous and non-homogeneous swarms.

Table 5. Most important regulations on the use of UAVs for setting up a swarm.

Category Article Regulation

Article 3 Uncrewed aircraft systems operational categories
UAV Article 4 The UAS operational category “open”
categories Article 5 The UAS operational category “special”
Article 6 The UAS operating category “requiring approval”
Article 7 Rules and procedures for the operation of UAS
Operational
Article 8 Rules and procedures related to the competency of long-distance pilots
requirements
Article 10 Rules and procedures related to the airworthiness of UAS
Article 11 Rules for operational risk assessment
Risk classes Article 12 Operating license for the “special” category
Article 14 Registration of the UAS operator and UAS subject to licensing
Long distance Article 13 Cross-border operation or operation outside of the state of registration
flight Article 15 Operating conditions for UAS geographic areas
Privacy policy Article 19 Security information
European Union regulation 2019/947 of 24 May 2019 (https://eur-lex.europa.eu/legal-content/EN/AUTO/?uri=
CELEX:02019R0947-20220404, accessed on 15 March 2023).

11. Conclusions
Multi-UAV systems are of great interest to the scientific community because they
represent a wide range of opportunities. Their usefulness in the industrial sector is no
longer in doubt. They are an effective means for carrying out complex missions that require
several types of UAVs. The question of the implementation of a multi-UAV system is a
subject that involves two major challenges: technical–functional and legal/ethical. At the
technical–functional level, the multi-UAV system must ensure interoperability between
these components and maintain a level of autonomy that allows it to make collective
decisions without human intervention. Using multi-UAVs for highly mobile network and
wide-area missions is a challenging task because the UAVs must not move too far away
from each other to avoid signal loss, which is not often the case. In addition, there are
constraints related to the system infrastructure, the communication architecture, and the
routing protocol. From a legal/ethical point of view, it is very important to consider people’s
privacy and safety, as well as to implement rigorously the regulations that guarantee the
responsible use of multi-UAV systems.
Drones 2023, 7, 263 33 of 37

Many scientific contributions have already explored countermeasures to these con-

cerns; nevertheless, certain areas, such as mathematical modeling, the comparative study
of communication architectures as well as routing protocols, and decision-making roadmap
concerns, have yet to be addressed. With the growing interest in self-organizing swarms,
these four challenges cannot be ignored. For this reason, this review paper offers a thor-
ough investigation of the challenges associated with deploying an autonomous multi-UAV
system that integrates collaborative actions, as well as the complexity of the collaborative
target detection problem in distributed environments. It is an NP-Hard problem [58],
which results in challenges related to the balancing of workloads in the swarm and the
minimization of the path costs to reach targets. The current challenges, as well as a number
of suggestions, are presented in this study, which emphasizes that the performance of
multi-UAV systems can be improved if collaborative decision making, collective navigation,
energy consumption, and regulation problems that remain to be addressed in this area are
dealt with seriously.

Author Contributions: Conceptualization, W.Y.H.A. and S.L.; methodology, W.Y.H.A.; software,

J.S.F.; validation, S.L., R.G. and M.B.; formal analysis, W.Y.H.A.; investigation, W.Y.H.A.; resources,
W.Y.H.A. and J.S.F.; writing—original draft preparation, W.Y.H.A.; writing—review and editing, S.L.;
visualization, R.G. and M.B.; supervision, S.L.; project administration, R.G. and M.B.; funding acqui-
sition, R.G. and M.B. All authors have read and agreed to the published version of the manuscript.
Funding: This research was funded by CASUS Open X Projects.
Data Availability Statement: Not applicable.
Conflicts of Interest: The authors declare no conflicts of interest.

Abbreviations
The following abbreviations are used in this manuscript:

UAV Uncrewed Aerial Vehicle

UAS Uncrewed Aircraft System
RPA Remotely Piloted Aircraft
RPAS Remotely Piloted Aircraft System
M-UAV Multi Uncrewed Aerial Vehicle
VTOL Vertical Takeoff and Landing
IoT Internet of Things
AI Artificial Intelligence
MANET Mobile Ad-Hoc Network
VANET Vehicular Ad-Hoc Network
FANET Flying Ad-Hoc Network
RC Remote Controller
GCS Ground Control Station
A2A Air to Air
G2G Group to Group
A2G Air to Ground
OLSC Oriented Line Segment Coverage
LCAD Load Carry And Deliver
DCR Data-Centric Routing
MLHR Multi-Level Hierarchical Routing
OLSR Optimized Link State Routing
DSDV Destination Sequence Distance Vector
BATMAN Better Approach to Mobile Ad-Hoc Network
DSR Dynamic Source Routing
AODV Ad-Hod On-Demand Vector
TSOD Time-Slotted On-Demand
Drones 2023, 7, 263 34 of 37

ZRP Zone Routing Protocol

TORA Temporarily Ordered Routing Algorithm
GPSR Greedy Perimeter Stateless Routing
GLSR Geographic Load Share Routing
GPMOR Geographic Position Mobility-Oriented Routing
RGR Reactive Greedy Reactive
CAN Clustering Algorithm for Networking
MPC Mobility Prediction Clustering
IMU Inertial Measurement Unit

References
1. Chen, X.; Tang, J.; Lao, S. Review of Unmanned Aerial Vehicle Swarm Communication Architectures and Routing Protocols.
Appl. Sci. 2020, 10, 3661. [CrossRef]
2. Lachow, I. The upside and downside of swarming drones. Bull. At. Sci. 2017, 73, 96–101. [CrossRef]
3. Scharre, P. Robotics on the Battlefield Part II: The Coming Swarm; Technical Report; Center for a New American Security: Washington,
DC, USA 2014.
4. Shakhatreh, H.; Sawalmeh, A.H.; Al-Fuqaha, A.; Dou, Z.; Almaita, E.; Khalil, I.; Othman, N.S.; Khreishah, A.; Guizani, M.
Unmanned Aerial Vehicles (UAVs): A Survey on Civil Applications and Key Research Challenges. IEEE Access 2019, 7,
48572–48634. [CrossRef]
5. Thiele, S.T.; Varley, N.; James, M.R. Thermal photogrammetric imaging: A new technique for monitoring dome eruptions. J.
Volcanol. Geotherm. Res. 2017, 337, 140–145. [CrossRef]
6. Chriki, A.; Touati, H.; Snoussi, H.; Kamoun, F. UAV-GCS Centralized Data-Oriented Communication Architecture for Crowd
Surveillance Applications. In Proceedings of the 2019 15th International Wireless Communications and Mobile Computing
Conference (IWCMC), Tangier, Morocco, 24–28 June 2019 ; IEEE: Tangier, Morocco, 2019; pp. 2064–2069. [CrossRef]
7. Scherer, J.; Yahyanejad, S.; Hayat, S.; Yanmaz, E.; Andre, T.; Khan, A.; Vukadinovic, V.; Bettstetter, C.; Hellwagner, H.; Rinner,
B. An Autonomous Multi-UAV System for Search and Rescue. In Proceedings of the First Workshop on Micro Aerial Vehicle
Networks, Systems, and Applications for Civilian Use, DroNet ’15, Florence, Italy, 18 May 2015; Association for Computing
Machinery: New York, NY, USA, 2015; pp. 33–38. [CrossRef]
8. Adamopoulos, E.; Rinaudo, F. UAS-Based Archaeological Remote Sensing: Review, Meta-Analysis and State-of-the-Art. Drones
2020, 4, 46. [CrossRef]
9. Patel, M.; Bandopadhyay, A.; Ahmad, A. Collaborative Mapping of Archaeological Sites Using Multiple UAVs. In Proceedings of
the Intelligent Autonomous Systems 16; Ang, M.H., Jr., Asama, H., Lin, W., Foong, S., Eds.; Lecture Notes in Networks and Systems;
Springer International Publishing: Cham, Switzerland, 2022; pp. 54–70. [CrossRef]
10. Rejeb, A.; Abdollahi, A.; Rejeb, K.; Treiblmaier, H. Drones in agriculture: A review and bibliometric analysis. Comput. Electron.
Agric. 2022, 198, 107017. [CrossRef]
11. Goodrich, P.; Betancourt, O.; Arias, A.C.; Zohdi, T. Placement and drone flight path mapping of agricultural soil sensors using
machine learning. Comput. Electron. Agric. 2023, 205, 107591. [CrossRef]
12. Febria, J.; Dewi, C.; Mailoa, E. Comparison of Capacitated Vehicle Routing Problem Using Initial Route and Without Initial
Route for Pharmaceuticals Distribution. In Proceedings of the 2021 2nd International Conference on Innovative and Creative
Information Technology (ICITech), Salatiga, Indonesia, 23–25 September 2021; pp. 94–98. [CrossRef]
13. Mellinger, D.; Shomin, M.; Michael, N.; Kumar, V. Cooperative Grasping and Transport Using Multiple Quadrotors. In Distributed
Autonomous Robotic Systems: The 10th International Symposium; Martinoli, A., Mondada, F., Correll, N., Mermoud, G., Egerstedt, M.,
Hsieh, M.A., Parker, L.E., Støy, K., Eds.; Springer Tracts in Advanced Robotics; Springer: Berlin/Heidelberg, Germany, 2013;
pp. 545–558. [CrossRef]
14. Ritz, R.; D’Andrea, R. Carrying a flexible payload with multiple flying vehicles. In Proceedings of the 2013 IEEE/RSJ International
Conference on Intelligent Robots and Systems, Tokyo, Japan, 3–7 November 2013; IEEE: Tokyo, Japan, 2013; pp. 3465–3471.
[CrossRef]
15. Loianno, G.; Kumar, V. Cooperative Transportation Using Small Quadrotors Using Monocular Vision and Inertial Sensing. IEEE
Robot. Autom. Lett. 2018, 3, 680–687. [CrossRef]
16. Menouar, H.; Guvenc, I.; Akkaya, K.; Uluagac, A.S.; Kadri, A.; Tuncer, A. UAV-Enabled Intelligent Transportation Systems for the
Smart City: Applications and Challenges. IEEE Commun. Mag. 2017, 55, 22–28. [CrossRef]
17. Stan, A.C. A decentralised control method for unknown environment exploration using Turtlebot 3 multi-robot system. In
Proceedings of the 2022 14th International Conference on Electronics, Computers and Artificial Intelligence (ECAI), Ploiesti,
Romania, 30 June–1 July 2022; pp. 1–6. [CrossRef]
18. Choudhary, G.; Sharma, V.; Gupta, T.; Kim, J.; You, I. Internet of Drones (IoD): Threats, Vulnerability, and Security Perspectives.
Res. Briefs Inf. Commun. Technol. Evol. 2018, 4, 64–77.
19. Al-Emadi, S.; Al-Mohannadi, A. Towards Enhancement of Network Communication Architectures and Routing Protocols for
FANETs: A Survey. In Proceedings of the 2020 3rd International Conference on Advanced Communication Technologies and
Networking (CommNet), Marrakech, Morocco, 4–6 September 2020; IEEE: Marrakech, Morocco, 2020; pp. 1–10. [CrossRef]
Drones 2023, 7, 263 35 of 37

20. Agrawal, J.; Kapoor, M. A comparative study on geographic-based routing algorithms for flying ad-hoc networks. Concurr.
Comput. Pract. Exp. 2021, 33, e6253. [CrossRef]
21. Rizwan Ghori, M.; Safa Sadiq, A.; Ghani, A. VANET Routing Protocols: Review, Implementation and Analysis. J. Phys. Conf. Ser.
2018, 1049, 012064. [CrossRef]
22. Hogie, L.; Bouvry, P.; Guinand, F. An Overview of MANETs Simulation. Electron. Notes Theor. Comput. Sci. 2006, 150, 81–101.
[CrossRef]
23. Khan, M.A.; Khan, I.U.; Safi, A.; Quershi, I.M. Dynamic Routing in Flying Ad-Hoc Networks Using Topology-Based Routing
Protocols. Drones 2018, 2, 27. [CrossRef]
24. Digulescu, A.; Despina-Stoian, C.; Stănescu, D.; Popescu, F.; Enache, F.; Ioana, C.; Rădoi, E.; Rîncu, I.; S, erbănescu, A. New
Approach of UAV Movement Detection and Characterization Using Advanced Signal Processing Methods Based on UWB Sensing.
Sensors 2020, 20, 5904. [CrossRef] [PubMed]
25. Jabbar, R.; Dhib, E.; ben Said, A.; Krichen, M.; Fetais, N.; Zaidan, E.; Barkaoui, K. Blockchain technology for intelligent
transportation systems: A systematic literature review. IEEE Access 2022, 10, 20995–21031. [CrossRef]
26. Ahmadzadeh, A.; Jadbabaie, A.; Kumar, V.; Pappas, G.J. Multi-UAV Cooperative Surveillance with Spatio-Temporal Specifications.
In Proceedings of the 45th IEEE Conference on Decision and Control, San Diego, CA, USA, 13–15 December 2006; IEEE: San
Diego, CA, USA, 2006; pp. 5293–5298. [CrossRef]
27. Abdelkader, M.; Güler, S.; Jaleel, H.; Shamma, J.S. Aerial Swarms: Recent Applications and Challenges. Curr. Robot. Rep. 2021,
2, 309–320. [CrossRef]
28. Khalil, H.; Rahman, S.U.; Ullah, I.; Khan, I.; Alghadhban, A.J.; Al-Adhaileh, M.H.; Ali, G.; ElAffendi, M. A UAV-Swarm-
Communication Model Using a Machine-Learning Approach for Search-and-Rescue Applications. Drones 2022, 6, 372. [CrossRef]
29. Petrlik, M.; Vonasek, V.; Saska, M. Coverage optimization in the Cooperative Surveillance Task using Multiple Micro Aerial
Vehicles. In Proceedings of the 2019 IEEE International Conference on Systems, Man and Cybernetics (SMC), Bari, Italy, 6–9
October 2019; IEEE: Bari, Italy, 2019; pp. 4373–4380. [CrossRef]
30. Perez, D.; Maza, I.; Caballero, F.; Scarlatti, D.; Casado, E.; Ollero, A. A Ground Control Station for a Multi-UAV Surveillance
System: Design and Validation in Field Experiments. J. Intell. Robot. Syst. 2013, 69, 119–130. [CrossRef]
31. Aljehani, M.; Inoue, M. Communication and Autonomous Control of Multi-UAV System in Disaster Response Tasks. In Agent
and Multi-Agent Systems: Technology and Applications; Jezic, G., Kusek, M., Chen-Burger, Y.H.J., Howlett, R.J., Jain, L.C., Eds.; Series
Title: Smart Innovation, Systems and Technologies; Springer International Publishing: Cham, Switzerland, 2018; Volume 74,
pp. 123–132. [CrossRef]
32. Skorobogatov, G.; Barrado, C.; Salamí, E. Multiple UAV Systems: A Survey. Unmanned Syst. 2020, 08, 149–169. [CrossRef]
33. Maza, I.; Ollero, A.; Casado, E.; Scarlatti, D. Classification of Multi-UAV Architectures. In Handbook of Unmanned Aerial Vehicles;
Valavanis, K.P., Vachtsevanos, G.J., Eds.; Springer: Dordrecht, The Netherlands, 2015; pp. 953–975. [CrossRef]
34. Krichen, M.; Adoni, W.Y.H.; Mihoub, A.; Alzahrani, M.Y.; Nahhal, T. Security Challenges for Drone Communications: Possible
Threats, Attacks and Countermeasures. In Proceedings of the 2022 2nd International Conference of Smart Systems and Emerging
Technologies (SMARTTECH), Riyadh, Saudi Arabia, 9–11 May 2022; pp. 184–189.
35. Citroni, R.; Di Paolo, F.; Livreri, P. A Novel Energy Harvester for Powering Small UAVs: Performance Analysis, Model Validation
and Flight Results. Sensors 2019, 19, 1771. [CrossRef] [PubMed]
36. Hajijamali Arani, A.; Azari, M.M.; Hu, P.; Zhu, Y.; Yanikomeroglu, H.; Safavi-Naeini, S. Reinforcement Learning for Energy-
Efficient Trajectory Design of UAVs. IEEE Internet Things J. 2022, 9, 9060–9070. [CrossRef]
37. Chen, Y.; Baek, D.; Bocca, A.; Macii, A.; Macii, E.; Poncino, M. A Case for a Battery-Aware Model of Drone Energy Consumption.
In Proceedings of the 2018 IEEE International Telecommunications Energy Conference (INTEL EC), Turino, Italy, 7–11 October
2018; pp. 1–8. ISSN: 0275-0473. [CrossRef]
38. Zhang, J.; Campbell, J.F.; Sweeney II, D.C.; Hupman, A.C. Energy consumption models for delivery drones: A comparison and
assessment. Transp. Res. Part D Transp. Environ. 2021, 90, 102668. [CrossRef]
39. Dorling, K.; Heinrichs, J.; Messier, G.G.; Magierowski, S. Vehicle Routing Problems for Drone Delivery. IEEE Trans. Syst. Man
Cybern. Syst. 2017, 47, 70–85. [CrossRef]
40. Chethan, R.; Kar, I. Multi-Agent Coverage Path Planning using a Swarm of Unmanned Aerial Vehicles. In Proceedings of the 2022
IEEE 19th India Council International Conference (INDICON), Kochi, India, 24–26 November 2022; ISSN: 2325-9418. [CrossRef]
41. Zhang, S.; Cheng, T. Optimizing Energy Consumption of Rotor UAV by Path Planning. In Proceedings of the 2022 Global
Conference on Robotics, Artificial Intelligence and Information Technology (GCRAIT), Chicago, IL, USA, 30–31 July 2022; pp.
54–58. [CrossRef]
42. Galkin, B.; Kibilda, J.; DaSilva, L.A. UAVs as Mobile Infrastructure: Addressing Battery Lifetime. IEEE Commun. Mag. 2019,
57, 132–137. [CrossRef]
43. Pasek, P.; Kaniewski, P. A review of consensus algorithms used in Distributed State Estimation for UAV swarms. In Proceedings
of the 2022 IEEE 16th International Conference on Advanced Trends in Radioelectronics, Telecommunications and Computer
Engineering (TCSET), Lviv-Slavske, Ukraine, 22–26 February 2022; pp. 472–477. [CrossRef]
44. Safavi, S.; Khan, U.A. Dynamic leader-follower algorithms in mobile multi-agent networks. In Proceedings of the 2015 IEEE
International Conference on Digital Signal Processing (DSP), Singapore, 21–24 July 2015; pp. 10–13. ISSN: 2165-3577. [CrossRef]
Drones 2023, 7, 263 36 of 37

45. Cao, Y.; Yu, W.; Ren, W.; Chen, G. An Overview of Recent Progress in the Study of Distributed Multi-Agent Coordination. IEEE
Trans. Ind. Inform. 2013, 9, 427–438. [CrossRef]
46. Kanellakis, C.; Nikolakopoulos, G. Survey on Computer Vision for UAVs: Current Developments and Trends. J. Intell. Robot.
Syst. 2017, 87, 141–168. [CrossRef]
47. Waharte, S.; Trigoni, N.; Julier, S. Coordinated Search with a Swarm of UAVs. In Proceedings of the 2009 6th IEEE Annual
Communications Society Conference on Sensor, Mesh and Ad Hoc Communications and Networks Workshops, Rome, Italy,
22–26 June 2009; pp. 1–3. ISSN: 2155-5494. [CrossRef]
48. Sharma, V.; Chen, H.C.; Kumar, R. Driver behaviour detection and vehicle rating using multi-UAV coordinated vehicular
networks. J. Comput. Syst. Sci. 2017, 86, 3–32. [CrossRef]
49. Liu, Y.; Luo, Z.; Liu, Z.; Shi, J.; Cheng, G. Cooperative Routing Problem for Ground Vehicle and Unmanned Aerial Vehicle: The
Application on Intelligence, Surveillance, and Reconnaissance Missions. IEEE Access 2019, 7, 63504–63518. [CrossRef]
50. Zhao, P.; Quan, Q.; Chen, S.; Tang, D.; Deng, Z. Experimental investigation on hover performance of a single-rotor system for
Mars helicopter. Aerosp. Sci. Technol. 2019, 86, 582–591. [CrossRef]
51. Segui-Gasco, P.; Al-Rihani, Y.; Shin, H.S.; Savvaris, A. A Novel Actuation Concept for a Multi Rotor UAV. J. Intell. Robot. Syst.
2014, 74, 173–191. [CrossRef]
52. Cai, G.; Lum, K.Y.; Chen, B.M.; Lee, T.H. A brief overview on miniature fixed-wing unmanned aerial vehicles. In Proceedings of
the IEEE ICCA 2010, Xiamen, China, 9–11 June 2010; pp. 285–290. ISSN: 1948-3457. [CrossRef]
53. Boon, M.A.; Drijfhout, A.P.; Tesfamichael, S. Comparison of a Fixed-Wing and Multi-Rotor UAV for Environmental Mapping
Applications: A Case Study. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 2017, XLII-2/W6, 47–54. [CrossRef]
54. Saeed, A.S.; Younes, A.B.; Cai, C.; Cai, G. A survey of hybrid Unmanned Aerial Vehicles. Prog. Aerosp. Sci. 2018, 98, 91–105.
[CrossRef]
55. DJI Mavic 3 Bestellen-DJI Store. Available online: https://store.dji.com/de/product/dji-mavic-3?vid=109821 (accessed on 15
March 2023).
56. Bentley, J.L. Multidimensional divide-and-conquer. Commun. ACM 1980, 23, 214–229. [CrossRef]
57. Maini, P.; Sundar, K.; Singh, M.; Rathinam, S.; Sujit, P.B. Cooperative Aerial–Ground Vehicle Route Planning With Fuel Constraints
for Coverage Applications. IEEE Trans. Aerosp. Electron. Syst. 2019, 55, 3016–3028. [CrossRef]
58. Garey, M.R.; Johnson, D.S.; Stockmeyer, L. Some simplified NP-complete problems. In Proceedings of the Sixth Annual ACM
Symposium on Theory of Computing, STOC ’74, Seattle, WA, USA, 30 April–2 May 1974; Association for Computing Machinery:
New York, NY, USA, 1974; pp. 47–63. [CrossRef]
59. Irnich, S.; Desaulniers, G. Shortest Path Problems with Resource Constraints. In Column Generation; Desaulniers, G., Desrosiers, J.,
Solomon, M.M., Eds.; Springer: Boston, MA, USA, 2005; pp. 33–65. [CrossRef]
60. Schrijver, A. On the history of the transportation and maximum flow problems. Math. Program. 2002, 91, 437–445. [CrossRef]
61. Bellman, R. On a routing problem. Q. Appl. Math. 1958, 16, 87–90. [CrossRef]
62. Dijkstra, E.W. A note on two problems in connexion with graphs. Numer. Math. 1959, 1, 269–271. [CrossRef]
63. Djidjev, H.; Chapuis, G.; Andonov, R.; Thulasidasan, S.; Lavenier, D. All-Pairs Shortest Path algorithms for planar graph for
GPU-accelerated clusters. J. Parallel Distrib. Comput. 2015, 85, 91–103. [CrossRef]
64. Zeng, W.; Church, R.L. Finding shortest paths on real road networks: The case for A*. Int. J. Geogr. Inf. Sci. 2009, 23, 531–543.
[CrossRef]
65. Adoni, W.Y.H.; Nahhal, T.; Aghezzaf, B.; Elbyed, A. The MapReduce-based approach to improve the shortest path computation
in large-scale road networks: The case of A* algorithm. J. Big Data 2018, 5, 16. [CrossRef]
66. Kerr, C.; Jaradat, R.; Ibne Hossain, N.U. Battlefield Mapping by an Unmanned Aerial Vehicle Swarm: Applied Systems
Engineering Processes and Architectural Considerations From System of Systems. IEEE Access 2020, 8, 20892–20903. [CrossRef]
67. Saeed, A.; Abdelkader, A.; Khan, M.; Neishaboori, A.; Harras, K.A.; Mohamed, A. On Realistic Target Coverage by Autonomous
Drones. arXiv 2018, arXiv:1702.03456.
68. Viseras, A.; Garcia, R. DeepIG: Multi-Robot Information Gathering With Deep Reinforcement Learning. IEEE Robot. Autom. Lett.
2019, 4, 3059–3066. [CrossRef]
69. Singh, A.; Krause, A.; Guestrin, C.; Kaiser, W.J. Efficient Informative Sensing using Multiple Robots. J. Artif. Intell. Res. 2009,
34, 707–755. [CrossRef]
70. Viseras, A.; Wiedemann, T.; Manss, C.; Magel, L.; Mueller, J.; Shutin, D.; Merino, L. Decentralized multi-agent exploration with
online-learning of Gaussian processes. In Proceedings of the 2016 IEEE International Conference on Robotics and Automation
(ICRA), Stockholm, Swede, 16–21 May 2016; pp. 4222–4229. [CrossRef]
71. Krichen, M.; Lahami, M.; Cheikhrouhou, O.; Alroobaea, R.; Maâlej, A.J. Security testing of internet of things for smart city
applications: A formal approach. In Smart Infrastructure and Applications; Springer: Cham, Switzerland, 2020; pp. 629–653.
72. Zhou, Y.; Cheng, N.; Lu, N.; Shen, X.S. Multi-UAV-Aided Networks: Aerial-Ground Cooperative Vehicular Networking
Architecture. IEEE Veh. Technol. Mag. 2015, 10, 36–44. [CrossRef]
73. Cheng, C.M.; Hsiao, P.H.; Kung, H.T.; Vlah, D. Maximizing Throughput of UAV-Relaying Networks with the Load-Carry-and-
Deliver Paradigm. In Proceedings of the 2007 IEEE Wireless Communications and Networking Conference, Hong Kong, China,
11–15 March 2007; pp. 4417–4424. ISSN: 1558-2612. [CrossRef]
Drones 2023, 7, 263 37 of 37

74. Ohta, Y.; Ohta, T.; Kohno, E.; Kakuda, Y. A Store-Carry-Forward-Based Data Transfer Scheme Using Positions and Moving
Direction of Vehicles for VANETs. In Proceedings of the 2011 Tenth International Symposium on Autonomous Decentralized
Systems, Tokyo, Japan, 23–27 March 2011; pp. 131–138. ISSN: 1541-0056. [CrossRef]
75. Sharma, V.; Ganpati, A. Comparison of Topology Based-Routing Protocols in Wireless Network. J. Multimed. Inf. Syst. 2019,
6, 61–66. [CrossRef]
76. Moniruzzaman, A.B.M.; Rahman, M.S. Analysis of Topology Based Routing Protocols for Vehicular Ad-Hoc Network (VANET).
arXiv 2014, arXiv:1411.7662.
77. Kaur, H.; Singh, H.; Sharma, A. Geographic Routing Protocol: A Review. Int. J. Grid Distrib. Comput. 2016, 9, 245–254. [CrossRef]
78. Shirani, R.; St-Hilaire, M.; Kunz, T.; Zhou, Y.; Li, J.; Lamont, L. On the Delay of Reactive-Greedy-Reactive Routing in Unmanned
Aeronautical Ad-hoc Networks. Procedia Comput. Sci. 2012, 10, 535–542. [CrossRef]
79. Zang, C.; Zang, S. Mobility prediction clustering algorithm for UAV networking. In Proceedings of the 2011 IEEE Globecom
Workshops (GC Wkshps), Houston, TX, USA, 5–9 December 2011; pp. 1158–1161.
80. Liu, K.; Zhang, J.; Zhang, T. The clustering algorithm of UAV networking in near-space. In Proceedings of the 2008 8th
International Symposium on Antennas, Propagation and EM Theory, Kunming, China, 2–5 November 2008; pp. 1550–1553.
81. Krishnamachari, B.; Estrin, D.; Wicker, S. Modelling Data-Centric Routing in Wireless Sensor Networks. IEEE INFOCOM 2002, 2,
39–44.
82. Belding-Royer, E.M. Multi-Level Hierarchies for Scalable Ad hoc Routing. Wirel. Netw. 2003, 9, 461–478. [CrossRef]
83. Jacquet, P.; Muhlethaler, P.; Clausen, T.; Laouiti, A.; Qayyum, A.; Viennot, L. Optimized link state routing protocol for ad hoc
networks. In Proceedings of the Proceedings. IEEE International Multi Topic Conference, 2001. IEEE INMIC 2001. Technology for
the 21st Century, Lahore, Pakistan, 30 December 2001; pp. 62–68. [CrossRef]
84. Perkins, C.E.; Bhagwat, P. Highly dynamic Destination-Sequenced Distance-Vector routing (DSDV) for mobile computers. ACM
SIGCOMM Comput. Commun. Rev. 1994, 24, 234–244. [CrossRef]
85. Johnson, D.B.; Maltz, D.A. Dynamic Source Routing in Ad Hoc Wireless Networks. In Mobile Computing; Imielinski, T., Korth,
H.F., Eds.; The Kluwer International Series in Engineering and Computer Science; Springer: Boston, MA, USA, 1996; pp. 153–181.
[CrossRef]
86. Zapata, M.G. Secure ad hoc on-demand distance vector routing. ACM SIGMOBILE Mob. Comput. Commun. Rev. 2002, 6, 106–107.
[CrossRef]
87. Forsmann, J.H.; Hiromoto, R.E.; Svoboda, J. A time-slotted on-demand routing protocol for mobile ad hoc unmanned vehicle
systems. In Proceedings of the Unmanned Systems Technology IX; SPIE: Orlando, Florida, USA, 2007; Volume 6561, pp. 530–540.
[CrossRef]
88. Park, V.; Macker, J.; Corson, M. Applicability of the temporally-ordered routing algorithm for use in mobile tactical networks. In
Proceedings of the IEEE Military Communications Conference. Proceedings. MILCOM 98 (Cat. No.98CH36201), Boston, MA,
USA, 19–21 October 1998; Volume 2, pp. 426–430. [CrossRef]
89. Seither, D.; König, A.; Hollick, M. Routing performance of Wireless Mesh Networks: A practical evaluation of BATMAN
advanced. In Proceedings of the 2011 IEEE 36th Conference on Local Computer Networks, Bonn, Germany, 4–7 October 2011;
pp. 897–904. ISSN: 0742-1303. [CrossRef]
90. Haas, Z.J.; Pearlman, M.R. The performance of query control schemes for the zone routing protocol. ACM SIGCOMM Comput.
Commun. Rev. 1998, 28, 167–177. [CrossRef]
91. Karp, B.; Kung, H.T. GPSR: Greedy perimeter stateless routing for wireless networks. In Proceedings of the 6th Annual
International Conference on Mobile Computing and Networking, MobiCom ’00, Boston, MA, USA, 6–11 August 2000; Association
for Computing Machinery: New York, NY, USA, 2000; pp. 243–254. [CrossRef]
92. Cardellini, V.; Colajanni, M.; Yu, P. Geographic load balancing for scalable distributed Web systems. In Proceedings of
the 8th International Symposium on Modeling, Analysis and Simulation of Computer and Telecommunication Systems (Cat.
No.PR00728), San Francisco, CA, USA, 29 August–1 September 2000; pp. 20–27. ISSN: 1526-7539. [CrossRef]
93. Lin, L.; Sun, Q.; Li, J.; Yang, F. A novel geographic position mobility oriented routing strategy for UAVs. J. Comput. Inf. Syst. 2012,
8, 709–716.
94. Zhang, D.G.; Zhao, P.Z.; Cui, Y.y.; Chen, L.; Zhang, T.; Wu, H. A New Method of Mobile Ad Hoc Network Routing Based on
Greed Forwarding Improvement Strategy. IEEE Access 2019, 7, 158514–158524. [CrossRef]
95. Schranz, M.; Umlauft, M.; Sende, M.; Elmenreich, W. Swarm Robotic Behaviors and Current Applications. Front. Robot. AI 2020,
7, 36. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.