A Survey on Unmanned Aerial Vehicles Emerging Applications

Communication Control, 3D Reconstruction, Challenges and

Federated Learning-Based Solutions

Abstract:

Unmanned Aerial Vehicles (UAVs) have quickly integrated themselves as fundamental

technologies in nearly every field, including surveillance, delivery, precision agriculture, and
even disaster management. This survey looks at the application of UAVs in communication
systems, control systems, and the 3D reconstruction of objects and utilizes image data
techniques. UAVs operate in real time on command which requires strong and dependable
communication for multiple UAVs to operate in swarming and beyond-visual-line-of-sight
(BVLOS) missions. Control systems incorporating AI-based autonomy need robust path planning
in dynamically and uncertainly hostile environments. Furthermore, precision 3D reconstruction
allows accurate environmental modeling needed in inspection for urban planning infrastructure.
Despite technological advancements, UAV systems face several challenges, including bandwidth
limitations, energy constraints, latency, and data privacy concerns. The survey analyzes the
growing importance of Federated Learning (FL) in UAV networks as an approach to model
training that ensures privacy and security of data. This paper represents different applications
and challenges in UAVs. This paper also describes different FL-based methods and techniques to
overcome different challenges in UAVs. This research not only analyzes emerging trends but also
addresses unfulfilled needs to provide an actionable guide to build upon for further
advancement in UAV networks, urging for new enhanced adaptability, flexible scalability, and
intelligent multifunctional solutions for wide-ranging future uses.

Keywords: Unmanned Aerial Vehicles (UAVs), 3D Reconstruction, Challenges, Federated

Learning (FL), Components, UAV imaging, Communication Channels, Applications

1. Introduction

An Unmanned Aerial Vehicle (UAV), also known as a drone, is an aircraft without a human pilot
and is operated remotely and autonomously by onboard computers and sensors. UAVs include
aerial control stations and system communication [1]. They are also equipped with sophisticated
sensors such as GPS, high-resolution camera, Light Detection and Ranging (LiDAR), and on-board
computing hardware, allowing them to execute complicated missions in both constrained and
unconstrained environments. UAVs were originally developed for military surveillance and
operations, which are too dangerous for humans. UAVs are now quickly becoming indispensable
tools across a wide range of civilian and industrial applications in areas such as agriculture,
logistics, disaster response, urban planning and infrastructure inspection [2]. They are valuable
because they are potentially lightweight, low-cost, and can traverse areas that are inaccessible
to people, providing high-resolution data in real time with minimal operational overhead [3].
UAVs have emerged as pivotal tools across numerous domains due to their operational
flexibility, cost-efficiency, and ability to access remote or hazardous environments.

The most significant applications of UAVs include aerial reconnaissance, environmental

surveillance and 3D reconstruction. In aerial remote sensing, UAV enablement of real-time
surveillance is essential in military applications, border patrolling, and disaster recovery for the
rapid collection of data with human personnel out of danger [4]. For environmental monitoring,
UAV are used to monitor changes in land use, vegetation vigor, air quality and wildlife activity
which significantly improves the ecological research and conservation tasks by providing
infrequent temporal images with spatial resolution difficult to achieve with satellite-based
platforms. Moreover, drones have transformed 3D reconstruction through accurate and time-
effective terrains, urban areas and historic building mapping by photogrammetry and LiDAR
sensors [5]. Such technologies would enable the acquisition of high-resolution digital surface
models for urban planning, heritage preservation, and infrastructure monitoring. Stable control
and reliable communication systems are of great importance for UAVs working in complex and
high-speed environments, such as 3D reconstruction and emergency aerial bases. The
applications require very accurate trajectory management and low-latency data transmission for
accurate and timely mission completion [6].

The integration of reliable communication protocols and decentralized control structures greatly
improves the UAVs flying performance in dynamic surroundings. The inherent 3D flight
trajectory and maneuverability of UAVs lead to great challenges of wireless communication,
control accuracy and trajectory optimization, especially in dynamic and complex environments
[7]. With their highly non-linear and quickly evolving trajectories, the maintenance of stable
communication links is challenging, particularly in scenarios with non-continuous traffic away
from the central Line of Sight (LOS) region and interference, which cause a significant impact on
data throughput and delay-sensitive applications like streaming of information to aerial base
stations and real-time 3D mapping [8].

Moreover, accurate UAV control is difficult to achieve in hazardous airspaces where external
inputs, such as wind turbulence or GPS drift, are influencing the stability, requiring adaptive
control systems and AI flight models [9]. In addition to the trajectory optimization, energy
limitations, obstacle avoidance, and real-time decision making further complicate the case for
swarm or cooperative UAV missions [10]. These challenges directly impact the progress of new
UAV applications and necessitate more reliable multi-UAV coordination frameworks, as well as
intelligent edge computing and secure communication protocols, capable of exploiting the full
potential of UAVs in promising application domains, smart cities, disaster response, and
autonomous delivery networks.

Federated Learning (FL) is a decentralized machine learning framework that enables multiple
UAVs to collaboratively train a shared model without transferring their raw local data to a
central server, thus enhancing data privacy, communication efficiency, and resilience in
distributed environments. In UAV networks, FL overcomes major challenges in communication
control by allowing UAVs to perform localized model updates and aggregate only encrypted or
masked model parameters, significantly reducing the bandwidth requirements and improving
secure coordination, especially in remote or infrastructure-limited areas [11]. For 3D
reconstruction, FL facilitates distributed image processing and mapping by enabling UAVs to
collaboratively improve 3D models without centralizing massive volumes of sensitive visual
data, thereby preserving privacy and accelerating reconstruction tasks [12]. Furthermore,
intelligent frameworks have integrated FL with deep reinforcement learning to jointly optimize
UAV trajectory planning, task offloading, and transmission security, ensuring stable 3D flight
control and robust data aggregation in urban and disaster-stricken zones. Figure 1 describes the
overall structure of the paper and table 1 is the List of abbreviations. Overall, FL has proven
essential for enhancing UAV autonomy, reliability, and scalability in real-world aerial missions
involving complex communication and 3D environment challenges.
 Figure 1: Over structure of the paper

The contributions of this article are outlined as follows

 The article systematically reviews the major challenges in UAV communication control,
including connectivity preservation, wireless security vulnerabilities, and mobility
management in dynamic 3D environments.
 The article explores recent advancements in UAV applications across many fields,
emphasizing the growing need for robust communication and control.
 The article aims to describe the importance of UAVs systems and the architecture of UAVs
components.
 The article aims to describe the UAVs imaging components and UAVs Communication
Channels Technologies.
 This article provides Federated learning-based techniques, methods, and models to
overcome different challenges.
 Table 1: List of abbreviations

Abbreviations Definition
UAVs Unmanned Aerial Vehicle
LiDAR Light Detection and Ranging
LOS Line of Sight
FL Federated Learning
DRL Deep reinforcement learning
U2G UAV-to-ground
U2U Inter-UAV
BVLOS Beyond-visual-line-of-sight
GCS Ground control station
VTOL Vertical take-off and landing
GPS Global positioning system
IMU Inertial measurement unit
DPU Data Processing Unit
FCS Flight Control System
MSS Multi-state systems
SAR Synthetic aperture radar

This article aims to provide a comprehensive overview of UAV technologies, with a focus on UAV
imaging systems, communication channels, and wireless communication protocols, alongside an
in-depth analysis of the challenges and emerging applications associated with these
technologies. The remainder of the article is summarized as follows. Section 2 provides an
overview of related work. Section 3 details the methodology for UAVs challenges, FL-based
techniques and research questions. Section 4 presents the experimental results and discussions
to answer the questions related to the applications, challenges and FL-based solution to these
challenges. Finally, Section 5 concludes the paper and discusses future work.

2. Related work

UAVs rely on a sophisticated system architecture comprising flight controllers, sensors,

communication modules, and payloads to enable autonomous operations. The imaging
capabilities of UAVs, equipped with high-resolution and multispectral cameras, have
revolutionized applications in precision agriculture, infrastructure inspection, and
environmental monitoring [14]. Efficient communication channels, including cellular (4G/5G),
satellite, and ad-hoc networks, are critical for real-time data transmission, swarm coordination,
and beyond-visual-line-of-sight (BVLOS) operations [15]. Recent advancements in lightweight
materials and edge computing have further enhanced UAV endurance and onboard processing
for imaging tasks [16]. However, challenges such as signal interference, bandwidth limitations,
and cybersecurity risks persist in UAV communication systems. Studies have also explored AI-
driven image processing techniques to improve object detection and 3D mapping accuracy from
 UAV-captured data. This related work examines these architectural, imaging, and
communication advancements, highlighting their interdependencies.

2.1. UAV System Architecture and Components:

UAVs include several basic units whose cooperation allows aerial, autonomously or remotely
controlled work. Modern UAVs usually consist of the airframe, propulsion engines, control and
navigation systems, communication devices, ground control station (GCS), and mission
payloads. UAVs airframes are classified into fixed-wing, rotary-wing, and hybrid vertical take-off
and landing (VTOL) types, each offering distinct aerodynamic and operational benefits for
various mission profiles [17]. The autopilot system plays a central role in the autonomy of UAV
and includes a global positioning system (GPS), an inertial measurement unit (IMU), and control
algorithms to coordinate flight dynamics and navigation [18]. Communication systems are
crucial for the transmission of telemetry and command signals between the UAV and the
ground station, and up-to-date UAVs now employ dual-link communication, including radio
frequency, cellular, and satellite communication, for reliability [19]. Battery System is one of the
most significant components which supply electric energy for all the on-board systems, such as
propulsion, avionics, sensors, communication devices, and flight controller. The Data
Processing Unit (DPU) is an essential onboard system responsible for data collection,
interpretation, and analysis from the flight mission. Figure 2 presents the UAVs architecture
components. It is the algorithmic heart of the system, dealing with payload data, sensor inputs
and mission-relevant decisions within real-time processing.

The payload systems, such as multispectral sensors, LiDAR, or high-resolution imaging sets of
the devices, are modular and customized for application fields from environmental monitoring
to tactical surveillance [20]. Another essential module of a UAV architecture is the Flight Control
System (FCS), the core subsystem which provides stability to the aircraft and implements motion
commands. It ensures that the UAV maintains controlled, stable, and accurate flight according
to mission parameters. GCS serve as an operational hub for mission planning and real-time UAV
control, with scalability ranging from compact tablet interfaces to comprehensive multi-display
systems [21]. When integrated with edge and cloud computing, GCS systems are augmented by
additional processing power, real-time analytics and additional data storage beyond the local
hardware. This connectivity provides tele-autonomy, mission flexibility and scalable data
processing. The recovery system refers to the group of mechanisms and procedures which have
been developed to ensure that the UAVs returns to its base safely at the end of a mission. Its
primary aim is to reduce the destructive impact on the UAV and its payload and systems,
especially in hazardous and unpredictable environments. Additionally, reliability is enhanced
through modelling UAVs as multi-state systems (MSS), enabling real-time assessment of
component degradation and system-level dependability [22]. Collectively, the modern UAV
 system architecture is a fusion of aerospace design with embedded systems and wireless
communication technology for more challenging and autonomous missions.

Figure 2: UAVs architecture components

2.2. The importance of UAV imaging

UAVs imaging is an established state-of-the-art technology employed in the military and civilian
domains that enhances situational awareness, precision targeting, and real-time information
collection. Military applications UAV cameras can be used for several military operations as
reconnaissance, surveillance, target recognition and continuous monitoring of the battlefield.
The UAV Hold Fast Imaging Platform has years of experience working with the capabilities and
limitations of our many advanced sensors to achieve optimal performance in both civilian and
military applications. Red green and blue (RGB) cameras record high-quality color images, which
are widely applied to mapping, monitoring, and real-time target detection tasks [23]. Thermal
infrared cameras sensing long-wave infrared radiation are vital when the visibility is low or in
the darkness, and are more and more widely applied among search and rescue, security, and
structural inspection [24]. Multispectral sensors gather image data across several discrete
spectral bands, making them particularly valuable in agriculture, forestry, and environmental
monitoring for assessing vegetation health.

Hyperspectral sensors build upon this capability by providing hundreds of such narrow spectral
bands, thus allowing further material identification and anomaly detection, such as mineral
identification and pollution assessment [25]. LiDAR uses pulsed laser beams to generate
accurate 3D point clouds of terrain and objects, and it’s an essential tool of the trade for
topographical mapping and infrastructure inspections. The synthetic aperture radar (SAR) is an
active remote sensing system known for its ability to generate high-resolution imagery
irrespective of illumination and weather conditions, and is cannily employed in defence
reconnaissance and disaster mapping processes. Figure 3 represents the UAVs imaging
components and importance

Finally, 3D reconstruction systems combine visual, thermal, and LiDAR data to build spatial
models that are used to produce detailed georeferenced environmental representations to
improve both mission planning and infrastructure analysis. These integrated imageries will
together enhance the coverage, accuracy and flexibility of the UAV-based remote sensing
platforms.
 Figure 3: UAVs imaging components and importance

2.3. The Importance of UAVs Communication Channels Technologies

Different wireless channel models are being employed to provide robust connectivity and
feasibility in diverse mission environments in the UAV communication systems. The Air-to-
Ground (A2G) channel is an indispensable link for UAV systems, over which most of the data
downlink and telemetry are exchanged between the airborne vehicles and the ground network.
It is highly sensitive to the height of UAVs, number of buildings and the type of environment, as
evidenced by the large-scale ray-tracing models that study path loss and LoS probabilities in
different urban scenarios [26]. Recent measurement-based studies have refined models of G2A
communication under 5G New Radio conditions, highlighting the importance of elevation angle,
signal strength, and Rician K-factor in achieving stable high-altitude UAV control [27]. The Air-to-
Air (A2A) channel, which is essential for coordinating UAV swarms or multi-drone missions,
supports collaborative mission execution and real-time data exchange; in urban environments,
it is most accurately modelled using probabilistic two-ray path loss models to account for the
altitude-dependent reflections and building-induced shadowing [28]. Drone-to-Drone (D2D)
communication improves the resilience and adaptability of swarms, mainly in some scenarios
such as military, where decentralized coordination is very important. Such LoS channels, which
typically exist in an open environment or at high altitude, offer higher throughput and better
 received signal quality and are also better suited for high-data-rate applications, such as live
video streaming.

However, he stability of the UAV is of paramount importance in the performance of the LoS
channel [29]. On the other hand, Non-Line-of-Sight (NLoS) type channels, which prevail in
obstructed and urban environments, suffer from substantial signal deterioration caused by
shadowing and the presence of reflections. Empirical measurements show that shadow fading
in NLoS environments can even reach twice the value found in the LoS ones, which rules out the
possibility of ignoring the transmission channel, and points towards a direction of complex
communications channel modelling and error correction techniques [30]. Figure 4 represents
UAVs Communication Channels Technologies, and Table 2 shows the research with methodology
and future directions.

Finally, multipath fading channels, caused by signal reflections from buildings, terrain, and other
UAVs, contribute to fluctuations in received signal quality. These effects are typically simulated
with a Rician or Rayleigh fading distribution and compensated using advanced modulation,
coding and antenna methods [31]. Collectively, these channels determine the performance
landscape of UAV communications and offer guidance on how to design robust and efficient
aerial networks for both civilian and military use cases.

Table 2: UAVs communications channels with future work and methodology

Ref.# Methodology UAVs Advantage Disadvantage future work

communication
channel
[26] Ray-tracing Air-to-Ground (A2G) Accurate Limited to Extend to
simulation in LoS/NLoS specific building dynamic
3D urban modeling in distribution environments
environments urban contexts assumptions and real-time
UAV motion
[27] Empirical 5G Ground-to-Air (G2A) Validated 5G Limited to one Multi-
NR signal signal behavior at frequency band frequency
measurements high altitudes and static analysis and
at varying scenarios adaptive G2A
altitudes modeling
[28] Probabilistic Air-to-Air (A2A) Effective urban Does not Integration
two-ray aerial link account for UAV with mobility
modeling for estimation mobility and trajectory
 A2A channels variations prediction
models
[29] Experimental Line-of-Sight (LoS) Quantifies jitter Focused only on Include
analysis of UAV impact on LoS hovering UAVs horizontal
jittering on LoS signal quality flight and
stability complex
maneuvers
[30] Ultra-wideband Non-Line-of-Sight Highlights Human-induced Incorporate
channel (NLoS) shadow fading blockage only vehicular and
measurements variations in NLoS partially structural
conditions analyzed blockages
[31] Outage A2G and Enhanced energy Assumes ideal Test in low-
probability backscatter hybrid efficiency via power and power and
analysis using backscatter fading interference-
probabilistic communication conditions prone
models environments
[32] Thermal and Only focus on High-resolution Computationall Real-time 3D
RGB-based 3D imaging thermal 3D maps y intensive reconstructio
reconstruction process n from UAVs
pipeline
[33] Comparison of Only focus on Robust 3D LiDAR cost and Lightweight
LiDAR and imaging detection under power 3D sensing
passive 3D occlusion and fog requirements integration on
imaging in UAVs
foggy settings
[34] Deep learning- mmWave Air-to- Superior accuracy Model Adaptable
based channel Ground over traditional generalizability learning
estimation for models to unseen models for
mmWave UAV environments is diverse
links limited scenarios
[35] Statistical UAV-to-Ground Closed-form Theoretical Test in real-
modeling of (Fading Channel) expressions for validation only; world
cascaded key performance lacks hardware deployments
shadowed UAV- metrics implementation with mobility
to-ground and fading
fading channels variation
using Meijer G
functions
[36] LoS-based Vehicular-UAV Improves Assumes Adapt to
coordinated Hybrid (LoS- throughput and accurate LoS dynamic
channel enhanced) reduces RSU load prediction and vehicular
allocation in high-density synchronized speed and
protocol for areas backoff trajectory
 VANETs using variations
UAVs

3. Methodology

The objective of this work is to construct an SLR that aims to capture real-world evidence
concerning UAVs their applications, challenges and federated learning-based solutions. UAVs
has a lot of application in communication systems, control systems, and the 3D reconstruction
of objects and utilizes image data techniques. The applications challenges and FL based
solutions have been examined in this paper. Also, the SLR focuses a comprehensive overview of
UAV technologies, with a focus on UAV imaging systems, communication channels, and wireless
communication protocols, alongside an in-depth analysis of the challenges and emerging
applications associated with these technologies. In addition, this paper addresses many of the
limitations posed on UAVs in different domains and the role of Federated learning in
overcoming these limitations. This goal will be achieved through the use of research questions
which will be substantiated in the review stage. Figure 4 illustrates the planning methodology,
while the research questions are outlined in table 3.

Figure 4: Proposed systematic literature review (SLR)

Table 3: Research questions

 Research Question Motivations
What are the applications of UAVs, including aerial The purpose of this question is to highlight
reconnaissance, environmental monitoring, and 3D different applications of UAVs and their impact
reconstruction, both in the military and civilian within different fields like civil and military.
fields?

What are the challenges in UAVs regarding 3D The purpose of this question is to highlight
trajectory, high maneuverability of UAVs, wireless different challenges of UAVs which need to be
communication performance, UAV control and solved with the latest techniques of FL-based
solutions.
trajectory optimization?
What is the Federated learning based optimal This includes an analysis of particular FL-based
multi-UAV control strategies for energy-efficient techniques which are thought as a solution to
and accurate 3D reconstruction and high UAVs challenges.
maneuverability of UAVs in large-scale urban and
disaster zones?

3.1. Select data sources

Actual source is defined as different libraries and repositories which serves as the source for
different search results. Springer Digital Library, Willey, ACM Digital Library, IEEE Explore and
Science Direct are the top digital libraries from which we will obtain the required results. To
extract results from preceding studies, the full text literature is mandatory and crucial. Any
digital library provides a multitude of means with which to seek the required information. To
gain the correct results, all research methods must be adapted. As for being in the table and
the figure, Table 4 and the and Figure 5 show different data sources and the number of studies,
respectively, resultant from the search queries.

Table 4: Query results from data sources

Library Initial Title and Keyword Abstract Introduction and Full Text
Conclusion
 Springer 220 200 80 54 28

Willey 205 150 75 44 22

ACM Digital 190 130 55 31 22
Library
IEEE Explore 195 140 40 22 24
Science 165 60 45 24 24
Direct
Results 975 680 295 175 120

Search Query
1200

1000

800

600

400

200

0
Initial Title and Keyword Abstract Introduction and Full Text
Conclusion

Springer Willey ACM Digital Library

IEEE Explore Science Direct Results

Figure 5: Search queries representation

3.2. Formulate search string

String generation based in on the focus of the study, keywords and search terms that are joined
to constitute a search string that eventually leads to the answers of the research questions. For
now, search strings are built by choosing terms related to the research questions as shown in
table 5 just below. To construct every next search string the above-mentioned keyword items
are used within the logical operation (AND or OR) order defined. Figure 6 provides a case study
depicting the factors which influence developing the search strings and illustrate some of the
most important terms in making this search list.

Table 5: Search string formulation

 Keyword Synonym /Alternative word
Unmanned Aerial (“UAVs OR “Drones”)
Vehicles
Federated Learning (“Federated Learning” OR “FL ”)
Techniques (“Techniques” OR “Methods”)
Applications (“Applications” OR “Domains ”)
Integration (“Combination,” “OR “merge ”)
Challenges (“Issues” OR “Problems”)

Figure 6: The process of formulating the search string.

3.3. Define inclusion and exclusion criteria

The specific studies that fulfil the criteria and parameters for inclusion provided above will
constitute the entire coverage involved in the literature review. The following are the inclusion
criteria for this review, all study included studies should be available in English study must be
conducted between 2021 to 2025. Integration of FL-based methods and techniques in UAVs is
the top priority of research. The focus of studies and its investigation is on how FL-based
methods and techniques can be incorporated to UAVs to overcome different challenges. All
 selected studies must evaluate UAVs and its applications and what are the challenges in UAVs
and what are the applications of a UAVs and why there is need of FL-based techniques to
overcome the challenges faced by the UAVs. Only articles published in reputable journals,
books, or conferences should be included and considered. Studies that will not be included in
the literature review are studies that do not meet the exclusion criteria. All the papers which
were published before 2021 are excluded.

3.4. Define quality assessment criteria

In systematic literature reviews SLRs and in meta-analyses, qualitative methods are a specified
set of techniques for judging the worthiness, credibility and quality of research attempts
Qualitative methods are accepted platforms or techniques for determining the validity
equivalency and the structure of the quality asserts. By introducing quality control methods and
proofs of results, we can ensure that the primary studies which we have selected are able to
provide sufficient information to answer our research questions. In this step, each research
topic is compared with the standard. The quality control system is listed in Table 6 and is
included starting from QCS.

Table 6: Quality assessment criteria.

Sr. No QA Questions
QCS1 Does the following study provide enough information on applications of
UAVs?
QCS2 Does the study suggest the challenges in UAVs?

QCS3 Does the study neglect the key issues related to UAVs and the solutions
with the integration with FL-based techniques and methods?

3.5. Primary Study Selection

The concept ‘primary study’ is used as a means of detecting specific passages and subsections
in a book that are relevant to certain research questions. To facilitate a well-structured review,
relevant articles were selectively included based on tollgate approach consisting of five
successive steps. In this way, 120 original studies were identified in strict accordance with pre-
determined quality criteria. The primary works are presented in Figure 7 and Table 7 and the
diagram in Figure 8 presents the PRISMA diagram.

Table 7: Final paper Selection

Year Final Selection

2021 17
2022 26
2023 31
2024 43
2025 3

Final Paper Selection

50
45
40
35
30
25
20
15
10
5
0
2021 2022 2023 2024 2025

Final Selection

Figure 6: Final paper selection

In the search phase, several academic databases such as Springer, ACM, IEEE, Web of Science,
Elsevier, Google Scholar and Wiley and are searched to cover the domains of the importance
and applications of UAVs, different challenges in UAVs and FL-based solutions to these
challenges. The PRISMA diagram reduces the possibility that high quality and relevant studies
can be missed due to carelessness or lack of structure, thereby strengthening the credibility
and validity of research.
 Figure 8: PRISMA diagram

4. Result and Discussions:

4.1. What are the applications of UAVs, including aerial reconnaissance, environmental
monitoring, and 3D reconstruction, both in the military and civilian fields?

UAVs are playing important roles in military and civil fields because of their flexibility,
effectiveness, and the ability to perform complex tasks in an autonomous mode. Their
importance lies in enhancing operational effectiveness, reducing human risk, and enabling high-
resolution data collection. In the military domain, UAVS are fundamental elements in
intelligence, surveillance, and reconnaissance (ISR) missions, which provide real-time situational
awareness and enable target tracking while minimizing human exposure in hostile areas [37].
On the civilian side, UAVs contribute significantly to public safety, infrastructure inspection,
disaster management, and environmental monitoring. Their use for surveillance in safety and
rescue operations has transformed the way emergency response can be carried out, enabling
access where it is difficult or dangerous terrains. Moreover, in the field of social sciences and
environmental studies, the usability of UAVs has been recognized to facilitate the geospatial
data collection and 3D mapping, which is indispensable for archaeological surveys, agricultural
planning, and urban planning [38]. Different applications of UAVs are presented in Figure 9.

The mission flexibility of UAV systems, with the superior persistent range, supports healthcare
logistics, border patrol, and communication networks, reinforcing their multipurpose utility [39].
As UAV technology advances, its importance across various fields will mature, and
interdisciplinary cooperation, cost efficiency, and rapid decision-making in a complex
operational milieu will be enhanced.

Figure 9: Applications of UAVs

4.1.1. Aerial Reconnaissance

Aerial surveillance is one of the essential applications of UAVs that plays a pivotal role in both
military as well as civilian applications. UAVs play an important role in aerial reconnaissance as
they supply situational understanding that depends on the capturing of high-resolution imagery
and obtaining required information in real time [40]. Its main role is the surveillance, detection
and reconnaissance over large areas with the minimum of human risk. Recent work has
concentrated on the optimization of UAV path planning via DRL and a variety of heuristic
methods to enhance mission efficiency in dynamic environments. Moreover, mobile ground
vehicles can be assisted by UAVs to stretch out their range of sight and improve navigation
decision-making with adaptive 3D path planning solutions [41]. These functions are crucial in
surveillance, disaster response, and tactical military operations, where UAVs enhance coverage,
reduce response time, and improve decision-making accuracy.

4.1.2. Environmental Monitoring

In environmental monitoring, UAVs can be applied to monitor environmental conditions in

large-scale, time-sensitive and remote and hazardous areas. UAVs systems use mobile aerial
sensing platforms with onboard environmental sensors for acquisition of air, land and water
quality data, thus offering much enhanced spatial and temporal resolution compared to
traditional techniques [42]. Integrating UAVs and edge computing wireless sensor networks can
improve data processing and reduce transmission latency, such that it can be applied to areas
with insufficient infrastructure. These systems are especially vital in climate-sensitive regions
where real-time data collection can guide immediate policy or conservation interventions.
Furthermore, as a result of the ability to take high-resolution images, UAVs can substantially
increase the accuracy of monitoring in environmental restoration and emissions tracking, with
AI-based analytics [43]. Overall, UAV-based environmental monitoring systems provide a game-
changing move towards proactive, cost-efficient ecological management and policy support.

4.1.3. 3D Reconstruction and Mapping.

3D Reconstruction and Mapping from UAV follows a transformative application in both civilian
and industrial domains, for accurate modelling of physical environments in urban planning,
infrastructure monitoring, and disaster management [44]. By taking overlapping high-
resolution snapshots during a flight, UAVs with powerful stereo vision systems and
Simultaneous Localization and Mapping (SLAM) techniques can support the dense modelling of
3D scenes. These systems utilize GPS and deep learning-based feature matching to minimize
drift and improve the spatial resolution [45]. Recent innovations also include the deployment of
multiple aerial robots using wide-baseline stereo vision to achieve long-range depth estimation
and broader terrain coverage. These types of methods render UAV-based 3D reconstruction
more reliable and more scalable, and would find extensive applications where efficient, low-cost
and high-fidelity modelling of complex and large-scale scenarios is needed [46].

4.1.4. Surveillance and Intelligence Gathering

Surveillance and intelligence are a cornerstone application of UAVs because they provide very
high-resolution data monitoring without endangering personnel. UAVs are used as independent
flying vehicles that can acquire data in real time using onboard cameras and sensors to detect
threats, observe enemy movements and keep track of sensitive installations [47]. These
capabilities are enhanced through integration with neural networks for automatic object
recognition and classification, significantly improving the speed and accuracy of mission-critical
decisions [48]. During these investigations, collaborative UAV surveillance schemes and swarm
intelligence methods have also been developed to enhance coverage and operational reliability
in complex scenarios [49]. Secure data transfer and cryptosystems also guarantee the security
of sensitive information collected during reconnaissance flights [50]. As a result, the significance
of UAVs is an important tool for real-time situational awareness, which could provide strategic
edges in military operations and civil surveillance.

4.1.5. Disaster Response and Search & Rescue

Disaster recovery and search and rescue (SAR) are among the most important life-saving roles of
UAVs, providing fast situation awareness, real-time search, and accurate victim positioning in
dangerous and inaccessible areas. UAVs serve as agile sensor-equipped platforms, which can
perform rapid searches, identify survivors and provide crucial information to emergency
responders, significantly reducing the time to locate and assist victims [50]. The most advanced
systems, which are integrated with thermal imaging, AI-based recognition and edge computing,
are employed to reinforce detection accuracy and autonomy even under GPS-denied and
communication-degraded areas [51]. Multi-level swarm architectures of UAV are also being
suggested to achieve post-disaster settings, fast area coverage and multimodal sensing, which is
better than the conventional traditional ground-based search methods [52]. Mobile edge and
fog computing can be integrated with UAVs to improve task offloading and resource allocation,
to meet the requirements of real-time response under energy and computing constraints [53].

4.1.6. Agricultural Monitoring and Precision Farming

The use of UAVs in agricultural surveillance and precision farming is a key aspect of shifting
traditional farming toward more efficient, data-driven and sustainable approaches. UAVs are
used for important purposes like crop health monitoring, weeds, pest identification, moisture in
soil moisture, and high precision spraying that result in increased productivity and decreased
resource utilization [54]. UAVs have multispectral, thermal and hyperspectral sensors, and they
are capable of providing real-time, high-resolution data for precision agriculture on a field-
specific basis [55]. UAVs also facilitate information-driven decision-making with AI and IoT
integration, which helps farmers in quick decision-making in response to field conditions [56].
Furthermore, cost savings and higher operational efficiency can be guaranteed by the use of a
heterogeneous swarm of UAVs with optimized flight path planning [57]. With an increasing
agricultural demand, UAVs play an increasingly pivotal role in ensuring food security through the
security of food by convenient and precise resource management.

4.1.7. Infrastructure Inspection

Infrastructure monitoring is one of the major applications for UAVs, which replaces the manual
inspection of important structures due to its safety, efficiency, and cost-effectiveness, such as
bridges, power lines, and pavements [58]. UAVs use tower-based platforms without any pilot,
and these UAVs use cameras, LiDAR, and thermal sensors that collect resolution data that can
be used to perform structural examination in real time and post-processed structural
assessment [59]. Their importance is particularly significant in hard-to-access or dangerous
areas, where they can reduce risks for human intervention and improve the number and quality
of inspections. Advanced planning schemes, such as receding horizon control and energy-aware
path optimization, have been introduced for the full and efficient coverage of points of interest
for the critical infrastructure [60].

4.1.8. Logistics and Package Delivery

The UAVs impact the logistics and package delivery domain by providing exceptional
transformation to contemporary supply chains and last-mile deliveries due to their ease and
rapid delivery of goods and services. Autonomous Rotorcraft UAVs are used for parcel
distribution in remote and time-sensitive delivery regions, providing immense value in terms of
time and emission reduction [61]. Enhanced by image-based authentication, GPS tracking, and
AI-driven route optimization, UAVs ensure secure and efficient deliveries to the intended
recipients [62]. Furthermore, the implementation of Blockchain alongside IoT frameworks adds
a layer of trust to the logistical systems [63]. With the continuous advancement of UAV
technologies, their importance in the healthcare and humanitarian sectors will only increase
along with smart city infrastructural development.

4.1.9. Wildlife Conservation and Tracking

UAVs are extremely helpful in wildlife conservation as they allow for monitoring in non-intrusive
ways, use high-definition imaging of the locations of animal populations, and behaviours across
vast or inaccessible terrains. They are used to gather visual and sensor information from
different kinds of species through machine learning algorithms for real-time detection and
classification [64]. Such systems facilitate uninterrupted monitoring even in the forests or
mountainous regions where GPS based systems do not work [65]. With advancements in
thermal imaging and adaptive flight path algorithms, UAVs can detect wildlife activity at night or
in visually obscured environments. Moreover, the ecological data gathering UAVs employ is
useful in considerably reducing the ecological impact, enabling accurate species counts, habitat
assessment, and poaching prevention with minimal disturbance to animals [66]. These benefits
reinforce the importance of use of UAVs in modern strategies aimed at conserving biodiversity
and formulating policies regarding environmental changes.

4.1.10. Border Patrol and Security

The applications of UAVs in border patrol and security have emerged as a technologically
advanced strategy for automating national defense systems, especially for monitoring and
controlling large remote regions [67]. UAVs are deployed to track and respond to intrusions
through automatic detection using stochastic and optimization navigation algorithms that adjust
to environmental uncertainties. Integration of UAVs with AI techniques such as convolutional
neural networks (CNN), augments real-time situational intelligence and object recognition in
maritime and land surveillance. Ground-air cooperative frameworks enhance system
performance with strong communication and dynamic cluster formation and control during
challenging security conditions [68]. UAVs have also been integrated to assist in force protection
by providing real-time information relay and targeting during border confrontations [69].

4.1.11. Urban Planning and Traffic Monitoring

The employment of UAVs in urban planning and traffic monitoring is increasingly important for
the development of smart adaptive cities. UAVs provide high-resolution aerial data that enable
real-time detection of traffic congestion and urban mobility issues, allowing authorities to
dynamically adjust traffic signals and optimize flow [70]. Their mobility allows them to respond
to unanticipated congestion situations much faster than fixed-position systems, which enhances
their coverage and responsiveness [71]. UAVs are also employed for urban object detection and
mapping, which improves the quality of decision-making for infrastructure development,
zoning, and land use planning. Moreover, traffic behavior monitoring can easily be achieved
using mobile platforms with UAVs that can perform lightweight real-time analytics [72].

4.1.12. Military Target Acquisition and Battle Damage Assessment

The use of UAVs for military target acquisition and battle damage assessment has proved to
facilitate the improvement of accuracy, speed, and precision in combat operations and
minimizing collateral damage. UAVs are capable of accurate target identification, tracking, trike
imaging and analysis confirmation. UAVs serve as real-time surveillance platforms, capable of
identifying and tracking targets with high accuracy, then confirming the effectiveness of strikes
through post-operation imaging and analysis [73]. These systems support artillery and missile
units by correcting fire in real-time, significantly improving targeting precision and response
 efficiency. UAVs have also been employed to evaluate material and structural damage on the
battlefield, which assists commanders in determining whether the mission was successful or if
they need to re-engage [74]. With advanced technologies like AI, swarm-based UAV platforms
provide persistent surveillance, automated terrain and risk zone analysis, allowing high
operational awareness [75]. All these functionalities highlight subordination in UAVs strategy as
critical instruments for modern warfare available in real-time intelligence and combat
assessment.

4.2. What are the challenges in UAVs regarding 3D trajectory, high maneuverability of UAVs,
wireless communication performance, UAV control and trajectory optimization?

Despite the high potential in various industries, UAVs face complex and unique challenges which
create barriers for integrating them into sophisticated technological infrastructures. Their
energy consumption level relative to their operational range is one of the main problems
because current battery technology severely limits the flight range [76]. Researchers are trying
to solve the problem with hybrid and dynamic propulsion systems for charging strategies [77].
Communication reliability also poses a substantial challenge, particularly in swarm and
collaborative UAV systems where real-time data exchange is vital. Table 8 shows the different
challenges with application domains and functions in UAVs, and the figure 10 also shows the
different challenges in UAVs in pictorial form. The complexity of the interference, mobility, and
environmental adaptability makes the problem of creating reliable, secure, low-latency
communication networks more difficult [78]. Navigation is another significant challenge,
especially when there is no GPS signal available, unlike other satellite systems that can be used.
There are new advancements in sensor fusion and AI-based localization that may provide the
answer to this problem.
 Figure 10: Challenges in UAVs technology

Moreover, trajectory management and resource distribution need to be optimized for uncertain
and ever-changing conditions which requires automation algorithms sophisticated enough for
independent decision making [79]. Challenges like, jamming, spoofing, and cyberattacks are
serious vulnerability threats to UAV operations during defense and emergency response
situations [80]. There is a need for Machine learning-based and Federated learning-based
anomaly detection algorithms for secure communications. Collectively, these challenges
underscore the need for continued interdisciplinary research to ensure the safe, efficient, and
scalable deployment of UAV technologies. UAVs face different kinds of challenges, such as 3D
trajectory, high maneuverability, wireless communication, control, and optimization.
Optimization of 3D trajectories is complex, and the addition of energy consumption,
communications and urban and obstructed environments complicates matters further [81].
Recent studies have addressed these issues using various advanced methodologies.

Real-time optimization is further complicated by the complex characteristics of UAV motion and
their integration within wireless networks [82]. Studies mention the problem of reliable
 wireless communication, which becomes even more problematic for high altitude
communications due to a UAV's susceptibility to signal weakening and eavesdropping dangers
from above perpendicular sight. These vulnerabilities necessitate advanced trajectory and
beamforming strategies, including those enhanced with machine learning algorithms to
optimize UAV positioning and power allocation in real-time scenarios. Moreover, manipulations
and optimizations of multiuser UAVs in highly clustered scenarios, such as at multi-user levels
and during disaster recovery, become complicated with energy restrictions, rapid changes in the
surroundings, and controlled connectivity-shared lines [83].

Furthermore, UAV-assisted edge computing systems face challenges of resource allocation,

managing latency, and working with multiple perspectives, which emphasises the need for
unified optimization frameworks that combine trajectory design with computation and
communication parameters [84]. The complex nature of these problems illustrates the UAV
performance enhancement needs multi-domain optimization strategies where control,
communication, and application machine learning require optimization design to work together.

Table 8: Challenges and threats in UAVs.

Ref.# Challenge Application The Function and Description Potential threats in

name domain of UAVs Role in UAVs UAVs
[85] 3D Trajectory Urban Wireless Ensures LoS Uses model predictive Navigation failure due
Optimization Communication connectivity for control and dynamic to non-convex terrain
smart vehicles programming for UAV constraints
navigation in dense areas.
[86] Wireless Military and Civil Maintains secure Addresses broadcast Eavesdropping and
Communication Surveillance data transmission vulnerability with physical signal interception
Security layer security approaches.
[87] Beamforming & Secure Supports Uses deep reinforcement Data leakage due to
Trajectory Communications clustered users learning for beamforming passive eavesdropping
Control via aerial relay and trajectory tuning under
optimization unknown CSI.
[88] Connectivity Energy Delivery Coordinates Applies the CP-MARL Loss of swarm
Preservation & IoT multiple UAVs for algorithm to maintain inter- coordination
Communication energy and data UAV connectivity in RIS-
relay assisted systems.
[83] LoS Disaster Provides reliable Utilizes genetic and Disrupted
Maintenance in Recovery and terahertz link for heuristic algorithms to communication in
3D Emergency ground users optimize trajectory amidst obstructed areas
Environments Response urban blockages.
[84] Trajectory- Edge Computing Supports Surveys techniques for Resource bottlenecks
Control in MEC and IoT computing integrating energy, latency, and mobility issues
Systems services through and control in trajectory
 optimized flight planning.
paths
[89] QoS-aware Industrial IoT Balances Age of Combines genetic Outdated or delayed
Data Collection Networks Information (AoI) algorithms and convex data transmission
and throughput tools for time and path
optimization.
[90] Joint Smart Acts as a mobile Highlights challenges in Delayed task
Optimization in Infrastructure computing node multi-objective UAV control execution and
Aerial MEC and Mobile and computation reduced energy
Cloud offloading. efficiency
[91] Urban Path Cellular- Supports MTTP framework Flight risk from
Planning Connected UAV connectivity integrates DRL and dynamic urban
Networks while avoiding heuristic algorithms for features
obstacles obstacle-aware pathing.
[92] Energy-efficient Wireless Sensor Reduces motion Introduces a control-based Excessive energy use
3D Pathing Networks energy costs model with realistic UAV is causing mission
while dynamics. failure
maintaining
throughput

4.3. What is the Federated learning based optimal multi-UAV control strategies for energy-
efficient and accurate 3D reconstruction and high maneuverability of UAVs in large-scale
urban and disaster zones?

Federated Learning (FL) is a decentralized approach to machine learning in which several

devices, like IoT devices and UAVs train a common model collaboratively while retaining their
local data confidential [93]. Instead of sending raw data to a central server, each device trains
the model on its data locally and shares model parameters with a central server that aggregates
the updates to improve the global model. FL is very effective in improving the performance,
efficiency, and security of UAV systems. One of its primary benefits is privacy preservation, as FL
allows UAVs to collaboratively train models without transmitting sensitive raw data, which is
critical for surveillance, disaster response, and military operations [94]. Moreover, it decreases
communication overhead in bandwidth-constrained environments since only model updates
instead of voluminous imagery or sensor data are exchanged. Figure 11 represent the
integration of FL-based techniques with UAVs. FL increases effectiveness in multi-UAV systems
by promoting decentralized learning under different operating conditions and mission types,
which enhances the system's scalability and adaptability, thus improving the model's robustness
and generalization in heterogeneous fleets [95]. Below are the different techniques and
methods of FL which can overcome different challenges in UAVs.

Figure 11: Integration of FL-based techniques and method with UAVs to overcome challenges

4.3.1. Personalized Federated Deep Reinforcement Learning (PF-DRL) for 3D Trajectory

Optimization.

Personalized Federated Deep Reinforcement Learning (PF-DRL) has proven useful for addressing
the challenge of optimization of 3D trajectories in UAV networks within dynamic,
heterogeneous settings. The integration of federated learning with deep reinforcement learning
(RL) through PF-DRL enables each UAV to train on-board trajectory control models that capture
local observations while contributing to a globally shared model. This mitigates the
heterogeneity of data and preserves privacy [96]. Furthermore, under the constraints of limited
energy, connectivity maintenance, and urban obstacle avoidance, PF-DRL improves learning
efficiency through reward-shaping and experience replay, termed prioritized experience replay
[97]. PF-DRL also improves the energy efficiency and 3D mapping accuracy of swarm UAVs
operating in disaster and urban environments.
4.3.2. Secure Aggregation FL (SecAgg-FL) for Wireless Communication Security.

The Secure Aggregation Federated Learning model (SecAgg-FL) is especially important when
attempting to integrate security measures into wireless communications in UAV Networks since
it focuses on privacy during model updating in federated learning processes, individually
masking each model update. The classical Google's Secure Aggregation (SecAgg) protocol
employs masking techniques and secret sharing to prevent even the central aggregator from
accessing raw model updates, but recent innovations have aimed to improve its efficiency and
scalability [98]. Newer models such as Communication-Efficient Secure Aggregation (CESA) and
ACCESS-FL were developed to optimize the computational and communicative burden of
SecAgg with regards to shared secrets between UAV nodes, making them ideal for large-scale
aerial swarms. Moreover, some frameworks, such as Robust Federated Learning with Pairwise
Attention (RFLPA) combine secure aggregation with robust cosine similarity-based aggregation
to defend against model poisoning attacks in UAV networks [99]. These improved techniques of
SecAgg-FL ensure that UAVs can securely and collaboratively execute flight path optimization,
beamforming, and data collection in adversarial settings while sensitive information is secure,
thereby enhancing privacy, security, and resiliency in wireless UAV systems.

4.3.3. Federated Multi-Task Learning (FedMTL) for Beamforming & Trajectory Control.

Federated Multi-Task Learning (FedMTL), stands out as a promising solution to the challenges of
dual beamforming and trajectory control in UAV networks, enabling task-adaptive learning to
specific tasks that is decentralized and preserves user privacy. It allows sharing knowledge
among UAVs to perform tasks like optimal beam selection, signal transmission power
modification, and 3D flight path tracking without raw data exchange and overcomes the data-
sharing and privacy challenges [100]. Parameters of specific tasks and low-overhead secure
aggregation techniques guarantee model personalization for optimal performance by UAVs with
disparate communication and mobility capabilities. In UAV-augmented mobile edge computing
(MEC) systems, frameworks like FairLearn utilize federated deep reinforcement learning in
conjunction with multi-task learning to optimize power transmission, scheduling, and 3D spatial
trajectory timing, improving energy-efficient flight planning and beamforming precision [13].
Additionally, hierarchical multi-task federated learning (HFedMTL) reduces communications
overhead and accelerates convergence, lowering the feasibility threshold for swarm UAV
deployment in urban and disaster regions [101]. These models significantly enhance
connectivity maintenance, energy efficiency, and real-time adaptability for UAV-based
communication and sensing missions.

4.3.4. Hierarchical Federated Learning (HFL).

The challenge of connection preservation in UAV networks can be addressed using Hierarchical
Federated Learning (HFL) in which multi-level model aggregation is performed via
communication topology-dependent aggregation. Through HFL, UAVs are divided into clusters
and groups where each group contains a local aggregator that collects model updates from
subordinates and sends them to a central aggregator, which reduces communication and
increases reliability [102]. The authors of the study [103] developed over-the-air hierarchical
aggregation techniques combined with differential privacy preserving communication efficiency
along with confidentiality, through user privacy protection. HFL models such as UAV-assisted
two-tier hierarchical federated learning have significantly enhanced the reliable communication
preservation during the linkages between UAVs and ground devices by strategically optimizing
UAVs positioning and resources allocation, as well as relay resources controlling [104]. These
approaches ensure continual inclusion of UAV swarms into federated learning processes despite
dynamic node mobility, environmental disturbances.

4.3.5. Federated Deep Deterministic Policy Gradient (Fed-DDPG).

Federated Deep Deterministic Policy Gradient (Fed-DDPG) has emerged as one of the effective
frameworks for UAVs interfacing using federated reinforcement learning techniques to deal with
LoS communication challenges. Fed-DDPG enables cross-UAV coordination in advanced urban
multi-rotor UAVs to optimize trajectories for maintaining appropriate LoS, especially in dynamic
urban and disaster scenarios, without sharing sensitive sensor data and mission parameters
[105]. Global critic-local actor structures along with secured gradient aggregation techniques
are used to modulate the computation privacy trade-off which allows UAVs to actively
reposition in 3D space to maintain LoS links with ground nodes. Furthermore, robust variants
like Byzantine-robust Fed-DDPG introduce geometric median aggregation to protect against
malicious updates during learning, ensuring reliability even in partially compromised networks
[106]. Using Fed-DDPG for the coordination of multiple UAVs has shown to enhance precision
for controlling navigation during flight, reducing the frequency of collisions, and improving the
quality of communications relative to other vehicles within complex multi-agent environments
[107].

4.3.6. Federated Soft Actor-Critic (Fed-SAC).

Federated Soft Actor-Critic (Fed-SAC) is a sophisticated method that merges the concepts of
federated learning and deep reinforcement learning to address the trajectory control problem
in mobile edge computing (MEC) systems for UAVs. In Fed-SAC, communication is minimized as
UAVs collaboratively policy learn while preserving decentralized model training, thus guarding
data privacy. [108] proposed a constrained soft actor-critic (CSAC) algorithm where UAVS
optimize their trajectories under strict energy budgets as a constrained Markov Decision Process
(CMDP), enhancing mission effectiveness while enabling federated model updates. Additionally,
[109] presented a federated SAC framework that aims to jointly optimize UAVs’ trajectory
planning alongside SAC training hyperparameters to reduce flight energy expenditure within
dynamically MEC-shifted environments. Furthermore, [110] developed a multi-agent SAC
(MASAC) model which allows for the decentralized coordination of UAVs in route power
allocation and path control to optimize overall network spectral efficiency. These fed-SAC
innovations greatly augmented UAV responsiveness, resiliency, and energy consumption in
sophisticated MEC-integrated settings, ensuring dependable service provisioning and
computation offloading.

4.3.7. Federated Reinforcement Learning (FedRL) QoS-aware Data Collection.

Privacy preserving policy learning along with communication-efficient information exchange

rendered Federated Reinforcement Learning (FedRL) a strong technique for solving the
challenge of QoS-aware data collection in UAV networks. Different methods such as
Uncertainty-Aware FedRL (UA-FedRL) enhance heterogenous UAV fleets’ data accuracy and
energy efficiency by optimizing local training epoch and communication schedule in regards to
the uncertainty of the UAV’s observations [111]. Additionally, Other frameworks, like the one
proposed in Federated Double Deep Reinforcement Learning (FedDdrl), incorporate early client
termination and epoch-adaptive policies introducing lower latency and higher quality data
collection by UAVs operating under diverse conditions [112]. Further advancements of these
frameworks, such as in Federated Heterogeneous Q-Learning (FedHQL) enable collaboration
among UAVs with diverse capabilities and operational environments to formulate reliable and
timely optimized policies for contingent data collection and aggregation [113]. These models of
FedRL improve the UAV fleets’ QoS capability by reducing the operational bandwidth, data-loss,
and providing adaptive and resilient collection strategies for heavily populated urban and large-
scale disaster regions.

4.3.8. Model-Aided Federated QMIX (FedQMIX).

Framework for Model-Aided Federated QMIX (FedQMIX) is a newly proposed technique in

federated learning to address the problem of joint optimization in aerial Mobile Edge
Computing (MEC) systems for UAVs with enhanced efficiency and scalability in multi-UAV
collaboration. In the study of [114] federated QMIX is implemented to train multiple UAVs which
locally train QMIX-based multi-agent reinforcement learning models using simulated
environment models derived from real-world measurements like radio channels and IoT device
placements, measurably decreasing the extensive training data needed. This alternation
between simulation updates and federated training enables UAVs to optimize complex
objectives, including data harvesting, communication relay, and computation offloading, while
maintaining low latency and energy consumption. Techniques like model augmentation,
federated averaging with weighted updates, and dynamic environment simulation refinement
are employed to enhance convergence speed and adaptability. In contrast to conventional
MARL framework, FedQMIX is able to approximate three orders of magnitude real-world
experience while UAV swarms retain similar and superior optimization capabilities on
cooperative task performance. This strengthens the claim that FedQMIX is a critical approach
towards effective, adaptive, and robust solutions to the control and operation of UAVs in fast
changing aerial MEC settings.

4.3.9. Clustered Federated Learning (CFL).

The strategy of Clustered Federated Learning (CFL) alongside dividing UAVs into clusters
depending on environments and data distributions solve the planning challenge for UAV
networks, provides and personalized efficient model training. Genetic CFL is a method which
overrides hyperparameters in a certain cluster with Genetic Algorithms (GA) to increase
convergence and accuracy of models over IID-Non environments (Non Independent and
Identically Distributed Data), essential for rapidly changing urban landscapes [115]. Additionally,
FedAC introduces adaptive cluster fine-tuning allowing UAVs to adjust to complex urban path
dynamics and optimize trajectory decisions with the use of trajectory fine-tuning, thus enabling
more effectiveness with regards to urban path interpolation [116]. More advanced frameworks
like the Three-Stage Stackelberg Game-based CFL model formulate the interaction between
UAVs, cluster heads, and model owners as a hierarchical reinforcement learning problem,
optimizing resource allocation and path planning cooperatively in heterogeneous swarms [117].
These CFL frameworks ensure reliable, adaptable, and scalable path optimization for the UAV to
traverse dynamically modified routes adaptively adjust control locked in chronically complex
obstructed urban zones.

4.3.10. Asynchronous Federated Learning (Async-FL).

Asynchronous Federated Learning (Async-FL) accomplished the defined goal of energy-efficient

3D pathing in UAVs permitting availability and network conditions to dictate model update
processes, and minimizing idle time and energy waste. In contrast to traditional synchronous FL
where UAVs must wait for slow participants (stragglers), Async-FL enables continuous model
updates, significantly improving convergence speed and energy efficiency [118]. Further
improvements have been made to communication and computation burdens imposed on
resource limited UAVs by proposing model update majority voting and dynamic aggregation
threshold techniques. Furthermore, the Relaxed Asynchronous Heterogeneous Federated
Learning (Async-HFL) model presents the local aggregator hierarchical structure which allows
grouped UAVs to optimize their 3D primary flight paths based on local network topologies while
still upholding global model parity [119]. Additionally, the Dynamic Asynchronous Federated
Learning (DAFL) framework applies deep reinforcement learning for real-time estimation of
energy states and network parameters to autonomously select UAVs for aggregation, increasing
energy efficiency and accelerating convergence time for path planning [120]. These
advancements in Async-FL make it a highly scalable and energy-efficient solution for 3D UAV
trajectory control in large-scale, dynamic, and resource-constrained environments.

5. Conclusion and Future Work

The rapid advancement of UAVs has unlocked transformative applications across both military
and civilian domains, particularly in aerial reconnaissance, environmental monitoring, and 3D
reconstruction. These missions require reliable UAV control and stable wireless services are
required to provide fast-response operations within GPS-denied and emergency environments.
However, the combination of modern mobile network user's demands and UAV's 3D mobility
coupled with the need to concurrently optimize communication reliability, route, and UAV
control in short time frames is extremely problematic due to the ever-changing conditions of
wireless channels. With the growth of UAV networks towards autonomously cooperative
frameworks, the demand for resource-aware, flexible, efficient, and robust security policies and
control boundaries escalates rapidly. This paper provided a comprehensive review on UAV
imaging schemes, communication frameworks, channel models, and new UAV applications
focuses on these technologies along with the gap analysis of the most critical challenges in this
field. Federated Learning (FL) is one of such decentralized approaches which increase privacy
and decrease the need for communication by modifying control parameters in real-time
responsive to the conditions, providing robust solutions to UAVs control, communication and
optimization challenges.

Model training and deployment along with energy expenditure during long-endurance missions
should be dealt with on resource constrained UAVs. Therefore, FL frameworks that are
asynchronous and lightweight need to be developed. The integration of realistic 3D air-to-
ground and UAV-to-UAV channel dynamics into the federated learning processes will enhance
the trajectory prediction accuracy along with the communicational reliability.

Reference

1. Laghari, A.A., et al., Unmanned aerial vehicles: A review. Cognitive Robotics, 2023. 3: p.
8-22.
2. Mohsan, S.A.H., et al., Unmanned aerial vehicles (UAVs): Practical aspects, applications,
open challenges, security issues, and future trends. Intelligent service robotics, 2023.
16(1): p. 109-137.
3. Mohsan, S.A.H., et al., Towards the unmanned aerial vehicles (UAVs): A comprehensive
review. Drones, 2022. 6(6): p. 147.
4. Keyvanfar, A. and A. Shafaghat, Emerging dimensions of unmanned aerial vehicle’s (uav)
3d reconstruction modeling and photogrammetry in architecture and construction
management. ACE: Arquitectura, Ciudad y Entorno, 2022.
5. Deliry, S.I. and U. Avdan, Accuracy of unmanned aerial systems photogrammetry and
structure from motion in surveying and mapping: a review. Journal of the Indian Society
of Remote Sensing, 2021. 49(8): p. 1997-2017.
6. Javaid, S., et al., Communication and control in collaborative UAVs: Recent advances and
future trends. IEEE Transactions on Intelligent Transportation Systems, 2023. 24(6): p.
5719-5739.
7. Li, B., et al., 3D trajectory optimization for energy-efficient UAV communication: A
control design perspective. IEEE Transactions on Wireless Communications, 2021. 21(6):
p. 4579-4593.
8. Carvajal-Rodríguez, J., et al., 3D placement optimization in UAV-enabled
communications: A systematic mapping study. IEEE Open Journal of Vehicular
Technology, 2024.
9. Abbas, N., et al., A survey: Future smart cities based on advance control of Unmanned
Aerial Vehicles (UAVs). Applied Sciences, 2023. 13(17): p. 9881.
10. Iqbal, M.M., et al., Motion planning of UAV swarm: Recent challenges and approaches.
Aeronautics-New Advances, 2022.
11. Bian, J., C. Shen, and J. Xu, Joint Client Assignment and UAV Route Planning for Indirect-
Communication Federated Learning. ArXiv, 2023. abs/2304.10744.
12. Tan, Z., et al., A FL-Based Radio Map Reconstruction Approach for UAV-Aided Wireless
Networks. Electronics, 2023.
13. Karmakar, R., G. Kaddoum, and O. Akhrif, A Novel Federated Learning-Based Smart
Power and 3D Trajectory Control for Fairness Optimization in Secure UAV-Assisted MEC
Services. IEEE Transactions on Mobile Computing, 2024. 23: p. 4832-4848.
14. Telli, K., et al., A comprehensive review of recent research trends on unmanned aerial
vehicles (uavs). Systems, 2023. 11(8): p. 400.
15. Politi, E., et al., Enabling Technologies for the Navigation and Communication of UAS
Operating in the Context of BVLOS. Electronics, 2024. 13(2): p. 340.
16. Hussain, A., et al., Computing Challenges of UAV Networks: A Comprehensive Survey.
Computers, Materials & Continua, 2024. 81(2).
17. Alghamdi, Y., A. Munir, and H. La, Architecture, Classification, and Applications of
Contemporary Unmanned Aerial Vehicles. IEEE Consumer Electronics Magazine, 2021.
10: p. 9-20.
18. Isgandarova, S.A., Relationship between the automatic control system and its integration
on unmanned aircraft vehicle. Bulletin of the National Engineering Academy of the
Republic of Kazakhstan, 2024.
19. Gura, D., et al., Automated system for dispatching the movement of unmanned aerial
vehicles with a distributed survey of flight tasks. Journal of Intelligent Systems, 2021. 30:
p. 728-738.
20. Zhou, Y., et al., Research on UAV Multi-Source Information Monitoring System for Indoor
Space. Journal of Physics: Conference Series, 2024.
21. Mohammed, A., et al., Challenges of Ground Control System in Ensuring Safe Flights for
Unmanned Aerial Vehicles. MEKATRONIKA, 2022.
22. Kliushnikov, I., V. Kharchenko, and H. Fesenko, An Unmanned Aerial Vehicle as a Multi-
State System. 2022 IEEE 16th International Conference on Advanced Trends in
Radioelectronics, Telecommunications and Computer Engineering (TCSET), 2022: p. 291-
296.
23. Ulloa, C.C., et al., Thermal, Multispectral, and RGB Vision Systems Analysis for Victim
Detection in SAR Robotics. Applied Sciences, 2024.
24. Lin, Y., et al., ThermalNeRF: Thermal Radiance Fields. 2024 IEEE International Conference
on Computational Photography (ICCP), 2024: p. 1-12.
25. Feng, K., et al., Visible and NIR microscopic hyperspectrum reconstruction from RGB
images with deep convolutional neural networks. Optics express, 2024. 32 3: p. 4400-
4412.
26. Song, M., et al., Air-to-Ground Large-Scale Channel Characterization by Ray Tracing. IEEE
Access, 2022. 10: p. 125930-125941.
27. Rostamikafaki, Z., F. Chan, and C. D’amours, 5G New Radio Signal Propagation and
Ground-to-Air Channel Modeling at 3.565 GHz Based on Extensive Measurements. IEEE
Access, 2024. 12: p. 143975-143984.
28. Cui, Z., S. Pollin, and Z. Zhong, Probabilistic Two-Ray Model for Air-to-Air Channel in
Urban Drone Networks. Proceedings of the Eighth Workshop on Micro Aerial Vehicle
Networks, Systems, and Applications, 2022.
29. Li, N., et al., The Impact of UAV Jittering at Different Altitudes on Air-to-Ground Wireless
Channel. 2023 IEEE 6th International Conference on Electronics and Communication
Engineering (ICECE), 2023: p. 95-99.
30. Lv, Y., et al., Ultra Wideband Channel Measurement and Analysis for Low Altitude UAV
Air-to-Ground Scenario. 2021 13th International Symposium on Antennas, Propagation
and EM Theory (ISAPE), 2021. Volume1: p. 1-3.
31. Jiang, X., et al., Outage Analysis of UAV-Aided Networks With Underlaid Ambient
Backscatter Communications. IEEE Transactions on Wireless Communications, 2023. 22:
p. 7492-7505.
32. Ferrari, E., et al., Automatic Generation of High-Resolution Thermal 3D Building Envelope
Models exploiting UAV Imagery. The International Archives of the Photogrammetry,
Remote Sensing and Spatial Information Sciences, 2024.
33. Usmani, K., et al., 3D object detection through fog and occlusion: passive integral
imaging vs active (LiDAR) sensing. Optics express, 2022. 31 1: p. 479-491.
34. Al-Gburi, A., et al., Channel Estimation for UAV Communication Systems Using Deep
Neural Networks. 2022.
35. Polus, R. and C. D’amours, Cascaded Shadowed UAV-to-Ground Channels. 2024 IEEE 10th
World Forum on Internet of Things (WF-IoT), 2024: p. 918-923.
36. Ismail, M., et al., Line-of-Sight-Based Coordinated Channel Resource Allocation
Management in UAV-Assisted Vehicular Ad Hoc Networks. IEEE Access, 2024. 12: p.
25245-25253.
37. Dehghan, M. and E. Khosravian, A Review of Cognitive UAVs: AI-Driven Situation
Awareness for Enhanced Operations. 2024.
38. Ergunsah, S. and S. Kosunalp, An Overview of Unmanned Aerial Vehicles Based
Environmental Applications. Uluslararası Yönetim Bilişim Sistemleri ve Bilgisayar Bilimleri
Dergisi, 2022.
39. Ervural, B.C., Determination of Strategic Location of UAV Stations. Lecture Notes in
Mechanical Engineering, 2021.
40. Munir, A., A. Aved, and E. Blasch, Situational awareness: techniques, challenges, and
prospects. AI, 2022. 3(1): p. 55-77.
41. Jayaweera, H. and S. Hanoun, UAV Path Planning for Reconnaissance and Look-Ahead
Coverage Support for Mobile Ground Vehicles. Sensors (Basel, Switzerland), 2021. 21.
42. Fascista, A., Toward integrated large-scale environmental monitoring using
WSN/UAV/Crowdsensing: A review of applications, signal processing, and future
perspectives. Sensors, 2022. 22(5): p. 1824.
43. Namdev, M., J. Dholaria, and M. Rathi, Harnessing Transformative Potential of UAV-
Based AI-Powered Data Analysis for Emission Monitoring. 2024 6th International
Conference on Computational Intelligence and Networks (CINE), 2024: p. 1-6.
44. Hu, D. and J. Minner, UAVs and 3D City Modeling to Aid Urban Planning and Historic
Preservation: A Systematic Review. Remote Sensing, 2023. 15(23): p. 5507.
45. Gupta, A. and X. Fernando, Simultaneous localization and mapping (slam) and data
fusion in unmanned aerial vehicles: Recent advances and challenges. Drones, 2022. 6(4):
p. 85.
46. Maboudi, M., et al., A review on viewpoints and path planning for UAV-based 3-D
reconstruction. IEEE Journal of Selected Topics in Applied Earth Observations and
Remote Sensing, 2023. 16: p. 5026-5048.
47. Anagnostis, I., P. Kotzanikolaou, and C. Douligeris, Understanding and Securing the Risks
of Unmanned Aerial Vehicle Services. IEEE Access, 2025.
48. Santos, N., et al., Automatic Detection of Civilian and Military Personnel in
Reconnaissance Missions using a UAV. 2023 IEEE International Conference on
Autonomous Robot Systems and Competitions (ICARSC), 2023: p. 157-162.
49. Fang, Z. and A. Savkin, Strategies for Optimized UAV Surveillance in Various Tasks and
Scenarios: A Review. Drones, 2024.
50. Gnatyuk, S., et al., FORMATION OF A DATASET OF CRYPTOGRAPHIC ALGORITHMS FOR
ENSURING DATA CONFIDENTIALITY TRANSFERRED FROM RECONNAISSANCE AND
SEARCH UAV. Cybersecurity: Education, Science, Technique, 2023.
51. Oh, D.-E. and J. Han, Smart Search System of Autonomous Flight UAVs for Disaster
Rescue. Sensors (Basel, Switzerland), 2021. 21.
52. Abdellatif, A., et al., PDSR: Efficient UAV Deployment for Swift and Accurate Post-Disaster
Search and Rescue. ArXiv, 2024. abs/2410.22982.
53. Sun, G., et al., Joint Task Offloading and Resource Allocation in Aerial-Terrestrial UAV
Networks With Edge and Fog Computing for Post-Disaster Rescue. IEEE Transactions on
Mobile Computing, 2023. 23: p. 8582-8600.
54. Vashishth, T.K., et al., Integration of Unmanned Aerial Vehicles (UAVs) and IoT for crop
monitoring and spraying, in Internet of Things applications and technology. 2024,
Auerbach Publications. p. 95-117.
55. Agrawal, J. and M. Arafat, Transforming Farming: A Review of AI-Powered UAV
Technologies in Precision Agriculture. Drones, 2024.
56. Ganesan, T., N. Jayarajan, and P. Sureshkumar, Review on the Real-time Implementation
of IoT-enabled UAV in Precision Agriculture and the Overview of Collision Avoidance
Strategies. Philippine Journal of Science, 2023.
57. Mukhamediev, R., et al., Coverage Path Planning Optimization of Heterogeneous UAVs
Group for Precision Agriculture. IEEE Access, 2023. 11: p. 5789-5803.
58. Villarino, A., et al., UAV Applications for Monitoring and Management of Civil
Infrastructures. Infrastructures, 2025. 10(5): p. 106.
59. Zhang, Z. and L. Zhu, A review on unmanned aerial vehicle remote sensing: Platforms,
sensors, data processing methods, and applications. Drones, 2023. 7(6): p. 398.
60. Wang, Y., et al., An Energy-Aware UAVs Path Coverage for Critical Infrastructure
Inspections. 2024 Twelfth International Conference on Advanced Cloud and Big Data
(CBD), 2024: p. 222-227.
61. Rodrigues, T., et al., Drone flight data reveal energy and greenhouse gas emissions
savings for very small package delivery. Patterns, 2022. 3.
62. Mathi, S., et al., Enhanced UAV Delivery System with Image-Based Authentication. 2024
4th International Conference on Sustainable Expert Systems (ICSES), 2024: p. 465-472.
63. Chen, C.-L., et al., An IoT and blockchain based logistics application of UAV. Multimedia
Tools and Applications, 2023. 83: p. 655-684.
64. Tripathi, R.N., et al., Conservation in action: Cost-effective UAVs and real-time detection
of the globally threatened swamp deer (Rucervus duvaucelii). Ecol. Informatics, 2024. 85:
p. 102913.
65. Qiao, H. and X. Bai, A Wildlife Tracking Scheme in Forest Area Based on UAV and Sensors.
Proceedings of the 2021 2nd International Conference on Computing, Networks and
Internet of Things, 2021.
66. Wich, S.A., et al., Drones for conservation. Conservation technology, 2021. 35.
67. Tyshchuk, V.V., A Review of Legal Regulation Regarding the Use of Unmanned Aerial
Vehicles for Border Security and the Impact of Global Technologies. International
Comparative Jurisprudence, 2024. 10(1): p. 61-81.
68. Shao, Y., et al., Clustering Algorithm Based on the Ground-Air Cooperative Architecture in
Border Patrol Scenarios. Electronics, 2022.
69. Reznik, О., et al., Analysis of Experience of Using Small UAVs in the Battlefield: Training
Enhancement Perspectives. Educological discourse, 2024.
70. Bhattarai, S., Drone-Enabled Aerial Monitoring for Dynamic Traffic Control Using Multi-
Modal Data Synthesis. Journal of Robotic Process Automation, AI Integration, and
Workflow Optimization, 2024. 9(12): p. 23-35.
71. Hu, Y., et al., Improving inadequate coverage of an unmanned aerial vehicle base
station: mobility strategies and delay-tolerant service requests. Telecommunication
Systems, 2022. 81(4): p. 527-537.
72. Lu, Y.-R. and D. Sun, Development of Real-Time Unmanned Aerial Vehicle Urban Object
Detection System with Federated Learning. J. Aerosp. Inf. Syst., 2024. 21: p. 547-553.
73. Lavers, C., Design in Engineering: An Evaluation of Civilian and Military Unmanned Aerial
Vehicle Platforms, Considering Smart Sensing with Ethical Design to Embody Mitigation
Against Asymmetric Hostile Actor Exploitation. 2021.
74. Cherednychenko, O., et al., Analysis of the experience of the combat use of unmanned
aerial vehicles against anti-aircraft missile systems in the military conflict in Nagorno-
Karabakh. Communication, informatization and cybersecurity systems and technologies,
2022.
75. Wang, E., et al., MADRL-based UAV swarm non-cooperative game under incomplete
information. Chinese Journal of Aeronautics, 2024.
76. Elmeseiry, N., N. Alshaer, and T. Ismail, A detailed survey and future directions of
unmanned aerial vehicles (uavs) with potential applications. Aerospace, 2021. 8(12): p.
363.
77. Basil, N., et al., Systematic Review of Unmanned Aerial Vehicles Control: Challenges,
Solutions, and Meta-Heuristic Optimization. International Journal of Robotics and
Control Systems, 2024.
78. Fratu, O., et al., RF Propagation and Interferences Challenges for UAVs Swarm Detection
and Positioning. 2022 25th International Symposium on Wireless Personal Multimedia
Communications (WPMC), 2022: p. 501-504.
79. Noorwali, A., M. Javed, and M.Z. Khan, Efficient UAV Communications: Recent Trends
and Challenges. Computers, Materials & Continua, 2021.
80. Oruc, A., Potential cyber threats, vulnerabilities, and protections of unmanned vehicles.
Drone Systems and Applications, 2022. 10(1): p. 51-58.
81. Gu, X. and G. Zhang, A survey on UAV-assisted wireless communications: Recent
advances and future trends. Computer Communications, 2023. 208: p. 44-78.
82. Basharat, M., et al., Resource optimization in UAV‐assisted wireless networks—A
comprehensive survey. Transactions on Emerging Telecommunications Technologies,
2022. 33(7): p. e4464.
83. Dabiri, M., et al., Optimizing UAV Trajectory for Emergency Response Operations under
Real 3D Environments: Integrating Priority Levels and LoS Constraints. 2024.
84. Adnan, M.H., Z. Zukarnain, and O.A. Amodu, Fundamental design aspects of UAV-
enabled MEC systems: A review on models, challenges, and future opportunities.
Comput. Sci. Rev., 2024. 51: p. 100615.
85. Eskandari, M., et al., Model predictive control-based 3D navigation of a RIS-equipped
UAV for LoS wireless communication with a ground intelligent vehicle. IEEE Transactions
on Intelligent Vehicles, 2022. 8(3): p. 2371-2384.
86. Li, J., et al., Wireless secure communication involving UAV: an overview of physical layer
security. MATEC Web of Conferences, 2021.
87. Abdalla, A., A. Behfarnia, and V. Marojevic, UAV Trajectory and Multi-User Beamforming
Optimization for Clustered Users Against Passive Eavesdropping Attacks With Unknown
CSI. IEEE Transactions on Vehicular Technology, 2023. 72: p. 14426-14442.
88. Chen, L. and Z. Wang, Multi-UAV Trajectory Planning For RIS-assisted SWIPT system
under connectivity preservation. Computer Networks, 2024.
89. Zhang, Y. and D. Mishra, Age of Information Minimization in QoS-Aware UAV-assisted
Wireless-Powered Data Collection. 2024 IEEE 35th International Symposium on Personal,
Indoor and Mobile Radio Communications (PIMRC), 2024: p. 1-6.
90. Song, Z., et al., A comprehensive survey on aerial mobile edge computing: Challenges,
state-of-the-art, and future directions. Comput. Commun., 2022. 191: p. 233-256.
91. Luo, X., et al., Multi-Tier 3D Trajectory Planning for Cellular-Connected UAVs in Complex
Urban Environments. Symmetry, 2023. 15: p. 1628.
92. Li, B., et al., 3D Trajectory Optimization for Energy-Efficient UAV Communication: A
Control Design Perspective. IEEE Transactions on Wireless Communications, 2021. PP: p.
1-1.
93. Nguyen, D.C., et al., Federated learning for internet of things: A comprehensive survey.
IEEE Communications Surveys & Tutorials, 2021. 23(3): p. 1622-1658.
94. Al Farsi, A.S., et al., Privacy and Security Challenges in Federated Learning for UAV
Systems: A Comprehensive Review. SECURITY AND PRIVACY, 2024.
95. Din, I.U., et al., Federated Learning for Trust Enhancement in UAV-Enabled IoT Networks:
A Unified Approach. IEEE Internet of Things Journal, 2025.
96. Song, Z., et al., Personalized Federated Deep Reinforcement Learning-based Trajectory
Optimization for Multi-UAV Assisted Edge Computing. 2023 IEEE/CIC International
Conference on Communications in China (ICCC), 2023: p. 1-6.
97. Gao, Y., et al., Federated deep reinforcement learning based trajectory design for UAV-
assisted networks with mobile ground devices. Scientific Reports, 2024. 14.
98. Nazemi, N., et al., Boosting Communication Efficiency of Federated Learning’s Secure
Aggregation. 2024 54th Annual IEEE/IFIP International Conference on Dependable
Systems and Networks - Supplemental Volume (DSN-S), 2024: p. 157-158.
99. Mai, P., R. Yan, and Y. Pang, RFLPA: A Robust Federated Learning Framework against
Poisoning Attacks with Secure Aggregation. ArXiv, 2024. abs/2405.15182.
100. Sen, P. and C. Borcea, FedMTL: Privacy-Preserving Federated Multi-Task Learning. 2024:
p. 1993-2002.
101. Yi, X., et al., HFedMTL: Hierarchical Federated Multi-Task Learning. 2022 IEEE 33rd
Annual International Symposium on Personal, Indoor and Mobile Radio Communications
(PIMRC), 2022: p. 1-6.
102. Xu, J., et al., Federated Learning Powered Semantic Communication for UAV Swarm
Cooperation. IEEE Wireless Communications, 2024. 31: p. 140-146.
103. Wang, Z., A. Grootveld, and M. Gursoy, Hierarchical Over-the-Air Federated Learning
with Differential Privacy. Proceedings of the 2023 ACM Workshop on Wireless Security
and Machine Learning, 2023.
104. Wang, T., et al., UAV Swarm-Assisted Two-Tier Hierarchical Federated Learning. IEEE
Transactions on Network Science and Engineering, 2024. 11: p. 943-956.
105. Adeogun, R., Federated Deep Deterministic Policy Gradient for Power Control in 6G in-X
Subnetworks. 2024 14th International Conference on Electrical Engineering (ICEENG),
2024: p. 199-201.
106. Lin, Q. and Q. Ling, Byzantine-Robust Federated Deep Deterministic Policy Gradient.
ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal
Processing (ICASSP), 2022: p. 4013-4017.
107. Pramuk, M., et al., Goal Driven Multi-Robot Navigation in Simulated Environments with
Federated Deep Reinforcement Learning. 2024 9th International Conference on Control
and Robotics Engineering (ICCRE), 2024: p. 114-120.
108. Zhou, X., et al., Constrained Soft Actor-Critic for Energy-Aware Trajectory Design in UAV-
Aided IoT Networks. IEEE Wireless Communications Letters, 2022. 11: p. 1414-1418.
109. Zhang, C., Y. Liu, and Z. Zhang, A Deep Reinforcement Learning Approach for Federated
Learning Optimization with UAV Trajectory Planning. 2023 IEEE 34th Annual
International Symposium on Personal, Indoor and Mobile Radio Communications
(PIMRC), 2023: p. 1-7.
110. Qin, Y., et al., Deep Reinforcement Learning Based Resource Allocation and Trajectory
Planning in Integrated Sensing and Communications UAV Network. IEEE Transactions on
Wireless Communications, 2023. 22: p. 8158-8169.
111. Sagar, A., et al., Uncertainty-Aware Federated Reinforcement Learning for Optimizing
Accuracy and Energy in Heterogeneous Industrial IoT. Applied Sciences, 2024.
112. Wong, Y., et al., FedDdrl: Federated Double Deep Reinforcement Learning for
Heterogeneous IoT with Adaptive Early Client Termination and Local Epoch Adjustment.
Sensors (Basel, Switzerland), 2023. 23.
113. Fan, F.X., et al., FedHQL: Federated Heterogeneous Q-Learning. ArXiv, 2023.
abs/2301.11135.
114. Chen, J., et al., Model-Aided Federated Reinforcement Learning for Multi-UAV Trajectory
Planning in IoT Networks. 2023 IEEE Globecom Workshops (GC Wkshps), 2023: p. 818-
823.
115. Agrawal, S., et al., Genetic CFL: Hyperparameter Optimization in Clustered Federated
Learning. Computational Intelligence and Neuroscience, 2021. 2021.
116. Zhang, Y.-X., et al., FedAC: An Adaptive Clustered Federated Learning Framework for
Heterogeneous Data. ArXiv, 2024. abs/2403.16460.
117. He, W., et al., Three-Stage Stackelberg Game Enabled Clustered Federated Learning in
Heterogeneous UAV Swarms. IEEE Transactions on Vehicular Technology, 2023. 72: p.
9366-9380.
118. Jang, S.-L. and H.-K. Lim, AsyncFL: Asynchronous Federated Learning Using Majority
Voting with Quantized Model Updates (Student Abstract). 2022: p. 12975-12976.
119. Yu, X., et al., Async-HFL: Efficient and Robust Asynchronous Federated Learning in
Hierarchical IoT Networks. Proceedings of the 8th ACM/IEEE Conference on Internet of
Things Design and Implementation, 2023.
120. Xu, G., et al., Energy-Efficient Dynamic Asynchronous Federated Learning in Mobile Edge
Computing Networks. ICC 2023 - IEEE International Conference on Communications,
2023: p. 160-165.