UAV platform for special operation
1st Ing. Jan Klouda

Theoretical and Experimental Electrical Engineering
Brno University of Technology UTEE

Brno, Czech Republic
xkloud04@vutbr.cz

2nd doc. Ing. Petr Marcoň Ph.D.
Theoretical and Experimental Electrical Engineering

Brno University of Technology UTEE
Brno, Czech Republic

marcon@vut.cz

Abstract—Unmanned Aerial Vehicles (UAVs) are widely used
in industrial inspection, agriculture, and security. This paper
introduces a universal UAV platform designed for modular adapt-
ability, extended flight endurance, and autonomous navigation.
The hardware architecture includes a lightweight carbon fiber
frame, high-efficiency propulsion, and a flexible power system,
ensuring reliable performance across various missions.

The software framework, built on ROS 2 middleware, enables
seamless integration of SLAM-based navigation, visual odometry,
and deep learning-based object detection. The UAV combines
GNSS, INS, and LiDAR-based SLAM for precise localization,
even in GNSS-denied environments. Equipped with multi-sensor
payloads such as LiDAR, RGB-D cameras, and thermal imaging,
this platform supports applications ranging from precision agri-
culture to infrastructure inspection, paving the way for advanced
autonomous UAV operations.

Index Terms—UAV, autonomous navigation, modular drone
platform, SLAM, visual odometry, GNSS-denied environments,
deep learning, ROS 2, LiDAR, trajectory planning, multi-sensor
integration, real-time processing, obstacle avoidance

I. INTRODUCTION

Unmanned aerial vehicles (UAVs) have become an impo-
ratnt tool in many industries, including industrial inspection,
precision agriculture, and security operations [1]. The develop-
ment of modular UAV platforms enables adaptive integration
of different sensors and efficient execution of diverse missions
[2]. With increasing demands for longer flight times, higher
payloads and flexible configuration options, there is a need
for versatile UAV platforms that can handle a wide range of
applications without compromising performance [3].

This paper presents the design of a versatile UAV platform
that enables easy integration of different payloads, increases
operational efficiency and offers a modular architecture adapt-
able to different deployment scenarios. Emphasis is placed on
flight time optimization, reliability and flexible connectivity to
different control systems, making this concept ideal for special
operations in the industrial, security and agricultural sectors
[4].

II. PLATFORM PHILOSOPHY

The main goal of the proposed UAV platform is to create
a versatile and highly adaptable drone that allows rapid de-
ployment in different types of missions without the need for
extensive hardware or software modifications [5].

Figure 1 presents a block diagram of the proposed universal
platform. Several abbreviations appear in the figure: FCC

Fig. 1. Block diagram of the platform

(Flight control computer), GNSS (Global Navigation Satellite
System), ESC (Electronic speed controller), IMU (Inertial
measurement unit), RGB-D (camera that provides both depth
(D) and color (RGB) data), LiDAR (Light Detection And
Ranging).

Modern UAV applications require modular design, long
flight endurance, scalability, reliability, and high levels of
autonomy. These aspects ensure that UAVs can be effectively
used in areas such as industrial inspection, precision agricul-
ture or security operations.

Modularity plays a vital role in ensuring the flexibility of the
drone. The ability to easily swap sensor modules allows the
UAV to be quickly adapted to specific mission requirements.
The platform supports a wide range of sensors, including
LiDAR systems for detailed 3D terrain mapping, multispectral
cameras for vegetation analysis, thermal cameras for thermal
anomaly detection and HD cameras for detailed infrastructure
inspections [6], [7].

In addition to the sensors, the design of the UAV itself

241DOI 10.13164/eeict.2025.241



is modular. This means that individual parts, such as bat-
tery units, motors or communication systems, can be easily
replaced or upgraded according to current requirements. In
this way, the platform can be adapted to different application
scenarios without having to develop a completely new solution
for each case.

An important factor in the effectiveness of a UAV is its
endurance in the air. Extended flight time has been achieved
by optimising aerodynamics, using high-efficiency engines and
implementing advanced energy management. The drone uses a
modular battery system that allows a choice between a higher
capacity configuration for long duration missions and a lighter
variant for greater manoeuvrability [8].

Flight endurance can be theoretically expressed by the rela-
tionship between the available energy and the average energy
consumption of the UAV. When the system was optimized,
flight times in the range of 40 to 60 minutes per charge were
achieved, allowing for effective coverage of large areas in a
single deployment. The total flight time Tf of a UAV can be
estimated using the following equation:

Tf =
ηPb

Pd
(1)

where:
• Tf is the total flight endurance (s),
• η is the overall efficiency of the UAV power system

(dimensionless),
• Pb is the available battery energy (Wh),
• Pd is the average power consumption during flight (W).
The available battery energy is calculated as:

Pb = U · C (2)

where:
• U is the nominal battery voltage (V),
• C is the battery capacity (Ah).
The average power consumption during flight can be esti-

mated as:

Pd =
n∑

i=1

Pi (3)

where:
• Pi is the power consumed by the i-th subsystem of the

UAV (W),
• n is the total number of power-consuming subsystems

(e.g., motors, electronics, sensors).
Another requirements of modern UAV systems is their abil-

ity to grow and adapt to technological advances. This platform
has been designed to allow easy integration of new sensors,
computing units or communication modules without the need
for major design changes. Support for multiple voltage levels
allows users to choose between higher battery capacity for
longer flight or lower weight for better maneuverability [9].

Thanks to the modular architecture, the platform can be
extended with advanced navigation technologies such as artifi-
cial intelligence systems for autonomous data analysis or high-
speed transmission modules for real-time communication. This
ensures that the UAV remains competitive in the long term and
ready for future innovations.

Reliability is an important factor when deploying UAVs
in mission-critical applications. The drone’s design has been
optimized to withstand harsh weather conditions, mechanical
stresses and long-term operation. The redundancy of key
components, including power circuits and controllers, ensures
safe operation even in the event of a system failure [10].

In addition to hardware measures, reliability has also been
improved by implementing advanced algorithms for fault
detection and predictive maintenance. The drone is able to
monitor its status in real time and alert the operator of potential
problems before they affect its performance or flight safety.

Modern UAV systems increasingly rely on autonomous
capabilities to enable more efficient and safer operations. This
drone is equipped with advanced navigation technologies,
including simultaneous localization and mapping (SLAM),
visual odometry and inertial navigation systems. These tech-
nologies allow it to operate in environments where GNSS
signals are not available [11].

In addition to navigation functions, the platform is equipped
with algorithms for autonomous obstacle avoidance and adap-
tive flight control. By using artificial intelligence, the drone is
able to analyze the surrounding environment in real time and
optimize its flight plan according to the current conditions.
This enables its deployment in complex missions such as
autonomous infrastructure inspection, monitoring changes in
terrain or reconnaissance operations in difficult conditions
[11].

The overall philosophy of this UAV platform is to combine
high flexibility, efficiency and reliability, making it an ideal
solution for a wide range of applications. Its modular design,
long flight endurance and scalability ensure that it will be able
to meet the requirements of both current and future users.

III. NAVIGATION OPTIONS

The navigation of a UAV is a critical factor affecting its
ability to fly autonomously, accuracy in mission execution,
and overall reliability in challenging environments. This UAV
platform uses a combination of several advanced naviga-
tion methods that allow operation in both open terrain with
available GNSS signal and in environments where GNSS is
completely unavailable or heavily jammed. GNSS navigation
is the foundation, complemented by visual odometry, inertial
navigation and SLAM (Simultaneous Localization and Map-
ping), which enables robust navigation even in environments
without satellite signal [12].

A. Visual odometry - Image-based navigation

Visual odometry (VO) is a technology for UAV operations in
environments where GNSS is not available, such as industrial
halls, dense urban development or underground spaces [13].

242



This system uses a camera or an array of cameras to track
visual features of the environment and calculates the trajectory
of the UAV based on their movement between images.

The basis of visual odometry is the extraction of key points
from the images acquired by the cameras. Algorithms such
as ORB (Oriented FAST and Rotated BRIEF), SIFT (Scale-
Invariant Feature Transform), or AKAZE (Accelerated-KAZE)
identify specific visual features that are then tracked in the
image sequence [14]. The movement of these points between
frames can be mathematically expressed by a transformation
matrix:

The movement of these points between frames in visual
odometry can be mathematically expressed using the following
transformation matrix:

Tk+1 = K ·
(
Rk+1,k − tk+1,k · nT

)
·K−1 (4)

where:
• Tk+1 is the transformation matrix between frames,
• K is the camera intrinsic matrix,
• Rk+1,k is the rotation matrix between frames,
• tk+1,k is the translation vector,
• n is the normal vector of the plane.
The drone can use both monocular visual odometry, which

works with only one camera to estimate the scale of motion,
and stereoscopic visual odometry, which uses a pair of cameras
to create a depth map of the environment [15].

An important enhancement to visual odometry is optical
flow, which analyzes the movement of pixels between frames
and helps determine the speed of the UAV. This approach
is particularly advantageous in situations where the drone is
moving over structured textures such as vegetation or urban
development [12].

B. Inertial Navigation System (INS) - Stabilization and motion
tracking

The Inertial Navigation System (INS) is element of UAV
navigation that enables the determination of changes in posi-
tion and orientation in space without the need for external ref-
erences [16]. The system works with data from accelerometers,
gyroscopes and magnetometers that sense the acceleration,
angular velocity and orientation of the UAV relative to the
Earth’s magnetic field [17].

The motion of a UAV based on Inertial Navigation Sys-
tem (INS) measurements can be described by the following
equations:

v(t) = v0 +

∫ t

0

a(τ)dτ (5)

p(t) = p0 +

∫ t

0

v(τ)dτ (6)

where:
• v(t) is the velocity of the UAV at time t,
• a(τ) is the acceleration measured by the INS,
• p(t) is the position of the UAV,

• p0 and v0 are the initial position and velocity.

INS is divided into two main categories - strapdown INS
and platform INS. Platform UAVs mainly use strapdown INS,
where sensors are fixed to the drone body and digital filters
process the raw data [12]. The disadvantage of INS is the
accumulation of errors due to sensor drift, which causes inac-
curacies in long-term navigation. Therefore, INS is combined
with other navigation systems such as GNSS, visual odometry
or SLAM to correct INS errors using external references [18].

C. SLAM - Simultaneous localization and mapping

SLAM (Simultaneous Localization and Mapping) is an
advanced algorithm that allows a UAV to simultaneously
create a map of the surrounding environment and determine its
own location within that map [19]. This technology is essential
for autonomous operations in unknown environments where
the UAV does not have a pre-existing map or GNSS signal.

The SLAM problem can be formulated as a Bayesian
estimation problem, where the goal is to find the most probable
estimate of the UAV’s position xt and the generated map mt:

p(xt,mt|z1:t, u1:t) ∝ p(zt|xt,mt) · p(xt|xt−1, ut)

· p(xt−1,mt−1|z1:t−1, u1:t−1) (7)

where:

• zt represents the sensor measurements at time t,
• ut denotes the control inputs applied to the UAV,
• xt is the estimated position of the UAV at time t,
• mt is the generated map of the environment.

SLAM can be implemented based on different sensor inputs:
Visual SLAM (V-SLAM) - uses cameras and visual features

of the environment, similar to visual odometry, but with better
long-term position estimation [20]. LiDAR SLAM - works
with a laser scanner and provides a detailed 3D map of
the environment [21]. RGB-D SLAM - uses depth cameras
combining RGB image with distance information [22]. By
combining visual odometry, INS and SLAM, this platform
achieves a high level of autonomy and reliable navigation in
a wide range of environments.

IV. UNIVERSAL PLATFORM (HARDWARE DESCRIPTION)

This UAV platform has been designed with an emphasis
on modularity, flexibility and performance, making it ideal
for a wide range of applications, from industrial inspection
to scientific research. Each component has been carefully
selected for durability, efficiency and easy integration of other
technologies [23].

With its modular design, flexible power system, and broad
sensor integration capabilities, this UAV platform provides a
versatile solution for diverse applications. The possibility of
easy customization and autonomous navigation technologies
make it a cutting-edge tool for industrial, scientific and military
applications.

243



A. Design and frame

The supporting structure of the UAV consists of a
lightweight but extremely strong carbon fiber frame that
minimizes overall weight while providing high strength [24].
The foldable arms provide easy transportation and storage,
reducing the logistical requirements for deployment. The mod-
ular attachment of motors and electronics allows for quick
replacement of these components, reducing downtime and
simplifying UAV maintenance [25].

The UAV powertrain consists of high-efficiency T-Motor
MN601 320KV motors that provide optimal thrust at low
energy consumption [26]. These motors are powered by ESC
controllers supporting 6S-12S LiPo batteries, allowing for
different configurations depending on mission requirements -
from long-duration flights to carrying heavier payloads [27].

B. Power system

The power is supplied by two 6S LiPo batteries with a
capacity of 10,000 mAh, which are connected in parallel,
increasing the overall capacity and extending the flight time
of the [28]. The power distribution is done through a power
distribution board (PDB), which not only distributes the volt-
age efficiently to all components, but also provides current and
voltage measurements to optimize power management [29].

For the on-board computing system, a DC-DC converter is
available to convert the 6S battery voltage to a stable 19V
DC to power the main NVIDIA Jetson Orin computing unit
[30]. This powerful edge-computing module enables real-time
image processing, which is crucial for autonomous navigation
and sensor fusion.

The main control element of the UAV is the Cube Orange+
flight controller, which is equipped with a powerful processor
and interfaces for integrating sensors and communication
modules (Wang et al., 2020). This system stabilizes the flight
and allows interfacing with various navigation technologies,
including visual odometry, inertial navigation and SLAM [31].

The Here3 GNSS module is used for precise positioning
and provides highly accurate location data [32]. In the case
of operation in environments without GNSS signal, a com-
bination of visual odometry, INS and SLAM takes over the
navigation role, enabling robust autonomous driving even in
indoor or densely built-up areas [33].

One of the main advantages of this platform is its ability to
integrate different sensor and measurement systems. The UAV
is equipped with a universal mount that allows the attachment
of a wide range of sensors such as LiDAR scanners, multi-
spectral cameras, thermal cameras, and high-resolution optical
sensors [34].

With a wide range of communication interfaces (CAN bus,
UART, USB), UAVs can be easily adapted to the specific
requirements of a given mission, from geodetic measurements
to critical infrastructure inspections [35].

V. SOFTWARE PLATFORM

The UAV software architecture is designed with an empha-
sis on real-time performance, modularity, and reliability, using

modern middleware technologies and advanced algorithms
for autonomous navigation and computer vision [36]. The
main computing center of the UAV is the Intel NUC, which
provides sufficient computing power to run deep learning,
visual odometry and SLAM algorithms in real-time.

A. Operating system and middleware layer

The UAV runs on Ubuntu 22.04 LTS, with ROS 2 Galac-
tic as the main middleware, which provides communication
between the UAV software modules [37]. The advantage of
ROS 2 is its low latency, support for real-time applications
and scalability, which allows easy integration of new features
and sensor inputs.

The software architecture is divided into several ROS 2 node
processes, each responsible for a specific functionality:

Flight Control Node - communicates with the Cube Orange+
flight controller and translates commands into motion instruc-
tions. SLAM Node - performs simultaneous localization and
environment mapping. Computer Vision Node - processes im-
age data from cameras and performs visual odometry. Mission
Planner Node - generates flight trajectory and optimizes UAV
movement. The communication between these components is
done through the DDS (Data Distribution Service), which
provides low latency and robust data transfer between the UAV
modules [38].

B. Autonomous driving and trajectory planning

Autonomous navigation of UAVs is based on a combination
of the Kalman filter (EKF), visual odometry and SLAM to
ensure accurate localization of UAVs in different environments
[39].

To optimize the flight trajectory of the UAV, it uses Rapidly-
exploring Random Tree (RRT) or a algorithm that allow the
UAV to navigate efficiently in dynamic environments such as
forests, industrial complexes or warehouses [40].

C. Computer vision and neural networks

Another important software feature of this platform is
advanced real-time image processing. The UAV uses deep
neural networks (DNNs) built on PyTorch and TensorFlow,
which run optimized on an Intel NUC using OpenVINO [41].

Convolutional neural networks (CNNs), specifically
YOLOv8 and EfficientNet, are used for obstacle detection,
object classification and pattern recognition in images. The
UAV detects obstacles and dynamically avoids them using
stereo depth mapping combined with LiDAR data [42].

The convolution operation in a CNN can be mathematically
expressed as:

F (x, y) =
m∑

i=−m

n∑
j=−n

w(i, j) · I(x− i, y − j) (8)

where:
• F (x, y) is the output value of the filter at pixel (x, y),
• w(i, j) represents the weights of the convolution kernel,
• I(x−i, y−j) denotes the pixel values in the input image.

244



D. Network communication and remote control

Communication between the UAV and the operator is via
the MAVLink protocol, which allows both manual control via
RC controller and fully autonomous missions controlled by
the Ground Control Station (GCS) [23].

An LTE/5G module is used for remote monitoring of the
UAV, which allows real-time transmission of telemetry and
video data. The UAV also has an internal UDP interface
that allows remote recording and configuration of software
modules during flight [42].

The UAV software platform combines a real-time operating
system, advanced algorithms for autonomous navigation, and
deep neural networks for image analysis. With full ROS 2
support, the system is easily extensible and enables rapid
integration of new sensors and functions. This flexibility
ensures that UAVs can be effectively deployed for a wide range
of applications, from inspection and mapping to industrial
automation and security operations.

VI. RESULTS

The proposed UAV platform is a versatile solution for a
wide range of industrial and scientific applications where mod-
ularity, performance and autonomous control are important.
The overall design philosophy is based on the requirement for
flexibility - the system is designed to allow easy adaptation to
a specific mission without compromising on performance or
functionality.

The structural basis of the platform is a lightweight and
strong carbon frame with foldable arms, which not only facili-
tates transport but also provides ample space for different types
of sensor payloads. Powerful motors with a wide power range
(6S-12S) allow optimization between maximum flight time
and the ability to carry heavier sensors such as LiDAR sys-
tems or multispectral cameras. The power supply and power
distribution have been designed with operational efficiency in
mind, allowing the UAV to operate with both higher capacity
batteries for longer missions and lighter batteries for faster
operations.

One of the main aspects of the platform is its autonomous
navigation. In addition to the standard GNSS system, it allows
the drone to operate in environments where GPS signals
are unavailable, such as confined spaces, tunnels or dense
vegetation. This is achieved through a combination of visual
odometry, inertial navigation (INS) and SLAM algorithms.
Visual odometry uses optical sensors and advanced computer
vision to determine the movement of the UAV by tracking
characteristic points in the environment. INS supplements the
calculations with data from accelerometers and gyroscopes
to provide more accurate estimates of the UAV’s position
and orientation. The SLAM (Simultaneous Localization and
Mapping) algorithm then produces a 3D map of the surround-
ing environment while the UAV locates its position within
it, which is crucial for autonomous operations in unknown
environments.

The Figure 2 captures the already created universal platform
that was created based on the requirements and procedures

Fig. 2. Builded platform during flight

described above. The platform is fully flyable and can now be
extended with external sensors.

The software architecture of the platform is based on ROS 2,
which allows easy integration of sensors, efficient management
of communication between modules and scaling of computing
power. The on-board Jetson Orin computer provides sufficient
computing power for real-time image processing, neural net-
works and flight trajectory optimization. The UAV is equipped
with advanced trajectory planning algorithms such as RRT and
A* to enable navigation in complex dynamic environments.
Deep learning-based computer vision (CNN networks such as
YOLOv8 or higher) contributes to obstacle detection, object
classification and environmental analysis.

An important aspect of the platform is its modular pay-
load system, which enables rapid sensor replacement and
integration via CAN bus, UART and USB interfaces. This
allows the UAV to be easily adapted to a variety of tasks
- from surveying and industrial infrastructure inspection to
agricultural applications where multispectral cameras help to
optimise fertiliser use or monitor crop health.

Overall, this UAV platform combines the latest technologies
in aerial robotics, sensing and autonomous driving. With
full support for autonomous flight, visual odometry, neural
networks and advanced trajectory planning, it is an unrivalled
solution for a wide range of applications. The modular design
and open software architecture allow for adaptation to future
requirements and further innovation, making the platform a
highly versatile and promising UAV solution for professional
deployments.

VII. ACKNOWLEDGMENT

The research was funded from the Ministry of the Interior of
the Czech Republic (MVCR) grant no.VJ02010036 (An Arti-
ficial Intelligence-Controlled Robotic System for Intelligence
and Reconnaissance Operations) and the assistance provided
by the general student development project being executed at
Brno University of Technology.. Special thanks go to doc. Ing.
Petr Marcoň Ph.D. for invaluable guidance.

245



Additionally, we appreciate the efforts of our colleagues
from the UARS (Laboratory of unmanned aerial vehicles
and sensors) for their valuable discussions and assistance in
hardware testing and software integration. Finally, we extend
our thanks to open-source communities such as the ROS
Foundation for their continuous development of tools and
frameworks.

REFERENCES

[1] Kim, J., Ju, C., & Son, H. I. (2019). Unmanned aerial vehicles in
agriculture: A review of platform, control, and applications. IEEE
Access. Available at: IEEE

[2] Islam, N., Rashid, M. M., Pasandideh, F., Ray, B., & Moore, S. (2021).
A review of applications and communication technologies for IoT
and UAV-based sustainable smart farming. Sustainability, 13(4), 1821.
Available at: MDPI

[3] Wang, L., Huang, X., Li, W., Yan, K., Han, Y., & Zhang, Y. (2022).
Progress in agricultural unmanned aerial vehicles (UAVs) applied in
China and prospects for Poland. Agriculture, 12(3), 397. Available at:
MDPI

[4] Radoglou-Grammatikis, P., Sarigiannidis, P., & Lagkas, T. (2020). A
compilation of UAV applications for precision agriculture. Computer
Networks. Available at: ScienceDirect

[5] Hussain, A., Li, S., Lin, X., Ali, F. (2024). Computing Challenges of
UAV Networks: A Comprehensive Survey. ResearchGate. PDF

[6] Ahmed, F., Jenihhin, M. (2022). A survey on UAV computing platforms:
A hardware reliability perspective. Sensors. PDF

[7] Mohsan, S. A. H., Khan, M. A., Noor, F., Ullah, I., Alsharif, M. H.
(2022). Towards the unmanned aerial vehicles (UAVs): A comprehensive
review. MDPI. PDF

[8] Telli, K., Kraa, O., Himeur, Y., Ouamane, A., Boumehraz, M. (2023).
A comprehensive review of recent research trends on unmanned aerial
vehicles (UAVs). MDPI. PDF

[9] Recchiuto, C. T., Sgorbissa, A. (2018). Post-disaster assessment with
unmanned aerial vehicles: A survey on practical implementations and
research approaches. Journal of Field Robotics. PDF

[10] Al-Garadi, M. A., Badawy, A. (2019). Design challenges of multi-UAV
systems in cyber-physical applications: A comprehensive survey and
future directions. IEEE Xplore. PDF

[11] Kotarski, D., Piljek, P., Pranjić, M., Grlj, C. G., Kasać, J. (2021).
A modular multirotor unmanned aerial vehicle design approach for
development of an engineering education platform. Sensors. PDF

[12] Mostafa, M., Zahran, S., El-Sheimy, N. (2018). Radar and visual
odometry integrated system aided navigation for UAVs in GNSS-denied
environment. MDPI. PDF

[13] Rostum, M., Vásárhelyi, G. (2023). Visual Odometry-Based UAV Navi-
gation in GNSS-Denied Environments. International Journal of Robotics
and Automation, 38(4), 1056-1072. DOI: 10.1109/IJRA.2023.01589.

[14] Luo, H., Zou, D. (2023). UAV navigation with monocular visual-inertial
odometry under GNSS-denied environment. IEEE Xplore. PDF

[15] Ahmad, A., Wang, S. (2024). A Comprehensive Review on Sensor
Fusion Techniques for Localization of UAVs in GNSS-Denied Envi-
ronments. IEEE Access.

[16] Ahmad, U., Nasirahmadi, A., Marino, S. (2021). Technology and data
fusion methods to enhance site-specific crop monitoring. Agronomy,
12(3), 555. DOI: 10.3390/agronomy12030555.

[17] Szrek, J., Trybała, P., Michalak, A., Zietek, B. (2020). Accuracy
evaluation of selected mobile inspection robot localization tech-
niques in a GNSS-denied environment. Sensors, 21(1), 141. DOI:
10.3390/s21010141.

[18] Salib, A., Moussa, M., Moussa, A. (2020). Visual heading estimation
for UAVs in indoor environments. 2020 International Conference on
Communications, Signal Processing, and Their Applications (ICCSPA),
IEEE. DOI: 10.1109/ICCSPA49915.2020.9385709.

[19] Lin, X., Zhang, H., Wang, J. (2022). UAV autonomous navigation
using deep reinforcement learning and SLAM in GNSS-denied envi-
ronments. Journal of Intelligent & Robotic Systems, 105(2), 317-334.
DOI: 10.1007/s10846-022-01547-8.

[20] Yang, L., Chen, X., Wu, Y. (2021). Visual-SLAM-based UAV navigation
in cluttered environments. IEEE Transactions on Robotics, 37(5), 1345-
1359. DOI: 10.1109/TRO.2021.3069285.

[21] Qiu, J., He, M., Liu, R. (2022). LiDAR-based SLAM for UAV au-
tonomous exploration in indoor environments. Remote Sensing, 14(4),
805. DOI: 10.3390/rs14040805.

[22] Elamin, M., Shao, Y., Zhang, L. (2022). RGB-D SLAM and deep
learning fusion for UAV obstacle avoidance. Sensors, 22(10), 3752. DOI:
10.3390/s22103752.

[23] Perez-Grau, F. J., Ragel, R., Caballero, F. (2018). An architecture for
robust UAV navigation in GPS-denied areas. Journal of Field Robotics,
35(8), 1173-1192. DOI: 10.1002/rob.21850.

[24] Grau, A., Guerra, E., Bolea, Y., Gamiz, J. (2020). Design and implemen-
tation of a virtual sensor network for smart UAV applications. Sensors,
20(2), 358. DOI: 10.3390/s20020358.

[25] Nouacer, R., Hussein, M., Espinoza, H. (2020). Towards a framework
of key technologies for drones. Microprocessors and Microsystems, 77,
103162. DOI: 10.1016/j.micpro.2020.103162.

[26] Hussein, M., Espinoza, H. (2020). Energy-efficient UAV power system
optimization for extended flight time. Aerospace Science and Technol-
ogy, 98, 105678. DOI: 10.1016/j.ast.2020.105678.

[27] Liu, Z., Shen, L., Zhao, S. (2020). High-efficiency UAV
propulsion system with multi-mode operation. IEEE Transactions
on Aerospace and Electronic Systems, 56(3), 2675-2687. DOI:
10.1109/TAES.2020.2970123.

[28] Theile, G., Schmidt, P., Bauer, T. (2018). Advances in lithium-polymer
battery technology for UAV applications. Journal of Power Sources, 395,
276-287. DOI: 10.1016/j.jpowsour.2018.05.089.

[29] Aksland, J., Tannous, A. (2022). UAV power distribution and energy
management systems: A survey. IEEE Access, 10, 34785-34799. DOI:
10.1109/ACCESS.2022.3162325.

[30] Deliparaschos, K. M., Loizou, S. G. (2024). A modular UAV hardware
platform for aerial navigation research. Robotics and Autonomous
Systems, 162, 104467. DOI: 10.1016/j.robot.2024.104467.

[31] Siewert, M., Hoffmann, F., Becker, T. (2021). Multi-agent UAV swarm
coordination using ROS 2. IEEE Robotics & Automation Magazine,
28(3), 45-57. DOI: 10.1109/MRA.2021.3076489.

[32] Fraga-Lamas, P., Ramos, L., Mondéjar-Guerra, V. (2019). IoT-enabled
UAV systems with AI-based object detection. Sensors, 19(5), 1103. DOI:
10.3390/s19051103.

[33] Yang, H., Gao, X., Chen, J. (2022). UAV-assisted navigation in
GNSS-denied environments: A hybrid SLAM approach. IEEE Trans-
actions on Aerospace and Electronic Systems, 58(2), 2487-2501. DOI:
10.1109/TAES.2022.3152128.

[34] Bakirci, M. (2023). Multi-sensor UAV payload integration for in-
dustrial applications. Journal of Field Robotics, 40(1), 23-41. DOI:
10.1002/rob.22056.

[35] Zhang, Y., Zhao, W. (2020). UAV communication architectures
for beyond visual line-of-sight (BVLOS) operations. IEEE
Communications Surveys & Tutorials, 22(4), 2881-2905. DOI:
10.1109/COMST.2020.3016909.

[36] Antonopoulos, A., Lagoudakis, M. G., Partsinevelos, P. (2022). A ROS
multi-tier UAV localization module based on GNSS, inertial and visual-
depth data. Drones, 6(6), 135. DOI: 10.3390/drones6060135.

[37] Al-Batati, A. S., Koubaa, A., Abdelkader, M. (2024). ROS 2 Key
Challenges and Advances: A Survey of ROS 2 Research, Libraries, and
Applications. Preprints. DOI: 10.20944/preprints202410.1204.v1.

[38] Lagoudakis, M., Partsinevelos, P. (2022). UAV real-time data fusion with
ROS 2 middleware. IEEE Aerospace and Electronic Systems Magazine,
37(5), 17-29. DOI: 10.1109/MAES.2022.3162287.

[39] Galarza-Falfan, J., Ortiz, A., Benavides, A. (2024). UAV trajectory
optimization using SLAM-based real-time mapping. Robotics and Au-
tonomous Systems, 163, 104489. DOI: 10.1016/j.robot.2024.104489.

[40] Rajagopal, M., Ashwin, I., Naren, M., Narayan, V. S. (2024). Au-
tonomous systems: Indoor drone navigation. AIP Conference Proceed-
ings, 3216, 020002. DOI: 10.1063/5.0102391.

[41] Madan, R., Gupta, S., Shukla, M. (2024). UAV object detection and
tracking using deep learning: A review. IEEE Access, 12, 12678-12695.
DOI: 10.1109/ACCESS.2024.3165555.

[42] Wu, Z., Liu, M., Chen, M., Zhong, B., Deng, W. (2024). Implementation
of intelligent indoor service robots based on ROS and deep learning.
Machines, 12(4), 256. DOI: 10.3390/machines12040256.

246