1.

Obstacle Avoidance-Based Autonomous Navigation of a Quadrotor

System
Reference: Mohammed A. Alanezi , Zaharuddeen Haruna , Yusuf A. Sha’aban, Houssem R.
E. H. Bouchekara , Mouaaz Nahas and Mohammad S. Shahriar

Challenges/problem: Developing an obstacle avoidance-based autonomous navigation

of a quadrotor suitable for outdoor applications in livestock management. To accomplish this,
a Simulink model of the UAV is created and its transient and steady-state performance is
measured.

Strategy
Two genetic algorithm-based PID controllers were designed for quadrotor altitude and
attitude control, and an obstacle avoidance algorithm is used to ensure the quadrotor's
autonomous navigation. The implementation of the obstacle avoidance scheme demonstrates
that a distance threshold of 1 m is sufficient for the quadrotor's autonomous navigation.

Method:
Generally, the design algorithm for the autonomous navigation of the quadrotor system with
obstacle avoidance comprises three key components -

1. Perception
2. Planning
3. Control

1. Perception: The current state of the quadrotor (in terms of position and orientation) as well
as the state of the surrounding environment (in terms of obstacle and target positions) are
determined in this step.

2.Planning: This step makes decisions based on the information gathered during the
perception stage. This step involves the development of route planning and path planning
algorithms.

3. Control: In this step, the quadrotor's movements and actions are controlled to ensure that it
follows the path defined in the planning stage. This step involves the development of
algorithms such as path following, obstacle avoidance, path tracking, and stability control.

The quadrotor navigational system model is made up of the following subsystems. The path
generator generates the desired trajectory, the dynamic model of the quadrotor determines the
actual position of the quadrotor, and the altitude and attitude controllers generate the
manipulated signals that modify the system's behaviour to achieve the desired state.

RESULTS
In this research MATLAB Simulink was used to model and simulate the obstacle avoidance-
based autonomous navigation of a quadrotor system. Two genetic algorithm-based
Proportional Integral Derivative (PID) controllers were developed to control the quadrotor's
altitude and attitude using integral square error (ISE) and integral time absolute error (ITAE)
(ITAE). The designed controllers' performance was evaluated using transient parameters
(settling time, rise time, and overshoot) and steady-state error.

Conclusion
Collisions in quadrotor autonomous navigation could occur for a variety of reasons, including
poor weather, the nature of the deployment environment, and moving obstacles. As a result,
obstacle avoidance is critical in quadrotor autonomous navigation.

My learnings:

Through this research paper, I have learned the following things:

Due to their diverse applications in civilian and military applications, unmanned aerial
vehicles (UAVs) have attracted significant research attention from academic and industrial
disciplines.

Aerial and border surveillance, mapping and photography, and search and rescue operations
are among these applicationsUAVs can operate as drones (with rotary wings) or aircraft (with
fixed wings) depending on the application, and can be autonomous or teleoperated. Rotary-
wing UAVs are distinguished by their ability to hover, manoeuvre, and perform Vertical
Takeoff and Landing (VTOL) in confined spaces. Furthermore, fixed-wing UAVs have long-
range and endurance flight capabilities.
2. Sampling based path planning algorithm for UAV collision avoidance

Reference: A Saravanakumar, A Ksviyarasu and R Ashly jasmine

Challenges/problem:
a) If there are a lot of obstacles in the state space, the possibility of finding a path that isn't in
an obstacle region is slim.

b) A narrow path is also difficult to navigate.

c) RRT is time-consuming in large workspaces and inefficient with several unnecessary

branches.

Strategy:
To overcome these limitations, a modification is proposed by simplifying the node
connection strategy of randomly generated nodes in standard RRT and coupling it with
reachable sets to detect the dynamic obstacles position in order to avoid collision.

Method

1. Sampling-based path planning

These methods randomly sample the workspace as a set of nodes, then map the workspace
haphazardly to find the best path. Because these methods have a simple framework, any
modification can be made, and they are better suited for online planning conditions.

Under sampling-based methods, the RRT algorithm is chosen and modified to overcome the
limitations of the basic RRT algorithm.

2. Modified RRT algorithm

The RRT algorithm was improved by changing the node connecting strategy, which
generated a viable path in a short amount of time. Additional paths were added by
connecting the in-between waypoints in the developed paths to the goal, and the reachable
set concept is used to compute the states of obstacles in motion, allowing for more accurate
reactionary obstacle detection.

3. Simplified node connection strategy

To accelerate the path planning rate, a method for modifying the node linking strategy is
discovered. In contrast to the standard RRT technique, the sampled point is joined from the
beginning rather than the end of the RRT tree. A course is simulated from the start point to
the sampled point in each simulation sequence and checked against limitations. If the
requirements are met, a new path from the sampled point to the objective (goal) point is
simulated. Only if the formulated constraints are met can an entire trajectory course
connecting the start and goal point be completed.

4. Collision Avoidance with Reachable Sets (CARS)

Methods like standard RRT and simplified node connection strategy predict collisions by
assuming that obstacles move in a straight line. However, obstacles may not always take the
straight path; this can be dealt with by using reachable sets. A moving obstacle's reachable
set is the region defined by the set of all possible trajectories that can reach at any time
instance.

Result
The algorithm of randomly exploring random tree creates a path autonomously from the
Start to Goal State by randomly exploring the nodes in the given workspace, and the path
created is free of obstacles. Simplified node connection strategy is simulated for the same
start and goal node used in RRT algorithm simulation, and a path free of obstacles is
generated through a single node in a fraction of the time required by RRT.

The time taken by the algorithms to generate the feasible path using the same workspace
dimension, start and goal nodes is compared between the standard RRT and MRRT
algorithms. MRRT is more likely than standard RRT to generate the path in a short period of
time.

Conclusion
When simulated with static and dynamic obstacles, the modified RRT algorithm develops a
collision-free path in less time than RRT. The results show that the proposed method was
more robust in dealing with unexpected obstacle manoeuvres by autonomously defining the
trajectory and following up without colliding with the obstacles while predicting the
potential collision.
My learnings:

Through this research paper, I have learned the following things:

When it is discovered that the initial path is not feasible, the UAV reacts by moving along the
new path while avoiding collisions with moving plates in order to reach the goal. UAV path
planning focuses on automatically developing attainable paths from a predetermined start to a
target point while taking into account static and dynamic environmental constraints and
uncertainties.
3. An energy efficient IoD static and dynamic collision avoidance approach
based on gradient optimization
Reference: Authors- Mitch Campion, Prakash Ranganathan and Saleh Faruque

Challenges/problems: The main challenge in IoD formation is the need for an efficient
and effective collision avoidance protocol in static and dynamic environments, which is very
important in civilian and military applications. IoD can be used to efficiently complete tasks
that require highly complex tasks that individual drones cannot perform effectively. On the
other hand, the rapidly increasing popularity of IoD for consumer applications significantly
increases the number of drone accidents.

Strategy:
The obstacle must be detected during the mission, which necessitates real-time path
optimisation or dynamic path planning in the absence of a complete path. Rather, it provides
optimal path information within the drone's field of view. Furthermore, in IoD, each drone
must be equipped with an efficient detection mechanism that detects the speed and direction
of dynamic obstacles and then avoids them while continuing to plan its path.

Method
1. Energy model for drone path planning

Despite the benefits of drones and their growing popularity, inherent limitations continue to
limit their full potential.The main limitation is the limited lifespan of on-board batteries,
which is regarded as the primary disadvantage of drones. As a result, most applications are
unable to realise their full potential. As a result, reducing energy consumption during the
drone mission is a critical challenge. To mitigate this disadvantage, the planned trajectory
should use the least amount of energy.

2. Energy-efficient IoD path planning approach description

Some factors must be considered when planning the path of IoD formation, such as terrain
area, including obstacles in the terrain, and IoD formation. Mountains, buildings, radars, and
other threats may be encountered in the mission environment. Furthermore, the IoD
formation includes a large number of drones flying in this environment. As a result, the IoD
formation should be able to avoid colliding with these obstacles as well as other drones in the
formation. To accomplish this, the path should be constrained in order to avoid all of these
threats and obstacles. Furthermore, the path should achieve the goals while adhering to all
constraints.

3. System model of collision-free path generation approach for IoD

Because the flight terrain is assumed to be unknown in advance, prior knowledge of obstacles
is not required. As a result, an offline path planning phase is not required. The UAVs take off
from their starting points and fly in a straight line to their destinations until obstacles are
detected. When obstacles are detected, the collision avoidance algorithm is activated to avoid
them. The algorithm's concept is to divide the total time into small intervals. The path
planning generator receives parameters and environmental constraints at each interval and
returns a feasible path, which is the best partial path in which the objective function, i.e.
reducing path length or energy consumption, is achieved and constraints are satisfied.

Result
To assess the proposed approach's ability to avoid all obstacles while consuming the least
amount of energy, we run a simulation with different static and dynamic obstacles using the
Matlab tool. To demonstrate the effectiveness of thre algorithm, various scenarios are
simulated, including environments with only static obstacles, environments with only
dynamic obstacles, and dense environments with both static and dynamic obstacles.

The following collision avoidance scenarios are also evaluated:

 Drone to static obstacles

 Drone to dynamic obstacles
 Drone to drone.

We verify the safety distance required for avoiding collisions between drones in each
scenario, and then record the minimum separation distances between drones and dynamic
obstacles in dynamic obstacle and dense scenarios.

Conclusion
The proposed mechanism provided fast and accurate collision-free IoD formation paths with
minimal energy consumption. Energy consumption was calculated for various
environmental constraints. It can also generate safe and efficient energy-consuming
formation paths, and it can be scaled up to larger-scale problems.
My learning

Through this research paper, I have learned the following things:

The formation of the Internet of Drones (IoD) offers a wide range of applications in both
military and civilian environments. Dynamic and static obstacles have a significant impact on
IoD performance in congested terrain. One of the most difficult challenges in IoD missions is
avoiding obstacles in order to complete their missions successfully and safely. Another
difficulty is the limited flight time of a drone. To ensure drone safety, IoD must be provided
by an intelligent and accurate energy-efficient collision avoidance algorithm in which IoD
paths are modified online.
4. UAV swarm communication and control architectures
Reference: Authors- Mitch Campion, Prakash Ranganathan and Saleh Faruque

Challenges/problems: There is a risk of injury or fatality in manned aviation if a critical

error occurs in flight. These concerns are alleviated by an unmanned aircraft system. Manned
aviation is costly. The cost of purchasing or renting a general aviation aircraft is prohibitively
expensive. Pilot labour, fuel costs, and maintenance are prohibitive costs to the widespread
commercial use of general aviation aircraft.

Strategy:
Focused development of UAV swarms with autonomous UAV-to-UAV communication
capability is critical to advancing the utility of UAV swarms. The use of a cellular mobile
framework alleviates many of the constraints that limit the utility of UAVs, such as range of
communication, networking challenges, and size-weight-and-power considerations.
Furthermore, cellular networks make use of the robust and dependable infrastructure
proposed by 5G systems for machine-to-machine communication.

Method
1. Flying ad-hoc network (FANET) architecture

FANETs are networks of UAVs that communicate with one another without the use of an
access point, but at least one of them must be connected to a ground base or satellite. UAVs,
like autopilots, carry out their missions without the assistance of humans. Due to cheaper and
smaller wireless communicating devices, many research fields in academia and industry have
focused on FANETs in recent years. FANETs are now used in a variety of applications,
including military and civil applications, wildfire management, and disaster monitoring.
Because each network has its own set of specifications, it is critical to use a reliable protocol
based on those specifications and to test its performance using simulation.

Protocol simulation is influenced by two factors:

 The mobility model.

 Communicating traffic pattern

2. Proposed swarm architecture

The proposed architecture is an adaptation of an ad-hoc network made possible by
infrastructure support. The infrastructure, in particular, includes complete UAV-to-UAV
communication, with each UAV's telemetry communicated to every other UAV via cellular
mobile infrastructure. Decisions are distributed among the UAVs in this architecture, and the
infrastructure is only used to transmit data.

3. M2M and fifth generation (5G) networks

The maximum download speed of fourth generation (4G) cellular technology is 1 Gbps.
Maximum download speeds of 10 Gbps are expected, with network latency as low as 1 ms.
UAV communications typically use packet sizes ranging from 17 to 263 bytes. While 4G
speeds are adequate for these packets, 5G will enable additional data streaming, including
video from payload cameras and data from payload light detection and ranging systems. For
UAV swarm communication, the ability to achieve low latency is critical. M2M
communications is a key goal of 5G communications. 5G's M2M communication capabilities
would be an ideal backbone for UAV swarm environments.

4. UAV-to-UAV network communication test bed

The test bed that was constructed uses quadcopters that are custom-made. Flight controllers
connect the on-board companion computers and the quadcopters' mesh networking hardware.
When communicating with the on-board computer, the flight controller employs the Micro
Air Vehicle Link (MAVLink) protocol.

Result
The development of autonomous swarms with UAV-to-UAV communication and
coordination capability is critical to the advancement of UAV swarm utility. Though swarm
technology has yet to be used in commercial applications, it has enormous potential.
Traditional UAV swarm communication approaches are limited by the use of cellular mobile
frameworks. The use of cellular networks for UAV swarming would greatly increase swarm
efficiency and commercial utility, especially with the upcoming 5G networks with M2M
communication capabilities.

Conclusion
This paper presents a conceptual proposal, preliminary development, and literature review
for using cellular networks as the communication infrastructure for UAV swarms. It provides
an overview of the sUAS industry, UAV swarm applications, and in-house UAV swarm
development efforts. The paper examines preliminary test bed developments and provides
guidance for future work on UAV swarm systems.
My learnings:

Through this research paper, I have learned the following things:

Unmanned aerial vehicles (UAVs) have wreaked havoc on the aviation industry. As
technology and policy evolve, the magnitude of this disruption will only grow. UAV swarm
is one technology poised to accelerate this disruption. UAV swarming has the potential to
distribute tasks and coordinate the operation of a large number of UAVs with little to no
operator intervention. This paper reviews the literature on UAV swarms and proposes a
swarm architecture that will allow for greater swarm autonomy and reliability by utilising
cellular mobile wireless communication infrastructure.
5. Semi-autonomous drone control with safety analysis
Reference: Authors- Hirad Goudarzi and Arthur Richards

Challenges/problems: Drones will frequently operate in areas with good prior maps,
GPS coverage, and the ability to maintain visual line-of-sight (VLOS) to a pilot. These
industries are expected to have a significant economic impact. However, such operations can
still be difficult due to the need to monitor drone health and the flight environment, for
example, for passersby.

Strategy:
To propose an automation strategy for enabling pilot-only operation by automating GSO
functions and providing the pilot with an extremely simple flight management interface. It
should be noted that the operation of the payload is not considered in this work, but given
recent advances in fields such as computer vision, it is likely that it could be incorporated

Method
1. Finite-state model

The mission script is made from the following three commands:

I. Waypoints
II. Delays
III. Gotos

These commands are ordered so that each objective waypoint is preceded by a delay, a
SAFTI, another delay, and a goto. The delay command serves as the interface between the
GCA and the autopilot. The GCA performs safety checks on the drone's vital subsystems
before directing it to the next objective waypoint. If the GCA does not receive a command,
the autopilot will wait for a timeout in the delay state, increment the mission counter by
one, and navigate the drone to a SAFTI. Another attempt is made at SAFTI to fly to the next
objective waypoint. If no signal is received from the GCA, the goto command will skip the
mission to final approach and safely come back.

2. Human–machine interface

This work takes a system-centric approach in which the system's situation awareness (SA) is
shared between the human pilot and the automation. This method allows for task
distribution based on the capabilities of both agents (i.e., the pilot and automation). In
monitoring tasks, operator vigilance deteriorates over time. Due to the repetitive nature of
the task, monitoring of vital subsystems is delegated to GCA in order to maintain a high level
of SA across the system. Furthermore, this data is processed and reported to the pilot via a
traffic light system, establishing an implicit form of communication between the automation
and the pilot in which low-level details are filtered, reducing the impact on the pilot's
working memory.

3. Safety analysis of semi-autonomous design

This section will show how to use the claims-arguments-evidence (CAE) framework to assess
the safety of a semi-autonomous system. Cases for several example claims will be provided
following a review of the CAE approach. Exhaustive hazard identification and full system
safety analysis are beyond the scope of this document; instead, the examples will
demonstrate how claims typical of drone safety requirements can be supported in this
context.

A CAE block graphical safety case is represented by a hierarchy diagram with a "top claim" at
the top. The top claim is a logical statement of something that needs to be proven, usually a
safety claim resulting from a hazard identification process. Under that claim is an argument,
and beneath that are one or more subclaim or evidence blocks, arranged in such a way that
the top claim is verified by the subclaims and evidence, according to the argument. A "side
claim" justifying why the argument is applicable may be added to the argument at times.

3. Collision with a static obstacle

One of the primary ground risks identified by drone regulators is collision with static
obstacles such as buildings and infrastructure. This section examines this failure condition as
well as the system behaviour that protects against collisions with unmapped or unexpected
map locations of obstacles along the flight path. In this study, dynamic obstacles are
managed manually by the pilot and are classified as air risk, so they are excluded from the
following example.

Result
This paper describes a drone automation paradigm that allows a single pilot to operate in
VLOS over a sparsely populated area with a crew of just one. The framework reduces the
pilot's workload by automating mission execution and health monitoring, eventually taking
over the role of the GSO. The pilot's interaction with the drone is reduced to a single button
and a traffic light system, allowing them to use their working memory to identify people in
danger. The design can make use of existing commercial off-the-shelf autopilot technology
and is inherently resistant to automation failure and pilot incapacity. To maximise pilots'
situational awareness, the framework used a strict predictability requirement; however, this
limits automation's ability to respond to dynamic obstacles.
Conclusion
This paper describes the partial automation of drones (also known as unmanned aerial
vehicles, UAVs, or aerial robots) in populated areas within the pilot's visual line of sight.
Mission responsiveness is improved by reducing the number of human crew members and
avoiding the need for area clearance, while carefully managing the workload of those who
remain to ensure safety.

My learnings:

Through this research paper, I have learned the following things:

The work takes a system-centric approach to integrating human and automation tasks based
on their capabilities and the use of standard procedures, while prioritising the overall
system's predictability and simplicity. A structured safety case is used to present and
rigorously analyse safety claims about the proposed system. With simulation results and
scenario analysis, the proposed system is applied to a bridge inspection case study.
6. Modelling of unmanned aircraft visibility for see-and-avoid operations

Reference: Authors-

P. Highland,J. Williams, M. Yazvec, A. Dideriksen, N. Corcoran, K. Woodruff, C. Thompso

n, L. Kirby, E. Chun, H. Kousheh, J. Stoltz, and T. Schnell

Challenges/problems: With more unmanned aircraft (UA) taking to the skies every day,
the already high manned aircraft-UA exposure rate continues to rise. The majority of these
collisions and near-collisions occurred in good weather, during daylight hours, and with
manned aircraft that were much larger than the most common unmanned aircraft.

Strategy
To compensate for the increased exposure of manned aircraft and UA, a significant amount
of research and data-driven rule making is required. Under visual conditions, the see-and-
avoid concept allows aircraft to self-separate. There are see-and-avoid situations that are both
alerted and unalerted. Unalerted situations are more dangerous because it is entirely up to the
pilot to see the other aircraft.

Method
Pilots have used the see-and-avoid concept for right-of-way decisions since the dawn of
aviation. There are notable exceptions to the see-and-avoid system.

1. Pilot scanning and collision avoidance

Object recognition is crucial in the context of UA. An outside observer can usually predict
the flight paths of manned aircraft. UA aircraft, on the other hand, may employ autonomous
guidance logic, raising concerns about their motion predictability. Recognizing whether an
aircraft is manned or unmanned can help manned aircraft pilots.

2. Traffic detection and warning systems

For decades, traffic detection systems have caught the attention of researchers. Using a
simulator, we examine a pilot warning system to determine the relationship between distance
and detection of a target aircraft. The time required to avoid a collision is heavily dependent
on the pilot's alertness and detection rate. The more aware the pilot is of another object, the
more time he or she will have to manoeuvre around it to avoid a collision.

Result
This research demonstrated that UA visibility of 1300 m poses a serious threat to a midair
collision if a manned aircraft and a UA are on a collision course while operating under the
see-and-avoid concept.

This research also demonstrated that when UA visibility is 400 m and a manned aircraft and a
UA are on a collision course while operating under the see-and-avoid concept, a midair
collision may be unavoidable. This research also shows that small UA visibility distances of
1300 and 400 m occur frequently.

Conclusion
The findings of this study enable them to make data-driven decisions about when and how to
fly. Waiting for ideal visibility conditions or flying at slower speeds may increase safety
margins until improved alerting and UA visibility technology is available.

My learnings:

Through this research paper, I have learned the following things:

An already high manned aircraft to UA exposure rate keeps rising as more unmanned aircraft
(UA) take to the skies every day. Pilots and rule-making bodies are aware that the threat
posed by UA visibility to see-and-avoid operations is real but unquantifiable.
7. Development of Self-Synchronized Drones’ Network Using
Cluster-Based Swarm Intelligence Approach

Reference: Authors- Fawaz Alsolami; Fahad A. Alqurashi; Mohammad Kamrul

Hasan; Rashid A. Saeed; S. Abdel-Khalek; Anis Ben Ishak

Challenges/problems: The transceiver's function can be interrupted, interference can

occur, and spectrum precision can be affected by the drone system's inaccurate clock.
Reference Broadcast System (RBS) and other neighbour drone timing synchronisation
techniques, used in the drone network, produce good results. RBS is nonetheless impractical
to employ in some circumstances, such as an SDN network environment, due to the demand
of one super drone with a significant number of broadcasts.

Strategy
In a swarm of unmanned aerial vehicle (UAV) network or swarm of drones, timing
synchronisation is crucial. In order to significantly increase timing synchronisation precision
at this level, current timing synchronisation techniques concentrate on increasing single-hop
skews. By removing clock drift and improving timing synchronisation precision, the
appropriate studies and changes are required to achieve actual timing synchronisation. As a
result, a novel self-timing synchronisation strategy is suggested in this study, allowing several
MD drones to build swarm clusters on their own.

Method
By forming swarm clusters out of a random number of MD drones and implementing a
hybrid (one way and two ways) messaging system for both inter-swarm cluster and intra-
swarm cluster timing synchronisation approaches, the previous swarm drone timing
synchronisation techniques were quite different from the proposed method.

The Proposed Synchronization Strategy

The structural framework combines drone clustering and temporal synchronisation to provide
semi-flat synchronised SDN networks. The proposed method improved the control packets
overhead, which is essential for boosting the network performance and throughput, in
addition to synchronising the MD network. Since it is more applicable to three diminutions
large-scale master drones'(MD) distribution scenarios, this work will primarily focus on the
distributed approach.

Timing synchronisation for swarm drones in the old system is hampered by network
overhead, bandwidth use, misaligned packets, and offset. The network does not perform
inaccurate timing synchronisation in this case. Researchers have worked hard to create
algorithms and strategies to address these problems.

A. SDN Synchronization Model

The suggested method focuses primarily on self-timing synchronisation between master

drones in a situation where master drones (MDs) are dispersed randomly and without proper
planning. While some Master Drone can be gathered independently, some can be located next
to one another. The synchronised clock time will be communicated amongst members of a
cluster within each swarm cluster. First, upon the formation of a cluster, the nearby drones
will compare their local times and distances to choose a cluster head. A cluster head will
broadcast clock information inside the cluster as a beacon to all CLUSTER MEMBERS
(CM).

B. Timing Synchronization Structure

Here, the CH is calculated using the clock time and the distance between drones. In this
process, a drone not only transmits frames to nearby drones, but also fuzzy values. A CH, an
IMC, and Swarm cluster Members make up a swarm cluster (CMs). The temporal
synchronisation frame structure for the suggested method includes an Identification (id),
Sync, range, and multi-hop.

C. Swarm Cluster Inter-Cluster Synchronization

This process will begin with the initialization of swarm cluster creation by Cluster heads
(CHs), who will send a beacon packet to all Cluster members (CMs) to seek their
participation. The closest range and SYNC time will then be calculated by Cluster Members
(CMs) in order to create the clusters. Clustering is a critical topic for the drones' network in
order to make the network distributive. The temporal synchronisation process will be
expedited by the use of the aforementioned clustering approaches, and the network's overall
performance and throughput will be improved. With a large number of drone networks, the
ultimate objective of clustering is to segregate the network to reduce the offsets between the
drones over the entire network.

The swarm network has three levels of communication,

 Swarm drones’ control centers (SDCC) level

 Cluster head (CH) or master drones’ level

 Cluster members (CM) level

Result
The main objective of the self-organizing swarm clustering process is to dynamically
assemble the cluster from all of the MD drones to achieve better temporal synchronisation
when clock time and transmission range (range) are taken into account. The suggested self-
organizing approach is subjected to an analytical examination, with MATLAB software being
used to accurately depict the performance and throughput of the clustering approach.

Conclusion
In numerous literature reviews and industrial studies for drones/UAV swarm networks, clock
skew, drift, and offset have been evaluated as performance indicators, where it has been
demonstrated that they cause a critical disruption for the UAVs' operation and data exchange.
The semi-flat swarm drones network (SDN) architecture's MD is to be self-synchronized
using the suggested approach.

My learnings:

Through this research paper, I have learned the following things:

Time synchronisation is critical in a swarm drone network (SDN) or a network of unmanned
aerial vehicles (UAV). Modern time synchronisation methods concentrate on improving
single-hop skews, which significantly enhance timing synchronisation precision at this level.
An incorrect drone system clock can generate interference, decrease spectrum precision, and
disrupt transceiver functioning.
8. AGDS: adaptive goal-directed strategy for swarm drones
flying through unknown environments

Reference: Authors- Fakui Wang, Jialei Huang, Kin Huat Low, Zisen Nie & Tianjiang Hu

Challenges/problems: This research tries to address a challenging problem of a drone

swarm for a specific mission by reaching a target place in an unknown environment. Certain
drones in a swarm of drones have limited perceptual and computational skills, making it
challenging to collect plentiful information about the surroundings and their companions. By
depending just on the local knowledge gathered by each individual drone, it is a challenge to
create a drone swarm that does not collide while passing through an unfamiliar environment.

Strategy:
For the drones swarmed throughout unfamiliar surroundings, a bio-inspired flocking
algorithm with adaptive goal-directed strategy (AGDS) is proposed and developed. Each
drone has a biological optical mechanism that detects impediments within its local observable
range. Task information on the destination is only sent to a few designated drones (referred to
as informed agents), rather than to all other individual drones (uninformed agents).

Method
We present a self-organizing flocking algorithm that does not require every agent to be aware
of the desired region. This is especially useful for drone swarms working in interference or
electromagnetic conditions, when each aircraft's knowledge of the desired location cannot be
assured. A vision-based adaptive goal-oriented method is given based on employing the
mathematical bionic vision mechanism to sense the environment, which can effectively
improve the efficiency of the drone swarm navigating a difficult canyon-like environment.

First, a drone dynamics model with velocity and acceleration limitations is presented. The
drone swarm's communication architecture is then presented. Finally, many algorithm
evaluation measures are proposed. Initially, a flocking method for the drone swarm crossing
unknown environments, including non-AGDS and AGDS, is proposed. The flocking
algorithm's terms are then introduced in detail.

Being informed agents, these drones are provided with task information. The identities of
informed agents are concealed, and feedback from neighbours is required. As a result, all
agents have the same status. It also implies that the loss of some of the informed agents has
no effect on the flock's information interaction. Of course, informed agents can be added
dynamically, or existing regular agents can be identified as informed agents.

The drone's motion can be divided into two types: radial motion and normal motion. The
former can represent left and right steering behaviour, whereas the latter can characterise
acceleration and deceleration . As a result, the drone creates behaviour by perceiving the size
and position of obstacles or walls in its field of view.

Because a single drone has limited communication and sensory skills, it is difficult for the
aerial swarm to immediately avoid obstacles when they encounter them. Remarkably, when a
starling confronts natural enemies or barriers, it immediately spreads valuable information
throughout flocks by calls and other means to quickly escape or avoid obstacles. As a result,
in the real world, it is critical for robots to exchange valuable information as fast as possible.

However, drones operating in unknown environments often encounter serious conflicts

between task information and obstacle avoidance. This leads to a situation where the drone is
caught in a dilemma. In nature, migrating or foraging animals adjust their behaviour based on
real-time information about the environment. Therefore, we propose an adaptive goal-
directed strategy (AGDS) based on visual perception.

Result
Each drone has a biological optical mechanism that detects impediments within its local
observable range. The destination task information is only sent to a few designated drones
(referred to as informed agents), rather than to all other individual drones (uninformed
agents). In the suggested flocking swarm, the informed agents collaborate with the remaining
uninformed agents to complete a shared and overall purpose.

Conclusion
Each drone in the swarm functions autonomously as a computational node and collaborates
with others via local information interaction. Drones perceive the world in real time using
visual mechanisms rather than prior knowledge of the environment. The flocking swarm is
made up of both knowledgeable and uniformed agents who work together to complete the
assignments. Furthermore, each drone follows the same self-organizing laws.

My learnings:

Through this research paper, I have learned the following things:

Although the swarm has numerical and functional advantages, these advantages pose
significant obstacles to formation control. Nature's phenomenon of collective motion can be
used to help solve this challenge. The multiple unmanned aerial systems (MUAS) is an
unmanned intelligent cluster inspired by biological groups' self-organizing characteristics that
considerably enhances work efficiency through collaboration and complementing skills. As a
result, the aerial cluster cooperative control technology with dispersion, autonomy, and
robustness is a promising solution.
9. Online Path Generation and Navigation for Swarms of UAVs
Reference: Authors- Adnan Ashraf, Amin Majd and Elena Troubitsyna

Challenges/problems: The number of UAV mishaps is steadily growing. As a result,

UAV motion safety has become a top priority for UAV operators. A safe functioning of a
swarm of UAVs cannot be assured unless the UAVs avoid colliding with one another as well
as static and dynamically emerging, moving impediments in the flying zone. The number of
UAV mishaps is steadily growing.

As a result, UAV motion safety has become a top priority for UAV operators. A safe
functioning of a swarm of UAVs cannot be assured unless the UAVs avoid colliding with one
another as well as static and dynamically emerging, moving impediments in the flying zone.

Strategy:
The proposed system predicts and avoids the following by using the geographical positions of
the UAVs and the successfully identified, static, and moving obstacles:

i. UAV-to-UAV collisions

ii. UAV-to-static-obstacle collisions

iii. UAV-to-moving-object collisions.

The suggested system's capacity to forecast probable collisions and proactively determine
optimum strategies to prevent predicted accidents in order to maintain the safety of the entire
swarm is a distinguishing feature.

Method
1. Collision-Free Path Generation and Navigation

The main components of the proposed system include the following:

i. CEP and collision prediction module

ii. Mutually exclusive locking mechanism

iii. Collision avoidance mechanism.

It may not be possible to instantaneously compute a bypass route for all drones in a densely
populated and crowded flying zone. In such cases, the suggested system may instruct some of
the drones to hover in place until the situation improves and the pathways are clear. It may
also allow some UAVs to briefly retreat or reverse in order to discover more acceptable,
collision-free routes.
The proposed system follows a safety-first approach. As a result, hazard-free, safe swarm
operation takes precedence over all other objectives, including UAV route lengths, timely
arrival of UAVs at their destinations, and completion of any other mission-specific goals.

2. Complex Event Processing and Collision Prediction

Complex event processing (CEP) is a technique for deriving and identifying relevant complex
events and patterns in event streams in real-time, using a large number of events from one or
more event streams. CEP has been used successfully in a wide range of corporate disciplines,
including retail management, health care, and cloud computing. As a result, CEP offers the
same capabilities for real-time event streams as a relational database management system
does for persistent data.

The event sinks make use of a variety of characteristics and criteria to predict the three main
types of collisions. The criteria include the drones' (present) locations as well as the positions
of any impediments, both stationary and moving, as well as the drones' desired future
destinations.

3. Mutually Exclusive Locking Mechanism

The majority of the potential UAV collision scenarios are covered by the CEP and collision
prediction module discussed in the preceding section. Nonetheless, errors and accidents are
still possible because the UAVs may move quickly and arbitrarily. The collision avoidance
system might not have enough time to avoid the collision if the CEP and collision prediction
module take a little longer to forecast the UAV-to-UAV collision.

The proposed system prevents several UAVs from attempting to go to the same area at the
same time by using mutually exclusive locks on the present and next-to-immediate positions
of UAVs.

4. Collision Avoidance Mechanism

Every time the CEP and collision prediction module anticipates a collision, it activates our
collision avoidance mechanism, which seeks out the best ways to avoid the anticipated
collisions and online calculates collision-free routes for UAVs.

Collision avoidance mechanism uses one of the three techniques in the following order:

i. Redirecting the UAV into another direction

ii. Putting the UAV into the hover-in-place mode until the route is cleared

iii. Temporarily retreating or backtracking the UAV to explore some alternate collision-free
routes.

It should be noted that all three collision avoidance techniques incur some overhead, which
might extend the routes and increase the flight durations for some of the UAVs.
Result
The outcomes of Experiment 1 demonstrate the suitability of the suggested system for lesser
problem occurrences. The outcomes of Experiment 2, which generated drone routes for a
more complicated issue instance, provide more evidence of the suggested system's
effectiveness. In Experiment 3, the flying zone was crowded with both stationary and moving
obstacles, which made it more difficult for the drones to manoeuvre. Although it required
them to go farther to get there, the suggested strategy was able to avoid all snags and
collisions and direct all drones to their destinations. It demonstrates that the suggested system
is equally appropriate for crowded, densely inhabited flight zones. The findings of
Experiment 4 demonstrate that the suggested strategy is equally appropriate for complicated
problem scenarios with large probabilities of drone collisions.

Conclusion
The outcomes demonstrate that the suggested method successfully anticipates and averts all
three types of collisions in real time. Additionally, it easily grows to large-sized issue
instances, provides safe and effective UAV routes, and is appropriate for crowded flight
zones and for scenarios with a high risk of UAV collisions.

My learnings:

Through this research paper, I have learned the following things:

Several unmanned aerial vehicles (UAVs) working together can deliver significant
capabilities that a single UAV cannot. Consequently, several UAVs can be utilised together
in the form of a swarm or a fleet to execute larger and more complicated applications and
tasks that are either beyond the capability of a single UAV or cannot be done effectively if
only a single UAV is used.
10. Optimal path planning for drones based on swarm intelligence
algorithm
Reference: Authors- Rashid A. Saeed, Mohamed Omri, S. Abdel-Khalek, Elmustafa Sayed
Ali & Maged Faihan Alotaibi

Challenges/problems: Path planning is a challenge that extends beyond simply trying to

find a route that will get you from point A to point B. It also involves picking the best path
out of all those that are possible and creating a system for avoiding collisions. While deciding
how to build the optimum path, numerous relevant factors must be considered, such as safety,
obstacle avoidance, response speed to overtake obstacles, and so on.

Strategy
The ability of drones to deliver aerial viewpoints is an essential characteristic of drones that
promises cost-efficiency for many military and civilian applications because to their
flexibility to move freely in an operating environment. As a result of impediments and
hazards, the deployment of a drone to such applications faces challenges in ensuring the
safety and reliability of drone operations. The path planning problem has become one of the
most active research topics in order to address the aforementioned obstacles. Path planning
methods allow the drone flight path to be determined, allowing the drone to move in a safe
and unobstructed area.

Method
Drones must have the ability to fly autonomously and execute a variety of activities, as well
as the ability to implement advanced mechanisms that allow them to avoid obstacles and
achieve flight stability using a methodology that defines the course and movement
coordinates. Path identification techniques collaborate with intelligent decision-making
systems to recognise and assess situations around drones in flight, as well as identify key
locations that require quick action and coordinate an appropriate response.

Drones must be distinguished by their ability to fly autonomously and by their diversity. As a
result, intelligent management mechanisms in drones enable them to analyse dynamic and
complicated emergency data in order to assess the flying location and then choose the flight
path based on trajectory planning.

A. Particle swarm optimization (PSO)

Particle Swarm Optimization (PSO) is a swarm intelligence approach that uses social
behaviour as an optimisation scheme. PSO runs population-based stochastic optimisation
processes with randomization based on population member interactions. PSO is a powerful
technique that employs the concept of swarm intelligence to solve complicated problems with
high accuracy and speed.

B. Ant colony optimization (ACO)

Ant colony optimisation (ACO) is another sort of method that solves the path optimisation
problem using swarm intelligence [57]. To manage the operations of generating and
arranging pathways, this programme focuses on replicating the behaviour of real ants in their
colony. Based to research on ant interactions, the algorithm employs the same interaction
behaviour that ants use to discover paths between starting places and destinations by
depositing a chemical substance known as the pheromone.

C. Artificial bee colony (ABC)

Swarm intelligence is a new algorithm based on bee behaviour that allows for optimisation of
the search mechanism based on the nature of the bee agent's interaction in the colony. The
ABC algorithm is based on three types of artificial bee colonies: worker bees, spectator bees,
and scout bees. The path from the bee colony site to the food location is planned by
determining the optimal solution to follow the path.

D. Firefly algorithm (FA)

The firefly algorithm is a novel technique for increasing swarm intelligence that is based on
research methodologies based on firefly social behaviour. The Firefly algorithm uses
bioluminescence with varying flashing patterns to replicate the social behaviour of flying
fireflies, which communicate and search for mates.

Because the algorithm for fireflies enhances some of the flashing qualities of fireflies, the
improvement process is dependent on fireflies being attracted to each other regardless of
gender, the firefly brightness, and the goal function based on the firefly brightness.

E. Cuckoo Search Algorithm (CSA)

Cuckoo Search Algorithm (CSA) is an algorithm inspired by cuckoo bird behaviour. This
algorithm concept is based on basic criteria that cuckoo birds follow to lay only one egg in
each stage, hence cuckoos choose the nest to place the eggs at random. In addition, cuckoos
only carried high-quality eggs and nests. This algorithm is based on the cuckoo's behaviour
when hunting for a nest, and it changes the solution on a regular basis by adjusting the nest
position.

Result
Drones are playing a significant role in the development of a wide range of surveillance and
research applications. However, the use of drones in various situations poses a significant
barrier in terms of task planning and a secure approach to obtaining data rapidly, particularly
in complicated environments. we have discussed the features and characteristics of different
swarm optimization algorithms such as ant colony optimization (ACO), firefly algorithm
(FA), artificial bee colony (ABC) particle swarm optimization (PSO) and Cuckoo Search
Algorithm (CSA).
Conclusion
Path planning methods enable the drone flight path to be derived so that the drone
can move in a safe and unobstructed area .The ACO algorithm's performance is
compared to the upgraded ACO algorithm. The suggested approach provides fast
convergence, which speeds up the path planning process.

My learnings:

Through this research paper, I have learned the following things:

While deciding how to build the optimum path, numerous relevant factors must be
considered, such as safety, obstacle avoidance, response speed to overtake obstacles, and so
on. Swarm optimisation methods were employed to give intelligent modelling for drone path
planning, allowing the best path for each drone to be built. This is done in accordance with
the swarm members' planning and coordination dimensions.
`11. Optimal Collision-Free Navigation for Multi-Rotor UAV Swarms in
Urban Areas
Reference: Authors- Xiangpeng Wan; Hakim Ghazzai; Yehia Massoud; Hamid Menouar

Challenges/problems: UAVs for smart cities and other applications have numerous
problems that limit their practical deployment. For example, battery limitations impede
UAV] operation range and continuous activity, necessitating required consideration of UAV
energy management throughout operations. In addition, there are many high buildings in
metropolitan areas that may obstruct UAVs' direct routes. In other words, their mobility is
limited by a number of permanent and non-permanent barriers that must be considered for
optimal path selection and increased navigation.

Strategy
The goal is to reduce the total time spent by UAVs travelling from multiple areas to different
destinations. Path planning is carried out while ensuring that each UAV has enough energy to
complete the journey. Charging stations are set up throughout the city to allow UAVs to
recharge their batteries on their way to their destinations. To establish independent 3D routes
of a fleet of UAVs in urban locations, a generic framework is built. The goal is to identify the
fastest trajectories for all UAVs such that the overall time spent by the UAV swarms is as
short as possible.

Method
The proposed framework's goal is to reduce the total time that UAVs spend navigating from
one location to another within the city, while accounting for probable extra time due to
imposed stops for battery charge. In general, the routes chosen for some UAVs may not be
the fastest from an individual standpoint, but generally correlate to the optimum overall
system routes.

Initially, we must verify that the UAV is travelling in such a way that it is positioned at one
path segment or at a charging station throughout each step, or that it is hovering at an
intersection as long as it has not arrived at the destination. In the latter instance, the UAV
comes to a halt.

Second, when crossing an intersection, we must verify that the UAV goes sequentially. For
example, if the UAV enters an intersection, it must leave it, hover there, or proceed to the
nearest charging station if one is available.

Each UAV's battery should be sufficient to allow it to fly until it reaches its objective;
otherwise, it must be recharged along the way. In other words, the energy consumption for
each step s should be less than or equal to the quantity of stored energy.

The energy consumption for the first three actions depends on the speed of the UAV
Result
In this research, a programme for determining a collision-free navigation solution for a fleet
of UAVs in metropolitan locations is presented. This is done while keeping in mind the
UAVs' energy limitations and their ability to recharge their batteries in charging stations set
up around the region of interest. A MILP is created to find the best and fastest paths for the
UAVs to take. To avoid collisions and find speedier paths, static hovering, safety time gaps,
and altitude adjustments are commonly used.

Conclusion
Path planning is one of the most important UAV problems that needs to be addressed. The
main goal is to identify the best paths for a fleet of UAVs in populated areas that reduce the
sum of the arrival times of all UAV swarms while taking into account their energy
consumption limits and minimising collision risks.

My learnings:

Through this research paper, I have learned the following things:

It is important to jointly optimize the application performances of drones and ensure
collision-free operation while taking into account important factors such as the energy
consumption. The path planning is performed while ensuring that the energy level of each
UAV is sufficient to complete the trip.
12. A formation maintenance and reconstruction method of UAV
swarm based on distributed control
Reference: Authors- Xiaowei Fu, Jing Pan, Haixiang Wang, Xiaoguang Gao

Challenges/problems: There are two challenges faced here. Firstly, the problem of
formation maintenance and reconstruction of UAV swarm without obstacles. Secondly, the
problem of formation maintenance and reconstruction of UAV swarm with obstacle
avoidance.

Strategy
By changing the relative positional relationship between each UAV and the virtual leader, the
swarm formation transformation is realized. To decide whether each UAV has to avoid
obstacles, a collision prediction algorithm is provided. The artificial potential field method is
then integrated with the aforesaid control rule to realise swarm formation control and
maintenance with obstacle avoidance.

Method
To enable formation maintenance and reconstruction of UAV swarms, we first design a
cooperative control law based on the consensus method. The artificial potential field is then
integrated with the consensus method to enable swarm formation maintenance and rebuilding
with obstacle avoidance. Lastly, we use total distance cost as the optimisation goal and
propose a CBBA-based optimisation technique for the swarm formation transition. The
control rule established in this research can be well adapted to the formation transformation
problem of UAV swarm in a real-time distributed way by using CBBA to allot the position of
each UAV from the initial formation to the target formation.

1. A formation maintenance and reconstruction method with obstacle avoidance

When an obstacle is detected, the UAV swarm should adjust its formation to avoid it, and
then reconstruct its formation once the obstacle avoidance is complete. The formation control
law (4), which is based on the consensus method described in the previous section, can only
assure that the swarm proceeds in the direction of the virtual leader and cannot alter its path
to avoid obstacles in the real world.

2. Formation maintenance and reconstruction based on CBBA

The UAV swarm formation necessitates that each UAV maintain a precise geometric
positional connection in space. The position vector of the UAV relative to the swarm
geometry reference point is preset in control inputs. The problem of this is that the algorithm
lacks flexibility and is unable to generate an optimal allocation solution based on the various
situations, raising the flight cost.
Result
This research focuses on distributed control-based formation maintenance and reconstruction
strategy for UAV swarms. First, the consensus procedure is used to study the maintenance
and reconstruction of the swarm formation. By creating the control rule to manage the
relative positional relationship between each UAV and the virtual leader to a preset value and
the swarm speed to the appropriate speed, the desired formation shape may be generated and
maintained.

Conclusion
Because the position of each UAV in the target formation is fixed using the swarm formation
transformation approach, CBBA is used to allocate positions from the starting formation to
the target formation to achieve flexible and effective formation transformation. The
simulation findings demonstrate that the UAV swarm can generate, maintain, and reconstruct
the desired formation in real-time while effectively avoiding obstacles.

My learnings:

Through this research paper, I have learned the following things:

Formation control is the foundation for ensuring a UAV swarm's seamless performance of
varied tasks, and it has emerged as a new research hotspot in academia. UAV swarm
formation control refers to multi-UAVs that can generate, maintain, and reconstruct
corresponding formations based on mission requirements or realistic surroundings. Various
swarm forms have varied applications, and popular formations include the V-shape, straight-
shape, circle-shape, and so on.
13. Swarm-based counter UAV defense system

Reference: Authors- Matthias R. Brust, Grégoire Danoy, Daniel H. Stolfi & Pascal Bouvry

Challenges/problems: The expansion of UAVs will come at the expense of dense, high-
speed, and dynamic traffic, which is prone to UAVs going rogue or being deployed with
nefarious intent. Counter UAV systems (C-UAS) are thus required to ensure the safety of
their operations. Existing C-UAS, the most of which are from the military arena, lack
scalability or cause collateral damage.

Strategy
To ensure safe UAV operations, efficient and scalable counter-UAV systems (C-UAV) must
be developed. C-UAVs enable the detection and identification of malicious UAVs, as well as
the use of sensing and mitigation devices. Ground-based systems, such as radio frequency
(RF) of Global Navigation Satellite System (GNSS) signal jamming to damage UAV
communication and navigation systems, and net or laser guns, which impact the physical
integrity of the malicious UAV, are among the most prominent. They, on the other hand, may
cause collateral damage and are not scalable.

Method
Counter UAV defence system approach

The approach consists of a swarm of dUAVs that form a 3-dimensional cluster around the
mUAV, giving the mUAV a limited range of movement options. As a result, we infer that the
mUAV is attempting to avoid collisions with dUAVs in order to continue operating. The
dUAVs can then escort the mUAV outside of the flying zone by encircling it. Patrolling,
detection, interception, and counter-attack are the four phases used to realise the escort
manoeuvre (job, problem) in the proposed strategy.

The patrolling phase is easily stated as the defence UAVs surveying the fly zones quietly
until an abnormality is detected, which activates the following phase. The clustering and
formation algorithms for the detection and interception phases are conducted concurrently
throughout the mission, while the CH determines the transition between the interception and
counter-attack phases.

The cluster structure consists of the CH in the centre of the cluster, acting as the cluster's
coordinator, and a series of branches that arise from the CH. The formation is the rationale
for keeping a balanced structure. The goal is to build a clustering that is ideal for the intended
formation, which resembles a closed globe in which the CH seeks to contain its branches in
order to catch the mUAV. As a result, for the branches to be balanced, they should ideally be
the same length.

Clustering is done entirely locally at each UAV. We may differentiate three different states in
which a UAV can be: UAV, dUAV, and CH. The distinction between UAVs and dUAVs is
that UAVs have no parent and hence are not in a cluster and are looking for one, whereas
dUAVs are cluster members who are coordinated by the CH to fulfil the escort task.

1. Behaviour of cluster head (CH)

As soon as a UAV is chosen as CH, it begins accepting children. CHs then notify the children
that they can now begin accepting another child. We distinguish between two types of
messages:

i. The main message

These messages are issued by the CH to its branches in order to prompt them to accept
another child or to reject the existing child. Discarding a branch's child results in the
discarding of dUAVs because the message is delivered recursively to all children.

ii. Command message

CMs are sent recursively from dUAVs to the CH to notify of a new child.

Result
This study offers a novel autonomous defence system (C-UAV) made up of UAVs that act
together as a swarm. The most notable features and contributions of the current approach are
the balanced clustering and local formation control to enable intercept- and capture-
formation, hence avoiding the need for additional infrastructure such as GPS. The approach
also includes a modular design with the phases deployment, clustering, formation,
interception, and capture. Simulations showed empirical evidence of the swarm-based C-
ability UAV's to chase and escort a mUAV. Furthermore, a sensitivity analysis of the
system's characteristics was proposed, including the number of defence UAVs,
communication range, wobbling radius, and collision threshold.

Conclusion
In this research, a C-UAS capable of intercepting and escorting intruders is proposed. It is
based on an autonomous defence UAV swarm, which is capable of self-organizing their
defence configuration and intercepting the malevolent UAV. This entirely localised and GPS-
free technique employs a modular defence phase architecture and a newly created balanced
clustering to accomplish intercept- and capture-formations.

My learnings:

Through this research paper, I have learned the following things:

Unmanned Aerial Vehicles (UAVs) have quickly established themselves as one of the most
potential Internet-of-Things (IoT) gadgets for smart cities. UAVs have already showed
enormous potential for a wide range of commercial applications due to their mobility, agility,
and customizability of onboard sensors. The expansion of UAVs will come at the expense of
dense, high-speed, and dynamic traffic, which is prone to UAVs going rogue or being
deployed with nefarious intent. Counter UAV systems (C-UAS) are thus required to ensure
the safety of their operations. Existing C-UAS, the most of which are from the military arena,
lack scalability or cause collateral damage.
14. 3D Trajectory Planning Method for UAVs Swarm in Building
Emergencies
Reference: Authors- Ángel Madridano, Abdulla Al-Kaff, David Martín and Arturo de la
Escalera

Challenges/problems: It is critical to develop trajectories that allow safe mobility in the

environment, avoiding both static and dynamic barriers, for autonomous MRS capable of
carrying out a sequence of works in a specified scenario. At this point, trajectory planning
becomes critical for these types of systems to be able to complete more complex missions in
less time.

Strategy
There are two phases to the planning and set of actions that allow many vehicles to achieve a
goal:

i. Task Planning: For a given set of vehicles and destinations, the most optimal vehicle-
objective combination is determined, taking into account factors such as the vehicle's speed
and endurance, the distance to the target, and the task to be completed.

ii. Trajectory Planning (also known as needed movements): Once the objective has been
assigned, safe and efficient paths are generated for one or more vehicles to be able capable of
reaching this location .

Method
1. 3D Multi-Trajectory Planning Based on PRM

The goal of this study is to propose a collection of algorithms that can construct a variety of
optimal paths from 3D occupancy maps, allowing different drones to independently reach
different spots of interest inside the work area. As described in this section, algorithms based
on PRM are used in conjunction to explore a 3D environment and determine the ideal set of
paths for each aircraft to attain a certain objective in such a way that the mission as a whole is
created in the most efficient way feasible.

In addition to extending the work from 2D to 3D, all approaches are integrated in Gazebo's
simulation environment, allowing visualisation and testing of the algorithms prior to
deployment with real UAVs.

A MRS must plan safe and efficient paths in order to fulfil a series of activities. The
pathways generated in this scenario are targeted towards the deployment of a swarm of UAVs
to support ERT in metropolitan environments, particularly in cases of large structures with a
large area to cover. Thus, a centralised and scalable architecture is presented in which a
heterogeneous swarm of UAVs is established using the Robot Operating System Framework
(ROS) and the MavLink communication protocol (MAVROS) that can reach strategic places
in a short period of time through safe and efficient routes for the support and collection of
information required to increase the response and success of the ERT's tasks.

2. 3D Occupancy Map Generation

The initial stage in path planning is to create a 3D occupancy map of the work area where the
swarm will be operating. The information collected from Octree algorithms is used for this.
The octree data structure is a hierarchical data structure that subdivides space into cubes or
voxels until a specific resolution is reached. One of these volumes' qualities is the ability to
determine if it is free or occupied, and this information is used to guide the investigation of
the work area.

Another option for obtaining the occupation map is to deploy a set of vehicles equipped with
sensors throughout the work area in order to obtain a complete mapping, for which full
coverage and sensorial fusion algorithms that compile the dataset from each of the vehicles
and generate a complete map would be required.

3. Generation of Paths

The final stage is to determine the best way to connect the UAVs' initial position to the goals.
As stated in prior work, this final stage is directed to various situations, with the results
focusing on labelled and unlabeled examples. Despite the fact that there are various
conditions, path generation has a consistent evolution for all cases and adds a technique in the
case of designing trajectories for unlabeled cases. Once the PRM is formed, the common
component is to discover the best path. The rationale for this second phase is the PRM
algorithm's passive nature, which randomly investigates the surroundings but is incapable of
providing a solution.

4. Collision Avoidance

Once the ideal path combination has been determined, it is vital to ensure that the UAVs
follow these trajectories without colliding with one another.

If two or more UAVs approach the same spot, a mechanism is created to check their positions
and adjust their speeds. In this manner, the UAV moves autonomously at a speed determined
by the autopilot, until two or more UAVs come within a safety margin determined by the
user. In this instance, one of the two UAVs' speeds is reduced until the distance between
them is safe. In the case of more than two UAVs, each UAV except one has a separate speed
reduction.

Result
Simulations are critical in the field of autonomous vehicles, particularly UAVs, to test and
validate the implemented algorithms because even little errors in the methods created
frequently result in a failure in the control of the UAV and the loss of sensors or pieces of
high economic value.
As a result, a software architecture is created in Gazebo's simulation to allow testing of the
various MRS missions. In the field of robotics, and particularly in the field of unmanned
aerial vehicles (UAVs), simulation is an important stage before implementing an algorithm in
actual systems, because any minor error or failure of the methods created in mid-flight
frequently results in the loss of system control.

Conclusion
This research describes a method for 3D trajectory planning for a UAV swarm based on PRM
for urban situations, which may be applied to labelled and unlabeled scenarios while assisting
an Emergency Response Team. This work also contains a ROS architecture for simulation,
which establishes a communication bridge between ROS and the MavLink protocol.

My learnings:

Through this research paper, I have learned the following things:

Unmanned Aerial Vehicles (UAVs) and Unmanned Ground Vehicles (UGVs) have
established themselves at the forefront of significant fields of study aimed at a variety of
activities such as infrastructure inspection, surveillance, search and rescue, and delivery.
15. Collision-free trajectory planning based on Maneuver Selection-Particle
Swarm Optimization
Reference: Authors- D. Alejo; J. A. Cobano; G. Heredia; A. Ollero

Challenges/problems: In cooperative missions involving many aerial aircraft, safety is

critical. Several UAVs provide significant advantages for carrying out cooperative missions,
although the likelihood of vehicle crashes increases in these cases. Furthermore, while flying
in small-sized scenarios with multiple vehicles, the problem becomes more difficult to
address.

Strategy
As a result, a system that can plan collision-free trajectories and resolve detected conflicts
while taking into account as many potential manoeuvres as feasible is necessary and
important. Additionally, each vehicle should follow the mission trajectory as closely as
feasible, with no deviations from the original design.

Method
This work considers the challenge of collision-free trajectory planning of several UAVs
performing coordinated missions. Furthermore, the system must produce manoeuvres quickly
online since variations from the initial trajectories of each UAV may occur owing to
perturbations (e.g., wind) that could cause collisions. When a conflict is found during mission
execution, the system should re-plan collision-free trajectories. A sequence of waypoints
defines each 3D trajectory. UAVs share a common workspace and the separation between
them should be greater than a given safety distance.

The information that the system needs in order to solve the problem is the following:

1. Initial spatial trajectory of each UAV.

2. Parameters of the model of each UAV

3. Location of each UAV

To avoid collisions between UAVs, a security envelope is built around each one. Each UAV
security envelope consists of two boxes connected together: a horizontal box and a vertical
box that covers the aerial robot and its rotors. Each box is defined by the intersection of three
intervals, one of which is determined by an axis. The horizontal box measurement is related
to the minimum horizontal spacing between UAVs, and the vertical box measurement is
related to the vertical separation.
1. Continuous Collision Detection

It is critical to evaluate the detection algorithm's computational load because evaluating

probable solution trajectories is computationally expensive. A simple collision detection
technique would include simulating the trajectories of each UAV at a fixed sample rate. Next,
in each sampled instant, a collision detection is conducted. This strategy, however, has two
significant limitations. For starters, a full simulation of a UAV's motion can be quite
expensive because it comprises a system of multiple linked differential equations. Second, the
sampling rate parameter must be carefully set because lower values may necessitate a
disproportionate amount of processing effort, whilst higher values may leave some collisions
undiscovered.

2. Collision-Free Trajectory Planning Algorithm

The collision-free 3D trajectory planning system is based on Particle Swarm Optimization, a

heuristic global optimisation algorithm (PSO). PSO is an iterative process, and the result
improves over time. As a result, it is assured that a feasible solution is available at all times,
and that the quality of this solution will improve if more execution time is provided.

The PSO algorithm was created using swarm intelligence and is based on studies of bird and
fish flock movement behaviour. It works by keeping a swarm of particles moving around in
the search space, influenced by both the improvements identified by other particles and the
improvements achieved by the particle thus far.

Result

The proposed system automatically identifies conflicts among UAV and resolves them
cooperatively using a collision-free trajectory planning algorithm based on the named
Maneuver Selection Particle Swarm Optimization (MS-PSO). The improvements of this new
implementation of the PSO are:

i. The times of execution are reduced due to the reduction of the dimensionality of the
problem and the CCD algorithm.

ii. Different kinds of manoeuvres can be selected to solve the conflicts: course, velocity or
altitude change.

Conclusion
It is vital to note that the system has been validated by several simulations in various settings.
Also, the system may adapt to various cost functions. Furthermore, the ranges used to specify
when one manoeuvre is executed could be tailored to each UAV's preferences. For instance,
one manoeuvre may be promoted more than the others.
My learnings:

Through this research paper, I have learned the following things:

This work describes a system for collision-free trajectory planning with several Unmanned
Aerial Vehicles (UAVs) that detects conflicts automatically. When problems between UAVs
are detected, the system collaboratively resolves them using a collision-free trajectory
planning algorithm based on a stochastic optimisation technique known as Particle Swarm
Optimization (PSO). The novel Maneuver Selection Particle Swarm Optimization (MS-PSO)
implementation of the PSO algorithm outperforms prior implementations.
16. Using Optimization, Learning, and Drone Reflexes to Maximize Safety
of swarm drones
Reference: Authors- Amin Majd; Adnan Ashraf; Elena Troubitsyna; Masoud Daneshtalab

Challenges/problems: Despite the growing popularity of swarm-based applications of

drones, there is still a lack of approaches to maximize the safety of swarms of drones by
minimizing the risks of drone collisions. Maintaining a risk-free, safe flight is especially
critical for indoor applications.

Strategy
We offer a parallel and dynamic implementation of the imperialistic competition algorithm
(ICA) that allows us to find optimal collision-free paths for the drones in the swarm. The K-
nearest neighbour (KNN) learning algorithm is used in the learning component. To train the
system, each placement produced by our parallel ICA for a particular swarm and environment
state is evaluated. To compute alternate routing options, both the learning and optimisation
algorithms run in parallel.

Method
A drone swarm is an example of a complicated distributed networked system. Each drone can
be thought of as a mobile sensor node capable of gathering monitoring data and
communicating with other drones in the swarm as well as the cloud-based navigation centre.
It is a complex optimisation issue to find an efficient and safe route for each drone. As a
result, we must rely on heuristics to get the desired results. In this research, we compute
drone paths using a dynamic EA. The method proposed is based on the imperialistic
competition algorithm (ICA).

The algorithm in our approach takes as input K solutions developed for the same system
state. The average solution is then computed. Each drone in the swarm follows the K routes
that have been proposed to it. The average route is then calculated in terms of Euclidian
distance. The learning component computes the whole solution for the swarm by computing
such a path for each drone in the swarm.

Result
The subject of semi-autonomous robotic system motion safety is currently gaining a lot of
academic attention. The book provides a complete overview of the issues related with
autonomous mobile robots. According to the results of the investigation, the most popular
routing strategies do not ensure motion safety. Our method overcomes this problem by
ensuring not just safety but also effective online routing.

Conclusion
We developed a strategy in this research that leverages optimisation' learning and automatic
rapid responses (reflexes) of drones to assure safe and efficient operations of swarms of
drones. The proposed method combines a high-performance dynamic evolutionary algorithm
and a reinforcement learning algorithm to design safe and efficient drone routes, which are
then supplemented with dynamically computed drone reflexes to avoid collisions with
unforeseen obstructions in the flying zone. We also showed and assessed a parallel
implementation of the proposed technique against two benchmarks. According to the
findings, the suggested method maximises safety, creates highly efficient drone routes, and
has a minimal computation overhead.

My learnings:

Through this research paper, I have learned the following things:

The proposed method combines a high-performance dynamic evolutionary algorithm and a
reinforcement learning algorithm to design safe and efficient drone routes, which are then
supplemented with dynamically computed drone reflexes to avoid collisions with unforeseen
obstructions in the flying zone. In addition, we show a parallel implementation of the
suggested technique and compare it to two benchmarks. The results show that the suggested
method maximises safety while also producing highly efficient drone routes.
17. Brief analysis of drone swarms communication
Reference: Authors- Qiannan Cui; Peizhi Liu; Jinhua Wang; Jing Yu

Challenges/problems: Swarm networks can transition from sluggish to fast dynamic,

have intermittent connections, and have a fluid configuration.

Strategy
Efficient communication methods are a must for drone swarms and their effective
deployment. A big swarm need a completely dispersed communication system that scales
well and maximises "many-to-many" communication. A drone swarm is essentially a floating
dynamic wireless network, sometimes known as a wireless mesh network.

Method
1. Overview of Wireless Mesh Networks

Wireless Mesh Network (WMN) is marketed as a low-cost cooperative wireless connectivity

solution with multiple applications in areas such as enterprises, households, disaster,
emergency, and transportation. A mesh network is a network structure in which each node
relays network data. In the network, all nodes work together to distribute data. A completely
linked network is a mesh network with all of its nodes connected to each other.

2. Routing Technique
When a routing strategy is used, the message is transmitted along a path by hopping from
node to node until it reaches its target. A routing network must allow for continued
connections and reconfiguration around broken or blocked paths, utilising self-healing
algorithms, to assure the availability of all its paths.

When one node fails or a link fails, the self-healing capacity allows a routing-based network
to continue operating. As a result, because there is frequently more than one path between a
source and a destination in the network, the network is typically extremely reliable.

3. Flooding Technique

Instead of transmitting a message from one node to another via a defined route, the message
is broadcast to all nodes in the network. The Flooding approach is both simple and
dependable.

Because there is no routing, there are no advanced routing algorithms. No routing means no
network administration, no requirement for self-discovery, no need for self-repair procedures,
and no expense for delivering routing tables or routing information because the message is
the "payload."

Flooding technology has extra benefits in terms of propagation. Signals arriving at each node
via several propagation pathways benefit from the inherent space diversity, maximising
network resiliency in the face of blockages, interferences, and multipath fading resistance,
with virtually no single point of failure.

Result
We were able to gain an understanding of two common drone swarm communication
mechanisms. We will learn how to overcome the issues of drone swarm communication
design and how they are applied in different settings based on the comparison and study of
these two sample solutions.

The new research will allow for the development of a more secure and efficient drone
swarms communication network, giving users a better view of applications and making them
more secure and robust to use.

Conclusion
In recent years, swarm intelligence has gained a lot of study attention. Swarm of drones
systems, as a common application of swarm intelligence, often demonstrate decentralised
control achieved by basic agent behaviours and interactions, generating a self-organization
that is considered an emergence of order from the system. The system as a whole is larger
than the sum of its components, capable of addressing complicated issues that no single agent
could solve.

My learnings:

Through this research paper, I have learned the following things:

This thesis will quickly explain two common communication approaches and compare them,
including their relative merits and limitations as well as the many application contexts for
which they are suited.
18. A centralized control of movements using a collision avoidance
algorithm for a swarm of autonomous agents
Reference: Kleber Loayza, Pedro Lucas, Enrique Peláez

Challenges/problems: The use of UAVs (Unmanned Aerial Vehicles) in various domain

applications has increased in recent years, particularly to perform complex tasks; such
domain applications are taking advantage of the flexibility and robustness that swarm
behaviour can provide, where a group of individual agents perform simple tasks, but
collaboratively can overcome the complexity of difficult-to-solve problems.

Strategy
The method employs a centralised technique that takes into account trajectory calculation and
collision avoidance. The method was tested in both a simulated and real-world controlled
environment, employing a swarm of nano drones and a setup backed by a motion capture
system.

Method
The solution was first implemented in a simulation environment to test the effectiveness of
the centralised model in controlling and providing real-time feedback to the agents, as well as
the collision avoidance algorithm while increasing the number of agents to perform the task
of sowing seeds.

1. Centralized Swarm Method

To complete the task of sowing seeds, the swarm of drones must resolve the required planting
movements in a ground and modelled as a grid composed rows and columns, where each
element is a position in space.

Each point is separated by a distance. Its handled as a one-dimensional array that the agents
browse to determine the position where they need to go in order to collaborate in the seeding
process.

The strategy guides an agent along a straight line from its current position to a target position
at a speed.

When the agent completes its mission and returns, the target position is a selected position in
the set of points of the space, or the starting point .

When the is Near indication detects a prospective collision, the agent shifts its direction of
motion to the opposite side of the collision using the normalised vector. The directional
vector is defined to avoid the collision.

The collision detection drives the agent into a Moving Away state in order to move away
from the other agents using the directional vector defined in the equation. The agent employs
equation for moving away during a time defined by to achieve this displacement. When the
agent is no longer in the collision range, it resumes the trajectory to the target and returns to
the Moving state.

Result
The proposed solution was tested in two environments: a simulation and a real-world
configuration. The Unity3D simulator has a component-based programming paradigm;
therefore the agents are objects with unique behaviours that are implemented as one of its
components that takes into account the finite state machine. Because of space constraints and
the effectiveness of MOCAP, the field was configured with a size of in the real arrangement.
The agents were divided into groups of one to five drones, with 30 experiments for each
group and three obstacles assigned at random. The task is performed by a swarm of drones,
with the grid of holes for seeding designated as white areas on the floor, the drones circling,
and the obstacles enclosed in a square.

Conclusion
This work describes a centralised technique for directing a swarm of UAVs using an indirect
local communication mechanism that demonstrated low variability in accomplishing a
sowing seeding task in two settings, a simulation and a real-world controlled environment.

The results show that the time required completing the task decreases in a non-linear manner
as the number of individuals increases, demonstrating the behaviour and effectiveness of a
scalable system that improves or degrades in time as the number of agents changes and
possible collisions occur. Because the central control only updates the state for a location to
be worked on the ground, the system is also stable when an agent is separated from the
swarm owing to accidental events.

My learnings:

Through this research paper, I have learned the following things:

The method makes use of a centralised strategy that takes collision avoidance and trajectory
calculation into account. A swarm of nano drones and a motion capture system were used to
implement the solution both in a simulated scenario and in a real, controlled environment.
The solution was tested by sowing fictitious seeds on a field made up of a grid of points that
stand in for the actual sowing locations. In order to assess the efficacy, scalability, and
stability of the solution as well as the resilience of the collision avoidance algorithm while
expanding the number of agents to do the task, experiments were conducted for evaluating
completion times and attempts to prevent impacts.
19. IoD swarms collision avoidance via improved particle swarm
optimization
Reference: Gamil Ahmed, Tarek Sheltami, Ashraf Mahmoud, Ansar Yasar

Challenges/problems: Drones flights have been researched widely. In the presence of

high density and complex missions, collision avoidance among swarm of drones and with
environment obstacles becomes a tough task and vital.

Strategy
The goal is to improve the particle swarm optimisation (PSO) algorithm to improve the
optimality and speed of three-dimensional IoD path development. Among the enhancements
is the use of chaotic map logic to populate the PSO. Furthermore, adaptive mutation is used
to balance local and global search. The inactive particles are then replaced by new fresh
particles, pushing the solution closer to global optimum.

Method
1. Traditional Particle Swarm Optimization (PSO)

Particle swarm optimisation (PSO), a population-based meta-heuristic search technique, has

been widely applied to a wide range of issues. The procedure begins with a random candidate
solution of position and velocity, which is then modified iteratively until a global optimal
solution is identified.

2. IoD path planning description

Some elements must be considered when planning the course of IoD formation, such as
terrain area containing barriers in the terrain and IoD formation. Mountains, buildings, radars,
and other dangers may be encountered in the mission environment. Furthermore, the IoD
formation contains a large number of drones flying in this area. As a result, the IoD formation
should be able to avoid these impediments.

3. Path planning algorithm for IoD formation

The path planning generator takes parameters and environmental limitations and generates
possible paths with associated average fitness for all drones. Some enhancements to standard
PSO are implemented to increase search efficiency. Furthermore, it changes deceptively with
each repetition.

At the end of each repetition, the trail of all particles is examined, and the inactive particle is
replaced by a refreshed particle if its trail exceeds the threshold. The procedure is then
repeated until an ideal best is produced or time runs out.
Result
The IoD formation consists of multiple drones, each of which is subject to distinct
environmental limitations. Thus, analysing the entire formation is more effective than
arranging a single drone path. In this paper, the development of IoD is examined and
compared with SPSO and CIPSO utilising the suggested IPSO. The convergence speed is a
crucial feature to consider while evaluating the method, and it can be calculated using the
fitness value. IPSO has a faster convergence rate than SPSO and CIPSO.

Conclusion
The new strategy for IoD formation route planning was proposed in this research.
Furthermore, the goal of improved particle swarm optimisation (IPSO) is to improve ordinary
PSO's variety capabilities. The findings showed that the suggested method outperformed the
others, providing rapid and accurate IoD creation routes with obstacles and member collision
avoidance.

My learnings:

Through this research paper, I have learned the following things:

By enhancing the particle swarm optimisation (PSO) technique, this research attempts to
improve the optimality and speed of three-dimensional IoD path development. Among the
enhancements is the use of chaos map logic to start the PSO's population. In order to balance
local and global search, adaptive mutation is also used.
20. Drone Swarms as Networked Control Systems by Integration of
Networking and Computing
Reference: Godwin Asaamoning, Paulo Mendes, Denis Rosário, and Eduardo Cerqueira

Challenges/problems: Because of their cooperative behaviour, multi-agent systems such

as drone swarms have received increased attention. Nonetheless, automating swarm control is
difficult since each drone operates under varying wireless, networking, and environmental
restrictions.

Strategy
To address these issues, we consider drone swarms to be Networked Control Systems (NCS),
in which the total system is controlled within a wireless communication network. This is built
on a tight integration between the networking and computational systems, with the goal of
efficiently supporting basic control operations, such as data gathering and exchange,
decision-making, and actuation command distribution.

Method
1. Drone Swarms as Networked Control Systems

In most circumstances, drone systems are mission-oriented. The missions may be of varying
scales and complexities, necessitating the deployment of networks employing various
methodologies. A drone system in a simple and low-scale mission may consist of a single
drone or a number of non-interacting drones linked to a ground control centre. Multiple
interactive drones, or a drone swarm, may perform better on bigger scale tasks due to
consensus and swarm intelligence.

2. Networked Control System Model

A drone swarm can be treated as an NCS composed of a collection of computing systems

(i.e., the drones) linked by a communication network; an NCS is a computational system
operated in a closed loop via a communication network. Control and feedback messages are
specifically transferred among system computing units, or agents, in the form of an
information packet broadcast over a network.

A typical NCS's functionality is founded on two fundamental elements:

(i) A computer system capable of gathering data via sensors, making decisions, and carrying
out commands via actuators

(ii) A network based on communication modules, standards, and protocols (e.g., media access
control and routing) to permit information exchange.
3. Non-Interactive Deployment Strategy

Each drone is directly connected to a ground control station in a non-interactive deployment

method, which is used to monitor their status (e.g., location, sensor condition, networking
configurations), make choices, and send out new directives, such as new waypoints.

A non-interactive deployment technique results in a simple example of a network that cannot

be divided further. This method might be considered a single NCS because each drone can
construct a controlling closed loop to optimise its operation. The capacity to communicate is
also required because the drone must still engage with ground control to complete its system
operation of seeing, communicating, processing, and controlling.

4. Interactive Deployment Strategy

Drone swarms can be deployed totally autonomously in an interactive deployment strategy,

making their own decisions (e.g., about course adjustment) based on their assessment of the
surroundings, including neighbour drones. In an interactive deployment strategy, the drone
swarm will be able to coordinate their operation based on a cooperative approach to sensing,
monitoring, and information exchange, allowing them to reach an agreement on the best way
to accomplish a mission when confronted with detected obstacles.

Swarms rely on airborne networks with potentially strong line of sight attributes, which
facilitate data transmissions, as well as the ability to generate appropriate routing tables,
allowing drones to make decisions based on a good understanding of the nearby neighbour.
This localised operation, typical of self-organized systems, does not necessitate a networking
system capable of generating a complete image of the entire drone swarm, which is
advantageous in dynamic conditions.

Swarm Intelligence

Swarm intelligence aims to use drone system operation to demonstrate advanced and
complex swarm behaviours through collaboration, organisation, and information exchange. A
swarm intelligence system is made up of a group of simple agents (drones in this case) that
interact locally with one another and with the environment. Swarm intelligence systems are
designed to emulate biological systems in which agents (e.g., ants, bees) follow simple rules
that are not directed by any central body. Interactions between agents in these systems result
in the evolution of global behaviours.

5. Computing and Networking Integration

The necessity to generate a closed loop in drone swarms necessitates a close coupling
between the computational and networking systems, implying reciprocal influences and
dependencies. As a result, understanding such consolidated system impacts is critical for
making the most use of the limited resources that each drone may have to undertake sensing,
processing, and networking tasks.
Networking, on the surface, adds to the computational system by establishing a network
topology capable of supporting the essential data exchange between all drones in the swarm.
The data necessary for the computational system is transferred via the networking system,
and the results of each drone's computational system are typically shared with others via the
networking system, allowing the total system to establish consensus.

As a result, the networking system's performance may have a substantial impact on the
processing capability of the drone swarm, particularly for jobs that require real-time data.

Result
Based on the findings of this study, the next step would be to investigate a service-aware
networking framework based on an extension of the name function networking paradigm,
where services are tied to drone swarm operation, namely intelligent swarming and consensus
algorithms. The challenges involve designing the optimum networking and computing
coupling, such as selecting the ideal places (drones) to conduct the required computational
processes, and how to facilitate service chaining based on the dynamic features of a swarm of
drones.

Conclusion
Drone swarms and NCSs are gaining popularity. The development of drone swarms as NCS
is expected to aid in increasing their performance when confronted with operational and
environmental problems. As a result, this review contributes to a deeper understanding of
how drone swarms operate as NCSs. We began by analysing the topologies and technology
required to implement two types of deployment techniques.

My learnings:

Through this research paper, I have learned the following things:

In this paper, we provide a general overview of how to integrate a computing system and a
networking system to create self-organized drone swarms called NCS. As a result, we refer to
the characteristics of the suggested drone swarm components as an NCS in terms of
networking and processing systems. In order to improve the performance of a drone swarm,
we also examine how they are integrated. The integration of network and computers in a
drone swarm as an NCS is our final design option, along with a list of open research
challenges.