VISVESVARAYA TECHNOLOGICAL UNIVERSITY

Belagavi, Karnataka - 590018

Technical Seminar Report

on

“AI in Defence Sector”

Submitted in partial fulfilment of the requirement for the award of the Degree of
BACHELOR OF ENGINEERING
In

ELECTRONICS AND COMMUNICATION ENGINEERING

By

Vaishnavi P
1JS19EC155

Department of Electronics & Communication Engineering

JSS ACADEMY OF TECHNICAL EDUCATION
Dr. Vishnuvardhan Road, Bengaluru - 560060
 JSS ACADEMY OF TECHNICAL EDUCATION
JSS CAMPUS, DR.VISHNUVARDHAN ROAD, SRINIVASAPURA, BENGALURU - 560060
DEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING

CERTIFICATE

This is to certify that Ms. Vaishnavi P bearing the USN 1JS19EC155,

VIII semester B.E. in Electronics & Communication Engineering
has presented and successfully completed the Technical seminar titled
“AI in Defence Sector” for the partial fulfilment for the award of Bachelor of
Engineering Degree under Visvesvaraya Technological University Belagavi,
during the academic year 2022-23.

Seminar In charge Head of the Department

Dr. Veeramma Yatnalli Dr. P M sivakumaraswmy,
Asso.Professor. Professor and HOD,
Dept. of ECE, Dept. of ECE,
JSSATEB, Bengaluru JSSATEB, Bengaluru

Panel Members: Signature

Dr. Aravind H S, Professor
 JSS ACADEMY OF TECHNICAL EDUCATION
JSS CAMPUS, DR.VISHNUVARDHAN ROAD, SRINIVASAPURA, BENGALURU - 560060
DEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING

Declaration

I, Vaishnavi P, student of B.E. Electronics and Communication Engineering at

JSS Academy of Technical Education, Bengaluru, hereby declare that the

Technical Seminar entitled “AI in Defence Sector” has been presented

independently at JSS Academy of Technical Education, Bengaluru.

Vaishnavi P

1JS19EC155

(B.E. Electronics and Communication),

JSSATE, Bengaluru-5600060
 ACKNOWLEDGEMENT

The successful completion of any task would be incomplete without the

mention of people whose constant guidance and encouragement crowned my
effort with success.

I am grateful to our principal, Dr. Bhimasen Soragaon, for the support.

I convey my gratitude to Dr. P M Shivakumarasamy Prof. and HOD,

Department of Electronics & Communication Engineering for constant
encouragement.

I thank panel members Dr. Veeramma Yatnalli, and Dr. H S Aravind for their
immense support , guidance and ideas for the successful completion of the
Technical Seminar.
I thank teaching and non-teaching staff for their cooperation.

Finally, I thank my Parents for their support and encouragement.

Vaishnavi P

(1JS19EC155)
 ABSTRACT

The use of drones or Unmanned Aerial Vehicles (UAVs), both for

military and civilian purposes, has increased in India in the past
decade. At the same time, counterdrone systems are also being
developed to address the threats posed by UAVs. How effective are
these counter-drone mechanisms? This brief explores this question,
and offers suggestions for India to reduce the growing threat from
drones. Any evaluation of the efficacy of anti-drone systems has to be
conducted in view of current technologies such as Artificial
Intelligence (AI), cognitive Global Positioning System avoidance, and
hardware sandboxing—and such is the aim of this brief.
 Contents

Sl No Chapter Page No

1. Introduction 1
2. Literature Review 2
3. Practical Using of Neural Networks in Military Applications
4. Drone Feed Analysis 7
4.1 Introduction 7
4.2 Principle 8
4.2.1 Detection Techniques 8
4.2.2 Classification and Localisation Techniques 9
4.2.3 Jamming and Countering Techniques 11
4.3 Architecture and Implementation 12
5. Challenges of Drone Feed Analysis 18
6. Advantages of AI in defence sector 20
7. Disadvantages of AI in defence sector 21
8. Applications
9. Impact of Using Artificial Intelligence in Military on Society
10. Conclusion 22
11. References 23
 List of Figures

Figure No Name Page No

1 Object Identification and next actions to be taken 7
2.1 Micro-doppler Radar DroMiAn (Drone detection with 10
micro-Doppler Analysis)
2.2 Cepstrum of Four Rotor Blades (Second) 11
3 Envisioned Counter Drone System – Layered Architecture 14

4 Drone Dome Brochure (Open Source) – Array of sensors 15

and integration
5 Firefly Brochure (Open Source) – Tactical loitering weapon 16
to ‘Kill’ Drones
AI in Defence Sector

Chapter 1: Introduction

1.1 ARTIFICIAL
The simple definition of artificial is that objects that are made or produced by human
beings rather than occurring naturally.

1.2 INTELLIGENCE
The simple definition of intelligence is a process to entail a set of skills of problem
solving, enabling to resolve genuine problems or difficulties that encounters and to
create an effective product and must also entail the potential for finding or creating
problems and thereby laying the groundwork for the acquisition of new knowledge.
1.3 ARTIFICIAL INTELLIGENCE
Artificial intelligence is a branch of science which deals with helping machines
find solution to complex problems in a more human like fashion. This generally involves
borrowing characteristics from human intelligence, and applying them as algorithms in a
computer friendly way. A more or less or flexible or efficient approach can be taken
depending on the requirements established, which influences how artificial intelligent
behavior appears.
Artificial intelligence is generally associated with computer science, but it has
many important links with other fields such as maths, psychology, cognition , biology and
philosophy , among many others . Our ability to combine knowledge from all these fields
will ultimately benefits our progress in the quest of creating an intelligent artificial being.
A.I. is mainly concerned with the popular mind with the robotics development, but
also the main field of practical application has been as an embedded component in the
areas of software development which require computational understandings and modeling
such as such as finance and economics, data mining and physical science.
A.I in the fields of robotics is the make a computational models of human thought processes.
It is not enough to make a program that seems to behave the way human do. You want to
make a program that does it the way humans do it.In computer science they also the problems
bcoz we have to make a computer that are satisfy for understanding the high-level languages
and that was taken to be A.I. The intellectual roots of AI, and the concept of intelligent
machines, may be found in Greek mythology. Intelligent artifacts appear in literature since
then, with real mechanical devices actually demonstrating behaviour with some degree of
intelligence.
After modern computers became available following World War-II, it has become possible to
create programs that perform difficult intellectual tasks. We have witnessed the solving of a
number of intelligently hard problems by computers like playing good chess, for instance.
During the early days of computing the chess playing was considered a benchmark showing a
real intelligence. Even in seventies of the last century, when the computer chess was on the
masters level, it seemed almost impossible to make a program that could beat the world
champion. However, this happened sooner than expected. This had three reasons: increased
computing power, development of a good search algorithm (that can be used in many
applications beside chess,), and well organized knowledge bases that included all available
chess knowledge (first of all, opening and end games). In essence, the chess problem could be
solved because it was a specific intellectual problem belonging to so called narrow AI. A

ECE Department, JSSATEB Page 8

AI in Defence Sector

different case is translating from one language into another that requires general AI. In
sixties of the last century, especially after N. Chomski’s work in structural linguistics, it was
expected that the natural language translation problem will be solved soon. It has not
happened yet, although success is visible in some specific applications like, for instance,
Google’s AI linguistics. The reason is that this requires artificial general intelligence --
possessing of and ability to handle large amounts of knowledge in every field related to
human activities. It is generally accepted that AI can be considered in two ways: as a science
aimed at trying to discover the essence of intelligence and developing generally intelligent
machines, or as a science providing methods for solving complex problems that cannot be
solved without applying some intelligence like, for instance, playing good chess or making
right decisions based on large amounts of data. In the present seminar we will take the second
approach, advocate for applying specific AI methods to cyber defense and military problems.

1.4 AI in Defence Sector

Artificial Intelligence (AI) has been making significant contributions to the defence sector,
revolutionizing the way we think about national security. The use of AI in the military is not
new, but recent advancements in technology have made it more accessible and effective than
ever before. It is difficult to develop software with conventional fixed algorithms (hard-wired
logic on decision making level) for effectively defending against the dynamically evolving
attacks in networks.AI can be used in various areas of defence , from intelligence gathering to
decision-making in combat situations .With the ability to process vast amounts of data

ECE Department, JSSATEB Page 9

AI in Defence Sector

quickly and accurately, AI has the potential to enhance the capabilities of armed forces and
improve their effectiveness on the battle field. Some of the areas where AI is used are
Warfare platforms, Cyber security , Logistics and transportation , Target
Recognition ,Battlefield Healthcare ,Drone Feed Analysis , Autonomous Multi-Role Combat
Robot System , Unmanned round Vehicles , Military Aircraft "co-pilot" System , Automatic
Fault Diagnosis and Elimination System for Weapon Equipment , Automatic Intelligence and
Image Recognition System , Artificial Intelligence Weapons , Robot Army .
This brief outlines the significance of the various threats posed by the increasing
use of drones. It suggests a roadmap for addressing such threats. The brief covers the aspects
of detection, identification and localisation techniques, as well as jamming and other
countering measures. It also looks into the various existing anti-UAS solutions already in the
market.

Chapter 2: Literature Review

 Kerwin, K.R. and Bastian[1]

Predicting fraud is challenging due to inherent issues in the fraud data structure, since
the crimes are committed through trickery or deceit with an ever-present moving
target of changing modus operandi to circumvent human and system controls. As a
national security challenge, criminals continually exploit the electronic financial
system to defraud consumers and businesses by finding weaknesses in the system,
including in audit controls. This study uses stacked generalization using meta or

ECE Department, JSSATEB Page 10

AI in Defence Sector

super learners for improving the performance of algorithms in step one (minimizing
the algorithm error rate to reduce its bias in the learning set) and then in step two the
results are input into the meta learner with its stacked blended output (with the
weakest algorithms learning better). A fundamental key to fraud data is that it is
inherently not systematic, and an optimal resampling methodology has yet not been
identified. Building a test harness, for all permutations of algorithm sample set pairs,
demonstrates that the complex, intrinsic data structures are all thoroughly tested. A
comparative analysis on fraud data that applies stacked generalizations provides
useful insight to find the optimal mathematical formula for imbalanced fraud data
sets necessary to improve upon fraud detection for national security

 Clark IV, W.H., Hauser, S., Headley, W.C. and Michaels, A.J., [2]
Applications of machine learning are subject to three major components that
contribute to the final performance metrics. Within the category of neural networks,
and deep learning specifically, the first two are the architecture for the model being
trained and the training approach used. This work focuses on the third component,
the data used during training. The primary questions that arise are “what is in the
data” and “what within the data matters?” looking into the radio frequency machine
learning (RFML) field of automatic modulation classification (AMC) as an example
of a tool used for situational awareness, the use of synthetic, captured, and
augmented data are examined and compared to provide insights about the quantity
and quality of the available data necessary to achieve desired performance levels.
Three questions are discussed within this work: (1) how useful a synthetically trained
system is expected to be when deployed without considering the environment within
the synthesis, (2) how can augmentation be leveraged within the RFML domain, and,
lastly, (3) what impact knowledge of degradations to the signal caused by the
transmission channel contributes to the performance of a system. In general, the
examined data types each make useful contributions to a final application, but
captured data germane to the intended use case will always provide more significant
information and enable the greatest performance. Despite the benefit of captured
data, the difficulties and costs that arise from live collection often make the quantity
of data needed to achieve peak performance impractical. This paper helps quantify

ECE Department, JSSATEB Page 11

AI in Defence Sector

the balance between real and synthetic data, offering concrete examples where
training data is parametrically varied in size and source.

 Foster, K.L. and Petty, M.D., [3]

Militaries are developing autonomous robots to conduct missions such as
reconnaissance and surveillance. Some of those robots are intended to operate in
swarms. Because operational robot swarms are not yet available, doctrine developers
will initially use constructive entity-level combat models to develop and test tactics
for robot swarms. Design of experiments methods and retrodiction of the 1991 Battle
of 73 Easting between US and Iraqi forces were used to calibrate a semi-automated
forces system. The calibrated combat model was then used to estimate the tactical
impact of a notional Iraqi robot swarm conducting reconnaissance and surveillance in
that battle. The calibration ensured that the model’s parameters were accurate,
enabling a reliable estimate of the swarm’s tactical impact. Additionally, the design
of experiments methods produced estimates of the interaction of the robot swarm’s
effect with the technologies of the combatants’ weapon systems. Simulation trials
and statistical analysis showed that the tactical benefits of an Iraqi robot swarm were
overshadowed by the advantage provided by the US forces’ thermal sights. However,
additional trials indicated that if both sides had been equipped with optical sights
only, the early warning provided to the Iraqi forces by a robot swarm could have had
a significant effect on the battle’s outcome.

 E. Tyugu[4]
The speed of processes and the amount of data to be used in defending the cyber
space cannot be handled by humans without considerable automation. However, it is
difficult to develop software with conventional fixed algorithms (hard-wired logic on
decision making level) for effectively defending against the dynamically evolving
attacks in networks. This situation can be handled by applying methods of artificial
intelligence that provide flexibility and learning capability to software. This paper
presents a brief survey of artificial intelligence applications in cyber defense (CD),
and analyzes the prospects of enhancing the cyber defense capabilities by means of
increasing the intelligence of the defense systems. After surveying the papers
available about artificial intelligence applications in CD, we can conclude that useful

ECE Department, JSSATEB Page 12

AI in Defence Sector

applications already exist. They belong, first of all, to applications of artificial neural
nets in perimeter defense and some other CD areas. From the other side - it has
become obvious that many CD problems can be solved successfully only when
methods of artificial intelligence are being used. For example, wide knowledge usage
is necessary in decision making, and intelligent decision support is one of yet
unsolved problems in CD.

 Ali, A[7]
Artificial intelligence (AI) is trending in the military and safety-critical application
sectors. Currently, the private sector is helping the government sector to implement
new advanced techniques to bring a revolution for different government and public
sector management. It also helps to provide sustainable accountability in the
accounting field; at present, AI is bringing a revolution in concept building. It is
bringing potential revolutions by using novel approaches in such directions. This
paper is a novel approach in the same direction; our research aim of this paper is to
emphasize the AI in the militaries, what are the latest trend and usages recently
worldwide used for AI applications in militaries. In this paper, we not only discuss
the usage of AI applications in the military but also in the civil defense and health
industry. We review and discuss that AI has potential benefits in military
applications, HRMS, decision making, disaster prevention and response, GIS, service
personalization, interoperability, extensive data analysis, anomaly and pattern
recognition, intrusion detection, and new solution discovery using the highly
configurable system and real-time simulation.

 Zhang, Y., Dai, Z., Zhang, L., Wang, Z., Chen, L. and Zhou, Y.,[8]
The breakthroughs in artificial intelligence (AI) have led to a new race in military
throughout the world. Even though technologies in AI, like machine learning or big
data, are still at their juvenile stage, the technology experts in United States, Europe
Union or other countries, have been setting out to transfer AI technologies into the
military power, which is doomed to influence all spheres of warfare. To figure out
which domain has AI been studied to enhance military capability, the paper
summarized 7 major military applications of AI, based on the projects investigated by

ECE Department, JSSATEB Page 13

AI in Defence Sector

the defense department in US and Europe governments, finally analyses challenges

associated with military employment of AI techniques and draws a brief conclusion.

 Campion, M., Ranganathan, P. and Faruque, S.,[9]

The utility of unmanned aerial vehicles (UAVs) has significantly disrupted aviation-
related industries. As technology and policy continue to develop, this disruption is
likely to continue and become even larger in magnitude. A specific technology
poised to disrupt industry is UAV swarm. UAV swarm has the potential to distribute
tasks and coordinate operation of many drones with little to no operator intervention.
This paper surveys literature regarding UAV swarm and proposes a swarm
architecture that will allow for higher levels of swarm autonomy and reliability by
utilizing cellular mobile network infrastructure. Additionally, this paper chronicles
initial testbed development to meet this proposed architecture. Specific development
of higher levels of autonomous swarms with UAV-to-UAV communication and
coordination ability is central to advancing the utility of UAV swarms. The use of
cellular mobile framework alleviates many limiting factors for UAVs including range
of communication, networking challenges, size-weight-and-power (SWaP)
considerations, while leveraging a robust and reliable infrastructure for machine to
machine (M2M) communication proposed by 5G systems.

 Khalil, H., Rahman, S.U., Ullah, I., Khan, I., Alghadhban, A.J., Al-Adhaileh,
M.H., Ali, G. and ElAffendi, M.,[10]
This paper presents a UAV-swarm-communication model using a machine-learning
approach for search-and-rescue applications. Firstly, regarding the communication of
UAVs, the receive signal strength (RSS) and power loss have been modeled using
random forest regression, and the mathematical representation of the channel matrix
has also been discussed. The second part consisted of swarm control modeling of
UAVs; however, a dataset for five types of triangular swarm formations was
generated, and K-means clustering was applied to predict the cluster. In order to
obtain the correct swarm formation, the dendrogram of all types was investigated.
Finally, the heat map and contour were plotted for all kinds of swarm clusters.

ECE Department, JSSATEB Page 14

AI in Defence Sector

Furthermore, it was observed that the RSS of proposed swarms had good agreement
with swarm distances.

 Saatvik Awasthi,Balamurugan Balamurugan,V. Porkod[12]

The unmanned aerial vehicles (UAV) have great potential to support search tasks in
unstructured environments. These are agile, small and lightweight which can
incorporate many sensors that are suitable for detecting an object of interest across
various terrains. These UAVs or Drones are perfectly suited for reconnaissance and
perimeter sweeps. However, these have their limits and are vulnerable when operated
alone. The solution to the problem is using a number of these drones to form a swarm.
The swarm of drones can cover a larger area in a short period. The drones will be
resembling the flocking nature of birds and the hive nature of bees which means there
would never be a single drone operating alone. These drones are connected and the
main controller i.e. the hive. The Hive is a narrow AI which is a goal-based system.
The goals can be sweeping an area or periodic border patrols. The hive monitors the
drones, organizes them. The drones communicate in the swarm wirelessly creating a
Flying Adhoc Network. For navigation, this network utilizes GPS technology. This
project aims to develop an AI-based system that controls these drones and achieve the
goal that is assigned to it.

 The paper by Humphries, Parker, Jonas, Adams, and Clark

Investigates the problem of quickly and accurately identifying building and road
infrastructure via satellite imagery for the execution of tactical military operations in
an urban environment. Their work uses an object detection algorithm powered by
convolutional neural networks to predict both buildings and road intersections present
in an image, as well as use of a contourfinding algorithm for data labeling.

 The paper by Caelli, Mukerjee, McCabe, and Kirszenblat

Tackles the problem of integrated sensor and tactical information fusion from a
number of sources to enable rapid decision throughput based upon situation
awareness for maritime surveillance missions. Their work develops a method using a
hidden Markov model to objectively encode, summarize, and analyze airborne
maritime surveillance crew activities to gain insights into probabilistic relationships

ECE Department, JSSATEB Page 15

AI in Defence Sector

between the attention switching across sensor types and surveyed objects over the
entire mission.

 The paper by Lawley, Frey, Mullen and WissnerGross

Explores the sparse graph representation learning problem for network link prediction
and node classification tasks and whole-network reconstruction applicable to
militarily relevant graphs such as social and sensor networks. Their work employs a
novel graph-embedding model using joint sparsity biased variational graph
autoencoders capable of learning embedded representations of nodes, from which
both sparse network topologies and node features can be jointly and accurately
reconstructed

Chapter 3: Practical Using of Neural Networks in Military

Applications

ECE Department, JSSATEB Page 16

AI in Defence Sector

Chapter 4: Drone Feed Analysis

4.1: Introduction

Drone Feed Analysis system is an AI Deep Learning-Based real time/post flight military
object identification system. It identifies military objects from remotely piloted aircrafts It
builds a database repository and carries out AI-enabled analysis providing military pattern of
enemy operations and analysis/prediction of events.

Fig 1: Object Identification and next actions to be taken

ECE Department, JSSATEB Page 17

AI in Defence Sector

4.2:Principle
There are 4 stages that are detection, identification , localisation, techniques,
as well as jamming and other countering measures

4.2.1 Detection Techniques

 Radar.
Radar is a useful tool for detection of aircraft. There are various challenges, however.
These include a drone’s low altitude and velocity of flying, and very small radar
cross-section (RCS) which makes it extremely difficult to distinguish noise or clutter
from the actual target.

 RCS Study. Analysing the micro-doppler signature by multi-static radar can help in
accurate drone detection and tracking. Small drone RCS estimation obtained by
outdoor measurements has been analysed and proved that passive techniques can be
used to detect and track drones. Back scattering phenomenon associated with micro
motion has also been studied by scientists. Whereas the total RCS is important for
target detection, the energy backscattered from rotating parts like propeller and rotors
are crucial for extraction of useful micro-Doppler signatures.

 Multiple Input Single Output (MISO) Radar Systems. A single antenna is used
for transmission, whereas four antennae are used for reception. The concept of
estimating the minimum power requirement that must be transmitted to reveal a
target with a specific value of RCS is possible with the radar equation. Drones up to a
range of 150 metres can be detected using this technique.

 Multiple Input Multiple Output (MIMO) Radar Systems. Using MIMO radar
systems further increases the probability of detection of drones and UAVs. It
introduces high angular resolution and sensitivity for slow moving targets and covers
a wider area with lesser cost, as compared to conventional phased array systems. It
also exhibits enhanced discrimination from clutter and phase noise.

ECE Department, JSSATEB Page 18

AI in Defence Sector

 Ku–Band (12 to 18 Giga-hertz) Battlefield Radars. Theoretical analysis

of the detection probability has been carried out in addition to the relationship
between signal to noise ratio as well as small RCS of drones. This was based on a
study sponsored by MoD of the Slovak Republic. A 35 Giga hertz Frequency
Modulated Continuous Wave (FMCW) drone detection system has also been studied,
and the results obtained have been validated in theory as well as practice.

4.2.2 Classification and Localisation Techniques

 Audio. Although not a highly effective system in the stand-alone mode due to the
omnipresent background noise which increases the complexity of detection and
computation, it can be used as a system that can be superimposed with other detection
techniques. During the flight of the drone, the sound generated by the rotors can be
utilised in detection, classification and localisation of drones. Algorithms such as
Multiple
Signal Classification (MUSIC) can be used to estimate the direction of arrival.
Tetrahedron acoustic arrays have been used in systems to find out the direction of
arrival using received signal strength and time difference of arrival method in addition
to Kalman filter for tracking.

 Optical/Video/Infrared (IR). An object can be detected based on its appearance

features12 and/or its motion features across consecutive frames.13 For drone
detection, it is promising to combine both motion features and appearance features—
this gives higher accuracy to the system. A thermal imaging-based enhancement
algorithm for IR scanner system has also been proposed by the Military Institute of
Technology, Poland.14 Even in conditions of low visibility, the IR spectrum can be
discerned at considerable distances. However, there are certain disadvantages with IR
systems, including: low spatial resolution, low image contrast, diffused edges,
presence of noise, and pulse disturbances. The Inverse Synthetic Aperture Radar
(ISAR) technique has also been employed during a study in Korea Aerospace
Research Institute, for counter UAS systems to detect drones in urban areas. The
detailed structure, size of the drone and the number of rotors can also be determined
using this technique of ISAR as demonstrated in the study.

ECE Department, JSSATEB Page 19

AI in Defence Sector

 Micro-Doppler Analysis. The micro-doppler signature depends on parts of an object

moving and rotating in addition to the main body motion (e.g. rotor blades) and is a
characteristic of the type of object. Research has been centred around understanding
the micro-doppler spectra of specific commercial drones whose rotor blades or hubs
are uniformly displaced from the platform centre of mass. When superimposed with
radars in the X-Band (7-11 Giga-hertz) and Ku-Band, there is higher accuracy due to
the fusion of the radar sensor data as compared to a single radar.16,17 The cepstrum
reveals, using signal processing techniques, how the received signals can be used for
image recognition. As shown in Figure 1, the study carried out at Fraunhofer Institute
for High Frequency Physics and Radar Techniques (FHR), Germany, micro-Doppler
analysis when used in conjunction with radar data, provides good results when
determining the rotational speed of the rotors. The individual rotor speeds can
subsequently be read from the cepstrogram or the cepstrum. The accompanying graph
shows the cepstrum of the four rotor blades of a drone flight. Due to the flight
movement of the drone, two of the blades are rotating faster, therefore the higher
rotations per minute. Principal Component Analysis (PCA) is another method used to
reduce the number of variables in data by extracting important ones from a large pool.
It reduces the dimension (size) of the data with the aim of retaining as much
information as possible. The PCA technique has been used for feature extraction from
time-frequency spectrograms in addition to support vector machines (supervised
machine learning algorithms) to recognise the drones.

ECE Department, JSSATEB Page 20

AI in Defence Sector

Fig 2.1- Micro-doppler Radar DroMiAn (Drone detection with micro-Doppler

Analysis)

Fig.2.2 Cepstrum of Four Rotor Blades

ECE Department, JSSATEB Page 21

AI in Defence Sector

4.2.3 JAMMING AND COUNTERING TECHNIQUES

Defeating a drone can involve a plethora of techniques. To name a few, it may involve kinetic
means, electronic warfare (EW) and cyber warfare (CW) techniques, Drone vs drone, and
Directed Energy Weapons (DEW) using high-powered microwave or LASERs. Various
integrated solutions have been offered in the world market with respect to defeating drones or
sUAS. The noteworthy options include RF jamming, GPS jamming, GPS spoofing, and net
guns.18 While jamming the controller link is an option, the widely used concept involves
jamming the GPS link (bands L1 to L5) to make the drone lose control of its “auto-home”
option when the main controller link fails (i.e., jammed). One can also develop techniques to
jam the payload WiFi link which will also be in almost the same frequency range as the 5
Giga-hertz controller link (HP 47 counter UAV Jammer does exactly this by blocking the
video feeds). A layered approach is what will be preferable in a counter-drone system.

Being demonstrated the High Energy Laser Mobile Demonstrator in 2014, which fired a 10-
Kilowatt LASER and could function effectively in different climatic conditions like rain, fog
or wind. The Drone Catcher Gun is also a scalable option available to safeguard facilities as
has been researched and demonstrated.19 The jamming of drones can also be effected using a
3D MIMO radar.20 By using directional antennae, the power is confined spatially, thereby
allowing its use without interference to co-located RF devices/ equipment.

The two most potent and comprehensive repositories of counter-UAS have been studied as
part of preparing this brief.21, 22 The resulting database as part of the aforementioned
repositories is based on open-source research of technical and policy reports, news analyses,
manufacturers’ information, interviews with government officials, subject matter experts, and
participation in conferences. The sheer volume as well as the description of the counter UAS
weapons prohibits its inclusion in this brief.

The database consists of nearly 537 systems, sold by 277 vendors and fielded by 38 different
countries, some alone and others as part of joint ventures between two nations. RF & radar
detection have been found to be the most common detection techniques, followed by EO/IR
systems. Jamming the RF and GNSS signals is the most common method, followed by a few
which have spoofing capability. Some also mention the use of a ‘sacrificial drone’, or a drone
that may be used to deliberately collide with the detected enemy drone to destroy it.

ECE Department, JSSATEB Page 22

AI in Defence Sector

4.3:ARCHITECTURE AND IMPLEMENTATION

Of the 537 systems discussed earlier in this brief, unfortunately, none of the systems from
India find a mention. This may be because information may be classified, or there was no
participation by Indian vendors. However, during the Republic Day parade in 2020, there
were media reports of an anti-drone weapon having been fielded.Based on the various proof
of concepts as well as technology demonstrators, one can now elaborate on the system design
for a modular counter-drone weapon. The main purpose of these measures is to exploit the
vulnerabilities of drones.

 Heterogenous Sensing Unit – Detection, Localisation and Tracking

A heterogenous sensing unit will serve the purpose of detection and localisation of the
drone. The heterogenous system will comprise of various sensors which can be
utilised in any weather. There should be RF/ radar-based sensors as the first line of
detection, coupled with acoustic sensors. This can be followed by an EO/IR system
layer for enhancing the resolution. The range of detection should be a minimum of
three km and it should effectively be able to track the drone at minimum 1 km,
assuming the drone is carrying a destructive payload.

 Central Processing Unit

The Central Processing Unit should be able to collate the feed from all the active and passive
sensors. It should thereafter carry out analysis, derive drone features, and achieve
classification.

For the acoustic sensors, the Short Time Fourier Transform (STFT) can be adopted and the
histogram of oriented gradients (HOG) can be used 25 for extraction of image features. These
are in addition to the PCA methods described earlier.

In the case of the RF features to be extracted, it may be assumed that at least two out of the
following three links are active at any given point in time, viz., controller link or the ground
control link (to include all frequency bands as prevalent based on the manufacturer
specifications); GPS link (uplink & downlink frequency bands as also to cover GLONASS,
Bediou & possibly IRNSS); and the payload (assumed camera/ optical device for

ECE Department, JSSATEB Page 23

AI in Defence Sector

reconnaissance) WiFi link back to the ground control station which may transmit the data
captured in real time to the control station. The development and incorporation of neural
networks can be considered subsequently. By using the logical ‘OR’ operation, it will be
ensured that probabilities of detection are high despite the possibility of a false alarm, albeit
to a minimum.

 Jamming Unit
Based on the inputs and analysis carried out by the Central Processing Unit, the jamming unit
can be made to work in three modes akin to air defence systems viz., Weapon Hold, Weapon
Free, and Weapon Tight. Based on directional antennae to achieve better spatial economy as
well as meet power requirements, a 360-degree coverage is preferable (gimbalb based for
added stabilisation) with a slaved servo/proportional-integral-derivative (feedback based)
controller to the tracker. A spoofing unit should be made alongside the jamming unit to take
over the drone if desired for forensics at a later stage (Man in the middle attackc).

 Kill Unit
If the jamming unit is unable to affect the drone operations due to any reason, the high-power
microwave unit should kick into action to get the drone on the ground. The range of
engagement can be fixed with respect to various vulnerability aspects, such as the installation
being defended, the importance of the area, and the vital point. An alternate high-power
LASER may also be employed to destroy the drone. The close-in weapon support systems
(CIWSS) as used extensively by naval vessels the world over against hostile missiles is also
another option which may be explored.

 System Integration and Implementation

The proposed system should be fabricated in a modular fashion allowing sub-system

integration based on the felt need. The system at the same time should be made in two
versions, viz., manpack (portable) and vehicle-based, obviously scaling down the parts and
therefore, weight for the manpack system (SWaP considerations). The design of the system
may be adapted from the system as shown in Figure 3.

ECE Department, JSSATEB Page 24

AI in Defence Sector

Fig. 3 Envisioned Counter Drone System – Layered Architecture

While the system as envisioned in this brief has been trial-evaluated, stakeholders in the anti-
UAS campaign such as Rafael (Israel-based) have come up with the Drone-DomeTM drone
detection, neutralisation and interception system as well as the FireflyTM miniature tactical
loitering weapon.

ECE Department, JSSATEB Page 25

AI in Defence Sector

Fig. 4- Drone Dome Brochure (Open Source) – Array of sensors and integration

Fig. 5- Firefly Brochure (Open Source) – Tactical loitering weapon to ‘Kill’ Drones

ECE Department, JSSATEB Page 26

AI in Defence Sector

While the man-portable systems should be capable of being used by the dismounted soldier
or the infantry, the vehicle-based system may be utilised to guard static installations. It is
recommended that being a tri-service requirement, the fabricating process be undertaken
under the aegis of the HQ Integrated Defence Staff or a prototype be developed as an Army
Technology Board project. What must also be noted is that collaboration (internal civil
industry as well as global players) in this field will reap greater dividends rather than a purely
‘Make in India’ approach which may take more time.

Under the gambit of the draft DPP 2020, the innovation scheme under the ‘Make’ category/
Innovation in Defence Excellence (iDEX)/ Technology Development Fund initiatives may
help in the long run. DGCA will need to be incorporated to streamline the rules of
engagement (Civil Aviation Requirements or CAR). Such ‘Aerial Threat Reduction Teams
(ATRT)’ should be a tri-services unit or outfit, with a participation by personnel from all
three services, for effective and efficient utilisation.

ECE Department, JSSATEB Page 27

AI in Defence Sector

Chapter 5: Challenges of Drone Feed Analysis

The challenges being faced today with respect to counter-drone technology is manifold as
drone technology is itself growing by leaps and bounds. Reconnaissance, Intelligence,
Surveillance & Target Acquisition (RISTA) & Electronic Warfare systems will have to work
together to tackle the threat of both stand-alone and drone swarms. Some of the pertinent
technological challenges are being highlighted which will need immediate attention to be
considered when the anti-drone weapon is conceptualised.

 Heterogenous Sensor Fusion. The variety of protocols used for communication

between sensors poses a problem during their integration into a networked system.

 Energy Efficient Sensors. Power management considerations will always be

towering as the employment of such systems is going to be all-day and all-night. Solar
power sources and fuel cells are the technologies which may be tapped.
 Multiple Drone Detection and Localisation. Technical challenges will be
compounded when trying to detect and localise swarms wherein it will not be
necessary that all the drones are having a viable RF signature. Progress has been made
in using millimetre -wave radar as well as measuring the turbulence caused by the
drone rotors.
 Drone Signature Database. Extensive studies are to be conducted to have a
comprehensive database. Inter-agency cooperation may help mitigate the issues
involved.

 Identification Friend or Foe (IFF). The IFF shall always remain challenging with
the use of drones for all unethical purposes by non-state actors. However, research has
been carried out in this field using a relay drone with beacon system.

 Hardware Sandboxing. With systems in place to reduce the effectiveness of drone

jamming (sandboxing), novel kinetic kill methods will have to be developed.

 Prevention Against Jamming (JAM-ME) Techniques. Jamming is now being

leveraged for drone missions into completion.

ECE Department, JSSATEB Page 28

AI in Defence Sector

 Cognitive Avoidance in 4th Generation GPS. GPS systems are becoming

increasingly more interference-proof. Long Term Evolution or LTE may soon be used
to operate drones at theoretically unlimited ranges without RF links, i.e., cellular base
stations will provide theoretically unlimited range.

New flight modes that have been introduced in UAVs or drones which include
obstacle avoidance using ultrasonic sensors, terrain mapping cameras integrated as
SoC, follow me, tapfly, active track and sports mode. Better battery systems have
been incorporated giving the drone larger endurance (5230 mAh battery on board the
DJI Phantom 4).

ECE Department, JSSATEB Page 29

AI in Defence Sector

Chapter 6: Advantages of AI in defence sector

 Dealing with mundane tasks
One massive advantage of artificial intelligence is its potential to complete
mundane tasks through intricate automation that will increase productivity.
Theoretically this can even remove “boring” tasks from humans and free them up to
be increasingly creative.

 Faster decisions
Using artificial intelligence alongside cognitive technologies can help make faster
decisions and carry out actions quicker.

 Avoiding errors
The phrase “human error” was born because humans, naturally, make mistakes from
time to time. Computers however, do not make these mistakes – that is, of course,
assuming they are programmed properly. With artificial intelligence, data could be
processed error-free, no matter how big the dataset might be

 Taking risks on behalf of humans

With artificial intelligence, you can arguably lessen the risks you expose humans to in
the name of research. Take, for example, space exploration and the Mars rover,
known as Curiosity. It can travel across the landscape of Mars, exploring it and
determining the best paths to take, while learning to think for itself. Using artificial
intelligence in this manner could potentially lead to massive benefits in areas such as
demand forecasting, medical diagnosis and oil exploration.

 Difficult Exploration
Artificial intelligence and the science of robotics can be put to use in mining and other
fuel exploration processes. Not only that, these complex machines can be used for
exploring the ocean floor and hence overcoming the human limitations. Due to the
programming of the robots, they can perform more laborious and hard work with
greater responsibility. They do not wear out easily

ECE Department, JSSATEB Page 30

AI in Defence Sector

Chapter 7: Disadvantages of AI in defence sector

If robots start replacing human resources in every field, we will have to deal with
serious issues like unemployment in turn leading to mental depression, poverty and crime
in the society. Human beings deprived of their work life may not find any means to
channelize their energies and harness their expertise. Human beings will be left with
empty time.
Secondly, replacing human beings with robots in every field may not be a right
decision to make. There are many jobs that require the human touch. Intelligent machines
will surely not be able to substitute for the caring behavior of hospital nurses or the
promising voice of a doctor. Intelligent machines may not be the right choice for customer
service.
One of the major disadvantages of intelligent machines is that they cannot be
'human'. We might be able to make them think. But will we be able to make them feel?
Intelligent machines will definitely be able to work for long hours. But will they do it with
dedication? Will they work with devotion? How will intelligent machines work
wholeheartedly when they don't have a heart?

ECE Department, JSSATEB Page 31

AI in Defence Sector

Chapter 8: Applications
A. Intelligent target identification and monitoring The intelligent target identification and monitoring
systems use computer vision to identify and track targets for the autonomous weapon platform,
which several military research institutions and defense firms have already worked on. The most
well-known project is Maven of the US Defense Advanced Research Projects Agency (DARPA), which
could highly efficiently categorize and identify 38 different kinds of objects among the huge sums of
surveillance footage taken by Unmanned Aerial Vehicle (UAV) or other surveillance equipment.
Similarly, DARPA’s project Target Recognition and Adaption in Contested Environments (TRACE)
using computer vision to automatically locate and identify targets based on the images of Synthetic-
Aperture Radar (SAR). B. Autonomous Weapon Platforms Based on the target indication of the
intelligent target identification and monitoring module, autonomous weapon platforms can react
appropriately and always lock the target even which deploys countermeasures or evasive
maneuvers. And advanced algorithms can locate the problem before it happens, like ordering a
return when vehicles lost GPS signals or is facing fuel shortage. According to Israeli defense
contractor, Rafael Advanced Defense Systems, in virtue of intelligent inferred tracking modules
based on computer vision techniques, their GIL-2 handheld rocket system could fire without line of
sight to the target, and take advantage of UAV to lock the target and guide the missile. Another
valuable project is conducted by the French defense contractor, Thales, which is partially owned by
the French Department of Defense. Thales presented a weapons platform for fixed point defense
called RAPIDFire, which is equipped with 6 short-range anti-aircraft missiles and a 40mm cannon. It
could be inferred that RAPIDFire applies computer vision to automatically acquire hostile targets,
and provides it to human operators who must approve any discharge of defensive ordnance. Also,
Thales claims there is an option of operates fully autonomously. DARPA came up with a deeper and
more comprehensive research in aerial and underwater space, which aims to build the unmanned
combat team for the specific mission. AI modules in Unmanned Aerial Systems (UAS) can not only
recognize the target, but also use multi-objective evolutionary algorithms to solve allocation
problems in the planning phase of an operation, which optimize the number, payloads, and path
planning of drones. There is also a plan for building submarine-locating drones, aimed at making the
ocean

Chapter 9: Impact of Using Artificial Intelligence in

Military on Society

Chapter 10: Conclusion

AI is at the centre of a new enterprise to build computational models of

intelligence, and to solve complex problems in military which is not possible by
human. The main assumption is that intelligence (human or otherwise) can be represented
in terms of symbol structures and symbolic operations which can be programmed in a
digital computer. There is much debate as to whether such an appropriately programmed
computer would be a mind, or would merely simulate one, but AI researchers need not

ECE Department, JSSATEB Page 32

AI in Defence Sector

wait for the conclusion to that debate, nor for the hypothetical computer that could model
all of human intelligence. Aspects of intelligent behaviour, such as solving problems,
making inferences, learning, and understanding language, have already been coded as
computer programs, and within very limited domains, such as identifying diseases of
soybean plants, AI programs can outperform human experts. Now the great challenge of
AI is to find ways of representing the common sense knowledge and experience that enable
people to carry out everyday activities such as holding a wide-ranging conversation, or
finding their way along a busy street. Conventional digital computers may be capable of
running such programs, or we may need to develop new machines that can support the
complexity of human thought.
Weapons with artificial intelligence will be one of the most strongest weapon in
military, which can identify the targets itself and can make a decision weather to destroy
or not. Overall artificial intelligence is one of the most advantageous technology in
defense sector.

Chapter 11: References

[1]. Kerwin, K.R. and Bastian, N.D., 2021. Stacked generalizations in imbalanced fraud data
sets using resampling methods. The Journal of Defense Modeling and Simulation, 18(3),
pp.175-192.
[2]. Clark IV, W.H., Hauser, S., Headley, W.C. and Michaels, A.J., 2021. Training data
augmentation for deep learning radio frequency systems. The Journal of Defense Modeling
and Simulation, 18(3), pp.217-237.

ECE Department, JSSATEB Page 33

AI in Defence Sector

[3]. Foster, K.L. and Petty, M.D., 2021. Estimating the tactical impact of robot swarms using
a semi-automated forces system and design of experiments methods. The Journal of Defense
Modeling and Simulation, 18(3), pp.247-269.
[4]. E. Tyugu, "Artificial intelligence in cyber defense," 2011 3rd International Conference
on Cyber Conflict, Tallinn, Estonia, 2011, pp. 1-11.
[5]. Bastian, N.D., 2021. Artificial intelligence for defense applications. The Journal of
Defense Modeling and Simulation, 18(3), pp.173-174.
[6]. Gopal, V., 2020. Developing an Effective Anti-drone System for India's Armed
Forces (pp. 1-20). Observer Research Foundation.
[7]. Ali, A., 2021. Artificial Intelligence Potential Trends in Military. Foundation University
Journal of Engineering and Applied Sciences (HEC Recognized Y Category, ISSN 2706-
7351), 2(1), pp.20-30.
[8]. Zhang, Y., Dai, Z., Zhang, L., Wang, Z., Chen, L. and Zhou, Y., 2020, December.
Application of Artificial Intelligence in Military: From Projects View. In 2020 6th
International Conference on Big Data and Information Analytics (BigDIA) (pp. 113-116).
IEEE.
[9]. Campion, M., Ranganathan, P. and Faruque, S., 2018, May. A review and future
directions of UAV swarm communication architectures. In 2018 IEEE international
conference on electro/information technology (EIT) (pp. 0903-0908). IEEE.
[10]. Tahir, A., Böling, J., Haghbayan, M.H., Toivonen, H.T. and Plosila, J., 2019. Swarms of
unmanned aerial vehicles—a survey. Journal of Industrial Information Integration, 16,
p.100106.
[11]. Khalil, H., Rahman, S.U., Ullah, I., Khan, I., Alghadhban, A.J., Al-Adhaileh, M.H., Ali,
G. and ElAffendi, M., 2022. A UAV-Swarm-Communication Model Using a Machine-
Learning Approach for Search-and-Rescue Applications. Drones, 6(12), p.372.
[12]. Saatvik Awasthi,Balamurugan Balamurugan,V. Porkod,Artificial Intelligence
Supervised Swarm UAVs for Reconnaissance May 2020,DOI: 10.1007/978-981-15-5827-
6_33

ECE Department, JSSATEB Page 34