2022 International Conference on Data Science, Agents & Artificial Intelligence (ICDSAAI)

Intelligent Military Robot for Intruder Detection

Using Matlab with Machine Learning Technique
2022 International Conference on Data Science, Agents & Artificial Intelligence (ICDSAAI) | 979-8-3503-3384-8/22/$31.00 ©2022 IEEE | DOI: 10.1109/ICDSAAI55433.2022.10028923

1st Poornima Jayaraman 2nd K.P.Aswathi 3rd SureandarM P

Electronics and Communication Electronics and Communication Electronics and communication
Engineering Engineering Engineering
Chennai Institute of Technology Chennai Institute of Technology Chennai Institute of Technology
Chennai, India Chennai, India Chennai, India
poornimaj@citchennai.net aswathikp@citchennai.net mpsurenthar@gmail.com

4th Rithul.M
Electronics and communication
Engineering
Chennai Institute of Technology
Chennai, India
rithul0206@gmail.com

Abstract— In India most of the autonomous Robots are fate of future battles. Regarding the application of ML in the
implemented for soldier safety purposes.Still it requires self context of intrusion detection systems, mention is made of
reliant and Aritificial intelligence features in robots . The the supervised learning paradigm and the classification and
Indian Navy has taken the initiative and split implementation regression problem. In actuality, the enemy will be finalised
of AI usage into short-, medium-, and long-term objectives.The using the attributes that were retrieved from the analysed
use of a low-cost surveillance robot for the purpose of border image. Consequently, we refer to a dataset as "labelled"
surveillance and the prevention of terrorist infiltration. when the correct responses for the model to learn are known
Machine learning approaches are being used increasingly (obviously only for the learning phase).
frequently to address all the issues where it is challenging to
understand the process of interest. The robot consists of a II. LITERATURE REVIEW
surveillance unit and an arm unit, both of which are separate
modules. The robot will patrol certain predefined paths along Assuming implicitly that the learning algorithm has the
the border as it moves along. Intruders can be located with the chance to compare the prediction with the actual response at
help of the ultrasonic sensor, and if any are found, the least during the learning phase, we shall refer to supervised
information is relayed to the operator with the help of the RF learning throughout this paper. The concept of remotely
Transmitter. Real-time video from the robot is recorded and operated vehicles is not widely accepted in India [1-6], with
stored on the phone using the camera that is already installed the exception of some applications involving the detection
in the device. Through the use of Bluetooth technology, the
of explosive ordnance. The advanced robotics technology
robotic device may be controlled by the user's smartphone.
The operator in the control room is being provided with
available today can be put to use in a wide variety of
continuous live video feed from the camera. When an intruder additional tasks. The use of robots to collect information and
is spotted, the operator brings the robot near to the intruder to even execute jobs on the ground during a battle can be very
obtain additional information before removing the threat with advantageous in today's increasingly digital conflicts. This
a weapon. This system is required to reduce the army's is especially true when the robots are used in conjunction
remaining casualties, and it may be run at any time with a with modern warfare. On the other hand, the Indian military
larger workforce than regular soldiers and with less operator and paramilitary forces only deploy these remotely piloted
requirements..In this work we implemented the identification vehicles for a select few operations [7-11], largely for the
of intruder by capturing the image and processed by MATLAB
purpose of explosives detection. The employment of
and using machine learning techniques,the robot finalizes
whether the person is authorized or unauthorized person and
military robots often has an impact and benefits that can be
passing the LASER upon the enemy.As all the database stored described as "lifesaving." The ability of robots to carry out
in the cloud.The Regression algorithm effectively finalizes the tasks that are analogous to those performed by humans
enemy with better accuracy.The experimental results are without putting the lives of humans in danger is the first
simulated and explained. major benefit and significance of robots. These robots, in
contrast to human lives, are easily replaceable, albeit at a
Keywords— Military Robot, Intruder detection cost. Robots are able to endure the effects of bombs and
system,MATLAB, Machine Learning,Computer vision other weapons that would normally destroy a human body.
I. INTRODUCTION This is a significant advantage over humans. The fact that
robots are available in a range of sizes is just another
The Indian armed services have not yet made extensive advantage they offer. Some robots are capable of fitting into
use of robotics, despite the apparent benefits that such usage tight locations that are difficult for humans to access. Robots
would provide. India is now making progress in ensuring the
are ideally suited for work in environments that are
effective application of artificial intelligence (AI) in conflict
frequently confined and poorly lit. A further advantage of
resolution, which is quickly evolving into a crucial
operational requirement that may very well determine the employing robots is that in addition to neutralizing bombs

979-8-3503-3384-8/22/$31.00 © 2022 IEEE

Authorized licensed use limited to: VTU Consortium. Downloaded on October 10,2024 at 08:54:42 UTC from IEEE Xplore. Restrictions apply.
and weapons, they are also able to hold their position and XGBoost Extreme Gradient Boosting
fire [12]. Robots do not require internal drives such as self- SVM Support Vector Machine
preservation, fury, vengeance, hunger, weariness, or
SOM Self-Organizing Map
resentment in order to be driven. Because their senses are
more wide and capable, they are able to make decisions and LSTM Long Short-Term Memory
react to changing conditions more quickly than humans. On kNN k-nearest neighbour
the other hand, it's possible that autonomous weapons lack
LR Linear Regression
empathy. We are unable to make use of autonomous robots
until we can provide evidence that we have achieved a DNN Deep Neural Network
higher ethical standard than humans. MLP Multilayer Perceptron Neural Network
Researchers have recently shown an increased interest in
monitoring and surveillance technology. On the other hand, MPR Multiple Polynomial Regression
the construction of a small robot for the purposes of testing MLR Multiple Linear Regression
and research turns out to be an extremely expensive PCA Principal Component Analysis
endeavour. This is due to the fact that [13] primarily a
security robot would require certain components, such as a PCR Principal Component Regression
GPS module (Global Positioning System), high resolution MPR Multiple Polynomial Regression
cameras, radios for satellite connectivity, and so on. Each of PCA Principal Component Analysis
these components is fairly pricey, and putting them all
together in order to make a robot is not only expensive but PCR Principal Component Regression
also time consuming and labour intensive. In addition, a RBF Radial Basis Function
considerable amount of effort is lost in the process of
RF Random Forest
building driver code in order to connect all of these different
components. The answer to this seemingly intractable RLR Ridge Linear Regressor
problem is uncomplicated. In recent years, there has been a RPART Recursive Partitioning and Regression Trees
rise in demand for smartphones that come equipped with a
MLNN Multilayer Neural Network
multitude of capabilities. These mobile devices come
equipped with essential functions such as a GPS module, a MLP Multilayer Perceptron Neural Network
camera with a high quality, and access to the internet [14-
15]. In addition, the operating system on these smartphones
is equipped with Application Programming Interfaces III. DESCRIPTION OF PROPOSED SYSTEM
(APIs), which make it simple to make use of the numerous A personal computer, a mobile phone, and a robot all work
sensors. Without having to resort to writing driver code, we
together to form this system. The robot can be controlled by
are able to quickly and easily create applications in high-
level programming languages such as Java by utilising the a person who is located elsewhere using Bluetooth
APIs that are offered. In our system, we made use of a technology and a remote PC. The robot consists of a
smartphone that was powered by the Android operating smartphone that is powered by the Android operating
system. Android is currently one of the most well-liked system, an Arduino microcontroller that is used to control
mobile operating systems available. Consequently, the the movement of the robot, and any other required pieces of
creation of a fully-functional surveillance robot that can be
hardware (motors, chassis, power supply, etc.).
operated from a remote location by means of a commonly
available Android phone is our primary objective.Various
machine learning algorithms are mentioned in table 1. The user exerts control over the robot by sending control
Table 1 List of machine learning algorithms, signals to the Arduino board through Bluetooth. The
models and abbreviations Arduino board then directs the robot in the appropriate
direction after receiving the control signals. It is possible to
ANFIS Adaptive Network-Based Fuzzy Inference System
provide real-time video feedback to a distant user on an
ANN Artificial Neural Network Android smartphone by utilising the camera on the device
BART Bayesian Additive Regression Tree and sending it over the internet. This gives the user the
BRT Boosted Regression Tree ability to control the robot from a remote location. On the
GA Genetic Algorithm computer that is located remotely, supplementary processing
can be carried out on the video feed. In this study, we offer a
FFNN Feed-Forward Neural Network
prototype for a tank-based military robot that has turrets for
ELM Extreme Learning Machine
simulating firing at an enemy target using computer vision,
ET Extremely Randomized Tree object detection, and tracking.
GBDT Gradient Boosting Decision Tree
GBM Gradient Boosting Machine
GMM Gaussian Mixture Model
HMM Hidden Markov Model

Authorized licensed use limited to: VTU Consortium. Downloaded on October 10,2024 at 08:54:42 UTC from IEEE Xplore. Restrictions apply.
 photographs. Using the RIJNDAEL algorithm for increased
security, it is possible to record and remotely broadcast both
the video and still photographs that have been acquired. The
Android operating system, which is open source and helps
enable support and ease of use, is running on the smart
phone.

Figure 1:Proposed Hardware architecture

III. CONNECTION POINTS AND MODULE

We utilised an Arduino Uno, which is a board for

microcontrollers that is based on the ATmega328. The
project is broken up into three distinct parts, or modules, and
we will go into greater detail about each one below.Figure 1
shows the hardware architecture of the proposed model.

III. A) Transmission of information from the sensor to the

Figure 3:Proposed Intelligent Real time Robot for
command and control facility intruder detection
We utilised an Arduino that was connected to an ultrasonic
sensor. When an intruder is discovered by the sensor, a At any time, the camera-equipped smartphone that
signal is transmitted to an additional Arduino that is has been attached to the robot can be detached from it.
stationed at the command center. At one end, a radio Higher image quality is achievable through the use of more
frequency (RF) transmitter is used to send the signal, while capable cameras. The Android operating system, which is
at the other end, an RF receiver is used. The ultrasonic available for free download, was utilised in the development
sensor detects an intruder crossing the border and triggers of the programme.[17] A 32-bit encryption and decryption
the buzzer, which in turn alerts the operator who is in format is utilised by the application, which contributes to the
control of the robot.The optimal path detection of Robot is program's high level of dependability and safety. Through
detected.In figure 2 the optimal path detection of robot is the utilisation of our very own RIJNDAEL algorithm, the
mentioned. straightforward 128-bit encryption and decryption process
has been improved and rendered infallible.The intelligent
real time robot monitoring system is mentioned in Figure 3.

IV MACHINE LEARNING TECHNIQUES USED

Machine learning approaches are being used to increase

frequently to address all the issues where it is challenging to
understand the process of interest. The issue of intrusion
detection in networking systems .It is a one which is not
essential to understand the intercept measures.
It is difficult to get a classification's response in order to
get the information that we can get from a process
Algorithm. if the network traffic exhibits (there is therefore
a more likelihood of a breach). The unauthentication is
Figure 2:Optimal path detection rising,due to its consequences, networking is distributed and
autonomous systems are defined. System availability,
III. A) Communication between Android and its users dependability, and even safety. The K nearest
The camera-equipped smart phone the robot carries about neighbor(KNN) is implemented to realize the intruder
with it is capable of recording videos and taking still detection[20]

Authorized licensed use limited to: VTU Consortium. Downloaded on October 10,2024 at 08:54:42 UTC from IEEE Xplore. Restrictions apply.
 Distributed
network V COMMUNICATION BETWEEN THE ANDROID
AND THE ROBOT (ARDUINO):

The Android phone and the Arduino may communicate

Network Analyzer with one another thanks to the Bluetooth module. The
Arduino is powered by a rechargeable battery that is 12
volts. In order to communicate between the command centre
and the Arduino, the Android phone acts as a link. In order
Data sets to accomplish this, the phone is currently utilising not one
but two distinct apps at the same time[18]. The video feed is
transmitted from one of these to a computer that is located
elsewhere. In order to accomplish this, we made use of a
Pre processing algorithm programme that broadcasts live video and sends it to the
computer that was located remotely. The second app's job is
to transmit control signals from the Remote User to the
Arduino microcontroller, and it does this by acting as a
relay. A wheeled robot with a predetermined path patrols the
border of the territory.
Testing data set Training data sets There are ultrasonic sensors deployed along the border,
which are in constant communication with the command
and control centre. The robot is equipped with a camera that
is capable of continuously sending live stream video through
an application that was developed using the RIJNDAEL
Machine learning models algorithm. This programme provides the controller with live
video, and the controller provides it with a permanent IP
address. An indication that an incursion has taken place is
provided by an ultrasonic sensor whenever it receives an
echo; this signal is then sent wirelessly to the controller by
Class means of an RF transmitter and an RF receiver. The human
operator takes control of the robot and instructs it to proceed
to the location of the break-in. When it is determined that an
Figure 4: Workflow diagram for intruder detection using intruder poses a danger, the laser pistol that is attached to
Machine learning
the robot will be deployed to eradicate the danger. Control
of the laser cannon is accomplished through the use of a
After the analysis the KNN Model has been estimated
servo motor. An application for a Bluetooth controller is
the accuracy of 99%after the observation from the test data.
being created so that the robot can be controlled and fired at
the appropriate target.

VI RESULTS AND DISCUSSION

The Mega 2560 board can be programmed with the

Arduino Software (IDE). For details, see the reference and
tutorials. The ATmega2560 on the Mega 2560 comes pre-
programmed with a boot loader that allows you to upload
new code to it without the use of an external hardware
programmer. It communicates using the original STK500
Figure 5: Graphical representation using MATLAB
protocol (reference, C header files). we can also bypass the
Figure 4 shows the workflow model of machine learning boot loader and program the microcontroller through the
to detect the intruder.As the data gathered from the different ICSP (In-Circuit Serial Programming) header using Arduino
input devices it is processed as a testing and training ISP or similar; see these instructions for details.Figure 6
data.Using machine learning techniques K-NN Algorithm shows the Arduino output window
the accuracy has been detected around 99% shown in the
figure 5.

Authorized licensed use limited to: VTU Consortium. Downloaded on October 10,2024 at 08:54:42 UTC from IEEE Xplore. Restrictions apply.
 If (distance 3<20)
{
digitalWrite(3,0);digitalWrite(4,1);digitalWrite(5,1);digital
Write(6,0);delay(200);
digitalWrite(3,0);digitalWrite(4,0);digitalWrite(5,0);digital
Write(6,0);delay(200);
}
else
{
digitalWrite(3,1);digitalWrite(4,0);digitalWrite(5,1);digital
Write(6,0);delay(200);
digitalWrite(3,0);digitalWrite(4,0);digitalWrite(5,0);digital
Write(6,0);delay(200);
}
if(distance2<20)
{
digitalWrite(3,0);digitalWrite(4,1);digitalWrite(5,1);digital
Write(6,0);delay(200);
Figure 6: Programming in Aurdino IDE Software digitalWrite(3,0);digitalWrite(4,0);digitalWrite(5,0);digital
Write(6,0);delay(200);
VII ALGORITHM(source code)
}
#include <LiquidCrystal.h> Else
LiquidCrystal lcd(42,44,46,48,50,52); {
#include <ultrasonic.h> digitalWrite(3,1);digitalWrite(4,0);digitalWrite(5,1);digital
Ultrasonic(22, 23); Write(6,0);delay(200);
Ultrasonic ultrasonic2(24, 25); digitalWrite(3,0);digitalWrite(4,0);digitalWrite(5,0);digital
Ultrasonic ultrasonic3(26, 27); Write(6,0);delay(200);
int distance; }
int distance2;int distance3; delay(500); lcd.clear();
void setup()
{ VIII EXPERIMENTAL RESULTS
lcd.begin(16, 2);
The Experimental results are shown in figure 6.Based on
lcd.print("SMART ROBO"); the simulation using MATLAB the image was analysed and
delay(2500); gives the results to the authorized or unauthorized
lcd.clear(); person.This Experimental analysis shows accurate results on
pinMode(3,OUTPUT); estimating the distance of intruder and tracking the person
pinMode(4,OUTPUT); by capturing the image using stereo camera.It analyses the
pinMode(5,OUTPUT); authorized person and gives the acknowledgement.Figure 8
shows the unauthorized person figure and gives the
pinMode(6,O UTPUT);
mismatch command.
}
void loop()
{
distance = ultrasonic.read();
distance2 = ultrasonic2.read();
distance3 = ultrasonic3.read();
lcd.setCursor(0, 0);
lcd.print("U1:");
lcd.print(distance);
lcd.setCursor(8, 0);
lcd.print("U2:");
lcd.print(distance2);
lcd.setCursor(0, 1);
Figure 7: Results using MATLAB(Authorized person)
lcd.print("U3:");
lcd.print(distance3);

Authorized licensed use limited to: VTU Consortium. Downloaded on October 10,2024 at 08:54:42 UTC from IEEE Xplore. Restrictions apply.
 C. Pacchierotti published "Decentralized Control of a
Heterogeneous Human-Robot Team for Exploration and
Patrolling."
[4].J. Bae and M. Park, "A Heuristic for Efficient
Coordination of Multiple Heterogeneous Mobile Robots
Considering Workload Balance," IEEE Robotics and
Automation Letters, vol. 6, no. 2, pp.
[5]."Robust Formation Coordination of Robot Swarms With
Nonlinear Dynamics and Unknown Disturbances: Design
and Experiments," in IEEE Transactions on Circuits and
Figure 8:Results using MATLAB(Unauthorized person)
Systems II: Express Briefs, vol. 69, no. 1, pp. 114-118, Jan.
IX CONCLUSION AND FUTURE WORK 2022.
[6].M. H. Sharif, "Laser-Based Algorithms Meeting Privacy
This Paper denotes lot of opportunity for more in Surveillance: A Survey," IEEE Access, vol. 9, 2021, p.
recent features to be added. If we programme the [7].Y. Wu, S. Wu, and X. Hu, "Cooperative Path Planning
surveillance vehicle with an algorithm for on-board image of UAVs & UGVs for a Persistent Surveillance Task in
analysis, it will be able to recognise potential threats without Urban Environments," IEEE Internet of Things Journal, vol.
human intervention. We are able to teach the robot to 8, no. 6, pp.
independently locate and approach objects via [8].Modeling and Path Planning for Persistent Surveillance
programming. As a direct consequence of this, we are able by Unmanned Ground Vehicle, IEEE Transactions on
to fully automate it[17]. Automation Science and Engineering, vol. 18, no. 4, pages
In addition, because it will be equipped with GPS 1615-1625, October 2021.
navigation and mapping software, the robot will be able to [9].V. Savkin and H. Huang, "Navigation of a UAV
choose the path that will lead it to its destination in the most Network for Optimal Surveillance of a Group of Moving
expedient manner. Even when it's dark out, increased Ground Targets Along a Road," IEEE Transactions on
security brought on by the installation of a camera with Intelligent Transportation Systems (2009).
night vision is possible. We can also make it an all-terrain [10].Failure-Resilient Coverage Maximization With
robot by fortifying it and providing additional protection, Multiple Robots, M. Ishat-E-Rabban and P. Tokekar, IEEE
which would make it an excellent candidate for a Robotics and Automation Letters, vol. 6, no. 2, pp. 3894-
surveillance robot. If it is able to take flight, the surveillance 3901, April 2021.
vehicle can also be put to use for the purpose of monitoring [11].H. Huang and A. V. Savkin, "Navigating UAVs for
India's coastline. There is an infinite number of possibilities. Optimal Monitoring of Moving Groups of Pedestrians or
This robot, in its current configuration, functions as a Vehicles," IEEE Transactions on Vehicular Technology,
platform for further research into the enhancement of its vol. 70, no. 4, April 2021, pp. 3891-3896.
capabilities. [12]. Real-Time Optical-Wireless Video Surveillance
Warfare in the future will be fought between System for High Visual-Fidelity Underwater Monitoring,
operators and machines rather than between warriors on the IEEE Photonics Journal, volume 14, number 2, pages 1-9,
front lines. In our upcoming study, we'll suggest a technique April 2022, article number 7315609.
for shooting a target more accurately and precisely while [13].Energy-Aware Stochastic UAV-Assisted Surveillance,
utilising wireless technology. S. Hosseinalipour, A. Rahmati, D. Y. Eun, and H. Dai, IEEE
Transactions on Wireless Communications, vol. 20, no. 5,
X.REFERENCE pp. 2820-2837, May 2021.
[14]."Visibility-Based Persistent Monitoring of Piecewise
[1]. H. -C. Chang, Y. -L. Hsu, C. -Y. Hsiao, and Y. -F. Linear Features on a Terrain Using Multiple Aerial and
Chen, "Design and Implementation of an Intelligent Ground Robots," in IEEE Transactions on Automation
Autonomous Surveillance System for Indoor Science and Engineering, volume 18, issue 4, pages 1692-
Environments," IEEE Sensors Journal, vol. 21, no. 15, pp. 1704, October 2021.
17335-17349, August 1, 2021. [15]. X. Lin, Y. Yazcolu, and D. Aksaray, "Robust
[2] Machine Learning for Military Trauma: Novel Massive Planning for Persistent Surveillance With Energy-
Transfusion Predictive Models in Combat Zones, Constrained UAVs and Mobile Charging Stations," IEEE
ChristopherMarencoMDahttps://doi.org/10.1016/j.jss.2021. Robotics and Automation Letters, vol. 7, no. 2, April 2022,
09.017 pp. 4157-4164.
[3].In IEEE Transactions on Automation Science and [16]. Ahmed, S. Din, G. Jeon, F. Piccialli, and G.
Engineering, M. Aggravi, G. Sirignano, P. R. Giordano, and Fortino, "Towards Collaborative Robotics in Top View

Authorized licensed use limited to: VTU Consortium. Downloaded on October 10,2024 at 08:54:42 UTC from IEEE Xplore. Restrictions apply.
Surveillance: A Framework for Multiple Object Tracking by Alberto Carrasco 2 and Antonio LaTorre 3.
Detection Using Deep Learning," IEEE/CAA Journal of https://doi.org/10.3390/math10091397
Automatica Sinica, vol. 8, no. 7, July 2021, pp. 1253-1270. [20] Taher, K.A.; Jisan, B.M.Y.; Rahman, M. Network
[17] A Survey of Selected Algorithms Used in Military intrusion detection using supervised machine learning
Applications from the Viewpoints of Dataflow and GaAs. technique with feature selection. In Proceedings of the 2019
Ilir Capuni1 and Veljko Milutinovi´ International Conference on Robotics, Electrical and Signal
[18] Analysis, Design, and Comparison of Machine- Processing Techniques (ICREST), Dhaka, Bangladesh, 10–
Learning Techniques for Networking Intrusion Detection. 12 January 2019
Pierpaolo Dini and Sergio Saponara
https://doi.org/10.3390/designs5010009
[19] Military Applications of Machine Learning: A
Bibliometric Perspective José Javier Galán 1,*, Ramón

Authorized licensed use limited to: VTU Consortium. Downloaded on October 10,2024 at 08:54:42 UTC from IEEE Xplore. Restrictions apply.