Volume 5, Issue 8, August – 2020 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Implementation of Autonomous Control System of

The Chain Wheel Robot Using the Backpropagation
Artificial Neural Network (ANN) Methods
Nur Rachman Supadmana Muda1, Nugraha Gumilar1, R.Djoko Andreas.Navalino²,Tirton.N1, M.Iman Hidayat1
1
Poltekad Kodiklatad, Batu Indonesia
2
Defence University,Jakarta, Indonesia

Abstract:- The purpose of this research is to implement design of combat robot control systems using Artificial
the Artificial Neural Network (ANN) method in combat Neural Network methods.
robots so it can be directed to shoot targets well. The
robot control system uses remote control and II. MATERIALS AND METHODS
autonomous. In the autonomous robot system, ANN
back propagation method is applied, where the weight A. Materials
value variable depends on ultrasonic sensor, GPS and The brushless motor is a moving motor depends on
camera. The microcontroller system will process the "PWM" pulse generated from the pulse generator. The
automatically depending on the sensor input. Output speed of the motor depends on the PWM pulse setting [6].
data is used to direct the robot to the target, tracking
and shooting. Robot is used chain wheel systems and
weapons that used pistol types. The riffle is mounted on
the robot can be moved mechanically azimuth and the
elevation towards the target then triggered
mechanically by the riffle through the activation of data
relays from the microcontroller. Thus, the
backpropagation method can be applied to robots so it
can be functioned autonomously.

Keywords:- Camera, Robot control system, ANN

I. INTRODUCTION Fig 1:- DC Motor Brushless

Robotics is a manifestation of technological This DC motor driver circuit is used to adjust the
developments implementation that exist at this time, indeed direction of rotor rotation on DC motor which consists of
the development of scientific robotics needs to be pursued, several electronic components such as transistors and
especially for those who works in the electronics field [1]. relays.
It needs to know and create new innovations regarding
robots, can be done by applying various methods both
hardware and software. One way to create innovation in
robotics is patched Artificial Intelligence (AI) in robot
which we designed. Using artificial intelligence, it is
expected that the robot control system will be better and
can reduce the error value compared to conventional robot
control systems [2].

Basically, artificial intelligence can not only be used

for control system design, but can also be used for various
purposes such as analysis, diagnosis, predicting future
based on past data, and etc. Artificial intelligence can be
implemented in various fields including electronics, Fig 2:- Driver Motor DC L298
biomedicine, economics, marine, agriculture, geophysics
and etc. On robotics field, artificial intelligence has The current is able to be received or gave by very
produced many kinds of robots with excellent abilities such small microcontroller (in milliampere units) so
as recognizing sound, touch, avoiding obstacles, microcontroller can drive a DC motor it requires a series of
recognizing patterns or paths through which. In this motor drivers that are capable to flow currents up to several
research, artificial intelligence, namely artificial neural ampere.
networks, is used to analyze movements or controls in the

IJISRT20AUG688 www.ijisrt.com 1230

Volume 5, Issue 8, August – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
An Arduino Board that uses ATmega 2560 information which flows through the network. Because of
Microcontroller IC. This board has relatively many Pin I / the ANN characteristic is adaptive, ANN is also often
O, 54 digital inputs / outputs, 15 of which can be used as referred as adaptive tissue.[2]
PWM outputs, 16 analog inputs, 4 UART. Arduino Mega
2560 has 16 Mhz crystal [3]. Simply stated, ANN is a non-linear statistical data
modeling tool. ANN can be used to model complex
relationships between inputs and outputs to find patterns in
data. According to a theorem called the "universal
estimation theorem", ANN with minimum of hidden layer
with a non-linear activation function can model any
measurable Boreal function from one dimension to another.

Fig 3:- Arduino Mega

The abbreviation of Arduino IDE is Integrated

Development Environment, or in a simple language is an
integrated environment that is used for development. It is
referred as environment because the software is used to
program Arduino to do functions which patched using
syntax program. Arduino uses its own programming Fig 5:- Model Artificial Neural Network (ANN)
language which resembles C. The Arduino programming
language (Sketch) has been made changing to facilitate III. RESULT
beginners in programming from original language. Before
being sold to the market, the Arduino microcontroller IC The output block is part of the system tasked with
has embedded a program called Bootlader which the running the system according to the function of the
functions is as an intermediary between the Arduino designed equipment, based on the conditions given by the
compiler and the microcontroller.[3] process block. The modeling robot combat based on
Artificial Neural Network (ANN) is described as Fig6.
MPU-6050 Gyro Sensor 2 + Accelerometer module
uses the MPU-6050 which is a small part of the motion
processing technology. It combines 3-axis gyroscope and 3-
axis accelerometer on the same off-silicon stick together
with an onboard Digital Motion Processor which is capable
to process the complex 9-axis Motion Fusion algorithm, the
MPU-6050 is not far off with the problem of cross-axis
alignment that can creep in parts separate section [4].

Fig 6:- Module placement chart on robots

and controller

To facilitate the reading of the work system of the

Fig 4:- MPU-6050 Gyro Sensor program to be made or Block diagram on Artificial Neural
Network (ANN) combat robot, it will be described as
B. Methods below.
Artificial Neural Network (ANN), or also called
simulated neural network (SNN), or generally just called
neural network (NN), is a network of small group
processing units that are modeled based on human nervous
system. ANN is an adaptive system that can change its
structure to solve problems based on external or internal

IJISRT20AUG688 www.ijisrt.com 1231

Volume 5, Issue 8, August – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

Fig 9:- Firring Robot

 Software Planning
Fig 7:- Data Communication Block Diagram Software design is the design of a programming
language that will be included or embedded in an Artificial
Figure 7 can be known from the working system of Neural Network (ANN) combat robot. Software (program
the tool which is as follows: language) will be programmed on laptop. As Shown Figure
 Arduino Uno functions as the brain processing data 10, a software or sender flow diagram is explained in the
which is received from the wifi module into a command controller.
to control combat robots.
 Motor Driver is used to drive the motor and as a
current amplifier from Arduino UNO so that it can
control the motor.
 Motor Elevation is used to adjust the elevation of the
weapon that is moving up and down.
 Trigger motor is used to pull the trigger of a weapon
(trigger) so weapons in a city combat car can shoot.
 Right Motor and Left Motor are used to drive the
right and left wheels of a city combat car.
 Gradient Sensor/Gyro is used when combat robot
deals terrain conditions so at a certain slope the robot
moves slightly slow which the combat robot does not
reverse.

 Robot Control System

Robot control system is a robot system that is
controlled manually and automatically for maneuvering and
firing movements as shown in Fig 8 and Fig 9

Fig 8:- Maneuver Robot

Fig 10:- Flowchart

IJISRT20AUG688 www.ijisrt.com 1232

Volume 5, Issue 8, August – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
IV. RESULT AND ANALYSIS

A. Result of Arduino Nano and Gyro Testing

The goals of Arduino nano testing and the gyro testing
are determining whether the gyro sensor works well on
Arduino nano and measures the position of the robot.

 Tool
The tools which used on test process are written
below:
 Laptop/PC
 Arduino IDE Software
 Gyro Sensor
 Arc
 Matlab R2014a Software Fig 11:- Arduino dan Gyro picture

 Experiment Procedure  The circuit is placed in the middle of the arc, then
On testing, the experiment steps as follows: connect to the data communication in the Arduino IDE
 Arduino circuit and Gyro sensor as a determinant of program.
position sensors. Shown in Fig 11  Determine the position of the gyro sensor in the arc then
the value can be seen in the Arduino IDE program in
serial monitor.

From the Arduino Uno and gyro testing experiments

above, the data is obtained by displaying it on the Arduino
IDE programming screen and experimenting until
experiment 3.

Value
Angle
Error Error Error
detection/position Trial 1 Trial 2 Trial 3
% % %
1 2 3 4 5 6 7

180° 181.07 0.5 180.36 0.2 181.09 0.6

160° 160.53 160.74 161.08
0.3 0.4 0.6
140° 141.17 0.8 141.20 0.8 141.09 0.7
120° 120.23 0.1 120.18 0.1 121.04 0.8
100° 100.98 0.9 100.88 0.8 100.65 0.6
80° 80.43 0.5 80.99 1.2 80.36 0.4
1 2 3 4 5 6 7

60° 61.14 1.9 60.79 1.3 60.57 0.9

40° 40.78 1.9 40.58 1.4 40.96 2.4
20° 20.69 0.05 20.17 0.8 20.27 1.3
Table 1:- Arduino test result on Gyro

So it can be formulated by error value: This test is done based on clockwise if the test is done
on counter-clockwise then the value that will come minus
sensor measurement − actual measurement in the Arduino program.
=
actual measurement
× 100%
181.07 − 180
= × 100%
180
= 0.5%

IJISRT20AUG688 www.ijisrt.com 1233

Volume 5, Issue 8, August – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
From the Arduino testing experiment, the motor and
DC motor drivers above enter into Matlab for Learning
Process and Prediction Using Bacpropagation Artificial
Neural Network (ANN) as learning on DC motor when it’s
implemented to design of combat robot control systems
using artificial neural networks. On this occasion the author
made 6 hidden layers and repetitions for learning phase.
This can be done if the predicted results produce undesired
values, such as same value or very large difference value.
Repetition is done by selecting the train toolbox and
clicking the train as much as desired.

Fig 12:- Monitor serial picture on 180 degree angle

B. Motor Driver and DC Motor Test

When testing DC Motor, here is the following steps:
 Arduino circuits, motor drivers and DC motors to
control the motor speed as shown in Fig13.

Fig 14 a:- Plan ANN test

As shown in figure 14a 14b and 14c, the result graph

of the test (blue) almost follows from the target graph of
training voltage (red). In the earlier data, the test graph
Fig 13:- Arduino, Driver motor and DC motor shows the adaptation process that will be corrected for the
next data. The adaptation process will continue at the
 Arrange the motor driver and DC motor according to beginning of every change of the voltage number. The test
the experiment of Figure 13. Connect with data results prove that, network training is good enough.
communication in Arduino IDE programming.
 Set the motor speed and stable the voltage on the
specified Arduino pins, then record the results of the
experiment as in Table 2.

Table 2:- Arduino, Motor Driver, and DC motor tests Fig 14.b:- MSE test

IJISRT20AUG688 www.ijisrt.com 1234

Volume 5, Issue 8, August – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
[6]. Roni Permana Saputra, Rizqi Andry Ardiansyah,
Midriem Mirdanies, Arif Santoso, Aditya Sukma
Nugraha, Anwar Muqorobin, Hendri Maja Saputra,
Vita Susanti, Estiko Rijanto, “Dc Brushless Motor
Control Design And Preliminary Testing For
Independent 4-Wheel Drive Rev-11 Robotic
Platform”, Journal of Mechatronics, Electrical Power,
and Vehicular Technology, Vol. 02, No 2, pp 85-94,
2011

Fig 14 c:- Matlab R2014a Simulink Graph

V. CONCLUSION

The results of planning, making and testing

applications, then conclusions can be drawn. These
conclusions include:
 Robot chain wheel control system can be controlled
manually and automatically according to the ANN
method with an error approaching 0.5%.
 The elevation and firing control system is in accordance
with the matlab test data with an accuracy of 0.5%
error.

REFERENCES

[1]. Et. Karthikrajan, “Multiterrain robotic vehicle

with step climbing mechanism,” International
Journal of Advanced Research in Basic Engineering
Sciences and Technology (IJARBEST), volume 3,
Special issue 33, march 2017.
[2]. Manav Raj, Robert Seamans “Primer on artificial
intelligence and robotics” Journal of Organization
Design,2019
[3]. K. M. Merlin Ruby, F. Anne Jenefer, D.
Vidhya,”Study of Arduino Controlled Robotic
System”, International Journal of Innovative Research
in Computer and Communication Engineering, Vol. 4,
Issue 6, June 2016
[4]. Arif Ainur Rafiq, Wahid Nur Rohman, Sugeng Dwi
Riyanto,” Development of a Simple and Low-cost
Smartphone Gimbal with MPU-6050 Sensor”, Journal
of Robotics and Control (JRC), Volume 1, Issue 4,
July 2020
[5]. Israr Hussain, Rocky Chakma, Shafqat Ali, Imran Ali,
”A Microcontroller Based Autonomous Robot For
Navigating And Delivering Purpose”, Transactions of
Nanjing University of Aeronautics and Astronautics ·
October 2018

IJISRT20AUG688 www.ijisrt.com 1235