Neural Networks and Behaviour Based Control For Education Botanical Robot Navigation
Neural Networks and Behaviour Based Control For Education Botanical Robot Navigation
Neural Networks and Behaviour Based Control For Education Botanical Robot Navigation
aan.burhanuddin@gmail.com
Abstract. The development of robots began when the military needed it as war equipment,
and then it was used by some industries to develop production until now widely used for
education and agriculture. Educational robots are usually more universal and simpler than
industrial or military robots because robots for education are made only for simulations or
prototypes. In this study, the authors surveyed prototypes of Educational botany robots,
namely robots used to distinguish fruit maturity. In this journal, the behaviour-based control
(BBC) algorithm will continue implemented into mobile robots. The movement of the mobile
robot prepared in advance, and then actually the mobile robot is wheeled compared to the
desired path. Besides, both kinematic and dynamic smelling mobile robots are derived and
considered. In this study, only focused on mobile robots three-wheel differential drive, which
will explore in a circular and straight path. In this journal, it can be found that image
processing techniques can be used to determine the maturity of watermelons which are shown
in different average values in each image obtained by the camera. The best models are
generated by layer 32 hidden with an accuracy value of 87% at the training dataset level
60:10:30. The Behaviour Based Control Method has directional movements. But there was an
error in each test so that I couldn't take the watermelon correctly. The error rate reaches 25%
from 40 tests or about ten failures in the form of deviating from the object point.
1. Introduction
The science of robotics comes from the development of automation technology [1]. Until now,
researchers have not kicked off the exact definition of a robot. Even in the relevant research domain,
the classification of robots ranges from "programmable automation" [2] to definitions and approaches
to human characteristics and morphology, such as complex motion and anthropomorphic features [3].
Automation can be interpreted as work carried out by machines from work previously done by
humans [4].
The definition of robots from other studies suggests the importance of the environment as a factor
in the classification of robots. A robot is a very complex and very dynamic system that shows a
degree of autonomy and cognition when performing actions in a real environment [5]. Robots will be
able to understand the situation through input from sensors that do not require direct human relations.
1
The sensor allows robots to be placed in an unstructured state with task-oriented goals and oriented to
human safety.
The development of robots began when the military needed it as war equipment, and then it was
used by some industries to develop production until now widely used for education and agriculture.
Educational robots are usually more universal and simpler than industrial or military robots because
robots for education are made only for simulations or prototypes. In this study, the authors surveyed
prototypes of Educational botany robots, namely robots used to distinguish fruit maturity.
2.1. Method
In this study, an experimental research method used because the use of botanical robots as
educational aids was new so that no research has referred to the topic.
2.2. Design
The design of the robot classified into three main parts, namely the design of algorithms,
mechanical design, and electronic design. Designing algorithms includes designing algorithms in the
control system and designing algorithms on sensors.
2
connect with many other neurons (there are about 1011 neurons in the human brain) to form a network,
an organism can learn to think and make decisions.[7]–[10]s.
a. b C d
Figure 4 Pre-processing Image and Sample Histogram
3
A–1 A-2 B-1 B-2 C-1 C–2 D-1 D-2
Figure 5 Color and Pattern Identification Process on Watermelon with ANN
Figure 7 shows that the color identification process is done by eliminating image color
segmentation to find the best image clusters before continuing the feature extraction process. The
intended image area of coffee is the area that separates the color object of the foreground image and
its background, leaving a transverse pattern on the watermelon. The letters A - 1 and B - 1 show good
and mature fruit quality because the evenly distributed bright green color and pattern are indicated by
the identification of patterns in letters A - 2 and B - 2. The letters C - 1 shows bright colors but
patterns uneven, and the letter D - 1 shows watermelon that is bad because of the dark color and
uneven pattern. This was caused by the fact that the watermelon was rotten.
Table 1 Percentage of testing between human experts and ANN
4
Unstable data makes the calculation method using a camera as a sensor will be difficult to use to
calculate distances but is used to detect watermelon objects as presented in Table 3.
Table 3 Calculate the distance of the camera with objects
Training x Y X y
1 2700 1200 24.26 8.17
2 2700 1200 2.30 33.33
3 2700 1200 0.56 0.83
4 2700 900 26.44 11.11
5 2700 900 21.04 1 33.11
4. Conclusion
In this journal, it can be found that image processing techniques can be used to determine the
maturity of watermelons which are shown in different average values in each image obtained by the
camera. The best models are generated by layer 32 hidden with an accuracy value of 87% at the
training dataset level 60:10:30. The Behavior Based Control Method has directional movements. But
there was an error in each test so that I couldn't take the watermelon correctly. The error rate reaches
25% from 40 tests or about ten failures in the form of deviating from the object point.
5. References
[1] M. Malik and others, “Sistem Otomasi Mesin Perajang Tembakau Skala Industri Kecil dan
Rumah Tangga Berbasis Mikrokontroler (Studi Kasus di Kabupaten Temanggung),”
[Yogyakarta]: Universitas Gadjah Mada, 2012.
[2] Z. Qian and Z. Bi, Recent Development of Rehabilitation Robots, vol. 7. 2014.
[3] T. R. Kurfess, Robotics and Automation Handbook - Thomas R. Kurfess. 2005.
[4] R. Parasuraman, K. A. Cosenzo, and E. De Visser, “Adaptive automation for human
supervision of multiple uninhabited vehicles: Effects on change detection, situation awareness,
and mental workload,” Mil. Psychol., vol. 21, no. 2, pp. 270–297, 2009.
[5] D. Steinfeld, Aaron;Fong, Terrence; Kaber, “Common Metrics for Human-Robot Interaction,”
HRI’06, vol. 42, no. 3, pp. 410–413, 2006.
[6] R. A. Brooks, “Reason, Intelligence Without,” Thought A Rev. Cult. Idea, no. 1293, 1991.
[7] K. Patan, “Neural Networks,” in Studies in Systems, Decision and Control, 2019.
[8] R. E. Neapolitan and R. E. Neapolitan, “Neural Networks and Deep Learning,” in Artificial
Intelligence, 2018.
[9] P. M. Buscema, G. Massini, M. Breda, W. A. Lodwick, F. Newman, and M. Asadi-Zeydabadi,
“Artificial neural networks,” in Studies in Systems, Decision and Control, 2018.
[10] A. Karpathy and L. Fei-Fei, “Deep Visual-Semantic Alignments for Generating Image
Descriptions,” IEEE Trans. Pattern Anal. Mach. Intell., 2017.