Penghindaran Halangan Pada Quadcopter Menggunakan Metode Local Mean K-Nearest Centroid Neighbor Dengan Modifikasi

Prasetyo, Hendy (2021) Penghindaran Halangan Pada Quadcopter Menggunakan Metode Local Mean K-Nearest Centroid Neighbor Dengan Modifikasi. Masters thesis, Institut Teknologi Sepuluh Nopember.

[img] Text
07111950025001-Master_Thesis.pdf - Accepted Version
Restricted to Repository staff only until 1 October 2023.

Download (4MB) | Request a copy
[img] Text
07111950025001-Master_Thesis.pdf - Accepted Version
Restricted to Repository staff only until 1 October 2023.

Download (4MB) | Request a copy

Abstract

Sistem penghindaran halangan sangat diperlukan saat quadcopter menjalankan tugasnya. Salah satu halangannya adalah quadcopter lain (quadcopter halangan) yang melintas pada jalur target. Kehadiran quadcopter halangan tidak diketahui sebelumnya dan memiliki kecepatan yang bervariasi. Pada penelitian ini, dibuat sistem penghindaran halangan pada quadcopter yang memiliki 5 aksi hindar, yaitu kanan, kiri, atas, bawah dan berhenti. Sistem penghindaran juga dilengkapi dengan prediksi kecepatan quadcopter halangan untuk mengatasi kecepatan quadcopter halangan yang bervariasi. Hasil dari prediksi halangan tersebut selanjutnya digunakan untuk menentukan aksi hindar. Dalam menentukan aksi hindar, juga mempertimbangkan jarak terdekat dan efisiensi energi karena keterbatasan daya baterai pada quadcopter. Maka dari itu, diperlukan algoritma machine learning untuk mempermudah dalam menentukan aksi hindar tersebut. Pada penelitian ini, algoritma machine learning yang digunakan adalah LMKNCN (Local Mean K-Nearest Centroid Neighbor) yang dimodifikasi. Pada LMKNCN membutuhkan waktu komputasi sebesar 0.191318 detik dalam proses learning antara data training dengan data testing, sedangkan modifikasi LMKNCN memerlukan 0.142341 detik. Modifikasi LMKNCN menghasilkan keputusan aksi hindar dengan akurasi 97,5 %. Dari hasil simulasi menunjukkan bahwa quadcopter dapat mencapai titik target tanpa menabrak quadcopter halangan yang memiliki kecepatan bervariasi. ====================================================================================================== Obstacle avoidance systems are indispensable when the quadcopter does its job. One of the obstacles is another quadcopter (quadcopter hitch) crossing the target path. The presence of the hitch quadcopter was not previously known and varied in speed. In this study, an obstacle avoidance system was developed on a quadcopter that has 5 avoidance actions, namely right, left, up, down and stop. The avoidance system is also equipped with obstacle quadcopter speed prediction to cope with varying obstacle quadcopter speeds. The results of the prediction of the obstacle are then used to determine the avoidance action. In determining the avoidance action, also consider the shortest distance and energy efficiency due to the limited battery power of the quadcopter. Therefore, a machine learning algorithm is needed to make it easier to determine the avoidance action. In this study, the machine learning algorithm used is a modified LMKNCN (Local Mean K-Nearest Centroid Neighbor). LMKNCN requires a computation time of 0.191318 seconds in the learning process between training data and testing data, while the LMKNCN modification requires 0.142341 seconds. The LMKNCN modification resulted in an avoidance action decision with an accuracy of 97.5%. The simulation results show that the quadcopter can reach the target point without hitting an obstacle quadcopter that has varying speeds.

Item Type: Thesis (Masters)
Uncontrolled Keywords: Clustering, Energy Efficient, Machine Learning, Modifikasi LMKNCN, Movement Trends, Navigasi Quadcopter, Penghindaran Halangan, Modified LMKNCN, Obstacle Avoidance, Quadcopter Navigation.
Subjects: T Technology > TL Motor vehicles. Aeronautics. Astronautics > TL776 .N67 Quadrotor helicopters--Automatic control
Divisions: Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20101-(S2) Master Thesis
Depositing User: Hendy Prasetyo
Date Deposited: 20 Aug 2021 13:41
Last Modified: 21 Aug 2021 12:33
URI: https://repository.its.ac.id/id/eprint/88106

Actions (login required)

View Item View Item