Sistem Multi-UAV untuk Pelacakan Multi-Target dalam Ruang Tiga Dimensi

Maynad, Vincentius Charles (2024) Sistem Multi-UAV untuk Pelacakan Multi-Target dalam Ruang Tiga Dimensi. Masters thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 6022222024-Master_Thesis.pdf] Text
6022222024-Master_Thesis.pdf - Accepted Version
Restricted to Repository staff only until 1 October 2026.

Download (4MB) | Request a copy

Abstract

Penelitian ini berkaitan dengan sistem multi-UAV untuk melacak multi-target yang dapat diamati sebagian di lingkungan tiga dimensi yang ber-noise. Permasalahan ini biasa ditemui dalam sistem pertahanan dan pengawasan. Penelitian yang dilakukan merupakan perluasan dari penelitian-penelitian terdahulu yang berfokus terutama pada pengaturan dua dimensi, dapat diamati sepenuhnya, dan atau terukur secara sempurna. Target dimodelkan sebagai sistem linear time-invariant dengan noise Gaussian dan UAV pengejar direpresentasikan dalam model standar enam derajat kebebasan. Persamaan yang diperlukan untuk menggambarkan hubungan antara observasi mengenai target dan state pengejar diturunkan dan direpresentasikan sebagai model Gauss-Markov. Target yang dapat diobservasi sebagian mengharuskan para pengejarnya untuk mempertahankan nilai-nilai keyakinan untuk posisi target. Di hadapan lingkungan yang ber-noise, extended Kalman filter digunakan untuk memperkirakan dan memperbarui keyakinan tersebut. Algoritma Multi-Agent Reinforcement Learning (MARL) terdesentralisasi yang dikenal sebagai Soft Double Q-Learning diusulkan untuk mempelajari kontrol koordinasi di antara para pengejar. Algoritma ini diperkaya dengan regulasi entropi untuk melatih kebijakan stokastik tertentu dan memungkinkan interaksi antar pengejar untuk mendorong perilaku kooperatif. Pengembangan ini mendorong algoritma untuk melakukan eksplorasi area pencarian yang lebih luas dan tidak diketahui yang penting untuk sistem pelacakan multi-target. Algoritma dilatih sebelum diterapkan untuk menyelesaikan beberapa skenario. Percobaan menggunakan berbagai kemampuan sensor menunjukkan bahwa algoritma yang diusulkan memiliki tingkat keberhasilan yang lebih tinggi dibandingkan dengan algoritma dasarnya, hingga 4 kali lipat pada skenario tertentu. Penjelasan tentang banyak perbedaan antara lingkungan dua dimensi dan tiga dimensi juga disediakan.
=======================================================================================================
This research deals with multi-UAV systems to track partially observable multi-targets in a noisy three-dimensional environment. This problem is commonly encountered in defense and surveillance systems.It is a far extension from previous research which focused primarily on two-dimensional, fully observable, and or perfect measurement settings. The targets are modeled as a linear time-invariant system with Gaussian noise and the pursuers UAV are represented in a standard six degrees of freedom model. The equations required to describe the relationship between observations regarding the targets and the pursuer’sstate are derived and represented as a Gauss-Markov model. Partially observable targets require pursuers to maintain belief values for the target position. In the presence of a noisy environment, an extended Kalman filter is used to imagine and describe the belief. A decentralized Multi-Agent Reinforcement Learning (MARL) algorithm known as Soft Double Q-Learning is proposed to study coordination control among pursuers. The algorithm is enriched with entropy regulation to train specific stochastic policies and allows interaction between pursuers to encourage cooperative behavior. This development encourages the algorithm to perform exploration of wider and unknown search areas which is important for multi-target tracking systems. The algorithm is trained before being applied to complete several scenarios. Experiments using various sensor capabilities show that the proposed algorithm has a higher success rate compared to the baseline algorithm, up to 4 times in certain scenarios. An explanation of the many differences between two-dimensional and three-dimensional environments is also provided

Item Type: Thesis (Masters)
Uncontrolled Keywords: coordination control, extended Kalman filter, Multi-Agent Reinforcement Learning, multi-target tracking, multi-UAV systems, kontrol koordinasi, extended Kalman filter, Multi-Agent Reinforcement Learning, pelacakan multi-target, sistem multi-UAV.
Subjects: T Technology > T Technology (General) > T57.5 Data Processing
T Technology > T Technology (General) > T57.6 Operations research--Mathematics. Goal programming
T Technology > T Technology (General) > T57.8 Nonlinear programming. Support vector machine. Wavelets. Hidden Markov models.
T Technology > T Technology (General) > T57.83 Dynamic programming
T Technology > T Technology (General) > T57.84 Heuristic algorithms.
U Military Science > UG1242 Drone aircraft--Control systems. (unmanned vehicle)
Divisions: Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20101-(S2) Master Thesis
Depositing User: Vincentius Charles Maynad
Date Deposited: 06 Aug 2024 23:11
Last Modified: 06 Aug 2024 23:11
URI: http://repository.its.ac.id/id/eprint/111865

Actions (login required)

View Item View Item