Analisis Performa Moving Horizon State Estimation dan Unscented Kalman Filter Untuk Estimasi Posisi dan Orientasi Autonomous Underwater Vehicle

Dzaky, Muhammad (2025) Analisis Performa Moving Horizon State Estimation dan Unscented Kalman Filter Untuk Estimasi Posisi dan Orientasi Autonomous Underwater Vehicle. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5022211127-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Sistem navigasi AUV mengandalkan sensor internal seperti Inertial Measurement Unit (IMU) dan Depth Pressure Sensor yang rentan terhadap akumulasi error, khususnya drift yang terjadi seiring waktu akibat bias sensor, noise acak, dan ketidaksempurnaan kalibrasi. Kondisi ini menyebabkan penurunan akurasi estimasi posisi dan orientasi yang dapat mengakibatkan kegagalan misi. Penelitian ini bertujuan menganalisis dan membandingkan performa dua algoritma estimasi keadaan nonlinier, yaitu Unscented Kalman Filter (UKF) dan Moving Horizon State Estimation (MHSE) dalam estimasi posisi dan orientasi AUV untuk menentukan metode optimal berdasarkan kriteria akurasi dan efisiensi komputasi. Penelitian menggunakan pendekatan simulasi dengan model AUV 5-DOF yang mencakup dinamika translasi, rotasi, dan gaya hidrodinamika. UKF diimplementasikan menggunakan transformasi unscented untuk menangkap statistik sistem nonlinier, sedangkan MHSE menggunakan formulasi optimasi pada sliding window dengan variasi horizon length N=3, 5, 7, dan 8. UKF menunjukkan konsistensi performa dengan RMSE posisi ≈0.22 m dengan waktu komputasi yang rendah. MHSE sensitif terhadap horizon: N = 3 kurang akurat RMSE ≈0,68 m, sementara N = 8 mem¬per¬kecil RMSE hingga ≈0,1 m namun meningkatkan beban komputasi yang lebih tinggi. Penelitian ini menunjukkan bahwa UKF optimal untuk aplikasi real-time, sedangkan MHSE cocok untuk misi presisi tinggi dengan sumber daya komputasi memadai.
======================================================================================================================================
AUV navigation systems rely on internal sensors such as Inertial Measurement Unit (IMU) and Depth Pressure Sensor that are susceptible to error accumulation, particularly drift that occurs over time due to sensor bias, random noise, and calibration imperfections. This condition causes decreased accuracy in position and orientation estimation that can lead to mission failure. This research aims to analyze and compare the performance of two nonlinear state estimation algorithms, namely Unscented Kalman Filter (UKF) and Moving Horizon State Estimation (MHSE) in estimating AUV position and orientation to determine the optimal method based on accuracy and computational efficiency criteria. The research uses a simulation approach with a 5-DOF AUV model that encompasses translational dynamics, rotational dynamics, and hydrodynamic forces. UKF is implemented using unscented transformation to capture nonlinear system statistics, while MHSE uses optimization formulation on a sliding window with horizon length variations N=3, 5, 7, and 8. UKF demonstrates performance consistency with position RMSE ≈0.22 m with low computational time. MHSE is sensitive to horizon length: N = 3 is less accurate with RMSE ≈0.68 m, while N = 8 reduces RMSE to ≈0.1 m but increases computational burden by more than three times.This research demonstrates that UKF is optimal for real-time applications, while MHSE is suitable for high-precision missions with adequate computational resources.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Autonomous Underwater Vehicle, Unscented Kalman Filter, Moving Horizon State Estimation, Nonlinear State Estimation, Underwater Navigation
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20201-(S1) Undergraduate Thesis
Depositing User:	Muhammad Dzaky
Date Deposited:	25 Jul 2025 04:26
Last Modified:	25 Jul 2025 04:26
URI:	http://repository.its.ac.id/id/eprint/121634

Actions (login required)

View Item