Prediksi Posisi Menggunakan K-Nearest Neighbor Berbasis Wi-Fi Location Fingerprint

Afif, Akhmad Thoriq (2024) Prediksi Posisi Menggunakan K-Nearest Neighbor Berbasis Wi-Fi Location Fingerprint. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5024201028-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (15MB) | Request a copy

Abstract

Pengembangan teknologi Indoor Positioning System (IPS) merupakan aspek penting dalam era digital saat ini, khususnya dengan meningkatnya kebutuhan akan layanan berbasis posisi di lingkungan dalam ruangan. Sistem penentuan posisi global (GPS) yang kurang efektif dalam ruangan mendorong kebutuhan akan sistem IPS yang lebih akurat dan efisien. Tantangan dalam pengembangan IPS meliputi ketidakakuratan sinyal, ketersediaan, dan stabilitas sistem di lingkungan indoor yang kompleks. Penelitian ini bertujuan menggabungkan teknologi Visual Inertial Odometry (VIO) dan WiFi-Based Location Fingerprinting untuk mengatasi batasan sistem IPS eksisting dan meningkatkan akurasi penentuan posisi indoor. Berdasarkan hasil pengujian, ditemukan bahwa waktu eksekusi algoritma metode penggabungan meningkat seiring dengan bertambahnya jumlah fitur yang digunakan, dengan peningkatan waktu eksekusi bersifat linear. Rata-rata waktu eksekusi untuk 4 fitur adalah 0.0485 detik dan untuk 12 fitur sebesar 0.1021 detik. Meskipun terjadi peningkatan jumlah data input, waktu eksekusi algoritma relatif stabil, berkisar antara 0.0461 detik untuk 20 data hingga 0.0478 detik untuk 100 data. Algoritma fusion menunjukkan akurasi yang baik dalam menentukan posisi pada skenario track lurus dengan rata-rata error sebesar 0.15 meter. Namun, pada skenario track belok kiri, rata-rata error mencapai 0.32 meter. Untuk skenario track mengitari ruangan, rata-rata error adalah 0.21 meter.
============================================================
The development of Indoor Positioning System (IPS) technology is an important aspect in the current digital era, especially with the increasing demand for location-based services in indoor environments. The inefficiency of global positioning systems (GPS) indoors has driven the need for more accurate and efficient IPS. Challenges in the development of IPS include signal inaccuracies, availability, and system stability in complex indoor environments. This research aims to combine Visual Inertial Odometry (VIO) and WiFi-Based Location Fingerprinting technologies to address the limitations of existing IPS and improve indoor positioning accuracy. Based on the test results, it was found that the execution time of the merging algorithm increases linearly with the number of features used. The average execution time for 4 features is 0.0485 seconds, and for 12 features, it is 0.1021 seconds. Despite the increase in input data, the algorithm’s execution time remains relatively stable, ranging from 0.0461 seconds for 20 data points to 0.0478 seconds for 100 data points. The merging method shows good accuracy in determining positions on a straight track scenario with an average error of 0.15 meters. However, in a left turn track scenario, the average error reaches 0.32 meters. For a circular track scenario, the average error is 0.21 meters.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Indoor Positioning System, Visual Inertial Odometry, WiFi-Based Location Fingerprinting, K-Nearest Neighbors
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T58.5 Information technology. IT--Auditing
Divisions:	Faculty of Electrical Technology > Computer Engineering > 90243-(S1) Undergraduate Thesis
Depositing User:	Akhmad Thoriq Afif
Date Deposited:	06 Aug 2024 04:43
Last Modified:	06 Aug 2024 04:43
URI:	http://repository.its.ac.id/id/eprint/111407

Actions (login required)

View Item