OPTIMASI ROUTING BROADCAST MENGGUNAKAN GRAPH NEURAL NETWORK PADA JARINGAN LORA MESH UNTUK KOMUNIKASI DARURAT KAPAL NELAYAN

Muchammad Dimas Mufti Baskara, . (2026) OPTIMASI ROUTING BROADCAST MENGGUNAKAN GRAPH NEURAL NETWORK PADA JARINGAN LORA MESH UNTUK KOMUNIKASI DARURAT KAPAL NELAYAN. Skripsi thesis, Universitas Pembangunan Nasional Veteran Jakarta.

[img] Text
ABSTRAK.pdf
Download (270kB)
[img] Text
AWAL.pdf
Download (1MB)
[img] Text
BAB I.pdf
Restricted to Repository UPNVJ Only
Download (308kB)
[img] Text
BAB II.pdf
Restricted to Repository UPNVJ Only
Download (865kB)
[img] Text
BAB III.pdf
Restricted to Repository UPNVJ Only
Download (792kB)
[img] Text
BAB IV.pdf
Restricted to Repository UPNVJ Only
Download (1MB)
[img] Text
BAB V.pdf
Download (298kB)
[img] Text
DAFTAR PUSTAKA.pdf
Download (268kB)
[img] Text
RIWAYAT HIDUP.pdf
Restricted to Repository UPNVJ Only
Download (127kB)
[img] Text
LAMPIRAN.pdf
Restricted to Repository UPNVJ Only
Download (1MB)
[img] Text
HASIL PLAGIARISME.pdf
Restricted to Repository staff only
Download (24MB)
[img] Text
ARTIKEL KI.pdf
Restricted to Repository staff only
Download (797kB)

Abstract

The fishing profession carries a high risk of occupational accidents and requires a reliable emergency communication infrastructure. LoRa Mesh technology has the potential to overcome range limitations in maritime environments; however, conventional broadcasting (flooding) mechanisms often lead to broadcast storms, high transmission overhead, and excessive device power consumption. This study proposes the optimization of broadcast routing into a selective forwarding policy using Graph Attention Network (GAT) modeling to predict relay transmission viability at the edge-level task. Maritime network simulations were deterministically executed using the ns-3.41 simulator, generating 1,602,864 hop transmission records utilized to train a 29,969-parameter GAT model, achieving an F1-Score of 90,70%. Performance evaluation was conducted using an off-policy approach on 1,023 emergency (SOS) events and statistically validated through the non-parametric Wilcoxon Signed-Rank test alongside the Probability of Superiority (PSdep) metric. The test results indicate that the GAT policy successfully increased the Packet Delivery Ratio (PDR) reliability from 85,00% to 89,20%. Furthermore, the GAT model proved to be efficient (PSdep 100%) by reducing transmission overhead by 4,8% and saving battery energy consumption by 33,8%, with a noted trade-off in end-to-end latency, which experienced an increase of 1,3% or approximately 30,1 ms. The impact of this efficiency further secures the network's compliance status with the AS923 duty cycle regulation by suppressing the average airtime to 0.89% (well below the 1% maximum limit). The application of graph-based deep learning in maritime LoRa Mesh networks is proven to provide a highly beneficial trade-off optimization between efficiency and the reliability of safety systems at maritime environment.

Item Type: Thesis (Skripsi)
Additional Information: [No. Panggil: 2210511099] [Pembimbing 1: Indra Permana Solihin] [Pembimbing 2: Hamonangan Kinantan Prabu] [Penguji 1: Ridwan Raafi'udin] [Penguji 2: Nurhuda Maulana]
Uncontrolled Keywords: Emergency Communication, LoRa Mesh, Selective Forwarding, Graph Attention Network, Packet Delivery Ratio
Subjects: Q Science > Q Science (General)
T Technology > T Technology (General)
Divisions: Fakultas Ilmu Komputer > Program Studi Informatika (S1)
Depositing User: MUCHAMMAD DIMAS MUFTI BASKARA
Date Deposited: 30 Jul 2026 07:29
Last Modified: 30 Jul 2026 07:31
URI: http://repository.upnvj.ac.id/id/eprint/51925

Actions (login required)

View Item View Item