Rico Apriyan Chaniago, . (2026) PEMODELAN ALGORITMA Q-LEARNING UNTUK PENENTUAN RUTE EFISIEN PADA SIMULASI KOMUNIKASI DARURAT JARINGAN LORA MESH KAPAL NELAYAN. Skripsi thesis, Universitas Pembangunan Nasional Veteran Jakarta.
![[img]](http://repository.upnvj.ac.id/style/images/fileicons/text.png) |
Text
ABSTRAK.pdf Download (221kB) |
|
Text
AWAL.pdf Download (858kB) |
|
Text
BAB 1.pdf Restricted to Repository UPNVJ Only Download (421kB) |
|
Text
BAB 2.pdf Restricted to Repository UPNVJ Only Download (1MB) |
|
Text
BAB 3.pdf Restricted to Repository UPNVJ Only Download (471kB) |
|
Text
BAB 4.pdf Restricted to Repository UPNVJ Only Download (2MB) |
|
Text
BAB 5.pdf Download (268kB) |
|
Text
DAFTAR PUSTAKA.pdf Download (243kB) |
|
Text
RIWAYAT HIDUP.pdf Restricted to Repository UPNVJ Only Download (120kB) |
|
Text
LAMPIRAN.pdf Restricted to Repository UPNVJ Only Download (3MB) |
|
Text
HASIL PLAGIARISME.pdf Restricted to Repository staff only Download (35MB) |
|
Text
ARTIKEL KI.pdf Restricted to Repository staff only Download (1MB) |
Abstract
The high rate of fishing vessel accidents in Indonesia underscores the urgent need for reliable inter-vessel emergency communication systems in open-water environments. LoRa Mesh multi-hop networks present a promising solution due to their low power consumption and independence from centralized infrastructure; however, conventional routing methods such as Greedy geographic routing frequently fail under deteriorating channel conditions or dynamic network topology. This study designs and evaluates a Q-learning algorithm as an adaptive routing policy for LoRa Mesh networks among fishing vessels. The agent is built upon a 28-dimensional state space representing seven RSSI categories and four distance zones to the shore station, four semantic relay-selection actions, and a per-hop retry mechanism with a maximum of two attempts. Simulations were conducted using NS-3.41 across three network density scenarios (15, 20, and 25 vessels) within a 20 km × 20 km maritime area, comprising 6,000 training episodes and 3,000 evaluation runs per method. Results demonstrate that Q-learning achieves a Packet Delivery Ratio (PDR) of 79.88%, outperforming Greedy geographic routing at 45.06%, representing an absolute improvement of 34.82 percentage points (77.28%) with a very large effect size (Cohen's d = 1.3698). Furthermore, Q-learning improves the average RSSI by 1.49 dBm, increases SNR by 14.41%, and increases Path Reliability from 0.4512 to 0.8947 (+98.30%) with a very large effect size (Cohen's d = 2.3942). As a trade-off, hop count, end-to-end latency, Time on air (ToA), and total energy consumption increase by 82.52%, 32.60%, 32.60%, and 32.60%, respectively. Nevertheless, the energy consumed per successfully delivered packet is 52.36% lower than that of Greedy geographic routing. Robustness testing across 14 distinct operational conditions consistently confirms the superiority of Q-learning with p < 0.001. These findings demonstrate that an RSSI-based Q-learning approach is a viable adaptive routing solution for improving the reliability of maritime emergency communication systems over LoRa Mesh networks.
| Item Type: | Thesis (Skripsi) |
|---|---|
| Additional Information: | [No.Panggil: 2210511103] [Pembimbing 1: Indra Permana Solihin] [Pembimbing 2: Hamonangan Kinantan Prabu] [Penguji 1: Widya Cholil] [Penguji 2: Nurhuda Maulana] |
| Uncontrolled Keywords: | Emergency Communication, Fishing Vessel, Greedy geographic routing , LoRa Mesh, Q-learning. |
| Subjects: | Q Science > Q Science (General) T Technology > T Technology (General) |
| Divisions: | Fakultas Ilmu Komputer > Program Studi Informatika (S1) |
| Depositing User: | RICO APRIYAN CHANIAGO |
| Date Deposited: | 29 Jul 2026 06:46 |
| Last Modified: | 29 Jul 2026 06:46 |
| URI: | http://repository.upnvj.ac.id/id/eprint/51968 |
Actions (login required)
 |
View Item |