ACA-ESRGAN: Metode Super-Resolusi untuk Transmisi Gambar Efisien pada Kendaraan Otonom

Muhammad Wildan Syaifullah, - and Galura Muhammad Suranegara, - and Endah Setyowati, - (2025) ACA-ESRGAN: Metode Super-Resolusi untuk Transmisi Gambar Efisien pada Kendaraan Otonom. S1 thesis, Universitas Pendidikan Indonesia.

Abstract

Integrasi Kendaraan Listrik Otonom (AEVs) dalam lingkungan terbatas memerlukan solusi komunikasi yang efisien dan hemat energi, terutama untuk mentransmisikan data berdimensi tinggi yang penting bagi navigasi dan pemantauan secara real-time. Teknologi komunikasi LoRa (Long Range), yang dikenal dengan jangkauan luas dan konsumsi daya rendah, menjadi opsi yang layak namun memiliki keterbatasan bandwidth rendah, yang membatasi transmisi data, terutama untuk data gambar. Penelitian ini memperkenalkan pendekatan inovatif yang menggabungkan komunikasi LoRa dengan teknik super-resolusi berbasis Generative Adversarial Network (GAN) untuk transmisi gambar yang efektif pada AEV. Gambar beresolusi rendah ditransmisikan melalui LoRa untuk menghemat bandwidth, dan kemudian di-upscale menjadi resolusi tinggi menggunakan Super-Resolution GAN (SRGAN) di jaringan pusat. Studi ini meningkatkan arsitektur SRGAN dengan menerapkan model Enhanced Super-Resolution GAN (ESRGAN), yang mencakup Residual-in-Residual Dense Block (RRDB) yang disesuaikan untuk skenario operasional AEV. Selain itu, modifikasi struktural seperti lapisan konvolusi yang dioptimalkan, pendekatan pelatihan progresif, dan mekanisme Adaptive Convolutional Attention (ACA) yang mengintegrasikan Channel Attention, Spatial Attention, dan Squeeze-and-Excitation (SE) blocks meningkatkan efisiensi model dalam rekonstruksi gambar. Hasil eksperimen menunjukkan efektivitas pendekatan ini dalam mengatasi keterbatasan bandwidth LoRa, mendukung transmisi gambar berkualitas tinggi untuk aplikasi AEV di lingkungan dengan ruang terbatas. Penelitian ini menjadi dasar bagi pengembangan sistem komunikasi hemat bandwidth dan efisiensi tinggi di masa depan dalam kendaraan otonom. ----- The integration of autonomous electric vehicles (AEVs) into constrained environments requires efficient, low-energy communication solutions, particularly for transmitting high-dimensional data essential to real-time navigation and monitoring. LoRa (Long Range) communication technology, known for its extended range and low power consumption, presents a viable option but is limited by its low bandwidth, which constrains data transmission, especially for image data. This research introduces a innovative approach combining LoRa communication with a Generative Adversarial Network (GAN)-based super-resolution technique for effective image transmission in AEVs. Low-resolution images are transmitted over LoRa to conserve bandwidth, and subsequently upscaled to high resolutions using a Super-Resolution GAN (SRGAN) at the central network. Our study enhances the SRGAN architecture by implementing the Enhanced Super-Resolution GAN (ESRGAN) model, which includes a Residual-in-Residual Dense Block (RRDB) tailored for AEV operational scenarios. Additionally, structural modifications, such as optimized convolution layers, a progressive training approach, and an Adaptive Convolutional Attention (ACA) mechanism integrating Channel Attention, Spatial Attention, and Squeeze-and-Excitation (SE) blocks, improve model efficiency in image reconstruction. Experimental results demonstrate the effectiveness of this approach in overcoming LoRa’s bandwidth constraints, supporting high-quality image transmission for AEV applications in limited-space environments. This research lays the groundwork for future advancements in low-bandwidth, high-efficiency communication systems within autonomous vehicles.

	Text S_SISTEL_2102507_Title.pdf Download (1MB)
	Text S_SISTEL_2102507_Chapter1.pdf Download (510kB)
	Text S_SISTEL_2102507_Chapter2.pdf Restricted to Staf Perpustakaan Download (881kB)
	Text S_SISTEL_2102507_Chapter3.pdf Download (708kB)
	Text S_SISTEL_2102507_Chapter4.pdf Restricted to Staf Perpustakaan Download (1MB)
	Text S_SISTEL_2102507_Chapter5.pdf Download (401kB)
	Text S_SISTEL_2102507_Appendix.pdf Restricted to Staf Perpustakaan Download (1MB)

Item Type:	Thesis (S1)
Additional Information:	https://scholar.google.com/citations?view_op=list_works&hl=en&authuser=2&user=SopsQwEAAAAJ Dosen Pembimbing: Galura Muhammad Suranegara: 6703764 Endah Setyowati: 6681149
Uncontrolled Keywords:	Kendaraan Listrik Otonom, Komunikasi LoRa, Jaringan Adversarial Generatif, Super-Resolution GAN, Transmisi Gambar. Autonomous Electric Vehicles, LoRa Communication, Generative Adversarial Networks, Super-Resolution GAN, Image Transmission.
Subjects:	T Technology > T Technology (General)
Divisions:	UPI Kampus Purwakarta > S1 Sistem Telekomunikasi
Depositing User:	Muhammad Wildan Syaifullah
Date Deposited:	21 Jul 2025 08:17
Last Modified:	21 Jul 2025 08:17
URI:	http://repository.upi.edu/id/eprint/134255

Actions (login required)

View Item