Electronic Theses and Dissertation

Peningkatan jumlah perangkat nirkabel di lingkungan Fakultas Teknik Universitas Syiah Kuala menyebabkan kepadatan lalu lintas jaringan dan meningkatnya interferensi antar Access Point (AP) pada standar IEEE 802.11ax (Wi-Fi 6), sehingga menurunkan performa jaringan akibat keterbatasan algoritma Rate Adaptation (RA) dalam membedakan collision dan penurunan kualitas kanal. Penelitian ini bertujuan melakukan analisis komparatif terhadap tiga algoritma RA berbasis Machine Learning, yaitu Thompson Sampling, Reinrate, dan FTMrate, untuk mengetahui efektivitas adaptasi Modulation and Coding Scheme (MCS) pada kondisi Overlapping Basic Service Set (OBSS). Metode yang digunakan adalah simulasi eksperimental menggunakan Network Simulator 3 (NS-3) dengan skenario multi-Access Point dan parameter evaluasi berupa throughput, delay, packet loss, dan waktu komputasi. Hasil penelitian menunjukkan bahwa algoritma FTMrate merupakan mekanisme yang paling optimal pada kondisi jaringan lengang, menghasilkan throughput tertinggi hingga 93,53 Mbps. Di sisi lain, pada kondisi jaringan padat (40 stasiun pengguna), algoritma Thompson Sampling menunjukkan tingkat stabilitas adaptasi yang sangat baik dengan beban waktu komputasi paling efisien (45,25 detik). Sementara itu, algoritma Reinrate terbukti sebagai arsitektur pelindung keutuhan data terbaik di zona interferensi ekstrem dengan menahan batas kecepatan transmisi untuk menekan packet loss hingga titik terendah mutlak 3,29%. Berdasarkan temuan tersebut, penelitian ini merekomendasikan penyesuaian implementasi penempatan algoritma berdasarkan karakteristik kepadatan dan prioritas lalu lintas jaringan di masing-masing area Gedung Fakultas Teknik USK.

The increasing number of wireless devices in the Faculty of Engineering at Universitas Syiah Kuala has led to higher network congestion and interference among Access Points (AP) operating under the IEEE 802.11ax (Wi-Fi 6) standard, resulting in network performance degradation due to the limitations of conventional Rate Adaptation (RA) mechanisms in distinguishing collisions from channel quality degradation. This study aims to conduct a comparative analysis of three Machine Learning-based RA algorithms, namely Thompson Sampling, Reinrate, and FTMrate, to evaluate their effectiveness in adapting the Modulation and Coding Scheme (MCS) under Overlapping Basic Service Set (OBSS) conditions. The method used is an experimental simulation using Network Simulator 3 (NS-3) with a multi-Access Point scenario and evaluation parameters including throughput, delay, packet loss, and computation time. The results show that the FTMrate algorithm is the most optimal mechanism in sparse network conditions, producing the highest throughput of up to 93.53 Mbps. On the other hand, in dense network conditions (40 user stations), the Thompson Sampling algorithm demonstrates excellent adaptation stability with the most efficient computation time (45.25 seconds). Meanwhile, the Reinrate algorithm is proven to be the best data integrity protection architecture in extreme interference zones by restraining transmission speed limits to suppress packet loss to an absolute minimum of 3.29%. Based on these findings, this study recommends adjusting the placement of the algorithms based on the density characteristics and traffic priorities in each area of the USK Faculty of Engineering Building.