低軌道衛星下行通訊的波束切換：基於信息幾何優化 的 Actor-Critic 方法設計;Beam-switching for LEO Satellite Downlink Communications: An Actor-critic Framework using Information-Geometric Optimization

NCU Institutional Repository > 資訊電機學院 > 通訊工程研究所 > 博碩士論文 > Item 987654321/98046

請使用永久網址來引用或連結此文件: https://ir.lib.ncu.edu.tw/handle/987654321/98046

題名:	低軌道衛星下行通訊的波束切換：基於信息幾何優化的 Actor-Critic 方法設計;Beam-switching for LEO Satellite Downlink Communications: An Actor-critic Framework using Information-Geometric Optimization
作者:	倪士閔;NI, SHIH-MIN
貢獻者:	通訊工程學系
關鍵詞:	低軌衛星;深度神經網路;igo 演算法;Low Earth Orbit (LEO) satellite;Deep Neural Network (DNN);InformationGeometric Optimization (IGO) algorithm
日期:	2025-06-06
上傳時間:	2025-10-17 12:17:22 (UTC+8)
出版者:	國立中央大學
摘要:	6G 通訊技術被視為下一世代移動通訊的核心，提供比 5G 更高速、低延遲且高效的數據傳輸能力。低軌衛星通訊（LEO）因其全球覆蓋性及低延遲特性，成為未來全球物聯網及高效通訊系統的關鍵技術之一。低軌衛星由於運行速度快，受到都普勒效應影響的影響巨大，導致通道狀態變化劇烈，使用波束成形技術進行波束追蹤成為應對挑戰的重要手段。然而，隨著用戶及波束數量的增加，地面下行鏈路的干擾問題逐漸顯現。本研究聚焦於混合波束成形控制的設計與優化，特別針對低軌衛星與地面用戶間的通訊環境進行建模。研究的核心目標是透過分析不同狀態下各動作的分佈機率，改善波束控制策略。因此吾人提出一種基於信息幾何優化（IGO）的 Actor-Critic 方法設計，將 Actor-Critic 中的 Actor 網路以IGO 演算法替換，用於優化各動作的分布機率，使整體模型簡單化，避免因訓練數據較為缺乏而導致兩個神經網絡的誤差相互影響。實驗結果顯示，所提出的方法能在資訊不充足的情況下快速且穩定的收斂。 ;6G communication technology is regarded as the core of next-generation mobile communications, offering higher speed,lower latency, and more efficient data transmission capabilities compared to 5G. Low Earth Orbit (LEO) satellite communication, with its global coverage and low-latency characteristics, has become one of the key technologies for future global IoT and high-efficiency communication systems. However, due to the high velocity of LEO satellites, they are significantly affected by the Doppler effect, causing rapid variations in channel conditions. Beamforming technology is thus a crucial approach for beam tracking to address these challenges. Nevertheless, as the number of users and beams increases, interference issues in the ground downlink become more prominent. This study focuses on the design and optimization of hybrid beamforming control, specifically modeling the communication environment between LEO satellites and ground users. The core objective of this research is to improve beam control strategies by analyzing the probability distribution of different actions under various conditions. Therefore, we propose an Actor-Critic method using Information-Geometric Optimization (IGO), in which the Actor network in the Actor-Critic framework is replaced with the IGO algorithm to optimize the probability distribution of each action. This simplifies the overall model and prevents the mutual influence of errors between the two neural networks due to limited training data. Experimental results show that the proposed method converges quickly and stably, even under insufficient information.
顯示於類別:	[通訊工程研究所] 博碩士論文

文件中的檔案:

檔案	描述	大小	格式	瀏覽次數
index.html		0Kb	HTML	17	檢視/開啟

在NCUIR中所有的資料項目都受到原著作權保護.

社群 sharing

資料載入中.....