DSpace collection: 博碩士論文

毫米波多用戶大規模多輸入多輸出正交分頻多工系統具子載波樣本選擇之混合波束成形設計;Hybrid Beamforming Design with Subcarrier Sample Selection for mmWave Multi-User Massive MIMO-OFDM Systems

title: 毫米波多用戶大規模多輸入多輸出正交分頻多工系統具子載波樣本選擇之混合波束成形設計;Hybrid Beamforming Design with Subcarrier Sample Selection for mmWave Multi-User Massive MIMO-OFDM Systems abstract: 毫米波(mmWave)大規模多輸入多輸出(MIMO)通訊是未來蜂巢式網路的一項有力技術。然而，傳統全數位波束成形需為每根天線配置專屬射頻(RF)鏈，導致難以負擔的硬體成本與功耗。混合波束成形結構在大規模 MIMO 中扮演關鍵角色，因其能提供高陣列與多工增益，同時大幅降低硬體成本。本論文考慮寬頻通道下、具動態串流指派的上行毫米波多用戶(MU)大規模 MIMO 系統，並採用正交分頻多工(OFDM)以對抗通道的頻率選擇性，目標為最大化整體頻譜效率。在此框架之上，我們提出一種低複雜度的類比波束成形方案，命名為多子載波選擇與合併(Multi-Subcarrier Selection and Combining， MSSC)。MSSC不依賴單一子載波，而是選取一小組代表性子載波並合併其通道協方差，以設計類比預編碼器與合併器，從而捕捉整個頻段的通道變化，同時維持低於全頻帶方法的複雜度。數位預編碼器與合併器則結合區塊對角化(BD)與協調收發處理推導而得，以消除各子載波的干擾。此外，我們提供一個理論分析，將所需的選中子載波數連結至通道的頻率選擇性。透過相干頻寬，我們推導所需子載波數隨通道延遲擴展線性成長，且與總子載波數無關，並將此關係以物理系統參數表示為 ?

基於雙模型混合深度強化學習之雙面主動式 STAR-RIS 輔助整合感測與語義通訊;Twin-Model based Hybrid Deep Reinforcement Learning for Dual-Sided Active STAR-RIS-Assisted Integrated Sensing and Semantic Communications

title: 基於雙模型混合深度強化學習之雙面主動式 STAR-RIS 輔助整合感測與語義通訊;Twin-Model based Hybrid Deep Reinforcement Learning for Dual-Sided Active STAR-RIS-Assisted Integrated Sensing and Semantic Communications abstract: 在本研究中，我們考慮一個由流體天線系統（FAS）與雙面主動式同時透射與反射重構智慧表面（DS-STAR-RIS）輔助的語意整合感知與通訊（semantic ISAC）系統，其中FAS 天線與DS-STAR-RIS 元件皆可於指定區域內調整位置。系統同時考慮通訊、感知與語意傳輸需求，並以語意能源效率（SEE）最大化為主要目標。為此，我們聯合最佳化通訊與感知波束成形、DS-STAR-RIS 相位位移與放大係數、語意壓縮比、FAS 與使用者關聯、FAS 天線位置，以及DS-STAR-RIS 元件位置，同時滿足傳輸功率、DS-STAR-RIS 功率、最低資料率、感知波束圖增益、相位耦合與最小元件間距等限制。為了解決上述高維度且同時包含連續與離散變數的最佳化問題，本研究提出雙模型多智能體混合近端策略最佳化（TM-MAHPPO）架構。RIS 智能體與FAS 智能體分別配置連續型與離散型近端策略最佳化，以處理連續與離散動作，並透過雙模型結構進一步產生與修正決策。此外，本研究在各智能體內部以及智能體之間導入資訊共享，使不同策略能利用其他決策分支所產生的動作資訊，以提升不同決策之間的協調能力。此外，本研究採用變分自編碼器（VAE）壓縮高維度的狀態與動作資訊，以降低學習與資訊交換的負擔，並利用元近端策略最佳化執行超參數最佳化（HO），使PPO 超參數能隨訓練過程動態調整。模擬結果顯示，所提出的方法在SEE 與收斂表現上優於所考慮的基準深度強化學習方法。結果亦顯示，智能體內資訊共享與智能體間資訊共享均能改善不同決策模組之間的協調，而VAE 與HO 則分別在資訊維度與訓練適應性方面提供效益。此外，提高 DS-STAR-RIS 元件的移動能力能增加空間自由度並提升SEE，但隨著元件數量持續增加，其效能增益會受到額外功率消耗與最佳化複雜度影響而逐漸趨緩。;In this thesis， we investigate a semantic ISAC system assisted by FASs and a dual-sided active STAR-RIS， where both the FAS antennas and DS-STAR elements can adjust their positions within predefined regions. The system jointly considers communication， sensing， and semantic transmission requirements， with SEE maximization as the main objective. Accordingly， communication and sensing beamforming， DS-STAR phase shifts and amplification coefficients， semantic compression ratios， FAS-user association， FAS antenna positions， and DS-STAR element positions are jointly optimized under transmit-power， active DS-STAR power， minimum-rate， sensing beampatterngain， phase-coupling， and minimum-spacing constraints. To handle the resulting highdimensional mixed continuous-discrete optimization problem， we propose a TM-MAHPPO framework. The RIS and FAS agents employ ConPPO and DisPPO to optimize continuous and discrete actions， respectively， while the twin-model structure is used to generate and refine the decisions. Intra-agent and inter-agent information sharing are further introduced so that different policies can exploit action information generated by other decision branches. In addition， VAE-based compression is employed to reduce the dimensions of the state and exchanged action information， while Meta-PPO-based HO adaptively adjusts the PPO hyperparameters during training. Simulation results show that the proposed method achieves higher SEE and favorable convergence performance compared with the considered benchmark DRL methods. The results also demonstrate that both intra-agent and inter-agent information sharing improve coordination among different decision modules. VAE compression reduces the information dimension while retaining most of the useful features， and HO improves the adaptability and stability of the learning process. Moreover， increasing the mobility of DS-STAR elements provides additional spatial DoF and improves SEE， although the performance gain gradually saturates as the number of elements increases because of additional power consumption and optimization complexity.

REL-Net: 基於自適應 U-Net 壓縮之資源高效率學習架構用於持續式低軌衛星語義通訊;REL-Net: Resource-Efficient Learning via Adaptive U-Net Compression for Continual LEO Satellite Semantic Communications

title: REL-Net: 基於自適應 U-Net 壓縮之資源高效率學習架構用於持續式低軌衛星語義通訊;REL-Net: Resource-Efficient Learning via Adaptive U-Net Compression for Continual LEO Satellite Semantic Communications abstract: 在低軌衛星遙測平台快速發展的推動下，遙測影像已成為地球觀測任務中不可或缺的資源。然而，要將深度學習模型真正部署到這類衛星平台上，困難之處並不只在於辨識精度：衛星端本身的運算能力有限、對地面下傳的頻寬受限，而且當衛星通過不同地區上空時，所觀測到的地貌也會持續改變；既有研究多半只單獨處理其中一項限制，少有同時兼顧者。為此，本研究提出資源高效率的語義衛星通訊框架 REL-Net，在分割學習（Split Learning）架構下同時處理上述三項限制。分割學習將模型拆成兩半，衛星端只執行前半段的編碼器，其餘的解碼器與分類器交由地面站完成，因此衛星端的運算負擔大幅降低，原始影像也不必離開衛星。針對傳輸量，本研究提出深度壓縮特徵傳輸（Deep Compressed Feature Transfer， DCFT）模組：先為每一個特徵通道評估其區分不同地物類別的能力，只挑選貢獻最大的少數通道進行傳送，再以 Lloyd–Max 量化器將這些數值以較少的位元表示，使每張影像所需傳送的位元組數大幅下降；同時以深度 Q 網路（Deep Q-Network， DQN）依據當下可用的頻寬，決定編碼器各層應採用哪一種特徵擷取方式，其獎勵設計同時考慮新地區的分割精度、傳輸量，以及舊地區知識的損失程度。針對跨地區的持續適應，本研究提出兩階段雙頭式持續增量地物分類（Continual Incremental Land-Cover Classification， CICC）流程：第一階段先在來源資料集上完成訓練，第二階段再為新地區加上一個新的分類頭，同時凍結舊地區的分類頭，並以第一階段的模型作為參照，使模型在學會新地區的同時不至於遺忘舊地區。於 DeepGlobe 與 LoveDA 兩組資料集上的實驗顯示，相較於多種地物分類模型，本研究所提架構位於「精度—運算量」權衡的柏拉圖前緣（Pareto front），亦即沒有任何比較對象能在兩項指標上同時更好，能在明顯降低運算成本與傳輸負擔的前提下維持具競爭力的分割效能。實驗另外得到兩項觀察：其一，第一階段並非訓練得越久越好，訓練程度適中的模型保有較大的調整空間，其在第二階段的遺忘程度反而遠低於過度訓練的模型；其二，兩個遷移方向所產生的遺忘程度差異極大，顯示評估此類系統時必須同時觀察分割精度、遺忘量與傳輸量三項指標，僅以單一精度指標判斷容易造成誤判。;With the rapid advancement of low-Earth-orbit satellite sensing plat- forms， remote sensing imagery has become an indispensable resource for Earth observation. Deploying deep models on such satellites， however， is difficult for reasons that go beyond accuracy: the onboard processor has lim- ited computing power， the downlink to the ground has limited bandwidth， and the land cover the satellite observes keeps changing as it passes over new regions. Most existing works address only one of these issues at a time. This thesis proposes REL-Net， a resource-efficient framework for semantic satellite communication that addresses all three within a split learning ar- chitecture. Split learning divides the network in two: the satellite runs only the shallow encoder， while the ground station runs the decoder and the clas- sifier， so that the onboard computing cost stays small and the raw image never leaves the satellite. To reduce what must be transmitted， we design a Deep Compressed Feature Transfer (DCFT) module that scores every fea- ture channel by how well it separates the land-cover classes， sends only the highest-scoring channels， and represents their values with few bits through a Lloyd–Max quantizer， which sharply lowers the bytes sent per image. A Deep Q-Network (DQN) then decides， according to the available bandwidth， which feature-extraction branch each encoder stage should use; its reward accounts jointly for the accuracy on the new region， the transmitted volume， and the knowledge lost on the old region. For adaptation across regions， we propose a two-stage dual-head Continual Incremental Land-Cover Classifica- tion (CICC) procedure: the model is first trained on the source dataset， and a second classification head is then added for the new region while the old head is frozen and the Stage 1 model is kept as a reference， so that the new region is learned without erasing the old one. Experiments on DeepGlobe and LoveDA show that， against a broad set of land-cover classification models， the pro- posed system lies on the Pareto front of the accuracy–computation trade-off， meaning that no compared model is better on both axes at once， and that it markedly reduces computation and transmission while keeping competitive accuracy. Two further observations are reported. Longer Stage 1 training is not always better， since a moderately trained checkpoint keeps more room to adjust and forgets far less in Stage 2 than an over-trained one; and forgetting differs greatly between the two transfer directions， which shows that accu- racy， forgetting， and bandwidth must be read together rather than accuracy alone.

應用於非地面網路中基於都卜勒頻率之換手策略;應用於非地面網路中基於都卜勒頻率之換手策略

title: 應用於非地面網路中基於都卜勒頻率之換手策略;應用於非地面網路中基於都卜勒頻率之換手策略 abstract: 在非地面網路（Non-Terrestrial Networks， NTN）架構中，低軌道（Low Earth Orbit， LEO）衛星因具備較低的傳播延遲與實現廣域覆蓋的潛力，已成為新世代行動通訊網路的重要發展方向。然而，LEO 衛星與地面使用者設備（User Equipment， UE）之間具有高速相對運動，使 UE 在通訊過程中需要頻繁進行換手（Handover， HO），以維持連線的連續性與穩定性。傳統 5G NR 換手機制主要針對地面網路環境設計，通常以參考訊號接收功率（Reference Signal Received Power， RSRP）作為主要判斷依據。然而，在 LEO 衛星網路場景中，RSRP 的空間變化可能相對平緩，難以判斷換手時機，造成過早換手、重複換手或換手時機不適當等問題。為了改善上述問題，本研究提出一種基於都卜勒頻率之換手策略，利用都卜勒頻移與都卜勒變化率隨衛星與地面相對幾何關係呈現規律變化的特性，判斷服務衛星相對於 UE 的接近或遠離趨勢，並估計候選衛星的預期服務時間，將此估測資訊應用於換手策略中。模擬結果顯示，相較於傳統以 RSRP 為主要判斷依據的換手機制，本研究所提出的策略可有效降低換手次數，減少不必要換手的發生次數，並延長 UE 與單一服務衛星之平均連線時間，減少因頻繁換手所產生的信令開銷。研究結果顯示，都卜勒頻率特徵可作為 LEO NTN 移動性管理的重要輔助資訊，有助於提升換手決策的穩定性。;In non-terrestrial network (NTN) systems， low Earth orbit (LEO) satellites have emerged as an important technology for next-generation mobile communication networks because of their relatively low propagation delay and potential for wide-area coverage. However， the high relative motion between LEO satellites and terrestrial user equipment (UE) requires frequent handovers (HOs) to maintain connection continuity and stability. Conventional 5G New Radio (NR) handover mechanisms are primarily designed for terrestrial network environments and typically rely on reference signal received power (RSRP) as the main decision criterion. However， in LEO satellite network scenarios， the spatial variation in RSRP may be relatively gradual， making accurate handover timing decisions difficult and potentially resulting in premature， repeated， or improperly timed handovers. To address these issues， this study proposes a Doppler-frequency-based handover strategy. The proposed strategy exploits the predictable variations in Doppler shift and Doppler rate resulting from changes in the relative geometry between the satellites and the ground. These Doppler characteristics are used to determine whether the serving satellite is approaching or moving away from the UE and to estimate the expected service time of candidate satellites. The estimated information is then incorporated into the handover decision process. Simulation results demonstrate that， compared with conventional handover mechanisms that primarily rely on RSRP， the proposed strategy can effectively reduce the number of handovers and unnecessary handovers while extending the average duration for which a UE remains connected to a single serving satellite. Consequently， the signaling overhead associated with frequent handovers can be reduced. These results indicate that Doppler-frequency characteristics can serve as important auxiliary information for mobility management in LEO NTN systems and contribute to improving the stability of handover decisions.