毫米波（mmWave）通訊具備極大的頻譜資源潛力，被視為解決新 世代無線系統高傳輸速率需求的關鍵技術。然而，毫米波系統存在 高路徑損耗與高硬體複雜度等問題，特別是在實現大規模 MIMO 系 統時更為嚴峻。為克服此限制，本論文採用 Cell-Free Massive MIMO 架構，透過大量分散式接取點（Access Points, APs）協同服 務用戶，不再受限於傳統小區邊界。此外，我們引入混合波束成形 （Hybrid Beamforming）架構，以降低所需 RF chain 的數量，同時 保有陣列與頻譜增益。 本研究的主要目標是最小化下行總發射功率，並滿足每位使用者的 服務品質（QoS）需求，包括資料速率與錯誤率等。我們提出一套動 態 AP 選擇演算法，結合路徑損耗與空間分集增益，動態決定各用 戶最佳的服務 AP 組合。為進一步提升系統的頻譜使用效率與通道 適應能力，亦整合了自適應調變與編碼機制（MCS）至波束設計流程 中。此外，本論文設計了基於通道狀態的子載波配演算法，使每位 ii 用戶得以根據其通道品質獲得最佳的子載波。模擬結果顯示，所提 方法可在滿足 QoS 的前提下有效降低總發射功率，並展現出在不同 用戶密度與 AP 拓撲結構下的穩健性。與傳統固定分群或全數位波 束法相比，我們的方法能在效能與實作成本間取得更佳平衡，並可 在 Cell-Free 毫米波大規模 MIMO 系統中實現接近全數位 BD 架構 的性能，卻大幅降低實作複雜度，具備良好的實用性。;Millimeter-wave (mmWave) communication, with its wide spectrum availability, is a promising solution for meeting the data rate demands of next-generation wireless networks. However, mmWave systems suffer from high path loss and hardware complexity, especially when deploying massive MIMO. To address these challenges, this thesis considers a cellfree massive MIMO-OFDM architecture, in which many distributed access points (APs) jointly serve users without cell boundaries. We adopt a hybrid beamforming structure to reduce the number of RF chains while maintaining array and spectral gain. The main objective of this work is to minimize the total downlink transmit power while satisfying per-user Quality-of-Service (QoS) constraints such as data rate and error probability. We propose a dynamic AP selection algorithm based on path loss and spatial diversity to determine the most energy-efficient AP-user associations. To further enhance robustness and spectrum usage, we integrate an adaptive modulation and coding scheme (MCS) into the beamforming design. Moreover, a subcarrier allocation algorithm is developed to adjust the number of subcarriers per user based on their channel state and QoS requirements. Simulation results verify that our proposed method can significantly reduce power consumption and provide a better trade-off between performance and complexity compared to iv conventional fixed-cluster or fully digital beamforming schemes. The proposed algorithm approaches the performance of block diagonalization (BD) under full digital architecture while achieving much lower implementation cost, making it suitable for practical deployment in largescale cell-free mmWave massive MIMO systems.
