在現代無線通訊系統中，如何於缺乏基礎建設、地形遮蔽明顯或通訊資源受限之環 境下，建立穩定且可靠之區域性通信鏈路，一直是重要研究課題。高頻（High Frequency， HF）通訊因具備可透過電離層反射進行超視距（beyond line-of-sight）傳輸之特性，在軍事通信、災害應變、緊急救援及偏遠地區通訊中具有重要應用價值。其中，近垂直入射天波（Near Vertical Incidence Skywave， NVIS）通訊模式特別適用於區域性中短距離覆蓋，可有效填補直射波與低仰角天波之間之跳躍區（skip zone）。然而，HF NVIS 通道易受電離層時變、多徑傳播、Doppler 頻移、頻率選擇性效應及外部干擾影響，使系統輸出品質與波束指向性能產生不穩定現象，進而限制整體通訊效能。 為提升 HF NVIS 通訊系統於複雜通道環境下之穩定性與抗干擾能力，本研究探討 智慧天線結合自適應波束成形技術之應用，並建立一套 MATLAB 模組化模擬架構，以整合 HF NVIS 通道模型、陣列訊號處理模型及自適應權重更新機制。於通道建模方面，本研究納入 NVIS 傳播特性、多徑延遲擴展 、 Rayleigh 與Rician 衰落 、 Doppler 頻移、電離層高度擾動及非理想天線因素，並進一步考慮動態干擾、IQ 不平衡、相位誤差及多頻操作等實際系統條件，以提高模擬環境之完整性與工程意義。於自適應演算法方面，本研究選用最小均方（Least Mean Square， LMS）、正規化最小均方（Normalized LMS， NLMS）及遞迴最小平方（Recursive Least Squares， RLS）等三種代表性方法，針對其在不同陣列規模與通道條件下之收斂特性、均方誤差（MSE）、輸出訊號品質、波束方向圖及干擾抑制能力進行系統性分析與比較。 模擬結果顯示，三種自適應演算法皆可透過權重向量之動態更新形成指向目標方向 之主波束，並對干擾方向產生抑制效果，驗證自適應波束成形應用於 HF NVIS 通訊系統之可行性。在收斂特性方面，RLS 演算法具有最快之收斂速度與最佳之追蹤能力，能於較短時間內達到穩態並維持較佳之干擾抑制表現，但其計算複雜度與記憶體需求亦相對較高；NLMS 演算法則透過輸入訊號能量正規化，有效改善 LMS 在輸入振幅變動下之收斂穩定性，於收斂速度、穩定性與計算量之間取得良好折衷；LMS 演算法雖具結構簡單、實作容易及低複雜度等優點，但其收斂速度較慢，且對步長設定較為敏感。進一步分析顯示，在動態干擾、硬體誤差及多頻 NVIS 條件下，自適應波束成形系統仍可維持一定程度之輸出訊號品質與方向選擇能力，其中 RLS 於干擾抑制能力與方向圖穩定性方面表現最為突出，NLMS 則展現較佳之整體平衡性。 綜合而言，本研究建立之 HF NVIS 智慧天線模擬架構，不僅可作為分析自適應波束成形演算法於時變通道中性能差異之有效平台，亦驗證自適應波束成形技術對提升 HF NVIS 通訊系統穩定性、抗干擾能力及訊號品質具有實質效益。本研究成果可作為未來 HF NVIS 智慧天線系統設計、自適應演算法選擇與模擬平台擴充之參考依據。 關鍵字： HF 通訊、NVIS 、智慧天線、自適應波束成形、LMS 、 NLMS 、 RLS 、 Doppler;In modern wireless communication systems， establishing stable and reliable regional communication links in environments lacking infrastructure， with significant terrain obstruction， or under limited communication resources remains a critical research challenge. High Frequency (HF) communication， which enables beyond line-of-sight (BLOS) transmission via ionospheric reflection， has significant application value in military communications， disaster response， emergency rescue， and communications in remote areas. Among these， Near Vertical Incidence Skywave (NVIS) communication is particularly suitable for regional short- to medium-range coverage， effectively filling the skip zone between ground wave propagation and low-elevation skywave transmission. However， HF NVIS channels are highly susceptible to ionospheric time variations， multipath propagation， Doppler shifts， frequency-selective fading， and external interference， leading to instability in system output quality and beamforming performance， thereby limiting overall communication efficiency. To enhance the stability and interference resistance of HF NVIS communication systems in complex channel environments， this study investigates the integration of smart antennas with adaptive beamforming techniques and develops a modular MATLAB-based simulation framework. This framework incorporates HF NVIS channel models， array signal processing models， and adaptive weight update mechanisms. In terms of channel modeling， this study considers NVIS propagation characteristics， multipath delay spread， Rayleigh and Rician fading， Doppler shift， ionospheric height variations， and non-ideal antenna effects. Furthermore， practical system conditions such as dynamic interference， IQ imbalance， phase errors， and multi-frequency operation are included to improve the realism and engineering relevance of the simulation environment. Regarding adaptive algorithms， three representative methods are employed: Least Mean Square (LMS)， Normalized Least Mean Square (NLMS)， and Recursive Least Squares (RLS). Their convergence behavior， mean square error (MSE)， output signal quality， beam patterns， and interference suppression capabilities are systematically analyzed and compared under different array sizes and channel conditions. Simulation results demonstrate that all three adaptive algorithms can dynamically update weight vectors to form a main beam toward the desired signal direction while suppressing interference， validating the feasibility of applying adaptive beamforming in HF NVIS communication systems. In terms of convergence performance， the RLS algorithm achieves the fastest convergence speed and superior tracking capability， reaching steady-state rapidly while maintaining excellent interference suppression performance; however， it also incurs higher computational complexity and memory requirements. The NLMS algorithm improves convergence stability under varying input signal amplitudes through normalization， achieving a good trade-off among convergence speed， stability， and computational complexity. Although the LMS algorithm features a simple structure， ease of implementation， and low computational cost， it suffers from slower convergence and sensitivity to step-size selection. Further analysis indicates that under dynamic interference， hardware impairments， and multi-frequency NVIS conditions， the adaptive beamforming system can still maintain a certain level of output signal quality and directional selectivity. Among the algorithms， RLS provides the best interference suppression and beam pattern stability， while NLMS offers the most balanced overall performance. In conclusion， the proposed HF NVIS smart antenna simulation framework serves as an effective platform for analyzing the performance differences of adaptive beamforming algorithms in time-varying channels. It also verifies that adaptive beamforming techniques can significantly enhance system stability， interference resistance， and signal quality in HF NVIS communication systems. The results of this study can serve as a reference for future HF NVIS smart antenna system design， adaptive algorithm selection， and simulation platform development. Keywords: HF communication， NVIS， smart antenna， adaptive beamforming， LMS， NLMS， RLS， Doppler shift
