| 摘要: | 毫米波(Millimeter wave, mmWave) 抑或是稱為太赫茲
(Terahertz) 通訊系統被認為是往後世代的具高潛力的無線通訊系統
技術，利用高頻傳輸下，擁有傳輸每秒十億位元甚至以上的資料
傳接收速度，但同時也有了高度的傳輸損耗這一項缺點，使通訊
品質大幅下降，而為了保持良好甚至是精進其品質，利用具高指
向性的波束成型(Beamforming) 架構被視為毫米波通訊系統中不可
或缺的關鍵技術，在這技術下，需要精確的訊號發射(出發) 角度
(Angle of Departure, AoD)、接收(入射) 角度(Angle of Arrival, AoA)
以及適當的路徑增益就變得十分重要。尤其在行動通訊(mobile
communications) 抑或是近幾年很熱門的低軌道衛星系統(Low Earth
orbit(LEO) satellite) 下的場景.
由於大氣中的多樣的變動以致於傳送端與接收端的波束發生
錯位，這使得訊號接收的品質明顯下降，所以角度與路徑增益的
估計與追蹤成為毫米波通訊系統的核心主題。在多數的文獻裡搜
索都是基於均勻線性陣列(Uniform Linear Array，ULA) 角度搜索
及估計的情形下，本論文著重探討如何快速並直覺地進行使用均
勻平面陣列(Uniform Planar Array，UPA) 進行3D 空間的初始角度
搜索，亦即對於初始快速定位或通訊設備失聯情況下的重新全域
定位，並考慮單一使用多輸入多輸出正交分頻多工(Multiple-Input
Multiple-Output Orthogonal Frequency-Division Multiplexing , MIMOOFDM)
的UPA 天線架構， 且採取多路徑3 維通道(multi-path
three-dimensional (3D) channel) 為通道環境.
這裡我們提出UPA 分層波束搜索，首先進行粗略的波束匹配
獲得最大接收訊號強度的區塊，由此區塊進行次詳細的波束匹
配，最後再進行詳細的波束匹配使獲得最佳的波束匹配。並進
行模擬，再透過結果進行分析和討論，後續則可以假設傳送端
與接收端的波束中心角為初始估計的出發角、入射角，經由最
小平方法(least squares) 求初始路徑增益則，再利用正交匹配追蹤
(Orthogonal Matching Pursuit, OMP) 取得混合波束成型架構之預編
碼器(precoder) 與結合器(combiner) 的最佳化權重設計，而二，三
維空間自適應波束追蹤，這方面則可以參考蔽實驗室教授及同仁
之論文。最後我們會利用機器/深度學習之模型來探討是否可以簡
單地藉由搜索到的訊號能量來進一步的精確我們的角度。
;Millimeter wave (mmWave) or terahertz (Terahertz) communication
system is considered to be a high-potential wireless communication
system technology for future generations. Under high-frequency
transmission, one billion bits per second or even The above-mentioned
high data transmission and reception rate, but also has the disadvantage of
huge transmission loss, which greatly reduces the communication quality.
In order to maintain a good or even improve the communication quality,
the use of high-directional beamforming (Beamforming) is regarded as
mm It is an indispensable key technology in wave communication systems.
Under this technology, precise signal transmission (departure) aniv
gle (Angle of Departure, AoD), reception (incidence) angle (Angle of
Arrival, AoA) and appropriate path gain are required appears to be particularly
important. Especially in mobile communications (mobile communications)
or in recent years very popular low-orbit satellite system
(Low Earth orbit (LEO) satellite) scene.
Due to various changes in the atmosphere, the beams at the transmitting
end and the receiving end are misaligned, which will significantly
degrade the quality of the received signal. Therefore, the estimation and
tracking of angle and path gain have become the core research topics of
millimeter wave communication systems. In the case that the search in
most papers is based on the uniform linear array (Uniform Linear Array,
ULA) angle search and estimation, this paper focuses on how to
quickly and intuitively use the uniform planar array (Uniform Planar Array,
UPA) for 3D space The initial angle search, that is, for the initial
fast positioning or re-global positioning in the case of satellite loss, and
consider the single use of multiple-input multiple-output orthogonal frequency
division multiplexing (Multiple-Input Multiple-Output Orthogonal
Frequency-Division Multiplexing, MIMO-OFDM) UPA antenna architecture,
and a multi-path three-dimensional channel (multi-path twov
dimensional (3D) channel) is used as the channel environment.
Therefore, we propose UPA hierarchical beam search, which first performs
rough beam matching to obtain the block with the maximum received
signal strength, then performs sub-detailed beam matching on this
block, and finally performs detailed beam matching to obtain the best
beam matching. And use the simulation results for analysis and discussion.
In the future, it can be assumed that the beam center angle between
the transmitting end and the receiving end is the initial estimated signal
departure angle and incident angle, and the initial path gain is calculated
by the least squares method, and then using the orthogonal Orthogonal
Matching Pursuit (OMP) obtains the precoder and combiner weight design
of the hybrid beamforming architecture, as well as the two or three -
dimensional space adaptive beam tracking. In this regard, you can refer
to the professors and colleagues in the hidden laboratory Thesis.
Finally, we will use the machine/deep learning model attached to
MATLAB to explore whether we can simply use the searched signal energy
to further refine our angle or solve the incorrect field pattern caused
by some defects on the antenna Incorrect angles were searched in case. |
