大規模多輸入多輸出系統在部分連接混合式架構下基於波束選擇方法之最佳化空間預編碼

Abstract

因應現今行動通訊與無線通訊的發展，我們針對降低硬體成本、減少耗能、提高系統容量及大規模多輸入多輸出系統等等在本實驗室先前提出的系統架構下繼續進行效能之優化。 我們延續以往所使用的波束成型選擇之空間預編碼(Beamforming-selection spatial precoding, BSSP)結合混合式預編碼(Hybrid Beamforming)方法做為架構，並以最佳化演算法調整預編碼矩陣之參數，此一作法能夠有效減少干擾，同時降低CSI回傳的overhead及硬體成本消耗。 在本篇論文中，除了先前本實驗室討論過的交互耦合(Mutual Coupling)與空間相關性(spatial correlation)影響，另外加入了Gain Phase Error跟Phase Noise兩種誤差進行模擬，讓通道更貼近真實的狀況，並調整天線擺放的位置讓理論平均錯誤率有更好的表現。在最佳化演算法的部分，我們以Grey Wolf Optimizer(GWO)取代了以往的第二代合作式共同粒子群演算法(Cooperatively Coevolving Particle Swarm Optimization, CCPSO2)，改善了原本運算時間太長及維度不夠的問題，因此能夠模擬天線數量較多的情形。 由於我們所提出的架構中每條天線都需要一條RF(Radio Frequency) chain，隨著天線數量的增加，在連接上耗損的功率也越來越多，因此我們在理論平均錯誤率表現及耗能上進行取捨，加入了部分連接(Partially Connected)架構以降低傳輸的功率耗損，並以離散型GWO進行連接方式的最佳化。為了優化整體的表現，我們除了線性陣列(Linear Array)以外，也加入了圓形陣列(Circular Array)來進行模擬，並加入半正交空時區塊碼 (Quasi-Orthogonal Space-Time Block Code, QOSTBC)技術，透過同時最佳化天線部分連接方式以及天線元件擺設位置來降低錯誤率，優化整體表現。

In response to the development of mobile communications and wireless communications, we continue to optimize performance under the system architecture previously proposed by our laboratory for reducing hardware costs, reducing energy consumption, increasing system capacity, and massive multiple-input-multiple-output systems. We continue to use Beamforming-selection spatial precoding (BSSP) combined with Hybrid Beamforming (Hybrid Beamforming) as the architecture used in the past, and use the optimized algorithm to adjust the parameters of the precoding matrix. This approach can effectively reduce interference, while reducing the hardware cost and the overhead of CSI feedback. In this paper, in addition to the effects of Mutual Coupling and Spatial Correlation discussed earlier in this laboratory, gain phase error and phase noise are also added for simulation to make the channel closer to reality. Moreover, we adjust the position of the antenna to make the theoretical average error rate have a better performance. In the optimization algorithm part, we replaced the previous second-generation Cooperatively Coevolving Particle Swarm Optimization (CCPSO2) with Grey Wolf Optimizer (GWO), which improved the original calculation time and dimensionality. Thus, it can simulate the scenario of a large number of antennas. Since each antenna in our proposed architecture requires an RF (Radio Frequency) chain, when the number of antennas increases, more and more power is consumed in the connection. Therefore, we make a trade-off between theoretical average error rate performance and energy consumption. A Partially Connected architecture was added to reduce the power loss of transmission, and the discrete GWO was used to optimize the connection mode. In order to optimize the overall performance, not only the linear array (Linear Array) has been employed, we also added a circular array (Circular Array) for simulation, and added a semi-orthogonal space-time block code (Quasi-Orthogonal Space-Time Block Code, QOSTBC) technology, which reduces the error rate and optimizes the overall performance by optimizing the connection method of the antenna part and the placement of the antenna element at the same time.