多段式最佳化之空間預編碼應用於大規模多輸入多輸出混合式架構系統

Abstract

延伸自本實驗室先前提出的波束成型選擇空間預編碼 (BSSP)，為了節省毫米波(mm-Wave)大規模多輸入多輸出系統(massive MIMO) 的傳輸負擔與硬體成本，我們結合了傳送端混合式預編碼與BSSP兩種方法。此外，我們使用灰狼演算法 (GWO) 對預編碼矩陣進行參數最佳化。 不同於先前所提出的天線選擇最佳化結合多段式最佳化方法，該方法所使用的數位預編碼器(Baseband precoder)是基於MMSE法則，在本論文中我們將其使用GWO做最佳化，最佳化第一段為調整天線選擇矩陣與類比預編碼器，最佳化第二段為調整接收器，最佳化第三段為再調整天線選擇矩陣與類比預編碼器，最佳化第四段為調整數位預編碼器，最佳化第五段為調整天線選擇矩陣與類比預編碼器，最佳化第六段為調整接收器，最佳化第七段為再調整類比天線選擇矩陣與類比預編碼器，並且重複一到七段最佳化的方法直到效能改進不多的時候，即達到類比預編碼器、數位預編碼器、天線選擇矩陣、接收器皆為最佳的，如此能使用相比於全連接架構更少的天線數目，並且又能得到更好錯誤率。

The goal of this thesis is mainly to improve the Beamforming-selection spatial precoding (BSSP) previously developed by our laboratory as well as reduce the millimeter-wave massive multiple-input multiple-output (mm-Wave massive MIMO) wireless system’s hardware cost. By incorporating two methods, namely hybrid precoding and BSSP at the transmission end, we then employ Gray Wolf Optimization (GWO) and its variants, namely binary gray wolf algorithm and discrete gray wolf algorithm to obtain the optimal parameters for the precoding matrix required for mm-Wave massive MIMO wireless communications. Different from the previously proposed antenna selection optimization combined with multi-stage optimization method, the digital precoder (Baseband precoder) used in this method is based on the MMSE rule. During the proposed seven-stage optimization process, we first optimize antenna selection matrix and the analog precoder. The second stage is to optimize the receiver, the third stage is to optimize antenna selection matrix and the analog precoder, the fourth stage is to optimize the digital precoder, the fifth stage is to optimize antenna selection matrix and the analog precoder, the sixth stage is to optimize the receiver, and the seventh stage is to optimize antenna selection matrix and the analog precoder. This optimization process is then repeated until the overall system’s performance achieves a highly satisfactory result. That is, the analog precoder, digital precoder, antenna selection matrix, and receiver obtained are all optimal. Hence, the resulting mm-Wave massive MIMO wireless system can use fewer antennas and achieve better bit error rate than the system which uses a fully connected architecture. Finally, computer simulation examples are also presented for confirmation and comparison.