多用戶多輸出入傳輸的修正型區塊對角化

Abstract

在過去二十年當中，多輸入出傳輸(MIMO)被廣泛的研究討論。從點對點(P2P)到多用戶輸入輸出傳輸(MU-MIMO)，用戶可以共享頻譜資源，但是會造成天線間的干擾(IAI)以及用戶間的干擾(MUI)。為了解決干擾的影響，基地台與用戶端勢必要設計一套傳輸方式抵銷干擾影響。 在過去的研究中，區塊對角化(BD)是一個廣為人知的技術，透過設計出兩層的預編碼矩陣去消除干擾影響。在第一層的預編碼矩陣設計中，基地台確保用戶間不會互相干擾。將目標用戶的訊號放在實際通道上，並且設計解碼程序給目標用戶為第二層編碼設計目標。雖然這一架構能夠獲得一些好處在整體傳輸過程中，但是許多問題造成實行上的困難。其一，整體設計程序未將雜訊加入考慮，使得整體表現在低訊雜比(SNR)時較差。其二，過高的運算複雜度造成硬體實行負擔。其三，在區塊對角化第二層目標用戶解碼程序設計中，必須知道所有用戶的通道訊息，使得在通訊負擔上過於嚴重。 這篇論文著重於這些問題並提出解決方案。利用平方根演算法[8]將雜訊加入預編碼設計流程，借此改善在低訊雜比時的表現，以及利用LQ分解取代原先在區塊對角化第一層預編碼矩陣設計中的奇異值分解(SVD)，使得整體複雜度降低。最後我們更改了第二預編碼矩陣設計，使得用戶僅需要得知自己的通道資訊，即可自行設計解碼程序，大幅降低通訊負擔。

For the past twenty years, multiple-input multiple-output (MIMO) had been widely researched. From point to point (P2P) to multi-user MIMO (MU-MIMO), the user terminals (UTs) can share the same resource block, but cause inter-antenna interference (IAI) and multi-user interference (MUI). In order to resolve influence of interference in multi-user communication systems, the base station and users have to design precoding and decoding scheme. Block Diagonalization (BD) is a well-known algorithm in past research which has two structure precoding process. In the first stage precoding design, it guarantees zero MUI. In the second stage, it separates the desired signal of target users into sub-channels, and designs demodulation matrix to users. Although this structure can get some of benefits in transmit scheme, but problems resist. First, design process does not consider noise, so the performance will poor in low SNR region. Second, high computational complexity causes hardware burden. Third, base station has to use extra bandwidth to transmit demodulation matrix or users joint design. Both methods cause communication overhead. Focus on the drawbacks of BD, we combine the square root algorithm into design process which can consider noise into design process to improve performance in low SNR region and use the LQ decomposition replace SVD operation to reduce computational complexity. Finally, we modify second stage precoding design of BD which can retard communication overhead in original BD algorithm.