有通道估計誤差下之多天線信號偵測

Abstract

相較於傳統的單輸入單輸出(SISO)系統，使用多傳送與接收天線的多輸入多輸出(MIMO)的無線通訊系統可大幅的增加系統容量(capacity)，因此在過去十年被廣泛的研究，相關技術並已應用於實際系統成為國際無線通訊標準規格不可或缺的一部份。大多數對於MIMO訊號偵測的研究，都假設接收端有百分之百正確的通道資訊(CSI)，但實際上，這是不可能的。通道資訊需透過某種估計演法得到，誤差在所難免，而基於此不正確CSI所偵測的數據之可靠性也必須打折扣。事實上許多研究都顯示接受器的效能將有顯著的降低。因此，如何在設計MIMO接收器時將通道估計誤差一併考慮以盡量減少因上述CSI不匹配的因素造成的效能損失是一項相當迫切的課題，也是本文的主要研究動機。我們首先探討通道估計誤差對兩種MIMO信號偵測法­即粒子群趨動交叉熵偵測法及QRD-M偵測法­的性能之影響。 粒子群趨動交叉熵法（Particle-Swarm-Driven Cross-Entropy）偵測法是在給定接收向量後，試圖尋找傳送信號位置的事後機率分布。這個機率分布是藉由在接收向量附近取樣並且反覆的更新使得機率分佈擁有最小交叉熵（相較於現有的機率分佈）。為了改善效能，粒子群趨動的概念則被引入來提供一組新的機率分布。另一方面，QRD-M則是一個以樹狀結構為基礎的低複雜度偵測法。在少許效能損失下，利用樹狀結構搜尋並且減少每一層的存活路徑。 由於最小歐式距離偵測法在有通道估計誤差的情形下對於前述兩個偵測器並不適用，我們提出考量通道估計誤差效應的新型最佳偵測結構，提出了新的解碼度量(decoding metric)。模擬結果顯示新法雖然增加了些許計算複雜度但卻有顯著的效能提昇。我們也發現將同樣的度量應用在有時空編碼的MIMO系統亦可得到類似的改善。

The multiple-input multiple output (MIMO) technology promises significant capacity increase over conventional single-input single-output systems. Most investigations on MIMO signal detection, however, assume perfect channel state information (CSI), which is difficult, if not impossible, to realize in practice. Earlier studies have shown that that the performance of MIMO detection schemes will suffer from severe degradation in the presence of channel estimation errors.

In this thesis, the effects of imperfect CSI on two MIMO signal detectors, namely, the Particle-Swarm-Driven Cross-Entropy (PSD-CE) based detector and the QRD-$M$ detector, are studied, and new detector structures that take into account the CSI error are proposed.

The PSD-CE detector tries to estimate and refine the `a posteriori probability distribution' of the transmitted signal location given the received vector. The distribution is estimated by sampling over the neighborhood of the received vector and is iteratively updated to the one which has the minimum cross entropy with respect to the current distribution. It is further modified by applying the concept of Particle Swarm Optimization to render a mixture of probability distribution. QRD-M, on the other hand, is an efficient tree-search based detector. It prunes the earch tree to reduce the number of surviving paths with minimum performance loss. Since the minimum Euclidean distance criterion is no longer suitable for both detectors in the presence of channel estimation errors, the proposed MIMO detectors take into account the imperfect CSI effect by averaging the estimation errors to obtain a new decoding metric. Numerical examples are given to demonstrate the performance improvement which is attained with insignificant complexity increase.