於多使用者多輸入多輸出系統中反覆式訓練序列設計

Abstract

下一代通訊標準制訂於使用多輸入多輸出傳輸技術以增進容量和連結可靠度。包含SU-MIMO、MU-MIMO和network MIMO(或稱CoMP)系統皆工作於OFDM和OFDMA使解碼和多重存取更容易。然而,降低解碼複雜度必須仰賴正確的通道資訊和同步(例如:訊框和載頻的位移)。可惜地,在目前標準中的前導序列和訓練序列並不一定針對通道估測的最佳化來設計,且通常對於提供正確訊框同步有偏差。因為這些序列沒有考慮或利用通道特性,如:傳送端和接收端的關係和空間上的關聯,所以影響了序列設計的最佳化。有色雜訊使得多使用者系統中的序列設計變成非凸面(nonconvex)的問題。因此,為了找到全域最佳解使得計算量更大。先前的研究假設傳送端和接收端之前存在一點或不存在關係,使得MIMO通道模型可用Kronecker模型來表示,但是這個假設在真實系統中是有問題的。另外,利用非線性求解的方法也有被提出,但不保證找到全域最佳解。 在這篇論文中,瀏覽了SU-MIMO傳輸技術,並且討論單一使用者和多使用者系統中的序列設計技術和通道估測。接著,提出一個反覆式序列設計的方法來估測有色雜訊中的通道。我們提出的方法沒有利用非線性求解和傳送端及輸出端之前不存在關係的假設。這個方法被證明至少能得到一組區域最佳解,且模擬的結果優於先前的方法。 這篇論文也有討論到MU-MIMO系統中的耐用(robust)訓練序列設計、多使用者MIMO-OFDM系統中同時訓練和減少PAPR的序列設計。

Next generation communication standards are set to use MIMO transmission techniques to improve capacity and link reliability. This methods, which include SU-MIMO, MUMIMO, and network MIMO (also known as CoMP), are expected to work seamlessly with OFDM and OFDMA to provide ease of decoding and multiple access. The reduced decoding complexity, however, is traded off by requiring accurate channel information, and synchronization (e.g. frame and carrier frequency offset). Unfortunately, proposed preambles and pilots, or training sequences, in these standards are not necessarily designed to render optimal channel estimation performance; rather, they are usually biased toward providing accurate frame synchronization. The optimality is destroyed because these sequences do not account and exploit physical channel attributes, such as transmitter-receiver coupling and spatial correlation, which can dramatically improve estimation accuracy if they are properly utilized during the design process. The sequence design problem for multiple user systems is exacerbated by the existence of colored noise, which can make the overall design problem nonconvex. Thus, finding the global optimal solution becomes computational expensive. Previous works have bypassed this problem by assuming little or no spatial coupling exists between the transmitter and receiver, such that the MIMO channel can be sufficiently modeled using the Kronecker model. This assumption, however, is questionable in most actual scenarios. Further extension has been made by using nonlinear optimization method to refine the solution obtained using the above technique. This, of course, does not guarantee that the global optimal solution can be obtained. In this thesis, an overview of SU-MIMO transmission techniques will be given, followed by a general discussion on sequence design techniques and MIMO channel estimation for single- and multiuser systems. Next, an iterative training sequence design scheme called iterative superimposed training sequence design with multiple interferers, or ISIMI, is presented for estimating MIMO channels with colored noise. The proposed approach does not utilized nonlinear optimization as used in previous literature, nor make any assumption about the lack of interdependence between the transmitter and receiver. The approach is proven to converge to at least a local optimal solution and is shown consistently by Monte Carlo simulation to outperform previously proposed MSE based approach for 2×2 and 4×4 MIMO systems, respectively, in terms of MSE. Analytical results illustrating efficacy and deficiency of ISIMI is also included. In addition, computational complexity of ISIMI is illustrated in terms of number of arithmetic operations. The thesis is concluded with a discussion possible modification for ISIMI in order to further improve its performance. This is followed by a discussion on how it can be extended to robust training sequence design for MIMO system with multiple interferers, and sequence design with limited feedback.