本文考慮在相關性通道情況下放大轉發 (amplify-and-forward, AF) 多重輸入多重輸出(multiple-input multiple-output, MIMO)多中繼(multiple relays)雙向(two-way)合作式通訊(cooperative communication)系統之聯合預編碼器設計,分別討論最小化均方誤差(minimum mean-square error, MSE)與最大化總和傳輸率(maximization sum-rate)為準則之最佳化,並加入中繼節點之中繼選擇(relay selection)與天線選擇(antenna selection)機制於系統中。將選擇結果構成之天線組合與中繼端預編碼矩陣(precoder)設計制定成一最佳化問題。選擇方式採用最大奇異值乘積,首先使用奇異值分解(singular value decomposition, SVD)架構制定預編碼之設計並引用均方誤差為準則處理中繼預編碼矩陣之最佳化。接著使用共軛梯度法(conjugate gradient method)架構制定預編碼之設計並引用總和傳輸率為準則之最佳化。由於兩個最佳化問題在中繼與來源端的功率限制下並非線性單凹或單凸,故本論文採用疊代計算的方式以求得最佳化中繼與來源端預編碼設計。最後在電腦模擬結果中驗證,考慮中繼與天線選擇並採用基於SVD架構聯合預編碼器設計之合作式通訊系統,比起沒有考慮的設計可大幅減少中繼天線的使用數仍具備優異的表現。以及最後考慮了通道估計誤差之強健型設計比非強健型設計能夠有效地抵抗在通道資訊缺陷的情況下殘餘自干擾項對總和傳輸率的影響。
In the thesis, a unified framework of relays/antennas subset selection and transceiver design is proposed based on the cumulative mean-squared error (CMSE) minimization and achievable sum-rate (ASR) maximization for a dual-hop two-way amplify-and-forward (AF) multiple-input multiple-output (MIMO) relaying system in correlated fading channels. By making use of the singular value decomposition (SVD) technique, the joint design of two-source transceivers and multiple-relay precoders is developed based on the minimum CMSE criterion. This MMSE based optimization leads to a highly nonlinear objective function of sources and relaying precoders under power constraints so that an iterative search algorithm is applied to explore the optimal power allocation. In addition, other joint designs of source and relay precoding matrices are developed based on the ASR maximization under incomplete self-interferences cancellation. Simulation results show that the SVD-based joint design of relay and source precoders in conjunction with the SVD-based relay-and-antenna selection (RAS) scheme is capable of achieving superior performance in terms of bit-error-rate (BER) and MSE with an alleviated demand on relays and associated antenna pairs. Also, results demonstrate that by taking correlated channel uncertainties and residual self-interferences into account, the robust precoding design is able to overwhelm the impacts on sum-rate effectively under correlated fading channels.