為了符合下世代行動通訊對於頻譜效率的需求,我們在本篇論文中,實踐非正交多工接取技術於多天線多使用者系統。除了多天線多使用者造成的波束間干擾,非正交多工接取所造成的波束內干擾也需要考慮以優化系統效能。不同於現今文獻假設傳送端完美的知道整個通道狀況,我們針對實際有限回授系統進行探討,因為傳送端只能得知非完美的通道資訊,所以波束間的干擾無法透過預編碼完全消除。因此,如何估計此干擾項來改善排程效能是我們著重的重點。首先,我們根據資料分布,建構一子碼本以表示因有限回授所造成的通道誤差資訊,讓基地台不用太多額外的回傳資訊即可利用此子碼本來估計干擾效應。接著,在得到較準確的SINR估計值後,我們提出了動態排程來決定是否在多天線多使用者系統中使用非正交多工接取技術,並採用比例公平排程來決定使用者組合、預編碼、波束內使用者能量分配、調變與編碼,尤其為了強化排程的信賴度,我們提出了一限制條件於使用者配對的步驟中。最後,模擬的結果顯示,我們所提出的估計方法在多天線多使用者系統中有76.83%的增益,且提出的限制條件可降低錯誤率29.59%,在考慮了非正交多工接取技術後,錯誤率更降低了39.54%。因此,解決估計及排程問題後,將非正交多工接取技術實踐於有限回授下多天線多使用者系統,可進一步提升下世代行動通訊之效能。
In this thesis, to accommodate growing demand for spectrum efficiency, we investigate non-orthogonal multiple access in multi-user multiple-input multiple-output systems (NOMA-MUMIMO). Besides inter-beam interference due to MU-MIMO, intra-beam interference should be taken into account because user equipments (UEs) are served in a single beam for NOMA. Unlike related work that assumes perfect channel state information at the transmitter (CSIT), we focus on NOMA-MUMIMO with limited feedback for practical systems, so the inter-beam interference cannot be totally eliminated because of the limited feedback. Therefore, how to estimate the interference to improve scheduling should be elaborated. First, without too much additional feedback, we propose an estimation method with additional feedback to capture error impact caused by the imperfect CSIT by constructing a sub-codebook to represent the error information based on statistical distribution. After obtaining more accurate estimated SINR, a dynamic NOMA-MUMIMO/MU-MIMO proportional fair-based scheduler to decide scheduled UE set, precoder, power, and modulation and coding scheme will be proposed. Specifically, a constraint is proposed for UE grouping to schedule reliably. Evaluation results show the gain of throughput is improved by 76.83% for proposed estimation method compared to a baseline approach in MU-MIMO. Moreover, BLER can be reduced by 29.59% when the constraint is applied. When considering NOMA-MUMIMO, BLER can be reduced by 39.54%. Therefore, after solving the estimation and scheduling problem, higher performance can be achieved with NOMA-MUMIMO for next generation communication systems with limited feedback.