在分頻多工巨量天線系統中免除通道資訊回授之多使用者下行路徑預編碼

Abstract

大規模多輸入多輸出系統在5G無線通訊中，因為具有突出的頻寬效率及能量效率，被視為非常有潛力的技術。 然而，在分頻多工下要獲取通道資訊需要極為大量的下行訓練及上行回傳，因此分頻多工大規模多輸入多輸出系統一直被視為不實際的方法。 本篇論文提出了一個適用於分頻多工大規模多輸入多輸出系統且不需使用通道回傳的多使用者傳輸機制，可以達到比以往低上許多的延遲時間。 該機制的下行預編碼設計使用了分頻多工的通道互易性及從上行獲得的部分通道資訊。 此外，該機制利用空時分組碼以避免通道回授，並採用了空間上的濾波設計達到抑制干擾，同時加強對部分通道資訊預估誤差的穩定性。 本文呈現了兩種功率分配的方案，分別為了最小化所有使用者之最大錯誤率及最大化使用者傳輸效率和。 模擬結果顯示了所提出的方法在使用者傳輸效率和及錯誤率效能方面的優勢。

Massive MIMO is a promising technique for the next-generation wireless communication systems due to its tremendous performances in spectral and energy efficiency. Channel state information (CSI) acquisition for massive MIMO operated under frequency-division duplex (FDD) is widely regarded as a challenging task due to enormous overhead of downlink training and uplink feedback. In this paper, a multiuser downlink precoding mechanism for FDD massive MIMO that does not require any explicit feedback of downlink CSI is proposed, which contributes to a much lower latency compared to the previous mechanisms. The proposed mechanism exploits the reciprocity of FDD systems, and uses only partial CSI obtained from uplink transmissions as the basis of downlink precoding design. Besides, the proposed mechanism employs space-time block code (STBC) techniques to avoid CSI feedback, and adopts a spatial filter design not only to suppress co-channel interference, but also to increase robustness against estimation error of partial CSI. Two power allocation schemes aiming at minimizing maximum symbol error rate (SER) of all UEs and maximizing sum rate are also presented. Simulation results demonstrate that the proposed mechanism achieves satisfactory SER and sum rate performances.