雲端無線接入網系統下之分散式大量多重天線中使用者移動性對分群演算法的影響

Abstract

在本論文中，為了降低雲端無線接入網系統下之分散式大量多重天線的運算量 而提出了一個固定重組週期的動態天線分群演算法。在分散式大量多重天線中，天 線和使用者都分散在一個細胞中，使得天線能夠更加地靠近使用者，降低路徑損耗 的影響並增強收到的訊號強度。根據第三代合作夥伴計劃的文獻，我們在模擬中觀 察使用者容量變化特性受到不同的天線重組週期及移動性的影響。其中屏蔽效應必 須要設定為兩點之間是相關的，因為如果移動的距離很短，屏蔽效應的變化量並不 會很大。我們使用了屏蔽效應的相關係數以及相關性的屏蔽效應模組來做模擬。雖 然分群演算法能夠降低運算量以及部份的干擾，但是群組間的干擾仍然對系統存在 威脅。我們的演算法中也處理了次要的干擾來提升系統的效能，但也因此增加了運 算量，因此利用了重組週期來降低分群演 算法的運算量。結果顯示我們的演算法 在運算量以及容量間會有個權衡，以比例來說犧牲的容量會比降低的運算量少，因 此是較為值得的。在我們考慮的三種情況中，在滿足最低容量門檻的限制下，犧牲 了一些容量，和不考慮週期的情況下相比，分別降低了11.2%、9.31%以及6.98%的 運算量。結果也顯示了當使用者移動距離超過10%的細胞半徑時，天線群組必須重 新進行分群才不會導致容量低於門檻。

In this thesis, we present a dynamic clustering algorithm with re-cluster period to reduce the computational complexity in Distributed Large-Scale MIMO (DLS MIMO). In DLS-MIMO, antennas and users are distributed in each cell, antennas are more close to the users, which decreases the path loss e?ects and enhances the intensity of received signal. According to the standard of 3rd Generation Partnership Project (3GPP), we take user's mobility in account to observe the characteristic of user's capacity with increasing the re-cluster period. We use the autocorrelation of shadowing and the correlated shadowing model are considered since the shadowing will not change signi?cantly if the moving distance is short. The idea of antenna clustering can reduce complexity of beamforming and intra-cell interference, but the inter-cluster interference will be harmful to the system. The second largest interference are managed to improve the system performance, and the re-cluster period are used to reduce the complexity of the antenna clustering algorithm. Our results show that the proposed clustering algorithm have a tradeo? between computational complexity and capacity, and the sacri?ce is considered operation cost. In the considered cases, the complexity are reduced to 32%, 13%, and 9.6%, respectively with the same capacity compared to the current antenna clustering algorithm. The simulation results also show that the cluster should be changed when user's moving distance is larger than 12%, 7.4% and 4.6% cell's radius in each case.