在WCDMA/TDD系統中一種以量測與預估為基礎之緩慢動態通道配置機制

Abstract

在一般傳統的中央控管式緩慢動態通道配置方案中，其目的便是希望能在上下連結不平衡的系統下找出最佳的時間格轉換點以達到最高的頻譜使用率與生產量，然卻分析於整個系統的細胞皆是一致的不平衡狀況下。在實際狀況中我們必須考慮各個細胞的不平衡狀況並不一致，且各個細胞的不平衡狀況會隨時間而變化。因此緩慢動態通道配置方案必須根基於現在的交通狀況才能得到最好的轉換點使得所得到的生產量為最高。而當交通不平衡狀況改變時，由於不佳的轉換點選擇所引起的瞬間效應，將會影響在所有連結的緩衝區長度中。再者，由於中央控管的緩慢通道配置方案會決定一個對整個系統而言折衷的轉換點，平均的緩衝區長度會反映在需要傳輸的總資料量或所需的額外資源上。因此在本篇論文中，我們提出一個所謂的”等價交通負擔”其中包含平均累積的交通速率以及所有的非即時連結中所剩餘的緩衝區長度，並且使用PRNN預測器預測平均累積的交通速率。然在我們想比較的對象中，他們所提出的緩慢動態通道配置只考慮到平均累積的交通速率且只是用過去一段時間的平均值當作預測結果。 從我們的模擬結果可以看出，藉由有考慮所有非即時連結中剩餘的緩衝區長度以及使用PRNN預測器來預估平均累積的交通速率，我們所提出的緩慢動態通道配置方案能夠改善緩衝區長度以及總生產量。尤其當交通狀況會劇烈變化時，我們的機制更能有效提升系統的容量。

In the conventional centralized SDCA algorithms, the goal is to maximize the spectrum utilization and thus to maximize the throughput under homogeneous, asymmetric traffic load condition among all the cells. However, in practical systems, the traffic load asymmetry conditions among all cells may vary along with time due to the change of the number of asymmetric connections and even the transient effects of each traffic source. Therefore, the SDCA should change the switching point (SP) based on the current traffic load asymmetry condition to maximize the throughput. As the traffic load asymmetry changes, the transient effects due to improper SP decision for this cell will impact on the queue lengths of all connections. Moreover, since the centralized SDCA scheme determines the overall SP with the compromise among all the cells, the mean queue length of connections in each cell reflects the amount of load to be transmitted or the extra resource required. In this thesis, the proposed centralized control SDCA algorithm decide the SP according to the “equivalent load” which is consisted of the mean aggregated traffic rate and the remaining queue length of all non-real time connections in each cell. And the PRNN predictor is used to predict the mean aggregated traffic rate. But in the paper that we want to compare with, their proposed centralized SDCA algorithm only consider the aggregated mean traffic rate which is a mean value in the past. According to the simulation results, by considering the remaining queue length of all non-real time connections in each cell and predict the mean aggregated traffic rate by PRNN, the queue length and the throughput can be improved in our proposed SDCA algorithm. Especially, when the traffic condition changes severely, our proposed SDCA algorithm has more capability to efficiently increase the system capacity.