高速下載封包擷取系統中的適應性連結調整演算法

Adaptive Link Adaptation Algorithm in High Speed Downlink Packet Access (HSDPA) System

曾明祥(Ming-xiang Zeng)

指導教授 ： 林宗男

國立臺灣大學/電機資訊學院/電信工程學研究所/碩士(2006年)

https://doi.org/10.6342/NTU.2006.02161

摘要

在高速下載封包擷取系統中提出了多項相對於全球行動通訊系統的技術改善來提升服務效能 ， 而它之所以能提升封包最大傳輸率達到10Mbps的主要原因在於高速下載封包擷取系統使用的連結調整是適應性調變編碼 ， 適應性調變編碼的功能在於節點B隨著通道情況的變化去選擇不同的調變與編碼組合來傳送封包 ， 如果傳送的封包使用了正確的調變與編碼組合 ， 則不僅生產量會提升、封包錯誤率會減少 ， 還有延遲問題亦會降低 ， 所以調變與編碼組合界線值的選取是一項影響其效能的重要因素 。 最佳化調變與編碼組合界線值的決定主要取決於目前手機用戶的無線通道環境 ， 但因為無線通道環境是隨著時間改變的 ， 所以我們並沒有辦法事先得知在哪種通道環境下的最佳化調變與編碼組合界線值 。 倘若今天手機用戶處於較差的通道環境中 ， 但卻使用了在較好通道環境中的最佳化調變與編碼組合界線值 ， 則它的生產量效能就會降低 ， 反之亦然 。 在這篇論文裡 ， 我們提出了一個利用背景查覺學習基準的方法來解決這個問題 ， 背景查覺的主要概念在於線上的學習通道環境的情況後做適應性的調整調變與編碼組合界線值來達到最佳化 ， 而我們做適應性連結調整的目標在於達到整體生產量的最大值 。 在我們的提出的作法中 ， 首先節點B會先收集從手機用戶回傳的通道環境訊息及封包接收情況的資訊後 ， 再利用類神經網路的概念去學習訊號雜訊比、調變與編碼組合界線值之間的複雜非線性封閉形式 ， 在獲得這個非線性的封閉型式後 ， 我們利用它來產生在某段訊號雜訊比區間內的累加生產量 ， 再來我們同樣利用類神經網路的概念去學習累加生產量和調變與編碼組合界線值之間的複雜非線性封閉型式 ， 而在獲得這個非線性封閉型式後 ， 我們可以根據累加生產量跟未調整前的調變與編碼組合界線值之間的梯度關係來做適應性的調整調變與編碼組合界線值達到最佳化的值 。 這篇論文的貢獻在於我們提出了一個在隨著時間改變的通道環境中做適應性的調整調變與編碼組合界線值達到最佳化值的方法 ， 進而達到生產量的最大值 。 模擬的結果證實了我們所提出的方法可以處理隨著時間改變的通道環境所造成的問題 ， 並且提昇了生產量 。

關鍵字

高速下載封包擷取 ； 適應性調變編碼 ； 連結調整 ； 類神經網路

並列摘要

In HSDPA system, a set of enhanced technique have been proposed to improve the service performance in proportion to the UMTS (Universal
Mobile Telecommunication System), and the primary reason that HSDPA could improve the peak rate of packet transmission up to 10Mbps is that the HSDPA utilize the AMC (Adaptive Modulation and Coding) as its link adaptation. The function of AMC is that Node B selects the variable MCSs (Modulation and Coding Schemes) to transmit the packet according to the changed channel conditions, if the packet transmits with the correct MCS then it will come out with a consequence that the throughput will be upgraded and the packet error rate and the delay requirement will be degraded, so the threshold of the MCSs is one of the most important factors that would affect the performance in HSDPA system. The optimal threshold depends on the user’s wireless channel environments. However, the channel condition is time-varying and it is difficult to decide optimal threshold in advance. If the user is in the worse channel condition but utilizes the thresholds suited to the better channel condition, then its throughput will be degraded, vice versa. In this thesis, we tackle this problem by a context-aware learning-based optimization approach. A context-aware framework is designed to learn the channel conditions on-line and optimize the threshold adaptively with the goal to maximize the overall throughput. In our approach, first, the node B collects the knowledge of the channel conditions and the information of the packet successful received or not those are feedback from users and uses these obtained knowledge and the capability of neural networks to learn the complex nonlinear function among the throughput, the threshold and the SNR (signal to noise rations), after acquiring the complex nonlinear function, we take advantage of it to get the aggregated throughput in certain SNR region. Then we use the capability of the neural network again to learn the complex nonlinear function between the aggregated throughput and the threshold and adjust the threshold adaptively to reach an optimal value according to the gradient which is derived from the aggregated throughput and the threshold. The contribution of this thesis is that we propose an approach that is applicable to the real situations to adaptively regulate the threshold to reach the optimal value and maximize the overall throughputs. The simulation results show that our approach can deal with the time-varying wireless channel conditions and improve the throughputs.

並列關鍵字

HSDPA ； AMC ； Link Adaptation ； Neural Network

參考文獻

[ 1 ] Theodore S. Rappaport, “Wireless Communications”.

Architectures”, John Wiley & Sons. 2002.

Connected Mode.

[ 10 ] 3GPP 25.877 High Speed Downlink Packet Access (HSDPA) –

Iub/Iur Protocol Aspects.