應用於WiMAX系統中配置二層需求之演算法

Abstract

由於移動式無線網路高服務品質的要求快速成長，寬頻無線存取成為一個有趣且受歡迎的網路架構。正交分頻多工存取(OFDMA) 是IEEE 802.16e的物理層中最常被討論的傳輸技術。在OFDMA的架構下，媒介擷取控制(MAC)訊框被展開成為兩個維度來看，一個是時間的維度，單位為一個OFDMA符號區間；另一個是頻率的維度，單位為一個邏輯上的次通道。OFDMA系統架構下，資源分配是很重要的一部分。一般來說 資源分配的模組包含兩部分： 排程和資料對映。排程部分，負責產生需求；資料對映部分，負責將需求放置在二維的MAC訊框內。由於所有的需求必需要以一個矩形的形狀被放置在訊框內，要找到一個下鏈路資料對映的最佳解為NP完全(NP-complete)問題。因此許多研究提出了多樣的啟發式演算法來達到低複雜且高效率的目的。 在本篇論文裡，我們提出了雙級需求資料對應(2L_DMA)演算法，主要目標有二：(1)提供雙級需求：必要(MUST)部分為高優先權的資料；期望(WISH)部分為低優先權的資料。(2) 退回最少的必要部分給排程器，同時維持資料對映機制的效率。我們將此想法實踐在當前擁有高效能的eOCSA演算法上，我們將會驗證所提出來的2L_DMA演算法能夠提高資料對映的效能。模擬結果顯示我們所提出的演算法可藉由最大化系統吞吐量來達到比eOCSA更好的效能。

Broadband wireless access has become a very interesting and popular networking infrastructure, because of the rapidly growing demands of high quality services over mobile wireless systems. Orthogonal Frequency Division Multiple Access (OFDMA) the physical transmission mode adopted by IEEE 802.16e WiMAX is one of the most intensively researched technologies. In OFDMA, the Medium Access Control (MAC) frame is extended over two dimensions; time in units of OFDMA symbol, and frequency in units of logical sub-channel. A very important component of OFDMA systems is resource allocation. In general, the resource allocation module consists of a scheduler and a data mapper. The scheduler generates requests while the data mapper maps those requests into the two-dimensional MAC frame. A constraint of downlink transmission is that every request has to be mapped as a rectangle. It was shown that due to this constraint finding an optimum solution for downlink data mapping is an NP-complete problem. Consequently, various heuristic algorithms were proposed to achieve high efficiency with acceptable complexity. In this thesis, we propose a data mapping algorithm for two-level requests, which we called 2L-DMA algorithm and it consists of two main targets: (1) apply a Two-Level request: a MUST part, for high priority data (urgent data); and a WISH part, for low priority data (non-urgent data); and (2) return as less possible MUST part to the scheduler, while keeping the mapping scheme efficient. We have implemented our idea on an existing algorithm called eOCSA, a high-performance packing algorithm recently presented. The goal of our proposed data mapping algorithm is to achieve high efficiency. Furthermore, the performance of our proposed algorithm is compared with that of eOCSA, a low-complexity algorithm with satisfactory performance. Experimental results show that our proposed algorithm yields much better performance than eOCSA.