| 摘要: | 要設計一套優質的無線行動網路系統，排程和資源分配的技術扮演著重要的角色。然而，由於行動用戶服務的多元化，不同類型的服務對網路品質有著不同要求，加上無線網路環境的多變性，使得排程和資源分配變得更具挑戰性。
本論文首先探討以排程技術實現行動裝置省電的議題。為了讓使用電池為電源的行動裝置有更長的續航力，它的省電能力是非常關鍵的特性。因此，我們提出了兩種以節省用戶行動裝置電能為中心的排程機制「全部重排和部分重排」演算法，它們係針對即時連線與非即時連線分別提供滿足其必須的品質服務(恰恰滿足的概念)，以達到有效節用電源並確保用戶服務品質的目的。模擬結果顯示：全部重排與傳統的方法比即時連線裝置的平均睡眠效率高8%，而比非即時連線裝置高11%，部分重排則對非即時連線裝置有比較好的省電表現，其睡眠效率比傳統方法高7.59%，而且兩種機制都滿足 QoS 的需求。。
其次，研究在可調性視訊群播環境下排程與資源分配的機制。我們的機制將服務群播群組(接收相同視訊節目服務者)分成數個遞送群播群組，且以「軟性分群」的概念來設計遞送群播群組，讓群組的分配更具彈性。並以此概念為基礎提出「跨層」及「品質感知」等兩種資源分配機制。「跨層」機制確實以頻域(實體層)與時域(MAC層)二維的角度去分配資源，由模擬結果顯示：「跨層」機制與我們之前所做研究「最佳用戶優先」方法相比，平均吞吐量為520 kbps優於445 kbps；「品質感知」機制以適應性的方式同時調整使用SVC的影像品質可調性、MCS的頻譜分配彈性以及UL-FEC的糾錯能力等特性，以達到資源有效分配的目的。模擬結果顯示：「品質感知」機制可獲得最高的傳輸效率為70% (傳輸效率定義為相對於系統使用最高MCS傳送速率之比值)，且同時滿足用戶真正要求的視訊品質。
最後，討論SC-FDMA LTE網路上行資源分配的演算法。我們提出的演算法將資源分配的程序分成匹配演算法和無線電資源分配演算法，以獲得較好的服務品質滿意度與系統產出量。在匹配部分，以通道狀況、頻寬分配的連續性及服務品質為考量，並採用Gale-Shapley演算法，找出RB(resource block)與UE(user equipment)的最佳匹配；在無線電資源分配方面則用前面找出的最佳匹配對為分配起點，考慮不同用戶的QoS要求，並在RB分配必須為連續的條件下，以啟發式演算法來分配UE的頻寬。模擬分析顯示：我們所提經過匹配的資源分配機制，其最小頻寬需求不滿足率平均而言小於傳統未經匹配的方法10%左右，確實提高了資源分配的效率。
Scheduling and resource allocation play important roles in designing effective wireless mobile networks. However, with diverse requirements of quality of service for mobile devices applications plus changing wireless enviroments, scheduling and resource allocation issues become more challenging. In this dissertation, we first investigate power saving scheduling and then resource allocation for multicasting scalable video coding (SVC) traffic in WiMAX network. Finally, resource allocation technique for uplink localized SC-FDMA LTE network is discussed.
To support battery-operated devices for longer use durations without recharging, power saving capability becomes a vital feature. Accordingly, we propose an energy-saving centric downlink scheduling (ESC-DS) scheme in WiMAX network. The proposed scheme treated real-time and non-real-time connections differently with respect to providing “just enough QoS” to support efficient power utilization and to satisfy the QoS requirements. Two rescheduling algorithms, whole–reschedule and partial-reschedule, are proposed and analyzed. Our simulation results indicated the whole-reschedule scheme exhibited 8 % higher sleep efficiency than the traditional scheme for RT MSs and higher than 11 % for NRT MSs while the partial-reschedule scheme obtained a minor 0.95% higher sleep efficiency than the traditional scheme for RT MSs and excellent 7.59% higher for NRT MSs.
For SVC multicasting scheduling and resource allocation, we treat a service multicast group, which receives the same video traffic simultaneously, can be divided into several delivery multicast groups in WiMAX network for better radio resource utilization. A delivery multicast group is designed in a “soft-group” manner, which can be flexibly arranged. Two radio resource allocation schemes, cross-layer resource allocation (CLRA) and quality-aware resource allocation (QARA), are presented. CLRA considers not only frequency domain (PHY layer) but also time domain (MAC layer) to obtain better resource allocation. Simulation results show that the average throughput per user of CLRA is 520 kbps better than 445 kbps of our previous presented BUF (best user first) scheme. QARA jointly uses SVC scalability, MCS flexibility, and UL-FEC capability to adaptively allocate radio resources for a video stream. Simulation results show that the proposed QARA scheme (with FEC) achieves the highest transmission efficiency 70% (transmission efficiency is defined as the ratio of average received data rate per slot using the selected MCS level to that of using the highest MCS level (i.e., MCS 6) for transmitting video data).
For uplink resource allocation in localized SC-FDMA LTE network, we propose a systematic approach, dividing resource block (RB) allocation process into matching algorithm and assignment algorithm, to achieve better QoS satisfacation and system throughput. The Gale-Shapley algorithm is applied to find the optimal matching between RBs and user equipments (UE) by considering channel conditions and the desired quality of service (QoS), and the resource assignment algorithm heuristically allocates bandwidth to UE by referring the matched RB under the constraint of carrier continuity. From the simulation analysis, it indicates that has the unsatisfied ratio (the percentage that the allocated bandwidth is less than the GBR minimum bit rate) of the proposed matching scheme is 10% less than that of the no-matching scheme on average. |
