IEEE 802.15.4協定下多點跳躍網路之建模與效能分析

Abstract

IEEE 802.15.4是針對低速率無線個人區域網路(low-rate wireless personal area networks, LR-WPANs)所發展之通訊協定，其中詳細定義了協定堆疊中之實體層(physical layer, PHY)與媒介存取控制層(medium access control layer, MAC)，為Zigbee之底層規範，適用於各種以感測器網路系統。 在802.15.4標準中，定義了兩種型態的網路拓樸，一種為星狀拓樸(Star topology)，其通訊鏈結建立在各裝置與一個中央控制器間(稱為coordinator)，裝置之間無法進行聯繫，僅能透過coordinator交換資料，適合較單純與簡易之應用；另一種為對等拓樸(peer-to-peer topology)，在通訊範圍之內，裝置可與任意其他裝置聯繫，屬於一種隨意網路(ad-hoc networks)，具有自我組織(self-organizing)與自我修復(self-healing)之能力，只要有任意裝置在通訊範圍內，便可以不斷延伸擴展，且支援以多點跳躍(multi-hop)之方式將資料傳遞至遠端。叢集樹狀網路(cluster-tree networks)便是以對等拓樸為基礎衍生出來的特殊網路，適合較為複雜之網路應用，如工業監控、WSN、貨物與庫存追蹤、環境氣候量測等。 一般對於IEEE 802.15.4之效能研究，多集中在星狀拓樸，且各種指標之分析已非常詳細，然而在對等拓樸的分析與探討卻相當缺乏。本論文則針對對等拓樸中之叢集樹狀網路，提出一種結合馬可夫鏈(Markov chain)模型與佇列理論(queueing theory)之架構，分析其多點跳躍環境下之效能變化。由於叢集樹狀網路的結構複雜度較星狀網路提升很多，因此以往用於分析星狀網路之模型不再適用。本論文將以往之模型加以修正，以一簡化型馬可夫鏈描述裝置節點與傳輸頻道之動態行為，進而估測出各種狀態之切換機率。而以往經常使用之無限容量(infinite capacity)佇列模型，也修正為有限容量(finite capacity)，以更加近似實際情況。 在效能指標方面，除了分析網路流通量( throughput)之外，也利用佇列模型估測網路之阻斷機率(blocking probability)。封包傳遞(packet delivery)方面則利用機率生成函數(probability generating function, PGF)描述佇列模型中之等候與服務時間，進而估計多點跳躍環境下難以預測之延遲時間。

IEEE 802.15.4 is a protocol developed for low-rate wireless personal area networks, LR-WPANs, which defined specifications to physical layer and medium access control (MAC) layer. It is the basis of Zigbee, and is suitable for sensor network systems. It defines two types of topology structure in IEEE 802.15.4 standard; one is star topology, which contains one central controller (also named as coordinator) and many device nodes. The communication links exist only between individual devices and coordinator, and therefore it could be used in some simple applications. The other one is peer-to-peer topology, belongs to ad hoc networks with self-organizing and self-healing characteristics. It supports the multi-hop function to relay data to remote devices, and there are no any communication constraints between individual devices in RF range. Therefore the network could be extended recursively. Cluster-tree network is derived by peer-to-peer topology, and is suitable for more complicated applications, such as industrial monitoring, WSN, environmental measurement. The research for IEEE 802.15.4 performance was primary focused on star topology, however, the peer-to-peer topology wasn’t be study in detail. This thesis combines Markov chain model and queueing theory to analyze the performance degradation under multi-hop networks. The complexity for cluster-tree networks is much more complicated than star networks, and therefore the simplified Markov chain model was proposed to estimate the transition probability between states. The infinite capacity queueing model was also modified as finite capacity to approximate real operation condition. Some performance indices are analyzed in this research, including throughput, blocking probability, and packet delivery delay. The throughput was estimated by Markov chain channel model; the blocking probability was predicted by queueing model in device and relay; and the packet delay was derived by probability generating function (PGF) to describe the waiting and service time in queueing models.