|Abstract: ||無線感測網路（Wireless Sensor Network）是由數量眾多且微小的無線感測器（Sensor）所構成，這些感測器均符合體積小、低成本、低耗電及傳輸距離短的特性。近幾年來，無線感測網路已經廣泛應用在許多領域，許多研究也已針對該網路的相關議題進行探討，其中感測器位置是一項重要的因子，也就是說，必須得知感測器的位置才能設計符合實際需求的方法。此外，無線感測網路的應用中，偵測和追蹤是兩項重要的應用，特別是使用者在意的緊急事件。在本論文中，我們將先說明上述研究議題的重要性與面臨的挑戰，並提出有效率的解決方案完成感測器的定址，及偵測並追蹤事件。|
無線感測網路中的許多應用必須靠感測器的精確位置標示事件發生的區域。然而，一些實際上的應用並不需要感測器的精確位置，因此本論文將提出一個以方向為主的定址方法（稱為DLS），感測器可以DLS方法得到相對於匯集點（Sink）的相對方向。基本上，DLS是以空間區域性（Spatial Locality Property）基礎，並透過我們設計的錨點佈設策略提高錨點附近感測器的定址正確性，藉此感測器可以根據接收到的封包以決定自己的方向。再者，我們還設計虛擬雙方向座標系統（Virtual Dual Direction Coordinate System, VDDC system）以改善靠近兩相鄰方向邊界附近的感測器之定址正確率。
以無線網路進行偵測與追蹤的方法中，所有的感測器均配備相同的感測元件，但迄今尚未有研究針對利用具備不同感測元件的感測器進行事件追蹤與偵測之可行性進行探討。此外，在一些日常的應用中，事件的偵測與追蹤可能必須透過不同種類的感測器方能完成，因此，本論文將提出一個分散式的事件追蹤與偵測方法（CollECT），該方法主要是應用在無線異質型感測網路（Wireless Heterogeneous Sensor Network, WHSN），即網路中所有的感測器並不相同。CollECT的核心概念是相同種類的感測器建立三角形結構（Triangulation），此外，CollECT亦考慮不同種類感測器的協同合作概念以決定事件是否發生。再者，CollECT亦選擇一些感測器代表事件發生區域的邊界。
The wireless sensor network (WSN) is a network comprising a huge number of tiny wireless devices called sensors, which promise communications in short distances. A sensor is characterized by its small size, low cost, low power, and short radio range. Typically, sensors are randomly deployed in an area of interest in either an ad hoc or manual pre-scheduled manner depending on the requirements of applications. Unlike other wireless networks, the WSN is application-specific and energy-constrained in essence.
Currently, the WSN is widely used for a variety of applications such as environmental monitoring, battlefield rescue, home automation, etc. Much research has paid attention to numerous attractive topics, among which sensor localization is essential but extremely crucial for many applications in WSNs. Additionally, in WSNs, detection and tracking are important operations especially for the urgent target of interest. In the dissertation, we primarily address these critical and attractive issues and aim at the development of mechanisms for such themes.
Numerous applications for WSNs require physical location to recognize more precise positions at which designated targets or events occur. Majority of the existing approaches assume that each sensor is aware of its location information (e.g., 2D coordinate) via either an installed GPS receiver or other GPS-less localization schemes. However, the precise location information may be unavailable due to the constraints in terrain. Additionally, several applications can tolerate the diverse level of accuracy in such geographic information, and in terms of efficiency, acquiring the direction of a sensor takes less effort than obtaining its physical location. As a result, in this dissertation, we propose a direction-based localization scheme, called DLS for a sensor to determine its direction related to the sink. Basically, DLS is motivated by the spatial locality property based on our comprehensive observations of received packets at a sensor. In addition, DLS adopts the anchor deployment strategy to improve the estimated accuracy of the sensor close to the sink. Furthermore, with the aid of a novel virtual dual direction coordinate (VDDC) system, DLS is able to efficiently and precisely position sensors around the boundary of two adjacent directional regions.
Recent research on detection and tracking has paid much attention to the WSN in which all sensors are identical in sensing units, but the potentiality of the utilization of different types of sensors has not been explored. In addition, the existing approaches largely focus on the event, which can be identified by the homogeneous sensors. However, the constraints in sensing and communication capabilities make these solutions difficult to apply to the network in which the event has to be detected and tracked via various types of sensors. In the dissertation, we propose a fully distributed protocol, CollECT, to event detection and tracking in Wireless Heterogeneous Sensor Networks (WHSNs), regarded as a network comprising various types of sensors. Basically, CollECT focuses on triangulation construction, in which the sensors with the same type construct the triangulation composed of multiple logical triangles to represent the attribute region. Specially, sensor collaboration of the different types of sensors is taken into account to determine the existence of the event. Additionally, CollECT also aims to select the sensor, called border sensor to represent the event boundary. In principle, CollECT includes the vicinity triangulation, event determination, and border sensor selection procedures to construct the attribute region, to determine the existence of the event, and to identify the event boundary, respectively.
Overall, the issues involved in the dissertation are really essential and important in wireless sensor networks. We not only propose a direction-based localization scheme for a sensor to obtain its geographic information, but also devise an efficient detection and tracking protocol for WHSNs to identify the event of interest. Experimental results demonstrate that the average estimated correct rates in DLS approximately reach 94%, 86%, and 81% for the networks with 4, 8, and 16 directions, respectively. The results also validate the practicality of DLS, and show that DLS effectively achieves direction estimation with regardless of sink placement or network density. In terms of performance of CollECT, approximately 94% sensors within the event region can be correctly identified on average. Moreover, the border sensors selected by CollECT also reasonably stand for the event boundary.