無線感測網路上省電之資料管理技術之研究

Abstract

隨著無線感測網路技術的進步，未來生活環境中，將佈滿感測器，用以監測並蒐集和生 活環境相關的資料。此外，感測器間亦會利用無線網路互相連結形成一無線感測網路。 然而無線感測網路具有低電量儲存、少量記憶單元、低無線網路通訊頻寬及較弱的硬體 運算能力限制。因無線感測器之限制，使得在無線感測網路上的資料處理、查詢與傳統 分散式系統有所不同，且這些限制使得如何在有限的資源內，提供低耗電，有效率之資 料處理與查詢成為日漸重要的課題。本計畫為一為期三年之計畫，探討在無線感測網路 下，如何有效率處理使用者的查詢。本計畫目前正在執行第一年計畫，我們初步設計出 一資料傳播機制，透過分配特定感測器特定的角色，以節省感測器之能源消耗並提高系 統的整體存活時間。 本計畫第二年的工作重點在於設計省電的位置管理(location management)機制，目的在於 設計省電的物計位置估計查詢處理方法。具體的工作內容包括(1)研發階層式之位置伺服 器(Hierarchical Location Servers)架構；(2)設計處理估計查詢之訊息流程：(3)研發位置伺 服器的移動與複製演算法及(4)設計更新訊息的多重播送(multicast)方法。在第三年裡， 我們將著重於感測器讀數的查詢。我們將考慮在使用者能接著某種程度誤差的情形下， 如何有效率地收集或查詢感測器的讀數。具體的工作內容包括(1)設計合適的關聯性評估 公式；(2)設計感測器群集演算法；(3)設計多代表點群集演算法及(4)設計群集維持 (maintenance)機制。近年來，無論工業界或是學術界對於無線感測網路均相當的重視並 且投入大量的人力及研究，無線感測網路於農業應用極為適合，我們相信此計畫之執 行，將可研發出適用於無線感測網路系統中的資料處理前瞻性技術。

The growing advance in wireless communications and electronics makes the development of low-cost and low-power sensors possible. These sensors are usually small in size and are able to communicate with other sensors in short distances wirelessly. A sensor network consists of a number of sensors which cooperates with one another to accomplish some tasks. Sensors can be deployed either in a random or in a predetermined manner. Since being self-organized, sensors are able to form a sensor network automatically. Due to the characteristics of wireless communication and configuration-free deployment, sensor networks are suitable for various application areas including inventory management, product quality monitoring and disaster area monitoring. Right now, we have developed a role-based data dissemination scheme to reduce energy consumption on large-scale sensor networks. In this project, we aim at designing energy-efficient data management schemes in wireless sensor networks. Our primary goals include (1) developing a hierarchical location management architecture, (2) devising approximate query processing schemes, and (3) designing location server moving and replicating algorithms and (4) devising location update mechanism. Next, by exploiting the spatial correlation among sensors, we will (1) develop a similarity measure metric suitable for wireless sensor networks, (2) design a hybrid sensor clustering algorithms, (3) design an algorithm to select multiple representatives for each cluster, and (4) devise a cluster maintenance mechanism. In view of the increasing attention on wireless sensor networks, we strongly believe this project is very timely and will deliver results of both theoretical and practical importance.