應用於無線感測器網路之低複雜度及分散式變化檢測

Abstract

無線感測器網路（Wireless Sensor Network）是藉由多個具有感測環境變化、處理感測 資訊、無線通訊傳輸功能的微小感測器將收集到的資料處理過後傳輸到基地台（Base Station）或資料融合中心（Data Fusion Center）。在無線感測器網路的應用中，感測器 往往需要偵測環境的改變並且將感測資訊傳送給資料融合中心，故如何有效益的偵測 改變的發生成為我們所關注的重點。於計畫第一年中，吾人將考慮在感測器警報機率 分佈下改變偵測的問題，在無參數改變偵測（Nonparametric Change Detection）的架構 下基於Rao test 發展一套演算法，利用將感測器分群（Grouping）估計出各群的期望值 來降低系統運算的複雜度，此外也對感測器與資訊融合中心之間發生干擾時來做進一 步的分析。因為能量的限制，整個網路的環境是非常不可靠的。因此許多的研究都著 重在如何有效的使用有限的資源，或是將傳統通訊網路的通訊協定概念加以修正、調 整到適用於大型感測器網路的模式，亦即分散式的處理。吾人發現就分散式偵測及估 計的感測器網路而言，最重要的就是設計Dynamical System 的函數形式。透過不同的 函數設計來決定系統傳播資訊的模式以及達到同步的方法，可以明顯的影響系統整體 的性能。於第二年中，吾人將針對無線感測器網路中偵測和估計（Detection and Estimation）的問題，利用非線性動態系統的觀念設計出分散式的解決方案。

A wireless sensor network is composed of many small sensors that are capable of sensing environment change, processing sensed information, performing wireless communication and transmitting collected processed data to base station or data fusion center. In the applications of wireless sensor network, the sensor node has to detect the changes in the environment and then send the information to the fusion center. Thus how to effectively detect the change in the environment becomes the focus of our concerns. In the first year of the project, we will consider the change detection issue under sensor alarm probability distribution, develop a Rao test based algorithm under the framework of nonparametric change detection, utilize grouping to estimate the group's expectations to reduce the computational complexity. In addition, the impact of interference occurred among the sensors and information fusion center will be further analyzed. Because of energy constraints, the entire network environment is not very reliable. Hence, many studies have focused on how to effectively use limited resources, or modify the conventional network protocol concept to apply to the large-scale model of sensor network, for example, decentralized processing. It is found that the most important thing for the decentralized detection and estimation of the sensor network is to design the dynamical system function. Through a variety of functions designed to determine the mode of information propagation as well as the method of synchronization could significantly impact the overall system performance. In the second year, we will focus on the detection and estimation problem in wireless sensor network and propose a decentralized solution by using the concept of nonlinear dynamic systems.