具有自我學習機制之無線感測網路室內定位演算法

Abstract

對於許多應用服務來說，若能提供物品位置的定位資訊，將能提供很高的附加價值。因此近幾年定位方法受到廣泛的討論，國內外學者紛紛針對不同的室內及室外環境需求，提出對應的方法來改善定位的準確度。在傳統定位方法中，常需要透過一些額外的設備來輔助計算位置，例如：GPS[21][22][23]、Infrared[24]和Ultrasound[25]；但是節點上額外的輔助設備會增加電力消耗及系統建置成本。本論文利用節點本身所具有的RF晶片進行室內定位，以常用於室內定位的樣式比對演算法為基礎，改善其訓練時間長、計算量大與定位準確度不夠等缺點。我們實際使用本計畫所開發的無線感測器來進行驗證，實驗結果顯示，我們所提出的方法與傳統樣式比對方法相比，不但可有效地提高最多達41%的室內定位精確度，在同樣室內定位精確度下亦能有效減少訓練pattern的數量。

Tracing the location of a specific item has induced abundant applications with the development of wireless sensor networks. In past years, different location algorithms have been proposed for indoor or outdoor environments. Some of these use additional equipments, such as GPS[21][22][23], Infrared[24] and Ultrasound[25]. However, these equipments will increase the power consumption and system cost. In this paper, we propose a Self-learning Indoor Locating Algorithm for Wireless Sensor Networks and use only the embedded RF chip to improve the PM (Pattern Matching) algorithm. We made some improvements on saving training time, computation, and higher the positioning precision. Experiments show that our proposed algorithm not only lower training time but also higher positioning precision in 41%.