無線感測網路錨節點之反向定位法

Abstract

無線感測網路定位已經成為各種新穎應用（如土石流偵測、精緻農業、健康照護等）高度需求的功能之一。 透過錨節點（anchor node）的參考位置值及與感測節點（sensor node）之間的訊號強度，即可定位無線感測網路中的感測節點。 在這類設計中，錨節點的位置愈精確，感測節點的定位結果就會愈正確。 為了要提供精確的錨節點位置資訊，許多研究利用 GPS 裝置來定位戶外的無線感測網路中的感測節點。 由於 GPS 裝置無法在室內環境中得到精確的位置資訊，因此，有部分定位方法會利用手動設定錨節點位置的方式來得到較精確的感測節點位置。然而，手動設定錨節點的方法卻無法應用於大型無線感測網路中，而且，手動設定的人工錯誤也可能會逐漸累積，並影響無線感測網路的定位結果。 在這篇論文中，我們研究多種反向定位802.11無線網路存取點的方法。有些方法仍舊採用GPS模組來定位錨節點（存取點）的位置；有一些則嚴格地限制訓練階段的條件（固定的啟始點與行進方向）。 為了改善這些問題與限制，我們提出一種新的無線感測網路反向定位錨節點的方法。我們的方法不需要GPS模組，也不需要嚴格的訓練限制。 考慮到無線通訊晶片的個體差異，在我們開始進入訓練階段之前，我們先對無線感測節點進行校正，以取得評估節點距離所需的 RSSI 係數，以利反向定位錨節點。 我們利用 MSP430FS5438/CC2500EMK 實驗板進行多種實驗，以便於精確性與成本之間取得平衡。 我們同時也說明如何應用我們的方法來估測錨節點位置，並將估測的位置應用於現有的定位方法中，以利定位感測節點。 實驗結果顯示：1) 如果先進行校正或移除訓練限制，我們就可以得到較高的精確度; 2) 比起真正的錨節點位置，我們所估測的錨節點位置會引入大約 9.25\% 的定位誤差。

Wireless sensor network (WSN) localization is demanded in many modern applications, like landslide detection, precision agriculture, health care, etc. By leveraging the anchor nodes, sensor nodes in a WSN can be localized. The more precise the position of an anchor node is, the more accurate the localization of a sensor node can be. To provide the accurate positions of an anchor node, many studies have taken advantage of a GPS device placed in the outdoor environment. Since the GPS device cannot work properly in an indoor environment, some existing localization methods also adopt the configuration of the anchor node in a manual fashion. However, manually configuring anchor nodes is not suitable for large-scale WSNs and artificial errors may be propagated and thus affect the result of the WSN localization. In this paper, we study several reverse localization methods regarding locating wireless access points (AP) in an 802.11 wireless network. Some of these methods still rely on the GPS modules for AP positioning, and some others strictly restrict the training conditions. As an improvement, we propose a novel approach to locate anchor nodes in a WSN without any GPS modules or strict traninig restrictions. Considering the individual differences of wireless chips, we apply the calibration before reversely locating the anchor nodes to obtain the RSSI (received signal strength indicator) coefficients for estimating the distance between two nodes. We conduct a series of experiments with MSP430FS5438/CC2500EMK devices to study the tradeoff between the accuracy and running cost. We also demonstrate how the proposed approach estimates the anchor nodes and applies the estimation to an existing localization method. The result shows that 1) better accuracy can be obtained if we apply the calibration or remove the training restrictions; 2) there is a gap of 9.25\% errors in average between the real and estimated anchor positions.