動態調整IMU並結合Wi-Fi資訊強化室內定位效果

Abstract

利用短距離無線技術(例如無線區域網路)來做室內定位系統已經被廣泛的研究，額外的資訊例如新興、便宜的微機電系統或是圖資也被用來增進室內定位的精準度。雖然很多將無線區域網路定位系統與微機電系統結合的方法很多，但他們假設使用者知道起始點跟方向，這樣的資訊並不是在任何情況下都可以容易的取得。因此，我們提出了一個室內定位系統來解決這個問題，同時也增加定位的精準度。在這篇論文中，我們將展示我們如何不一樣的利用粒子濾波將無線區域網路定位系統與微機電系統整合來提供行人追蹤的服務。我們將三種整合後會產生的問題進行妥善的處理，起始位置的問題、累積誤差的問題以及無線區域網路定位系統不穩定的問題。我們在南港展覽館四樓進行系統的評估，其結果與只使用指紋式無線區域網路定位系統以及粒子濾波方法，其效果都有顯著的提升。此外與直接將無線區域網路定位系統與微機電系統結合的方法相比也有不錯的表現，我們分別在三種不同的測試路徑上分別有206%、203%以及159%定位精準度的提升。

Indoor positioning systems based on short-range wireless technologies such as wireless local area network (WLAN) have been widely investigated. Additional information produced by the emerging low-cost micro-electro-mechanical system (MEMS) sensors or floor plans is usually used to improve the accuracy of indoor positioning system. Although several combining WLAN with MEMS based inertial measurement units (IMU) researches have been researched, they assume the user’s initial location and heading is given. But this information might not be available in some circumstances. Therefore, we proposed an indoor positioning system to solve such problem as well as improving the positioning accuracy. In this thesis, we show how to differently fuse IMU, map information and Wi-Fi by the utilization of particle filter to provide indoor pedestrian tracking service. Three drawbacks from Wi-Fi fingerprinting and IMU are handled which are the initial location problem, drift error problem and the unsteadiness of Wi-Fi positioning. The accuracy of the fused system was evaluated in Taipei NanGang Exhibition Center against ground truth. Our results show that accuracy is much higher than Wi-Fi fingerprinting alone and particle filter. Furthermore, we achieve better performance than fusing IMU and Wi-Fi fingerprinting directly. Up to 206%, 203% and 159 % enhancement are achieved in three distinct test paths respectively when compared with the latter approach.