使用無線區域網路頻道狀態資訊的室內定位系統實作與評估

Abstract

室內定位系統的需求在這幾年當中的需求逐漸提高，在室外的環境下，我們可以使用Global Positioning System(GPS)來知道目前所在的位置，但在室內的環境中，GPS訊號會受到建築物的影響而導至精準度不佳，所以室內的定位大多採用其他的方式來定位。本論文是以測量WiFi的訊號強度來定位，由於目前室內的WiFi AP普及，我們不需要額外花費其他的成本去安裝其他的設備，有鑒於此，我們採用分析WiFi channel的方式來定位。 定位主要分成兩個步驟分別是1.感測 2.使用定位演算法定位。在感測的部分，過去以測量WiFi的RSSI為主，但是單一的RSSI無法給予足夠的channel資訊，所以在這裡改以測量802.11n 實體層的頻道狀態資訊來提升定位的精準度。在定位演算法的部分，目前定位演算法主要以'Fingerprinting'或是使用'Propagation Model'為主，前者的精準度較高，但缺點在於測量所花費的成本較高，後者的精準度相對前者較低，但優點是測量成本較低，因此在考量到成本情況下，希望能夠增加用'Propagation Model'精準度，所以我們根據'Propagation Model'設計了一種定位演算法，在實測的結果當中發現這種演算法可以提升定位的精準度11.6百分比。

Demand for indoor positioning systems has increased in recent years. In outdoor environments, the Global Positioning System (GPS) can be used to obtain information on a person’s current position. However, GPS does not work very well in indoor environment because the GPS signals are affected by buildings. Therefore, we use other methods to implement indoor positioning. In this thesis, we measure the Wi-Fi signal strength to determine the indoor position. Most buildings contain numerous Wi-Fi access points, meaning that no additional deployment cost is required. Therefore, we implemented the indoor positioning system in this study by analyzing Wi-Fi channels. Indoor positioning comprises two steps: sensing, and using a local positioning algorithm to estimate the user’s location. In previous studies, Wi-Fi received signal strength indicator (RSSI) has typically been used in the sensing step. However, an RSSI measurement does not provide sufficient channel information. Therefore, in this study, we measured the channel state information of the 802.11n physical layer to improve the accuracy. Fingerprinting and propagation models were used in the local positioning step. Fingerprinting provides superior accuracy, but the deployment cost is high. The path loss model method provides inferior accuracy, but costs less to deploy compared with fingerprinting. We hoped to create a propagation model that was both accurate and economical. Thus, we referred to the propagation model to design a positioning algorithm, and determined that this algorithm can increase the position accuracy by approximately 11.6%.