在 Wi-Fi Monitor模式下通過Sniffers使用 RSSI 指紋進行被動室內定位

Abstract

本論文旨在實現Wi-Fi被動室內定位系統（IPS），無需待測物（DUT）安裝額外應用程序，也無需用戶主動協作。 Wi-Fi 偵測器 （Sniffers）部署在實驗區域，以掃描和收集 DUT 的 Wi-Fi 接收信號強度（RSSI）作為 Wi-Fi 指紋，應用加權k最近鄰域（WKNN）方法獲得DUT位置。為了每個RP可以接收的Wi-Fi RSSI最大化，本文使用具有增強信息熵特徵的目標函數，利用改進基因演算法（GA）優化Wi-Fi偵測器的部署，實驗上使用Wi-Fi測量機器人在每個 RP 上自動收集Wi-Fi 2.4 GHz和 5 GHz RSSI 數據。初步結果表明，僅使用 20 個 Wi-Fi Sniffer 作為模型訓練的特徵，離線定位精度可以達到可接受範圍為 2.2 m。此外，在 NTU 中實施了概念驗證的真實在線被動 IPS，以顯示檢測 DUT 在線存在並隨後獲得其 RSSI作為位置估計的在線指紋的可能性。

This thesis focused on the realisation of Wi-Fi based passive indoor positioning system (IPS) without any additional application installed on the device-under-target (DUT) and without active collaboration from the user. The Wi-Fi Sniffers are deployed in an area of interest to scan and collect DUT’s Wi-Fi received signal strength (RSIS) as Wi-Fi fingerprints for mapping vectors of RSSI to each reference point (RP) in the physical world by applying the weighted k-nearest neighbourhood (WKNN) method. To maximise the Wi-Fi RSSI that can be received in each RP, optimisation of the deployment of Wi-Fi Sniffers is considered using a modified Genetic Algorithm (GA) for an objective function with the enhanced feature of information entropy. Automate data collection of RSSI at each RP is done using a surveying robot for both Wi-Fi 2.4 GHz and 5 GHz. Preliminary result shows that the off-line positioning accuracy can achieve 2.2 m which is acceptable with only 20 Wi-Fi Sniffers as features for model training. In addition, a proof-of-concept real on-line passive IPS is implemented in NTU to show the possibility of detecting the on-line presence of DUT and subsequently obtain its RSSI as on-line fingerprints for position estimation.