Research On Indoor Localization Algorithm Based On Rotating Antenna Array

Accurate indoor localization is a hot topic of research.Wi-Fi technology,as a representative technology of WLAN with wide coverage and low cost,is widely used in indoor environment and can provide location information while communication.However,due to the closed and complex indoor environment,the existing indoor WiFi localization methods have the problems of complex phase calibration,high cost of fingerprint library construction and low resolution of localization,which are difficult to achieve the practical application requirements.Therefore,this thesis intensively analyzes the localization mechanism of Channel State Information(CSI)in the physical layer of Wi-Fi and carries out the research on indoor localization method based on rotating antenna array.The main research contents are as follows:(1)In view of the complex problem of CSI phase calibration,an Angle of arrival(AOA)localization algorithm based on the rotating antenna array is proposed.First,the CSI phase difference model of the rotating antenna array is constructed,and the CSI phase offset is removed by the feature that the clock frequency and the downsampling frequency of the wireless network card are the same.Further,the CSI phase difference optimization method is designed to unwrap,denoise and subcarrier select the phase difference to improve the accuracy of AOA estimation.Finally,the mapping relationship between the physical position of the rotating array antenna and the CSI phase difference change rate is analyzed to achieve the position estimation without phase calibration.(2)To reduce the influence of indoor multipath effect and solve the problem of high cost of fingerprint database construction,an AOA estimation correction model Fusion Rotation based on two-stream feature fusion neural network is designed.First,the spatio-temporal information implied in the angle-of-arrival sequence and CSI quotient matrix is extracted by convolutional neural network and feature fusion is performed.Then the attention mechanism is used to reduce the fused feature dimensionality.Finally,the localization task is converted from a regular classification problem to a regression problem,and the AOA estimation results are corrected.The method can achieve robust indoor localization using only a small number of location samples without building a fine-grained grid fingerprint library in the offline training phase.In summary,this thesis proposes a rotating array-based indoor localization algorithm and builds a prototype indoor localization system using TL-WDR4310 routers deployed with Atheros-CSI-Tool,rotating platforms,a remote control vehicle,and NOKOV 3D optical motion capture cameras.The experimental evaluation of the algorithm’s localization performance,rotation speed,sequence length and other parameters in two scenarios of indoor laboratory and underground tunnel,shows that the algorithm proposed in this thesis significantly outperforms existing algorithms.This thesis has 47 figures,4 tables and 87 references.