Research Of WiFi/PDR Indoor Fusion Positioning Based On Adaptive Kalman Filter

The demand for indoor location-based service(LBS)is growing.However,due to the serious signal occlusion,the mature outdoor satellite positioning technology cannot extract the effective information needed for positioning in the indoor environment.Currently,researchers have developed various positioning technologies with their own advantages,but with their own limitations.Radio frequency signals are limited by the hardware foundation and positioning environment,and inertial positioning is limited by the accumulative error.In this thesis,the fusion positioning method of Wi Fi fingerprint positioning technology and the technology based on the low-cost MEMS of smartphone is lucubrated.Taking Android smartphone as the experimental platform,and taking advantage of Wi Fi-wide indoor signals and mobile phone inertial sensor information,a fusion algorithm based on the improved adaptive Kalman filter is proposed and implemented.With the help of video monitoring,the Wi Fi fingerprint database information can be acquired quickly,which improves the collection efficiency of Wi Fi fingerprint database.An improved WKNN positioning algorithm is proposed to reduce the influence of device heterogeneity and fingerprint point mismatching,and improve the accuracy and stability of Wi Fi fingerprint positioning.Based on the low-cost MEMS of smartphone,this thesis mainly studies the sensor error correction,pedestrian frequency detection,step length estimation and heading update,and some improvement measures are proposed to improve the positioning accuracy and control the accumulative error.Combining the research results of Wi Fi and PDR,the fusion localization algorithm is discussed deeply,and the relevant localization test is conducted.The main content of this thesis is as follows:(1)Analyzing the reasons of low collection efficiency using traditional method,studying the method of moving target detection and plane constraint to calculate the coordinates of fingerprint points;Introducing the program for information continuous collection of fingerprint points;Exploring the unity of time reference between video recording equipment and signal acquisition equipment;to solve the problem of sparse sampling,different methods of fingerprint point interpolation are compared,and interpolation algorithm of kriging is choose to complete the construction of fine-particle fingerprint database;And comparing the difference of positioning effect and efficiency between single point acquisition and continuous acquisition.(2)Introducing various fingerprint positioning algorithms;Analyzing the influencing factors of Wi Fi fingerprint location;to improve the timeliness of positioning,the strategy of extracting regional fingerprint information by using Wi Fi information to complete rough positioning through machine learning is studied;Using the signal shape similarity distance to replace the Euclidean distance to solve the problem of device heterogeneity;To eliminate fingerprint mismatching points,the outlier detection and the geometric distance and signal similarity distance between adjacent points and unknown point to determine the weight together are used.Experimental results show that the improved WKNN algorithm is more accurate and more stable than the common fingerprint localization algorithm.(3)Introducing the principle of PDR;Through studying the data characteristics of sensors in different positions,an adaptive step-counting method based on mobile phone multi-source sensor is proposed.A method of step length estimation controlled by multi-variable is proposed to improve the estimation accuracy of step length.A quasistatic magnetic field identification method based on multiple parameters is proposed,and the zero-velocity detection method is studied.Based on the results of these two methods,the measurement equation is established and the pedestrian's heading is updated by EKF.(4)Studying the recognition method of pedestrians' fine-grained behavior.The behavior understanding methods based on the classification model and Key-DTW algorithm are introduced respectively.Focusing on using the inertial sensor data to complete the indoor behavior landmark recognition.(5)Introducing the principle and process of the improved SHAKF algorithm;And the fusion comparison experiment of Wi Fi positioning results and PDR positioning results is conducted by using EKF,UKF and the improved SHAKF algorithm,the experimental results show that the improved SHAKF algorithm has better fusion effect under certain conditions compared with EKF and UKF.