Research And Implementation Of EKF Indoor Positioning System Based On Fusing PDR/BLE

With the rapid development of mobile intelligent terminal computing capabilities and mobile Internet,Location Based Services(LBS)have penetrated into every corner of people's lives.The research focus of positioning shifts into indoor from outdoor gradually with the relative maturity of outdoor location technology.Indoor positioning systems based on high-precision and high-stability of intelligent terminals have become the current research hotspots and difficulties.In many positioning technologies,Pedestrian Dead Reckoning(PDR)is a continuous positioning method,but it has lower accuracy with cumulative errors.Bluetooth Low Energy(BLE)positioning technology has high positioning accuracy and simple hardware deployment.However wireless signal is vulnerable to environmental impact and the positioning is unstable.On the basis of analyzing the advantages and disadvantages of Pedestrian Dead Reckoning and Bluetooth Low Energy,this thesis studies the high stability and high precision indoor fusion positioning technology with complementary advantages based on EKF(Extended Kalman Filter).The author has implemented the fusion positioning system in intelligent terminal.The main contents lies in:(1)The problem of pedestrian's motion state judgment in PDR is analyzed.A sliding window whose length is same with the acceleration frequency is set up.The user's motion state is judged according to the standard deviation of acceleration in this window.A step detection algorithm based on finite state machine is proposed.The accuracy of the finite state machine method is better than peak detection method based on dynamic threshold and autocorrelation analysis method and it has strong anti-interference ability analyzing from user adaptability,walking cadence,holding mode and anti-interference performance.The step length model parameters were fitted using 54 sets of different pedestrians' non-synchronous walking data.The experiment of step length estimation shows that the nonlinear model that step length is related with height and step frequency is the most accurate when user's height and gender is clear.The accuracy of heading angle is estimated by gyroscope,digital compass and rotation vector sensor.The rotation vector sensor is the most accurate way to estimate heading angle.(2)In order to investigate the signal propagation characteristics of BLE in indoor space,the probability distribution of BLE in indoor environment was tested and it is a multi-peak distribution.The research of the Bluetooth signals propagation law in different scenarios shows that there is a large ranging error with a unified ranging model.According to the feature that the Bluetooth signal attenuation speed decreases with the increase of distance,a weighted Bluetooth fingerprinting algorithm based on signal strength is proposed and the positioning accuracy has improved compared to the WKNN algorithm.Aiming at the large workload of establishing and updating the fingerprint database,a rapid establishment of fingerprint library method is proposed.It uses the signal intensity sequence acquired by pedestrians walking at a uniform and lower speed as a reference point fingerprint.The experiment show that the method can save time by more than 70% with ensuring the positioning accuracy compared with the traditional static method.(3)The PDR and BLE positioning methods are fused with EKF based on analyzing their advantages and disadvantages.The fusing method can solve the problems of trajectory drift,sparse location and location back-jumping.The average error of fusing method is 1.31 m and the probability of accuracy superior than 2.5m is 92%.The stability of fusion positioning method is better than PDR and BLE positioning technology.Complete the design and implementation of indoor fusion positioning system.The basic functions and modules are analyzed and divided.The real-time positioning and map visualization are implemented based on Android platform and Arc GIS for Android development kit.