Research On Indoor Location Technology Based On WI-FI/PDR Fusion

A widely used high-tech technology,positioning technology,plays a huge role in commerce,transportation and military,and its development potential is obvious to all of society.With the improvement of the demand for the accuracy of individual location information in all walks of life,and the rapid development of portable intelligent mobile devices and Internet technology,indoor location technology research has become a hot spot in the field of location technology.In tall and complex buildings and underground places,GPS(Global Positioning System)positioning can not effectively provide positioning accuracy that meets the needs.Therefore,considering the efficiency,practicability and cost of positioning,Wi-Fi positioning technology and pedestrian dead reckoning(PDR)positioning technology are commonly used in the research of indoor positioning technology.PDR positioning technology is based on inertial sensor,which has a high positioning accuracy in a short period of time.However,with the passage of time,there will be accumulated errors,which will lead to the decline of accuracy.The error of Wi-Fi indoor location based on local wireless network is independent of the application time,and it can provide the absolute location information of the target point at a certain time.However,the signal will be affected by the structure of the building,resulting in the Wi-Fi location can not reach the ideal accuracy.How to combine the above two indoor positioning technologies is the focus of current research.Based on the combination of the above two methods,the indoor positioning method proposed in this paper focuses on the in-depth exploration of the Kalman filter model fusion positioning method,and finally achieves the purpose of optimizing the positioning effect.In the stage of Wi-Fi fingerprint location,fingerprint matching algorithm is the key to ensure the accuracy.This paper aims at the classical k-nearest neighbor(KNN)and weighted k-nearest(KNN)algorithms Neighbor,WKNN).After the experiment,we make a comparison and summarize the reason why the traditional algorithm is unstable.After the comparison,we propose to integrate the support vector machine algorithm into KNN algorithm.After that,through the experimental verification,the results show that KNN algorithm can remove the singular value of support vector machine,and support vector machine algorithm can also reduce the influence of KNN in complex building environment by multi-path problem,eliminate the situation of position jump,enhance the stability and reduce the positioning error as much as possible.After that,the positioning process of PDR algorithm is improved.Considering that the detection of heading angle is the most important factor that affects the positioning accuracy of PDR,this paper proposes that we can use magnetometer,building construction information and PDR / INS(internal navigation)to solve this problem System)to estimate the heading angle information of each step synthetically,and to provide the reliable heading angle data to the calculation method of pedestrian's heading,and to obtain a high accuracy.Combining the advantages and disadvantages of the two positioning methods,as well as the calculation error of the algorithm,this paper proposes two fusion algorithms to assist the positioning.The first is the Wi-Fi / PDR algorithm based on the adaptive extended Kalman filter,and the second is the Wi-Fi / PDR algorithm based on the adaptive unscented Kalman filter model.These two algorithms can complement each other to improve the positioning accuracy.After the actual comparison experiment,referring to the movement track simulated by various positioning methods in the experiment,the results and data of the experiment are analyzed.The adaptive unscented Kalman filter model can improve the error of the classical Wi-Fi / PDR fusion positioning method to a certain extent,improve the instability of the Wi-Fi positioning,and also improve the inertial navigation over time to a certain extent The accumulated error is corrected,which is better than the adaptive extended Kalman filter model and reduces the positioning error by 19%.Before and after the experiment,compared with the PDR positioning results before the fusion positioning,the adaptive unscented Kalman filter fusion algorithm reduces the positioning error by 70%.