Research On Indoor Positioning Algorithm For Pedestrian Dead Reckoning Under Complex Conditions

With the rapid development of mobile communication and inertial sensing technologies,indoor positioning is becoming more and more important.In daily life,people are used to using their smartphones for positioning and navigation,and Pedestrian Dead Reckoning(PDR)is a popular indoor positioning method based on smartphone.It does not rely on any pre-deployed infrastructure,only uses the inertial sensor built into the smartphone,and can provide real-time and continuous position estimation when the initial position is known.However,the complexity of indoor environment and the diversity of human motion patterns pose many challenges to traditional PDR localization method.Therefore,this thesis uses smartphones as the positioning device to deeply analyze and study the problems of the traditional PDR algorithm under complex conditions.The main work of this thesis is as follows:1)Aiming at the problem that the magnetometer sensor built into the smartphone is vulnerable to magnetic interference from metal facilities and electrical equipment in the complex indoor environment,thus reducing the accuracy of heading estimation,this thesis proposes a fusion heading estimation algorithm based on magnetic interference detection classification.Its basic idea is to detect whether the magnetometer is disturbed and apply different methods to fuse the heading according to the detection results.First,a support vector machine(SVM)is used to detect magnetic interference,and an improved Relief-F feature selection method based on maximum information coefficient and approximate Markov blanket,MIC-Relief F-MB,is proposed.Secondly,combined with the magnetic interference detection results output by SVM,the fusion heading estimation methods for nonmagnetic interference and magnetic interference scenarios are proposed respectively to reduce the error of heading estimation.Finally,the walking experiment under magnetic interference is designed to evaluate the overall performance of the algorithm.The experimental results show that the magnetic interference detection accuracy based on SVM reaches 98%,and the average error of the fusion heading estimation method is less than 4 degrees.In addition,when the experimental walking distance is 84 m and 252 m,the PDR average positioning error corresponding to the method in this thesis is0.67 m and 2.75 m,respectively.Compared with the traditional method,the positioning accuracy is improved by at least 76%.The above results prove the effectiveness of the method in improving the accuracy and environmental adaptability of the heading estimation algorithm in this thesis.2)Aiming at the disadvantage that the traditional PDR algorithm cannot adapt to different motion states and smartphone carrying postures,this thesis proposes a PDR algorithm based on motion pattern recognition assistance.First,we comprehensively analyzed and considered the five motion states of pedestrians(static,walking,running,going upstairs,and going downstairs)and four smartphone postures(holding,calling,pocket,and swinging),among which,going up/downstairs includes three ways of stairs,elevators,and escalators respectively.Aiming at various combination patterns of pedestrian motion state and smartphone posture,a motion pattern recognition algorithm based on a combination classifier,SVM-FSM-DT,is proposed.Secondly,based on the results of motion pattern recognition,the three main modules of the PDR algorithm are adaptively improved,so that each module can accurately estimate the number of steps,step length and heading of the pedestrian in the complex motion pattern.Finally,considering the movement of pedestrians up and down the stairs,combined with the improved PDR position update algorithm,the PDR localization in 3D space is achieved.The experimental results show that the average recognition accuracy of the SVM-FSM-DT algorithm for motion patterns is 94.6%.The average positioning error of the method in this thesis is 1.07 m in two-dimensional coordinates,while the average positioning error of the traditional PDR algorithm is 6.40 m.In contrast,this method achieves high-precision indoor positioning of pedestrians in multi-motion mode.