Research On Indoor Pedestrian Positioning Based On UWB/IMU Combination With Multi-source Constraints

With the rapid development of society and the Internet,people have higher and higher requirements for intelligent life,so the demand for Location Based Service(LBS)and indoor positioning is constantly improving.However,due to the shortcomings of each single positioning technology,it can not fully meet the needs of indoor positioning system,so combined positioning has become the main solution.Since the two positioning technologies based on ultra-Wideband(UWB)and Inertial Measurement Unit(IMU)have their own obvious advantages and inherent defects,this thesis takes advantage of the advantages of the two positioning technologies.Combined with prior error model,motion mode,Line of Sight(LOS)/Non-Light of Sight(NLOS)detection and map information,a combined indoor pedestrian location method based on multi-source constraints is proposed to realize the fusion of multi-source information and the two location methods.Improve the positioning accuracy and system stability of the integrated positioning system.The main contents include:1.Aiming at the problem that the traditional UWB positioning algorithm is prone to multipath and non-line-of-sight interference,the positioning accuracy is reduced.Based on the least square algorithm and Taylor algorithm,Extended Kalman Filter(EKF)algorithm is adopted to reduce ranging error and abnormal mutation points in positioning.Make the track more smooth,improve the positioning accuracy.2.In the indoor corridor with fixed non-line-of-sight scenes,an error correction algorithm based on Gradient Boosting Decision Tree(GBDT)is proposed for the non-line-of-sight errors of UWB positioning.Train the nonlinear relationship between UWB ranging value and positioning error,and correct the positioning error.Experimental results show that the proposed EKF+GBDT algorithm corrects the non-line-of-sight positioning error compared with the UWB filtering algorithm.3.An adaptive multi-threshold gait detection method is proposed to improve the accuracy of gait detection by machine learning which can recognize the Pedestrian movement pattern and adjust the threshold value adaptatively.A course Angle correction method based on moving average and limiting filter is proposed to reduce the course Angle error.4.In complex and changeable non-line-of-sight scenes,an improved particle swarm optimization algorithm based on multi-source constraints is proposed to solve the problems of multi-path UWB positioning,large non-line-of-sight error and PDR error accumulation.The line-of-sight coefficient was proposed based on LOS/NLOS detection,and the line-of-sight map was interpolated with natural adjacent points.Particle weights were adjusted adaptively according to the UWB positioning error threshold and line-of-sight map.LOS/NLOS detection,prior error model,motion mode,map information,UWB position estimation and PDR step size/heading Angle were combined.Experimental results show that the proposed indoor pedestrian combined location algorithm based on multi-source constraints can improve the location accuracy and reduce the local optimization problem of particle swarm optimization.It can accurately locate pedestrians in both LOS and NLOS.Compared with the single UWB,the average error of X-axis and Y-axis is reduced by 74.42% and 6.17% respectively.Compared with the single PDR positioning,the X-axis mean positioning error is reduced by 92.24%,and the Y-axis mean positioning error is reduced by 85.82%.There are 51 Figures,14 tables,and 86 references in this dissertation.