Research On Key Technologies Of Pedestrian Navigation System Based On Machine Learning

In recent years,pedestrian navigation has been gradually researched as the main field of navigation technology.The pedestrian positioning method based on inertial technology tends to be mature and has achieved good results.However,due to the accumulation of positioning errors,the performance of inertial navigation system is affected,so the pedestrian positioning technology based on multi-sensors becomes the main research trend in the future.This paper focuses on the inertial navigation system(INS)installed on foot and visual/inertial integrated navigation system(VINS)installed on trunk.This paper especially focuses on the micro-inertial pedestrian navigation and positioning method based on deep learning,and studies the two pedestrian navigation systems to help each other to improve the accuracy of pedestrian positioning.Firstly,this paper studies the micro-inertial pedestrian navigation system based on the inertial measurement unit(IMU)installed on foot.The navigation information of the system is mainly provided by strapdown inertial navigation.The Zero Velocity Update(ZUPT)based on Kalman filter is triggered by the zero speed detection of foot motion information,which compensates the navigation system error and realizes the navigation and positioning function with a certain accuracy.Secondly,aiming at the problem of malfunction or over-range of inertial measurement units installed on the foot of pedestrian,a gait recognition and robust autonomous location method based on deep learning is further studied.Support vector machine(SVM)is used to recognize various kinds of regular gait types of pedestrian,and different deep learning models is constructed according to different gait types,then the virtual inertial measurement units(VIMU)is constructed in the case of foot inertial system failure,thus forming a strong robust autonomous positioning method based on the reconstruction of the inertial navigation system.Finally,in view of the problem that the large heading angle error of micro-inertial pedestrian navigation system with IMU foot installation method,this paper studies the visual navigation and positioning method assisted by foot inertial information.Firstly,this paper studies the working principle of the visual/inertial integrated navigation system of the trunk installation.Secondly,in view of the problem that the heading angle error of the micro-inertial pedestrian navigation system cannot be corrected,this paper studies a method of using the heading angle of visual/inertial integrated navigation system to correct the heading angle of foot inertial navigation system under a certain walking mode.Aiming at the problem of large component error of the inertial measurement unit in the visual/inertial integrated navigation system,this paper estimates and corrects the component error through Kalman filter,so that it can be applied to the low-cost inertial measurement units.Finally,the micro-inertial pedestrian navigation system and vision/inertial integrated navigation system are validated in different real scenes.