Research On High Precision Indoor Navigation And Positioning Technology Based On Smartphones

At present,the public service construction facilities such as railway stations,hospitals,government affairs centers and shopping malls are getting larger and larger,so it is difficult to find destinations in large buildings,and people's demand for indoor positioning and navigation is increasing day by day.Indoor positioning technology realized by using basic hardware facilities such as bluetooth,router and UWB transmitter has been applied,but the purchase of hardware equipment will greatly increase the cost.When the power of the equipment is insufficient or there is a fault,positioning cannot be realized,resulting in poor indoor positioning stability;Hardware facilities need regular maintenance and replacement of the battery,the later workload is large.Sensors built into smartphones are becoming more accurate,historical sensor data for navigation can also be saved for use,and some of the building's location features are obvious,making indoor location independent of infrastructure possible.However,how to use the sensor data of mobile phone reasonably and achieve high precision indoor positioning and navigation still faces great challenges.An indoor positioning scheme that relies only on smart phones and not on infrastructure is proposed.The accelerometer,magnetometer and gyroscope data of the mobile phone are used for inertial navigation and positioning.Meanwhile,the location identification model is used to correct the error of inertial navigation and positioning and improve the positioning accuracy.Firstly,the pedestrian navigation position prediction algorithm was improved,the peak detection method was designed for the step detection,the dynamic step size estimation model was established for the step size estimation,and then the inertial navigation positioning was realized based on the course prediction.Then it analyzes the building location features,collects the sensor data of mobile phones when walking in the building staircases,elevators,corners and other places,designs a sliding window with a length of 128 moving steps and a length of 16 for data segmentation,and adds a location label to the obtained data unit to make a sensor reading data set for specific locations.Next,the CNN network structure dedicated to location recognition is designed,and the location recognition model is trained with the data set to improve the accuracy of location recognition by adjusting the super parameters.Finally,the location identification model is introduced into the inertial navigation and positioning system.In the process of inertial navigation and positioning,the location identification model is used to identify the location and correct the error of inertial navigation and positioning.The field test shows that there is a cumulative error in the simple inertial navigation positioning,and the positioning error is greater than 2.78 meters when the displacement is more than 30 meters,so it is not suitable for indoor positioning.After the introduction of the location recognition model,the positioning error of the fusion positioning scheme is kept below 2.31 meters,which can meet the positioning accuracy requirements of indoor positioning navigation.This scheme does not depend on any infrastructure,greatly reduces the system operation and maintenance costs,and provides a practical way for the use of smart phone sensor data for auxiliary positioning.