PDR Indoor Positioning System Based On Low-cost IMU Array

The role of location information in people’s daily work is increasing day by day.Due to factors such as building blocks,GNSS cannot provide accurate indoor location information.Therefore,scholars have studied many indoor positioning solutions.With the rise of MEMS technology,the accuracy of IMU based on MEMS technology is gradually improved,it has the advantages of low cost,small size.The PDR technology based on IMU is gradually applied in the field of indoor positioning.However,a single low-cost MEMS IMU(hereinafter referred to as a low-cost IMU)still has limited accuracy,and the error is large when used for PDR positioning.This subject designs a low-cost IMU array data synchronous acquisition system,fuses the measurement data of the array with the data fusion method,and applies the fused data to the PDR technology to further reduce the indoor positioning error.The main research work of this thesis is as follows:(1)Designed a set of IMU array data synchronization acquisition system.The system consists of two parts,a low-cost IMU array board and a high-precision IMU,and the sampling frequency of both parts is 100 HZ.The low-cost IMU array board uses the STM32 microcontroller to control four low-cost IMUs to synchronously collect data,and sends the collected measurement data to the upper computer through DMA.The measurement data of the high-precision IMU is directly sent to the upper computer through Wi Fi,two parts of data are synchronized through flag bits.(2)Produced a dataset including inertial data of multiple experimenters under various motion states,and proposed a low-cost IMU array data fusion using a MLP model.After comparing the accuracy and complexity of MLP models under different network structures,we selected the most suitable network structure.Under this network structure,the complexity of the MLP model is low,and the inertial data after model fusion is closer to the real inertial data of the carrier.(3)Studied the basic theory of PDR positioning technology and three links: cadence detection,step size estimation,and attitude calculation,and conducted Allan variance analysis experiments and multiple indoor pedestrian positioning experiments.In the Allan variance analysis experiment,compared with the angular velocity data of a single low-cost IMU,the random walk,bias instability,and rate ramp noises of the angular velocity data predicted by MLP are significantly reduced; indoor positioning with a total length of 592 meters In the experiment,compared with the error of using a single low-cost IMU data for PDR positioning,the error of using MLP model prediction data for PDR positioning is reduced by 37.5%.