Research On Time-based UWB Indoor Positioning Algorithm

As the pace of social modernization continues to accelerate,humans have more demands for space-time information.Aiming at the problem that indoor positioning is difficult to achieve high-precision positioning,based on the UWB positioning system,this paper studies on clock synchronization,error calibration,collaborative positioning and deep learning positioning.Mainly completed the following tasks.In a time-based indoor positioning system,clock synchronization is the basis for high-precision positioning.This paper proposes a clock synchronization method based on synchronization packets.The master base station periodically sends synchronization packets to the slave base stations to achieve clock synchronization between the base stations.In the case of line-of-sight,after using the synchronization packet to synchronize the base station clock,the maximum error between the base station clocks is about 0.75 ns,and the average error is 0.21 ns.Theoretically,the positioning accuracy within 10 cm can be achieved.Aiming at the environmental errors based on UWB indoor positioning system,such as clock error,multipath error and hardware error,etc.,this paper proposes an error processing scheme based on calibration method.Use the nodes with known positions to determine the error value and store it in the background server.This error value can be used to compensate the TDOA value in actual positioning.Experimental results show that applying the error calibration value to the existing collaborative positioning algorithm based on Chan and Taylor(C-T algorithm)can reduce the average positioning error from 14.5cm to 8.3cm,and the positioning performance has been greatly improved.In order to improve the accuracy of the UWB positioning algorithm,this paper proposes a weighted least squares and Taylor collaborative positioning algorithm based on error calibration(E-W-T algorithm).Based on the compensation of TDOA value by error calibration value,WLS is used to estimate the initial coordinates of the label and the Taylor algorithm is used to determine the final coordinates of the label.Experimental results show that the average error of the algorithm is 7.3cm,and the positioning performance is improved by 12.0% compared with the C-T algorithm.Based on the above positioning algorithm,this paper proposes an indoor positioning algorithm based on deep learning.Using deep learning to construct the mapping relationship between TDOA values and label coordinates.The deep learning model of the trained four-layer back propagation neural network is used to achieve high-precision indoor positioning.Experimental results show that the average error of the algorithm is about 5.1cm,and the positioning performance is improved by 30.1% compared with the E-W-T algorithm.