Research On Combined UWB/SINS Positioning Algorithm In Narrow And Complex Environment

In narrow and long scenarios such as tunnels and corridors,continuous,reliable and highprecision indoor positioning technology still faces many challenges:(1)a unified coordinate reference framework for indoor and outdoor is a prerequisite for seamless indoor and outdoor positioning,while the existing seamless reference mainly relies on manual completion,which is time-consuming and increases deployment costs,and is difficult to adapt to environments with high demand for temporary deployment,unknown time-varying and large-scale emergency rescue;(2)Under narrow space,as UWB(Ultra-Wide Band)base stations are generally laid on both sides,approximately two straight lines,all base station coordinates in the short axis difference is small,the geometric configuration of this base station layout is more unfavorable to UWB tag positioning,for planar positioning,directly lead to the traditional analysis method short axis error is large;(3)complex,non-Gaussian environment,the traditional filtering model,such as EKF(Extended Kalman Filter)and CKF(Cubature Kalman Filter)filtering models based on the minimum mean square error criterion often have the problem of decreasing estimation accuracy.In this paper,we conducted the research on the fast construction method of UWB/GNSS(Global Navigation Satellite System)reference,UWB static initialization positioning algorithm,and UWB/SINS(Strapdown Inertial Navigation System)dynamic combined positioning algorithm for complex and narrow environment.The main work and contributions of the paper are as follows:(1)In response to the current situation that the implementation of the existing seamless coordinate reference frame(CRF)in unknown and changing environments mainly relies on manual work and is difficult to deploy,a robust graph optimization-based GNSS/UWB indoor and outdoor unified reference frame fast construction method is constructed.The hybrid indoor and outdoor GNSS/UWB base stations and their formed ranging networks are abstracted as a graph model,where GNSS stations and UWB anchor locations are the nodes of the graph and the ranging between UWBs are the binary edges of the graph,based on the generic graph optimization framework(gerneral graphic optimization,g2o),by fixing the global GNSS nodes and using LM(Levenberg-Marquard)nonlinear optimization algorithm is used to achieve the fast solution of the global coordinates of indoor UWB base stations by fixing the global GNSS nodes.Taking into account the non-visual range environment,the robust g2 o optimization scheme proposed in this paper constructs the IGGIII anti-difference equivalent weight function to reconstruct the g2o information matrix based on the graph-optimized UWB ranging residuals.The positioning experimental results show that the g2 o optimization scheme improves 48.9%and 31.5% in the E direction and 14.7% and 6.8% in the N direction,respectively,compared with the conventional g2 o and Huber g2 o schemes.14.7% and 6.8% in the N direction,respectively.(2)A Chan-BFGS algorithm based on Kalman filter smoothing is constructed to address the problems of poor applicability and accuracy improvement of conventional UWB algorithms in narrow non-visual range environments.The algorithm reduces the coarse difference by Kalman filter smoothing ranging,and iterates with the BFGS algorithm using the solution result of Chan algorithm as the initial value.Taking the experimental results of tag T＿0 static localization as an example,the algorithm is reflected in the root mean square error,which is44.2%,about 41.0%,and 41.0% in the X-axis,and 30.1%,29.3%,and 39.7% in the Y-axis,respectively,compared with the least squares algorithm,Chan’s algorithm,and Taylor’s algorithm.(3)To address the problem that the traditional filtering algorithm based on the minimum mean square error criterion for UWB/SINS dynamic positioning is limited in accuracy in strongly nonlinear,non-Gaussian environments,the maximum correntropy square root cubature Kalman filter with the cost function consisting of weighted least squares and maximum correntropy criterion is applied.Gaussian and non-Gaussian simulation experiments show that the new maximum correntropy square root cubature Kalman filtering algorithm has the best tracking ability compared to other filtering algorithms.Taking the experimental results of nonGaussian environment as an example,the NMCSCKF(New Maximum Correntropy Square Root Cubature Kalman filter)improves the root mean square error in X and Y axes by 35.2%and 37.8% compared to the EKF tight combination algorithm;33.3% and Compared with the CKF tight combination algorithm,it improves 30.5% and 37.3%;compared with the SCKF(Square Root Cubature Kalman filter)tight combination algorithm,it reduces 13.6% and 16.6%,which effectively improves the accuracy and robustness of the combined positioning system.