Theory And Application Of Nonlinear Navigation Data Processing

In recent years,satellite navigation,underwater navigation and indoor navigation have developed rapidly.In these fields,various nonlinear problems are often encountered.If the model is linearized according to the traditional method,the adjustment and filtering methods are used to deal with these problems,which often bring serious high truncation error.Therefore,the adjustment and filtering method based on the theory of nonlinear numerical calculation should be considered.In this paper,the research literature on simultaneous localization and mapping(SLAM)at home and abroad is reviewed in detail,and the linear and nonlinear research methods and conclusions in the literature are summarized.In this paper,the nonlinear numerical calculation theory of static nonlinear and dynamic nonlinear problems at home and abroad is reviewed,and all kinds of classical nonlinear algorithms are derived.In addition,a Newton iteration method with damping correction and a nonlinear filter algorithm based on neural network algorithm are presented independently.The research data of this paper include two parts.The first part comes from the pseudo range observation data of the CORS station of Shandong University(Weihai)for a whole day.This part is mainly used for the research and algorithm test of dual station positioning with short baseline constraints.The second part of the data is mainly derived from the simulation data of MATLAB software.This part of the data is mainly used for experimental research and algorithm comparison of SLAM state filtering.Because the observation data of the GNSS observation station usually have all kinds of error components,the article first uses the GLAB software provided by ESA to correct the error of the observation data.Based on the MATLAB program of different numerical algorithms written independently,the article deals with the observation data for a whole day,and the calculated values of the results of the linear method,Newton method,simplex method and damping Newton method are obtained respectively,and then the external characters of the location results obtained by different algorithms are calculated.Finally,the advantages and disadvantages of the three nonlinear iterative algorithms are compared.For the simulation of the SLAM problem,a group of 35 road punctuation points and a motion trajectory with 17 turning points are set artificially,and the parameters of the wheel width,range range and velocity of the robot sensor,as well as the error model of the state transfer model and the observation model are set up.The MATLAB program of neural network filter is written independently,and the results of the simulation are processed by linearization(Calman filter),no trace Calman filter,particle filter and neural network filter.The filtering trajectory is obtained.Then the filtering trajectory and the simulated trajectory can be filtered.Finally,the accuracy,complexity and problems of the three filtering algorithms are analyzed.