Research On Error Compensation Method Of Geomagnetic Vector Measurement Based On Particle Filter

Geomagnetic field has many advantages,which set off the climax of geomagnetic navigation research.Geomagnetic measurement as the basis of geomagnetic navigation,measurement accuracy is particularly important.The geomagnetic measuring instrument mainly uses the optical pump magnetometer as the scalar magnetometer.It uses the optical pumping action of the atom and the magneto-magnetic resonance technique to measure the magnetic field intensity of the weak magnetic field.Fluxgate three-axis magnetometer for vector magnetometer,strapdown in the carrier of the three-axis magnetometer for geomagnetic field measurement with a large error.As many factors influencing the error of the three axis magnetometer,the influence factors are divided into two categories: the error of the tri-axial magnetometer and the error of the carrier magnetic field.Most of the scholars at home and abroad adopt different algorithms to analyze the two kinds of errors separately,and get good results in the estimation and compensation of the error parameters.In this paper,the particle filter algorithm is used to estimate and compensate the error parameters,and the compensation effect is obvious.This paper briefly introduces some knowledge of the geomagnetic field and analyzes the influencing factors of the measurement results of the tri-axial magnetometer.The mathematical models of the output of the tri-axial magnetometer are derived respectively.According to the principle of "attitude independence" calibration,two kinds of error observation models are deduced,and the model is used to estimate the error parameters.Because the observation model is a square term of the geomagnetic field vector model,it is a nonlinear function.In this paper,the particle filter algorithm is used to estimate the parameters.According to the two aspects of the basic particle filter algorithm,two improved methods are proposed: the first is the improved particle filter algorithm based on BP neural network and biodiversity entropy(improved PF1 algorithm);the second is an adaptive particle filter algorithm(improved PF2 algorithm).The improved PF1 algorithm is an improvement in the number of effective particles and the lack of particle diversity.It uses double threshold cutting method to improve the number of effective particles,while using BP neural network to improve the diversity of particles,and the introduction of biodiversity functions as a particle di-versity of the evaluation function.The improvement of the PF2 algorithm is based on the improvement of the low accuracy of the estimation of the particle and the estimation of the time with the use of more particles,and the combination of the likelihood distribution of the weights and the adaptive adjustment of the number of particles to reduce the estimation time and improve the estimation The purpose of precision.Simulation results show that the improved performance of the two improved algorithms is better than that of the basic particle filter.The results show that the improved performance of the two improved algorithms is better than that of the basic particle filter.Finally,two improved algorithms are compared,and it is concluded that although the improved PF1 algorithm is better than the improved PF2 algorithm,the estimation time is not dominant.The improved PF1 algorithm is applied to estimate and compensate errors of the carrier interference magnetic field.The estimated parameters are used as the state quantity,and the difference between the square of the vector model and the square of the real vector value is taken as the observation.The random noise is used as the independent variable to verify the robustness of the algorithm,and the result of the estimation of the parameters is relatively stable when the random noise is large.The improved PF2 algorithm is applied to the estimation and compensation of the self error parameters of the tri-axial magnetometer.Similarly,the state quantity is taken as the square of the state quantity and the real-time measurement vector value.The simulation experiment is carried out in the MATLAB environment.Finally,the two kinds of errors are analyzed synthetically,and the model is simplified.The simplified model is the same as the output model of the jamming magnetic field error,and the observation model with nine error parameters is deduced using the principle of "attitude independence".Using the improved PF1 algorithm to estimate the parameters and error compensation of the MATLAB simulation experiments,the estimated results and error compensation effect compared with the ones before compensation increase three orders of magnitude.