Research On Error Analysis And Calibration Of Three-Axial Magnetometers

Three-axial magnetometers is the most important pose estimation sensor in the area of geomagnetic navigation, and the accuracy of it influents the navigation directly. As geomagnetic navigation has become the hotspot of the research, scholars are paying more attention on the study of three-axial magnetometers correction. How to achieve the instrument error and the correction of magnetic deviation simply and accurately and how to realize the correction of strapdown three-axial magnetometers are the significant problems and hotspot in this domain at present. In the view of those problems mentioned above, the following aspects will be studied in this paper.Firstly, instrument error calibration of three-axial magnetometers is investigated. Three-axial magnetometers instrument errors include bias error, scale factor error and orthogonal error. A three-axial magnetometers instrument error correction model is established by analyzing the nature of it. In this paper, modulus correction method is used to transform the vector model to modulus model in order to obtain the linear equations of the parameters through the parameter transformation. Then, recursive least squares algorithm (RLS) is proposed in this paper to identify the parameters of the equations. Furthermore, the comparison of calibration results between RLS and neural networks (NN) algorithm is made. Simulation results show that the former are obviously superior to the latter on the convergence and the accuracy of calibration.Secondly, compass deviation correction of three-axial magnetometers is discussed. The compass deviation is composed of hard iron error, soft iron error and alignment error in this thesis. Detailed analysis has been carried out on the source and nature of the three errors and the compass deviation calibration model has been set up. Modulus correction method is also applied to obtain the linear equations of the compass deviation calibration parameters. The model parameters with RLS algorithm and Neural Networks algorithm are identified, respectively. Compared with the RLS algorithm, NN algorithm is worse on the accuracy. However, since the alignment error has nothing to do with the modulus, it can not be corrected in the modulus model. For this purpose, using Vertical Procrustes Problem to achieve alignment error correction is proposed in this paper.Finally, integrate error of srapdown three-axial magnetometers is deliberated. With the analysis of errors of three-axial magnetometers above, an error correction model is established. Correction problem can be divided into two cases. In the first condition, while the orthogonal error can be neglected, the calibration of three-axis magnetometer can be obtained by the algorithm what has been discussed in section 2 and 3. Otherwise, the singular value decomposition is used to decompose the parameters matrix. It is proved that the readings of magnetometer are on an ellipsoid when they lie in Cartesian coordinates. And the solution of the translation error is put forward. Then, the modulus correction model is presented in order to obtain the linear equations of substitution parameters with the parameters substitution and the substitution parameters are estimated by RLS. Finally, the error parameters matrix is calculated by the Cholesky decomposition of substitution parameters matrix, and the corrected readings of magnetometer can be obtained.