Research On Magnetometer Calibration Technology Based On MEMS Inertial Sensor

In recent years,with the popularization of inertial sensors such as embedded magnetic sensors and gyroscopes in mobile devices,indoor positioning technology based on geomagnetism has been widely used.However,due to the presence of airborne hard magnetic materials,soft magnetic materials,and the effects of different environments on the geomagnetic field,the measured values of magnetic sensors are often disturbed by various errors.In order to eliminate these deviations,many geomagnetic calibration methods have been proposed.However,most studies have only focused on the calibration of hard magnetic errors,ignoring the impact of soft magnetic errors on geomagnetic measurements,resulting in low accuracy of geomagnetic calibration in complex indoor environments and large fluctuations in performance with environmental changes,thus It has a greater impact on the performance of geomagnetic-based positioning systems.In order to solve the above magnetic calibration and positioning problems,this paper proposes a geomagnetic matching positioning algorithm based on magnetic sensor calibration.The algorithm takes into account both hard and soft magnetic errors in the calibration of geomagnetic observation results,and then performs error compensation on the collected geomagnetic data based on the calibration results,and uses a combined neural network to train the positioning model.Compared with the traditional geomagnetic positioning technology,the algorithm proposed in this paper further improves the positioning accuracy.This article uses the embedded magnetometer embedded in the smart terminal to collect data to avoid the additional overhead caused by the separate deployment of hardware devices.This article first proposes a hierarchical geomagnetic calibration framework,which greatly reduces the scope of the search space.The top layer preprocesses the magnetic observation data,removes the gross error,and uses the ellipsoid fitting algorithm to find the initial value of the calibration error,thereby obtaining the initial solution space.In the fine calibration layer,the cost function is constructed based on the Robust Least Squares(RLS)algorithm and the residuals are reasonably weighted.The genetic algorithm is used to replace the traditional gradient descent algorithm that is prone to local optimal dilemma.The loss of measurement accuracy caused by the soft magnetic error,and in the iterative process to achieve the security of effective information processing.The geomagnetic values after the final error compensation are used to refit the ellipsoid,and the improvement of the calibration performance is evaluated by the ellipsoid flatness before and after the fitting and the distance from the center of the sphere to the origin.Then,in the positioning stage,a combined neural network of Residual Network(ResNet)and Long Short-Term Memory(LSTM)is used to match the positioning algorithm of the geomagnetic sequence,and the geomagnetic fingerprint after error compensation is used as the data source.The combined network is used for sequence matching to obtain the final positioning result.Experimental results show that the geomagnetic positioning algorithm based on magnetic sensor calibration proposed in this paper can obtain positioning accuracy of 3 meters with 80%confidence,which is 20%higher than that of geomagnetic positioning method based on particle filtering and dynamic time warping method.This paper implements a complete calibration and positioning process based on geomagnetic information.