MAGNETOMETER CALIBRATION USING ONE-LAYER NEURAL NETWORK

Abstract

The paper presents a novel easy-to-use iterative calibration algorithm for a magnetic field estimation accuracy improvement, which can be successfully applied to the estimation of a 3-axis magnetometer biases and scale factors of the each axis, extended to estimate non-linearity and non-orthogonality corrections. The theory is based on the neural network that creates an inverse function the uncalibrated sensor’s transfer function. Learning process of the neural network uses a gradient methodology applying total differential on the scalar error equation. The analyzed theoretical principles are supplemented by simulations and experimental measurements. The performed simulations and experiments confirmed that the algorithm successfully converges to a good estimation of the calibration constants. Other advantage of this methodology is that the calibration procedure is based on the attitude independent sensor discrete random rotation in the 3D space without the need of any non-magnetic calibration platforms. Advantages of this method compared with others lie not only in the simplicity of the presented algorithm, sensor attitude independency, measurement repeatability and no need of non-magnetic calibration platform utilization, but also in the speed, precision, undemandingness and comfort of the presented calibration procedure, which lead to the effective magnetometer calibration constants determination and calibration errors reduction.