Efficient Simultaneous Calibration of a Magnetometer and an Accelerometer

Abstract

This paper describes a calibration algorithm to simultaneously calibrate a magnetometer and an accelerometer without any information besides the sensors readings. Using a linear sensor model and maximum likelihood cost, the algorithm is able to estimate both sensors' biases, gains, and covariances, besides sensor orientations and Earth's fields. Results show errors of less than 0.1 standard deviations in simulation, and high-quality estimates with real sensors even when the algorithm's assumptions are violated.