A new magnetic compass calibration algorithm using neural networks

摘要: The magnetic compass can provide heading direction by measuring the Earth’s field. In practical applications, there usually exists an unwanted local field that will distort magnetometer measurements; hence a calibration procedure is essential. Current methods are limited inaccurate error estimation when measurements deteriorated disturbances or large noises. This paper proposes new algorithm via modelling nonlinear relationship between and true using neural networks. When external reference available, networks be trained to properly model this input–output pattern even in presence of disturbances, subsequently applied convert into correct heading. proposed does not require declination information biases scale factor estimation. simulation test results have verified effectiveness robustness method also shown performance proportional quality training data.