A data-driven IMU/magnetometer fusion method for foot-mounted pedestrian localization in magnetically uncertain environments

Abstract Foot-mounted IMU/magnetometer positioning provides full autonomy and is particularly valuable in global navigation satellite system-denied environments. Conventional approaches rely on magnetometers for heading estimation in undisturbed fields or on magnetic fingerprint matching under significant interference. However, these methods perform reliably only when the magnetic field environment is well-characterized in advance. To enhance navigation performance in unknown magnetic interference conditions, this study proposes a data-driven IMU/magnetometer fusion technique for foot-mounted pedestrian localization. First, a magnetic interference detection strategy is developed based on dot-product consistency between gravity and geomagnetic vectors, enabling adaptive adjustment of the fusion process. Second, a drift reduction algorithm is introduced, where an optimized density-based clustering method identifies straight-line movement segments, from which position error observations are derived for navigation correction. The proposed method is particularly suited to navigation tasks in magnetically uncertain environments with prevalent linear structures. Experimental results demonstrate that, under significant magnetic interference, the proposed positioning method achieves an average relative error percentage of only 0.08%, substantially outperforming conventional approaches.