m-BMC: Exploration of Magnetic Field Measurements for Indoor Positioning Using mini-Batch Magnetometer Calibration.

Due to the ubiquity and lack of infrastructure, magnetic field-based (MF) indoor localization is garnering a lot of attention. However, there are still issues with discernability, interference from ferromagnetic materials, and heterogeneous devices for MF-based location signals. In this work, we investigate the importance of signal calibration for fingerprint development, showing how particle filtering can be used in conjunction with magnetometer calibration to predict and remove irregularities from MF signals. With this regard, we also evaluate the impact of the heterogeneity of device sensors on the performance of MF-based indoor localization. Finally, we apply and compare a set of machine learning classifiers for the sake of localization performance assessment on both homogeneous and heterogeneous setups. The results show that, in both scenarios, fuzzy KNN can outperform other classifiers by up to 85% and 78%, respectively.