Motion Data and Multiscene Feature Enhanced Magnetic/Inertia Positioning Method for Pedestrian

The development of the Internet of Things (IoT) relies on completing the ubiquitous positioning system, but the pedestrian-oriented indoor location-based services (POILBSs) have been an unresolved problem. One of the leading solutions for POILBS involves using smartphones to sense information and employing magnetic field matching (MFM) and pedestrian dead reckoning (PDR). However, this technique has several issues, such as a lack of robust orientation observation, the influence of equipment diversity difference, and the limitation of Kalman filtering (KF) based on constant noise, making the technique face difficulties in achieving high-precision positioning. This study performs experimental research to enhance the adaptive controllability of the positioning system in dynamic application scenes to address these issues, exploring more available information with limited information sources. We proposed four improved methods: 1) MFM improved by integrated weights; 2) adaptive KF enhanced by noise indicator; 3) orientation correction-based KF; and 4) recalculation trajectory recalculated by fitted orientation. The proposed method is conducted in four tailored tests using two public and one personal datasets, and compared to six other research methods to verify its performance. The results show that the root mean square error (RMSE) of three major proposed methods is decreased by 14.66%, 25.18%, and 31.54% compared to that of constant model method, respectively, proving that the proposed methods have better positioning performance and stronger robustness.