Fingerprint Accuracy Indication Geomagnetic/PDR Indoor Adaptive Integrated Positioning Based on Smartphones

Kefan Shao,Zengke Li,Qiang Guo,Zhenbin Liu,Qi Wu

DOI: https://doi.org/10.1109/jsen.2023.3327414

IF: 4.3

2023-01-01

IEEE Sensors Journal

Abstract:Geomagnetic (magnetic)/pedestrian dead reckoning (PDR) integrated positioning, used in smartphone-based indoor positioning of pedestrians, is popular owing to its autonomous positioning and lack of a need for additional equipment. Single-point magnetic matching (SPMM) involves minimal computation and localization but is limited by its poor adaptability to diverse environments, few magnetic field features, and device heterogeneity. To solve these problems, we propose an improved SPMM/PDR based on the extended Kalman filter (EKF). First, we present a dynamic ${K}$ estimation for the weighted ${k}$ -nearest neighbor (WKNN) without previous information to compensate for the shortage of the WKNN with fixed ${K}$ when different signal distributions exist in different locations. Then, we propose a stochastic measurement model construction method with combined fingerprint accuracy indicators (CFAIs) to improve the adaptability to diverse environments. Finally, we fully exploit PDR’s high-accuracy relative localization to achieve adaptive compensation of the stochastic measurement model—adaptive EKF (AEKF)—to improve the accuracy and robustness of positioning. We compare three common methods based on five different smartphones, and the decrease in the positioning root mean square error (RMSE) values of the proposed AEKF-CFAI exceeds 16%, with the RMSE being lower than 0.9 m. Moreover, we investigate the influencing factors of the stochastic model construction and the impact of walking speed on AEKF-CFAI, finding that AEKF-CFAI is less affected by device heterogeneity and that AEKF has better robustness for different walking speeds than the CFAI-based method.