Magnetic Semantic Extended Kalman Filtering for Inertial Navigation System

Systems combining inertial measurement unit and magnetic field information have been extensively studied for indoor positioning and navigation. However, existing research predominantly addresses static magnetic field distributions, with limited attention to user-environment interactions. Establishing the correlation between magnetic field changes and user behavior presents new challenges for enhancing spatial structure perception. This paper proposes an indoor positioning algorithm based on magnetic semantic extended Kalman filtering (EKF) for inertial navigation system. We analyze the projection characteristics of magnetic field vectors in the carrier coordinate system, revealing the inherent relationship between abrupt changes in magnetic field components and user's dynamic behavior. Based on this, we introduce fuzzy logic methods to model the relationship between abrupt changes in the magnetic field and angular velocity fluctuations, thereby accurately extracting the semantic information of the magnetic field environment. To fully utilize this semantic information, we design an EKF algorithm that integrates magnetic semantic perception information to assist the INS in state updating and error correction. The proposed algorithm not only improves positioning accuracy but also enhances the system's adaptability to dynamic environmental changes. The experimental results indicate that the maximum error of our indoor positioning system is 0.93 meters and the error is controlled within 1 meter.