A Particle Filtering-Based Magnetic Field Map Generation Method Using Smartphones

Abstract. Magnetic field matching has emerged as one of the mainstream indoor positioning methods due to its independence from base stations and its resilience to interference. A critical factor influencing magnetic field matching is the magnetic field map, and the challenge lies in generating a high-precision magnetic field map efficiently and cost-effectively. This paper proposes a method for magnetic field map generation based on particle filtering. This approach requires data collectors to traverse the same route repeatedly, using Pedestrian Dead Reckoning (PDR) to estimate trajectories and employing the horizontal and vertical components of the magnetic field to detect when users return to historical positions, thereby controlling position error accumulation. Furthermore, by leveraging the post-processing characteristics of the magnetic field map, all particles are retrospectively analyzed, significantly enhancing the accuracy of trajectory estimation. Subsequently, four non-colinear control points are used to calibrate the relative trajectories, and bilinear interpolation is employed to generate a grid magnetic field map. Experimental results demonstrate that the proposed method achieves a root mean square (RMS) positioning error of less than 1.3 m, meeting the requirements for meter-level magnetic field matching positioning. The generated magnetic field map exhibits errors of 30-40 mGauss in the northward, eastward, and vertical directions.