Magnetic Field Aided Vehicle Localization with Acceleration Correction

TL;DR

This work aims to showcase the potential utility of magnetic fields as supplementary aids to existing localization methods, particularly beneficial in scenarios where Global Positioning System (GPS) signal is restricted or where cost-effective navigation systems are required.

Abstract

This paper presents a novel approach for vehicle localization by leveraging the ambient magnetic field within a given environment. Our approach involves introducing a global mathematical function for magnetic field mapping, combined with Euclidean distance-based matching technique for accurately estimating vehicle position in suburban settings. The mathematical function based map structure ensures efficiency and scalability of the magnetic field map, while the batch processing based localization provides continuity in pose estimation. Additionally, we establish a bias estimation pipeline for an onboard accelerometer by utilizing the updated poses obtained through magnetic field matching. Our work aims to showcase the potential utility of magnetic fields as supplementary aids to existing localization methods, particularly beneficial in scenarios where Global Positioning System (GPS) signal is restricted or where cost-effective navigation systems are required.

Figures (8)

• Figure 1: System overview representing the magnetic field aided localization along with magnetic field data mapping and accelerometer bias and scale factor estimation
• Figure 2: Function fit using weighting technique: Local function fits are merged together by utilizing the weighting scheme.
• Figure 3: Magnetic field data matching technique along with the elimination of false positives. The magnetic field data batches are matched against the function fit map; where the possible matching sections are checked for having intersections with the predicted pose.
• Figure 4: Sensor Setup: The Magwalk magnetometer and VectorNav IMU are both places inside a vehicle as shown.
• Figure 5: Data Analysis: The first plot represents magnetic field data against the trajectory $S$ of the vehicle; the global function fit is obtained using segment-wise local fits. The second plot shows the standard deviation of the batches of magnetic field used for localization. The third plot illustrates the $X$ vs $S$ data and function fit. (As the pose data is sparse, it is not visible in the plot.) A similar plot for $Y$ vs $S$ is also a part of the map structure.