Real-time Uncertainty-Aware Motion Planning for Magnetic-based Navigation

TL;DR

Problem: navigational reliability in GPS-denied settings where SLAM generalizability is limited. Approach: a real-time uncertainty-aware MagNav framework that combines Tolles-Lawson magnetometer calibration, Bayesian localization with a map-informed sensor model, and expected entropy reduction ($EER$) planning, implemented via a particle-filter approximation. Contributions: integrated online planning that actively reduces localization uncertainty while balancing travel efficiency, validated in both simulations and real indoor hardware at 10 Hz. Significance: enables robust autonomous operation in GPS-denied environments using magnetic-field maps, with potential for long-duration missions.

Abstract

Localization in GPS-denied environments is critical for autonomous systems, and traditional methods like SLAM have limitations in generalizability across diverse environments. Magnetic-based navigation (MagNav) offers a robust solution by leveraging the ubiquity and unique anomalies of external magnetic fields. This paper proposes a real-time uncertainty-aware motion planning algorithm for MagNav, using onboard magnetometers and information-driven methodologies to adjust trajectories based on real-time localization confidence. This approach balances the trade-off between finding the shortest or most energy-efficient routes and reducing localization uncertainty, enhancing navigational accuracy and reliability. The novel algorithm integrates an uncertainty-driven framework with magnetic-based localization, creating a real-time adaptive system capable of minimizing localization errors in complex environments. Extensive simulations and real-world experiments validate the method, demonstrating significant reductions in localization uncertainty and the feasibility of real-time implementation. The paper also details the mathematical modeling of uncertainty, the algorithmic foundation of the planning approach, and the practical implications of using magnetic fields for localization. Future work includes incorporating a global path planner to address the local nature of the current guidance law, further enhancing the method's suitability for long-duration operations.

Figures (9)

• Figure 1: A representative scenario of uncertainty-aware guidance for magnetic-based navigation
• Figure 2: Schematic of the workspace and coordinate systems
• Figure 3: The Tolles-Lawson calibration result ($\hat{B}_{e}$) compared with the ground truth ($B_{e}$) and raw sensor readings ($B_{t}$)
• Figure 4: The magnetic field maps used for (a) simulation and (b) hardware demonstration
• Figure 5: Comparison of the robot trajectories executed by the presented algorithm with weights $w_{h}=0$, $0.5$, $1$, $5$ and $10$