Magnetic-field-based navigation: Empirical evidence of robust methods over contemporary methods

Hanley, David

Permalink

https://hdl.handle.net/2142/115310 Copy

Description

Title

Magnetic-field-based navigation: Empirical evidence of robust methods over contemporary methods

Author(s)

Hanley, David

Issue Date

2022-02-21

Director of Research (if dissertation) or Advisor (if thesis)

Bretl, Timothy

Doctoral Committee Chair(s)

Bretl, Timothy

Committee Member(s)

• Al-Hassanieh, Haitham
• Driggs-Campbell, Katherine R
• Roy Choudhury, Romit

Department of Study

Electrical & Computer Eng

Discipline

Electrical & Computer Engr

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Gaussian process
• indoor localization
• magnetic field mapping
• magnetic localization
• particle filter
• Dataset
• Comparison of Methods

Language

eng

Abstract

Steel studs, heating, ductwork, vents, air conditioning units, rebar, and many other building components produce spatially varying magnetic fields. Magnetometers can measure these fields and can be used in combination with inertial sensors for indoor positioning of robots and handheld devices, such as smartphones. These magnetic-field-based navigations systems do not require the installation of any special building infrastructure as would visible light communication (VLC) systems, nor do they require sensors (magnetometers and inertial sensors) to have an unobstructed field of view as would systems that rely on cameras. The goal of this dissertation is to critically examine two assumptions that have been made previously when designing magnetic-field-based navigation systems and to propose new systems that address these assumptions and work better in practice. The first assumption we examine is that magnetic fields do not vary with height. We show that out-of-plane variations in the magnetic field were significant at over half the locations where magnetometer measurements were taken during our survey. We also show that absolute trajectory error in positioning was low when both localization and mapping were based on magnetometer measurements taken at the same height, but that error increased significantly with even small differences between these heights. Next, we show that the choice of height at which to take measurements-if this height was kept the same for both localization and mapping-had no significant impact on absolute trajectory error when averaged across a given set of trajectories although some trajectories existed for which different measurement heights led to significantly different errors. Then, we show that absolute trajectory error decreased when magnetometer measurements were aggregated across a small range of heights to produce a single, planar map and when measurements at the median height were used for localization. The second assumption we examine is that magnetic fields do not change over time. Among the magnetic fields we consider, prior evidence has demonstrated that moving ferromagnetic objects (which we call live loads) are some of the biggest contributors to indoor magnetic fields that change over time. These live loads can possess magnetic field strengths at least as strong as the local magnetic field. In addition to out-of-plane variations, we show that the presence of live loads degrades the accuracy of magnetic-field-based navigation systems. We subsequently propose four strategies to make magnetic-field-based navigation systems robust to live loads and three strategies to make magnetic-field-based navigation systems robust to out-of-plane variations. Among the strategies considered, an approach that augments the state of a particle filter to include estimates of covariance of the magnetic field map performs best (in terms of absolute trajectory error) with respect to the number of live loads. Subsampling magnetometer measurements is easy to implement and works surprisingly well in making magnetic-field-based navigation systems robust to out-of-plane variations. Finally, we present a dataset for benchmarking magnetic-field-based navigation systems. This dataset includes live loads, was collected both with a pedestrian and with a ground robot, and was collected in three different buildings. We also present a hardware platform that could be used to more easily create new datasets that measure magnetic fields at different heights with pedestrians and robots. These datasets should provide the basis for rigorous characterization of future magnetic-field-based navigation systems in terms of both performance and robustness.

Graduation Semester

2022-05

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/115310

Copyright and License Information

Copyright 2021 David Hanley

Owning Collections