Promotion Frida Viset

Scalable magnetic field mapping and localization using Gaussian process regression

1e promotor: dr. Manon Kok, 2e promotor : dr. Rudy Helmons

Outdoors, the magnetic field can help navigation because it points towards the magnetic north, with the strength of the earth magnetic field. In environments that have elements that perturb the magnetic field, the field can point in various directions, and have various strengths depending on the precise location within that environment. These invisible fields therefore potentially provide information improving the localization of vehicles, people and robots in environments and situations where typical means of navigation such as satellite signals or pre-employed navigation infrastructure are unavailable. Gaussian process regression enables creation of maps of the nonlinear magnetic field variations by combining measurements with prior knowledge about the nature of the field. Existing literature provides solutions for creating maps of the nonlinear magnetic field variations, and using these maps to improve position estimates. This thesis works towards scaling magnetic field mapping and localization in five distinct ways: Making it faster, distributed, require less storage, cover larger areas and enable extrapolation based on repeated patterns.