Midterm colloquium Fedde van der Meer

Deep Learning Approaches for Noise Compensation in Magnetic Aided Navigation

Supervisor: Dr. Manon Kok

Abstract: Magnetic-aided aerial navigation leverages Earth’s magnetic anomalies for positioning, particularly in GNSS-denied environments. However, platform noise, interference from onboard electronics and structural components, poses significant challenges to measurement accuracy. Traditional methods for platform noise compensation, such as the Tolles-Lawson model and its extensions, are sufficient in case of ideal circumstances. However, limitations of these linear and time-invariant methods, such as multicollinearity and insufficient adaptability to complex, non-linear, time-varying dynamics, motivate the adoption of deep learning approaches. Recent advancements in using deep learning in combination with traditional models demonstrate improved accuracy and robustness in compensating for complex noise sources.