Machine Learning-Enhanced Magnetic Calibration for Airborne Magnetic Anomaly Navigation
Albert R. Gnadt
Abstract
View Video Presentation: https://doi.org/10.2514/6.2022-1760.vid Using the earth's magnetic field for navigation of aircraft has shown promise as a viable alternative to the Global Positioning System (GPS) and other navigation systems. An airborne magnetic navigation system collects real-time magnetic field data and uses predetermined magnetic maps of the earth to estimate location by aiding an inertial navigation system (INS), which would otherwise drift. Magnetic navigation has the benefits of being passive, globally available at all times and in all weather, and not reliant on sight of land or stars. Since the magnetic field strength of a dipole decreases with the inverse cube of distance, magnetic navigation is also nearly unjammable. A corrupting magnetic source would have to be near an aircraft to be effective. However, the magnetic components on the aircraft itself can interfere with the desired magnetic measurements that are required to navigate. When the measured data contains magnetic signals from both the (desired) earth field and (undesired) aircraft field it is difficult to separate the two signals. Previous work has proven the viability of magnetic navigation using exceedingly clean magnetic measurements taken by geo-survey aircraft. The most significant outstanding challenge for real-world, operational magnetic navigation is handling corruption of the measured magnetic signal by magnetic sources from aircraft components. The state-of-the-art linear calibration model, used to clean the magnetic signal, uses a single pair of scalar and vector magnetometers. Using additional magnetometer and other flight data with a neural network-based calibration model has been shown to outperform the state-of-the-art model when only in-cabin data is used. On held out testing data, magnetic signal errors of less than 6 nT and navigation position errors of less than 40 m are consistently achieved.