eess.SPOct 6, 2026

FlightMagNav: An Open Dataset and Probabilistic Map Learning and Validation Framework for Outdoor Magnetic Field-Based Positioning

Authors: Isaac Skog, Miguel Ramos Galrinho, Martin Gelin

Organizations: KTH Royal Institute of Technology Stockholm, Sweden · FOI Swedish Defence Research Agency Kista, Sweden

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Magnetic field-based positioning is a resilient positioning technology that requires no external infrastructure and is hard to jam at scale. To facilitate research on magnetic field-based positioning for aerial platforms operating close to the Earth's surface, an open dataset is presented with measurements collected using an unmanned aerial vehicle carrying two optically pumped magnetometers, a type of quantum magnetometer, and a global navigation satellite system-aided inertial navigation system. The dataset includes measurements for both magnetic-field map learning and validation. Along with the dataset, a probabilistic framework for magnetic-field map learning and validation is presented and used to illustrate how the dataset may be used. Finally, we outline research directions that may be explored using the dataset.

Figures & tables

Fig. 1 : Placement of the sensors on platform with distances in mm\mathrm{mm} . Also shown are the coordinate axis of the sensors.
Fig. 2 : Flight trajectories and height profiles of the map learning and validation datasets. Note that the validation flights follow nearly the same horizontal trajectory but at different heights and remain entirely within the mapped area.
FunctionSensorMeasurementSampling rate
Airborne magnetometers2×2\times QuSpin MAGPIETotal magnetic field125 Hz125\text{\,}\mathrm{Hz}
(front and rear)with vector add-onVector magnetic field125/3 Hz125/3\text{\,}\mathrm{Hz}
  Specific force and angular rate (from built-in ICM-20948 IMU)125/2 Hz125/2\text{\,}\mathrm{Hz}
  Position (from built-in GNSS receiver)1 Hz1\text{\,}\mathrm{Hz}
Airborne GNSS-aided INSXsens MTi-680Position, velocity, attitude, specific force, and angular rate100 Hz100\text{\,}\mathrm{Hz}
Local on-ground reference magnetometerQuSpin QTFM Gen-2Total magnetic field125 Hz125\text{\,}\mathrm{Hz}
TABLE I : Sensors and sampling rates used during data collection.
File nameSourceFlight coverageDescription
FrontMag.txtFront magnetometerFlights 1–8Raw output from the front QuSpin MAGPIE sensor module.
BackMag.txtBack magnetometerFlights 1–8Raw output from the back QuSpin MAGPIE sensor module.
GNSSINS_raw.mtbGNSS-aided INSFlights 1–8Native Xsens recording from the MTi-680 sensor module.
GNSSINS.txtGNSS-aided INSFlights 1–8Text export of the Xsens MTi-680 measurements used by the provided parsing software.
RefMag.txtLocal on-ground reference magnetometerFlights 2–8Measurements from the stationary QuSpin total-field magnetometer. The measurements are considered usable for Flights 3–6 and 8.
ups_ref_20260610.secUppsala reference stationFlights 1–3Magnetic-field measurements from the Uppsala geomagnetic reference station.
TABLE II : Raw data files included in FlightMagNav. Files with the same name occur in different acquisition folders. An acquisition may contain data from more than one flight; in particular, acquisition_06 contains Flights 6 and 7.
StructureFieldDescription
data(i)flight_idFlight identifier.
 start_timeStart time of the flight in UTC.
 stop_timeStop time of the flight in UTC.
 raw_folderRaw acquisition folder containing the corresponding measurements.
 purposePurpose of the flight: test, learning, or validation.
 GNSSaidedINSParsed measurements and navigation solution from the Xsens MTi-680.
TABLE III : Organization of the parsed FlightMagNav data. Each element data(i) contains the data and metadata corresponding to one flight.
Fig. 3 : Expected value μM(r)\mu_{\mathcal{M}}(r) of the learned probabilistic magnetic field anomaly map M(r)\mathcal{M}(r) . Only areas where the posterior standard deviation σM(r)<\sigma_{\mathcal{M}}(r)< 3.5 nT3.5\text{\,}\mathrm{nT} are shown. Also shown are the flight paths used in the training data collection.
HyperparameterEstimate [nT]
σ^n(2)\hat{\sigma}_{n^{(2)}}0.6
σ^n(3)\hat{\sigma}_{n^{(3)}}0.7
σ^n(4)\hat{\sigma}_{n^{(4)}}0.8
σ^n(5)\hat{\sigma}_{n^{(5)}}1.3
σ^ξ\hat{\sigma}_{\xi}9.6
σ^w\hat{\sigma}_{w}21.4
TABLE IV : Estimated hyperparameters for the map learning.
FlightAverage height [m]Effective noise σ^n\hat{\sigma}_{n} [nT]
6641.6
72712.9
8947.0
TABLE V : Estimated effective noise for the validation flights.

Explore similar work

CardsList

  1. Mag4D-SLAM Dataset: A Repeated-Traversal Multi-Modal 4D Geomagnetic Dataset for Localization and Mapping

    Jul 24, 2026Bibhutibhusan Nayak, Hyoseok Ju, Giseop KimOrb-Slam2Light Detection And Ranging Data

    Department of Robotics and Mechatronics Engineering, DGIST, Daegu, Republic of Korea

  2. Magnetic Indoor Localization through CNN Regression and Rotation Invariance

    Apr 24, 2026Helge Rosé, Konstantin Klipp, Tom Koubek +2Indoor LocalizationMagnetic Tunnel Junction

    Fraunhofer Institute for Open Communication Systems (FOKUS) Berlin, Germany · Daimler Center for Automotive Information Technology Innovations (DCAITI) Technical University Berlin Berlin, Germany

  3. Information-Theoretic Geometry Optimization and Physics-Aware Learning for Calibration-Free Magnetic Localization

    Apr 24, 2026Wenxuan Xie, Yuelin Zhang, Qingpeng Ding +4Optimal Sensor PlacementIndoor Localization

    Department of Mechanical and Automation Engineering and T Stone Robotics Institute, The Chinese University of Hong Kong, Hong Kong.