This repository contains magnetic field (MF) data collected in 2010-2011 with iRobot Create and MicroMag 3-axis magnetometer to perform MF-based SLAM. The data is used in experiments in my PhD research and the related MF SLAM papers (e.g. here, here, and here), and published for reproducibility.
| SLAM trajectories | Corresponding magnetometer readings | Validation data (loop closures) |
|---|---|---|
![]() |
![]() |
![]() |
- Original data converted to csv format easily readable by modern tools, such as Pandas.
- Conversion script in
notebooks/process_raw_data.ipynb
- Conversion script in
- Named columns and only relevant fields left:
t- relative (zero-based) time in secondsx,y- spatial coordinates in meterstheta- heading in radiansmx,my,mz- magnetometer readings in uTspacebar- manual keyboard press (for validation data)
- Different MF scaling factors corrected from the original data
- Only minor median filtering performed to remove MF outliers
- Should be easy to downsample as needed using Pandas, for example
- Raw mapping data (used to produce SLAM data by downsampling and filtering)
- Validation data (used to evaluate localization performance afterwards)
- Calibration data (used to find out the magnetometer bias)
- Matlab script to visualize the raw data and produce SLAM data
- Downsampled 2Hz data for selected data sets (contains produced robot control and MF readings)
Column numbering starting from 1 (for included MATLAB script).
- Odometry x and y, mm (1-2)
- Odometry heading, angles (4)
- Odometry distance, mm (5)
- Magnetometer data, uT scaled by 32 (or 8), and y and z multiplied by -1 (6-8)
- Space bar press indicator (18)
- Time stamp (19)
- Forward translation, m (1)
- Heading change, radians (2)
- Magnetic data, uT (3-5)
- Time stamp, s (6)


