SL(C)AMma: Simultaneous Localisation, (Calibration) and Mapping With a Magnetometer Array

Researchers Thomas Edridge and Manon Kok have introduced SL(C)AMma, a novel approach to Simultaneous Localisation and Mapping (SLAM) designed to solve a core robotics problem: accurate indoor navigation without GPS. Traditional methods suffer from signal attenuation and accumulating drift from inertial sensors. While magnetic field-based SLAM can reduce drift by recognizing revisited locations, exploring new areas remains a major challenge. The team's innovation uses an array of magnetometers, which offers a significant advantage over single sensors by enabling direct odometry estimation, though inconsistencies between sensor readings have historically complicated this process.

SL(C)AMma proposes two key filtering algorithms. The first, SLAMma, performs magnetic field-based SLAM using the array. The second, SLCAMma, extends this by jointly estimating the calibration parameters of the magnetometers in real-time, ensuring measurement consistency regardless of the robot's motion. Monte Carlo simulations and experimental validation on ten datasets confirm the system's robustness. The results are striking: in scenarios where single-magnetometer SLAM fails, SL(C)AMma provides reliable trajectory estimates with a reduction in positional drift of more than 80% compared to integrating data from standard proprioceptive sensors like accelerometers and gyroscopes.

The work, detailed in a 10-page paper on arXiv, represents a practical advance for autonomous systems operating in GPS-denied environments like warehouses, hospitals, or underground facilities. By making the sensor calibration an integral, online part of the localization process, the system maintains accuracy during extended exploration. The researchers have also contributed a Python implementation and their experimental datasets to the community, facilitating further development and application of this magnetometer-array-based navigation technique.