A dipole model-based approach for magnetic disturbance compensation of mobile robots using magnetometer arrays

Magnetic sensors are used for localization purposes in both outdoor and indoor applications. In indoor applications, three-axis magnetometers cannot be used as compasses due to the presence of hard and soft-iron effects that require to be neglected. This paper proposes a new method for magnetic compensation for an unknown metallic object. It considers the source of the disturbance as a dipole, uses an array of four no-coplanar magnetometers, and exploits an Artificial Neural Network (ANN). The ANN receives as inputs the measurement differences of each pairwise sensor and provides as output the disturbance components related to one sensor of the array. In such a way, by estimating the magnetic disturbances, it is possible to evaluate the real value of the Earth’s magnetic field. The ANN is trained by exploiting a database generated by using the dipole model and by evaluating the model for different intensities and orientations. The simulation results show how the method can effectively reduce the introduced disturbances in the surroundings of the object with a Mean Absolute Error (MAE) in the order of $0.3 \mu \mathrm{T}$, and a maximum error in the range from $1.8767 \mu \mathrm{T}$ to $2.5252 \mu \mathrm{T}$. The proposed method contributes to the development of improved localization algorithms, where the proposed disturbance rejection technique enables the utilization of reliable measurements for mobile robot pose estimation.