Neural Network Regression Analysis of Magnetic Sensor Data for Spatial Magnet Positioning

This paper deals with the magnet position estimation based on the measurements from multiple Hall sensors. Estimation of the position is obtained from trained artificial neural networks. Data are gathered using a custom sensor board that has eight three-axis Hall sensors on it, and correlation between magnetic field and position was determined to be able to estimate position in X and Y directions. First measurements were made to determine the optimal height of the magnet for later data gathering. The outcome of the research was evaluated by comparing the ground truth with the estimated values of the best-trained network to determine the error of estimation. Precise measurement of the position of the magnet is intended to be used for the development of a tactile sensor that estimates the contact force through the displacement of the magnet in the deformable layer.