# AI-Assisted Passive Magnetic Distance/Position Sensor

## TL;DR

This paper introduces an AI-based magnetic sensor that improves distance and position measurements with high accuracy.

## Contribution

A novel back propagation neural network approach for precise magnetic distance and position sensing.

## Key findings

- Using a single magnetic sensor, the BP neural network achieves a measurement error of −0.0268 mm to 0.0362 mm.
- With three sensors, the error is reduced to −0.0107 mm to 0.0093 mm.
- Positioning errors within a 60 mm × 60 mm area are confined to less than 1.13 mm along both axes.

## Abstract

Magnetic sensing technology is crucial for non-contact distance and position measurement. Due to the nonlinear characteristics of the magnetic fields from permanent magnets, conventional magnetic sensors struggle with accurate distance and position determination. To address this, we propose a distance/position sensor that employs a customized back propagation (BP) neural network. By detecting the magnetic field variations induced by a permanent magnet, the proposed sensor can effectively model the nonlinear mapping between magnetic field strength and distance, thereby enabling precise distance and position measurement. Experimental results demonstrate that the BP neural network approach, when employing a single magnetic sensor, exhibits a measurement error in the range of −0.0268 mm to 0.0362 mm over a distance of 0–70 mm, which is significantly lower than traditional methods based on the magnetic dipole model and the Levenberg–Marquardt (LM) algorithm. Increasing the number of sensors to three reduces the error further to −0.0107 mm to 0.0093 mm. Furthermore, when employing four magnetic sensors for position measurement within a 60 mm × 60 mm planar area, the positioning errors along the x-axis and y-axis are confined to the ranges of −0.6168 mm to 1.1312 mm and −0.6001 mm to 0.5133 mm, respectively.