Tracking the position of a magnetic object in real time from quantum diamond magnetometer signals with machine learning

1/ AI utilization method: a machine learning model trained only on experimental data
2/ Applied task: real-time tracking of the position of a magnetic object using quantum diamond magnetometer signals
3/ Reported results: training within 40 minutes, 10 Hz tracking, more than 80% accuracy, and 30 cm error tolerance were reported
4/ The researchers trained a machine learning model using only experimental data obtained with a quantum diamond magnetometer, and then used it to track changes in the position of an object emitting magnetic signals in real time.
5/ The target signal was a magnetic signal generated by elevator motion, and the model was reported to have been trained in about 40 minutes and to have performed tracking at a speed of 10 Hz.
6/ The reported performance includes more than 80% accuracy and a 30 cm level of error tolerance.
7/ This result is less a typical materials informatics problem that predicts composition or structure than a case of combining machine learning with a real tracking task that estimates object position from time-series signals produced by measurement equipment.
8/ From a materials research perspective, its significance lies in showing that ML inference can be attached in online fashion to experimental environments where sensitive magnetic sensor signals must be interpreted.
9/ However, this report concerns results under a specific experimental setup and data conditions, and it cannot be said that the same performance is guaranteed for other magnetic objects, other sensor configurations, or higher-noise environments. [1][2][3]