Automated Machine Learning Strategies for Multi-Parameter Optimisation of a Caesium-Based Portable Zero-Field Magnetometer

Q: Why is automated machine learning useful for multi-parameter optimisation in this magnetometer?

It finds optimal operational parameters more efficiently than manual methods, especially when multiple coupled parameters create a large configuration space that would make exhaustive search impractical.

Q: How is the magnetometer sensitivity optimised in the presented approach?

Sensitivity is optimized through direct measurement of the noise floor and indirectly via measurement of the on-resonance demodulated gradient of the zero-field resonance.

Q: Which automated optimisation techniques are used in the study?

The study uses three independent techniques: a genetic algorithm, a simplified gradient ascent optimisation method, and an open-source machine-learning package using predictive modelling.

Aug 3, 2026

Upload by Logic

Machine learning (ML) enables efficient interrogation of complex systems to identify optimal parameters without laborious manual tuning. This efficiency is crucial for devices with coupled dynamics across many parameters, where exhaustive searches become impractical. The work presents automated ML strategies for optimizing a single-beam caesium (Cs) SERF optically pumped magnetometer (OPM). Sensitivity is optimized by measuring the noise floor and the on-resonance demodulated gradient, yielding improved optimal sensitivity from 500 fT/√pHz to < 109 fT/√pHz, and supporting benchmarking of hardware design changes.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages16

Reading Time40 min

AccessFree to download

Introduction

Motivation for automated optimisation

Challenges in multi-parameter sensor tuning

Automated ML optimisation strategies

Genetic algorithm

Gradient ascent optimisation

Predictive modelling package

sensors Article Automated Machine Learning Strategies for Multi-Parameter Optimisation of a Caesium-Based Portable Zero-Field Magnetometer Rach Dawson *, Carolyn O'Dwyer *, Edward Irwin , Marcin S. Mrozowski , Dominic Hunter , Stuart Ingleby , Erling Riis and Paul F. Grifﬁn Citation: Dawson, R.; O'Dwyer, C.; Irwin, E.; Mrozowski, M.S.; Hunter, D.; Ingleby, S.; Riis, E.; Grifﬁn, P.F. Automated Machine Learning Strategies for Multi-Parameter Optimisation of a Caesium-Based Portable Zero-Field Magnetometer. Sensors 2023, 23, 4007 . [https://doi.org/10.3390/s23084007](https://doi.org/10.3390/s23084007) Academic Editor: Etienne Labyt Received: 27 February 2023 Revised: 11 April 2023 Accepted: 13 April 2023 Published: 15 April 2023 Copyright: © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license ([https://](https://)[ ](https://)[creativecommons.org/licenses/by/](creativecommons.org/licenses/by/)[ ](creativecommons.org/licenses/by/)[4.0/](4.0/)) . Department of Physics, Scottish Universities Physics Alliance SUPA, University of Strathclyde, Glasgow G4 0NG, UK * Correspondence: [rachel.dawson@strath.ac.uk](rachel.dawson@strath.ac.uk) (R.D.); [carolyn.odwyer@strath.ac.uk](carolyn.odwyer@strath.ac.uk) (C.O.) Abstract: Machine learning (ML) is an effective tool to interrogate complex systems to ﬁnd optimal parameters more efﬁciently than through manual methods. This efﬁciency is particularly important for systems with complex dynamics between multiple parameters and a subsequent high number of parameter conﬁgurations, where an exhaustive optimisation search would be impractical. Here we present a number of automated machine learning strategies utilised for optimisation of a single-beam caesium (Cs) spin exchange relaxation free (SERF) optically pumped magnetometer (OPM) . The sensitivity of the OPM (T/ pHz), is optimised through direct measurement of the noise ﬂoor, and indirectly through measurement of the on-resonance demodulated gradient (mV/nT) of the zero-ﬁeld resonance. Both methods provide a viable strategy for the optimisation of sensitivity through effective control of the OPM's operational parameters. Ultimately, this machine learning approach increased the optimal sensitivity from 500 fT/ pHz to < 109 fT/ pHz. The ﬂexibility and efﬁciency of the ML approaches can be utilised to benchmark SERF OPM sensor hardware improvements, such as cell geometry, alkali species and sensor topologies. Keywords: magnetometry; atomic; optimisation; machine learning; SERF; caesium 1. Introduction OPMs have shown impacts across many ﬁelds of magnetic sensing, with the potential perhaps being most transformative in the ﬁeld of magnetoencephalography (MEG) . The ﬂexible placement of sensing volumes and favourable operating temperature provide signiﬁcant advantages over superconducting quantum interference devices (SQUIDs) in many contexts. The sensitivity of commercial OPMs approaches that of SQUIDs while providing functional [1] and longitudinal [2] studies with an important new tool. SERF magnetometers demonstrate sensitivities that approach the low-femtoTesla regime, making this typeof zero-ﬁeld sensor ideal for MEG, although recent work has also demonstrated ﬁnite-ﬁeld sensors attaining the requisite sensitivity for these measurements in the Earth's ﬁeld [3,4] . The majority of reported work in SERF sensors for MEG utilise rubidium as the sensing species. Cs is attractive for MEG as the temperature needed to achieve a comparable vapour pressure is lower than that of other commonly used alkalis, rubidium or potassium. To date, few SERF sensors reported in the literature use Cs [5,6] and only a single sensor is known by the authors that operates in a single-beam conﬁguration [7] . As such, the optimal operation parameters of the sensor are not known a priori. The opti

FAQs

Why is automated machine learning useful for multi-parameter optimisation in this magnetometer?

It finds optimal operational parameters more efficiently than manual methods, especially when multiple coupled parameters create a large configuration space that would make exhaustive search impractical.

How is the magnetometer sensitivity optimised in the presented approach?

Sensitivity is optimized through direct measurement of the noise floor and indirectly via measurement of the on-resonance demodulated gradient of the zero-field resonance.

Which automated optimisation techniques are used in the study?

The study uses three independent techniques: a genetic algorithm, a simplified gradient ascent optimisation method, and an open-source machine-learning package using predictive modelling.

Aug 3, 2026

Upload by Logic