AI Science Tracker
EN

Simulation-trained network locates a physical magnetic resonator

Hamburg University of Technology; University Medical Center Hamburg-Eppendorf; Fraunhofer IMTE · 21 Oct 2026 · physics

Discovery

Simulation-trained network locates a physical magnetic resonator

Source check: Source checked. Not independent scientific confirmation.

Finding: A convolutional-transformer regressor trained on simulated signals with measured noise located a physical magneto-mechanical resonator on a robot-positioned grid. Across repeated static recordings, median position error was 7.26 millimeters and the 95th percentile was 23.7 millimeters. This validates a specific laboratory readout under fixed-orientation conditions, without demonstrating clinical tracking.

Technology: Convolutional-transformer regressor

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
The learned inverse mapping estimates three-dimensional position directly from measured multichannel time signals, enabling the reported robot-reference localization evaluation.
Test and result
A model trained on 100000 simulated signals was tested on physical recordings at75 grid positions,25 frames each, against robot positions.
Limitation / still unknown
One resonator and fixed orientation, repeated static frames and one coil system limit generalization. Frame-level bootstrap uncertainty is not independent hardware replication. Errors were greater near the coil surface; normalized inputs can lose information there. No moving-device clinical test, outcome benefit or general parameter-estimation claim was established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.29394v1 · https://arxiv.org/abs/2610.29394v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 21 Nov 2026. Review scope: Original v1 front, signal/data/model Methods, robot-reference experiment, evaluation Results and limitations.

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 1/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Open in the full catalog for filters, charts and related discoveries. Browse physics discoveries.