EMFieldGPR

Gaussian process regression-based magnetic field estimation for wireless power design

This repository provides code for the following preprint.

• Title: Magnetic field estimation using Gaussian process regression for interactive wireless power system design
• Authors: *Yuichi Honjo, Cedric Caremel, Ken Takaki, Yuta Noma, Yoshihiro Kawahara, *Takuya Sasatani
  • *Correspondence: honjo@akg.t.u-tokyo.ac.jp, sasatani@g.ecc.u-tokyo.ac.jp
• DOI: 10.48550/arXiv.2510.19277

Documentation: https://sasatanilab.github.io/EMFieldGPR/

📚 Research Code: This is academic research code accompanying a paper. No ongoing support, maintenance, or feature development guaranteed.

Installation

pip install git+https://github.com/SasataniLab/EMFieldGPR.git

Quick demo

1. After installation, the following command automatically downloads ~255 MB of data and launches an interactive 3D visualization.

# Run interactive 3D visualization demo
emml demo

2. Use the sliders in the right-hand window to change both the shield shape and the coil position.

Demo interface

This shows the magnetic field and efficiency of the wireless power system.
The application has three windows:

• Right window (system geometry & results): Change coil position and shield shape and see the efficiency. Toggle vectors with the checkbox.
• Polyscope window: Adjust visualization mode. Add a plane via View → Slice Plane → Add Plane. Switch Inspect → Color Map to visualize the magnetic field on that plane. See https://polyscope.run/py/ for details.
• Structures window: Toggle visibility and change object colors.

System Requirements

• Python 3.11+

• OpenGL 3.0+ support

• ~1 GB free space for demo data

• Tested with Macbook M4 Pro

Jupyter Notebooks for full workflow

• OverallFlow.ipynb: Complete process reproduction (requires FEKO)
• GenerateTestData.ipynb: Generate test data for evaluating predictions (requires FEKO)
• Figure.ipynb: Generate paper figures

Configuration

The EMFieldML code uses a modular configuration system:

• Settings: EMFieldML.config.settings — Core parameters (dimensions, prediction points, etc.)
• Paths: EMFieldML.config.paths — Directory paths for data, models, and outputs
• Templates: EMFieldML.config.templates — File naming patterns and templates
• Logging: EMFieldML.config.logger — Logging configuration
• CADFEKO executable: Set path in .env

Full Dataset (Optional)

For complete data:

• Zenodo: https://doi.org/10.5281/zenodo.17397941

Extract all_data and test_data to the repository root. Please note that large downloads (over 30 GB) are required.

Visual Overview

System Architecture

Active Learning Process

PolyCube Mesh Generation

Development (Internal)

We do not guarantee responses to issues or pull requests. The contributor instructions below are for internal use only.

Contributor setup

# Install dependencies
poetry install

# Run demo
poetry run emml demo

Code Quality

All code must pass these checks before merging. CI runs these on pull requests.

# run tests
make test

# run linters (ruff, black, isort)
make format

Documentation

Run make docs at the repository root to build HTML documentation in docs/_build/html/.
Documentation is also built and deployed via GitHub Actions.