Colloidal Magnetoelectric Shape Recognition Based on Machine Learning

Functionalized particles ranging from nanoscale to microscale and their assemblies have facilitated a wide variety of sensing concepts, from molecular-scale chemical and biological detection to large-scale engineering defect testing. Related to macroscopic object shape sensing, visual recognition is generally the most versatile approach whenever possible. However, under certain conditions where visual perception is hindered, for example, dark space or underwater, electrosensing can serve as an alternative sensation manner. Inspired by this concept, the sensing of rudimentary object shapes using electrically conductive, soft ferromagnetic Ni particles is demonstrated, herein denoted as colloidal magnetoelectric shape recognition. By confining the target and sensory particles between two planar electrodes and using a magnetic field to drive the particles toward object edges, changes in electrical conductivity are monitored. Machine learning is then used to resolve the exact object shapes with high fidelity. This study introduces a colloidal magnetoelectric shape recognition strategy for short-range shape sensing, with potential applications suggested for the fields such as soft robotics, drug delivery, and biomedical diagnostics.