AI-guided robotic in situ fabrication of curvature- and tissue-adaptive wireless bioelectronics

Hiroyuki Tetsuka, Jiaju Ma, Naomi Kumano, Minoru HiranoarXiv 2026 · arXiv

TL;DR

Researchers have developed an AI-guided robotic platform that doesn't just print electronics onto the body, but actually redesigns them on the fly to account for a patient's unique anatomy. By integrating 3D vision, deep learning, and electromagnetic optimization, the system fabricates wireless electrotherapy devices directly onto wound dressings, ensuring they maintain peak performance despite the complex curves and electrical properties of human tissue.

What Problem Does This Paper Address?

Bioelectronic "smart bandages" hold immense promise for accelerating wound healing through localized electrical stimulation. However, moving these devices from the lab bench to the patient’s body presents a major engineering hurdle: detuning.

Most wireless bioelectronics are designed and optimized for flat, air-filled environments. When you wrap a flexible coil around a curved limb or place it against conductive human tissue, its electrical properties—specifically its resonant frequency—shift significantly. This "detuning" can render wireless power transfer inefficient or non-functional. Current state-of-the-art methods typically use 3D scanning only to help a robot follow the "path" of a surface. This paper goes a step further, asking: How can we use anatomical data to fundamentally redesign the device's circuitry before the first drop of ink is even laid down?

Methods

The authors propose a "perception-to-design-to-fabrication" framework. The workflow is entirely automated and integrated under Python control:

  1. Perception: A 3D vision camera captures RGB and depth data of a wound area covered by a transparent polyurethane dressing.
  2. AI Segmentation: A deep learning model (YOLO-seg) identifies the wound boundaries and calculates its dimensions (length, width, area). Simultaneously, the system estimates the local radius of curvature from the 3D point cloud.
  3. Geometry-Aware Design: These physical parameters are fed into a MATLAB-based RF optimization module. The module adjusts the number of turns in the receiving coil and the configuration of the interdigitated electrodes (IDEs) to ensure the device resonates at exactly 25 MHz, accounting for both the surface curve and the "dielectric loading" (electrical interference) of the underlying tissue.
  4. Robotic Fabrication: A six-degree-of-freedom (6-DOF) robotic arm uses direct-ink-writing (DIW) to print a multilayer device—consisting of a silver conductive ink and an insulating layer—directly onto the dressing.
The integrated automated workflow, from 3D scanning to robotic in situ printing.

Datasets

To enable the vision system, the researchers created a custom wound-image dataset. This included 16 different wound phantoms with varying shapes, sizes, skin tones, and curvatures. The model was trained to provide pixel-level segmentation masks, which were then registered with depth data to provide high-accuracy physical measurements.

Key Insights and Ideas

  • Geometry as a Design Variable: The most significant "aha moment" is treating anatomical curvature not just as a physical obstacle for the robot to avoid, but as a critical input for the electrical engineering of the device.
  • The "On-Dressing" Strategy: Instead of printing on a flat substrate and then trying to stick it to a patient (which introduces mechanical stress and potential misalignment), the researchers print directly onto a pre-applied dressing. This ensures the device is "born" in its final, functional shape.
  • Material Synergy: The team developed a specialized silver ink (silver flakes + nanoparticles + PEO) that exhibits shear-thinning behavior. This means it flows easily through a needle under pressure but stays put once deposited, maintaining electrical continuity even when stretched or compressed by 50%.
RF design optimization showing how the Q-factor and resonant frequency are balanced based on the IDE finger number and coil turns.

Key Results

  • Precision Tuning: On an anthropomorphic forearm phantom, the system achieved a resonant frequency of 25.68 MHz, a mere 2.07% deviation from the simulated target of 25.16 MHz. Without this geometry-aware optimization, the device would have been significantly detuned.
  • Speed: The entire computational process—from scanning to generating the optimized design and robot path—takes only about 65 seconds. The total workflow, including four-pass printing and drying, takes roughly 11 minutes.
  • Robustness: The AI-based vision system was significantly more reliable than traditional color-based methods, maintaining a 100% detection rate across various lighting temperatures (from 2,700 K to 6,500 K), whereas color-based methods dropped as low as 12.5% detection in warm light.
Experimental validation on a tissue-equivalent forearm phantom, showing successful wireless signal transfer.

Limitations and Future Work

While a major step forward for "Physical AI," several challenges remain:

  • Static Targets: The current system is "open-loop" during the printing phase. If the patient moves their arm after the initial scan, the robot won't automatically adjust. Future versions will need real-time visual tracking or force feedback.
  • Biological Complexity: The study used tissue-equivalent phantoms. Real human skin is more complex, involving sweat, hair, and spatially heterogeneous dielectric properties that change over time.
  • Drying Time: Drying the ink currently accounts for 62% of the total workflow time. Improving ink chemistry or adding localized curing (like UV or heat) could bring the total time down to just a few minutes.

This work provides a foundational framework for a future where medical devices aren't just "one size fits all," but are computationally designed and robotically manufactured on-demand, right at the patient's bedside.

Summary updated

Abstract

Personalized bioelectronic devices for wound therapy require not only conformal fabrication on complex body surfaces but also functional adaptation of their electrical characteristics to individual wound geometries. Here, we present an AI-guided robotic in situ printing platform for fabricating wireless wound electrotherapy devices directly on transparent wound dressings. The system integrates a 3D vision camera, a six-degree-of-freedom robotic arm, direct-ink-writing, and geometry-aware RF design optimization. AI-based wound segmentation is used to determine wound dimensions and local surface curvature, while the reconstructed three-dimensional geometry is incorporated into the electromagnetic design of the receiving coil and interdigitated electrodes. In contrast to conventional approaches that use surface geometry only for trajectory generation, the proposed method accounts for both anatomical curvature and tissue dielectric loading to compensate for geometry-induced shifts in resonant behavior. A printable silver ink was characterized in terms of rheology, dimensional reproducibility, electrical resistance, and electromechanical stability, and was used to fabricate multilayer wireless devices directly on curved polyurethane dressings. Wireless power transfer, impedance characteristics, and electric-field delivery were evaluated experimentally and numerically. Finally, robotic in situ fabrication was demonstrated on an anthropomorphic tissue-equivalent forearm phantom, where the measured dielectric properties of the phantom were incorporated into the RF optimization, enabling resonant operation near the target frequency of 25 MHz. This work establishes a perception-to-design-to-fabrication framework for geometry-adaptive wireless bioelectronics and provides a technological foundation for future physical-AI-enabled manufacturing of personalized on-body electrotherapy devices.
BibTeX
@misc{tetsuka2026ai,
  author = {Hiroyuki Tetsuka and Jiaju Ma and Naomi Kumano and Minoru Hirano},
  title = {{AI-guided robotic in situ fabrication of curvature- and tissue-adaptive wireless bioelectronics}},
  year = {2026},
  eprint = {2611.17638},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2611.17638},
}
Figure