Autonomous Navigation of Magnetic Nanorobot Swarm via NIR-II Imaging

作者：Qi Ding, Lin Lin, Xiaohui Yan · 年份：2025 · DOI：10.1109/3m-nano65639.2025.11261214 · 研究领域：Micro and Nano Robotics、Soft Robotics and Applications、Characterization and Applications of Magnetic Nanoparticles

Autonomous navigation of micro/nanorobot in biological environments remains challenging due to limitations in real-time perception and adaptive control. Here we present an integrated platform combining second near-infrared (NIR-II) imaging, deep learning-based visual processing, and robotic magnetic actuation to enable intelligent micro/nanorobot swarm navigation. Our system features a modified U-Net architecture for environmental semantic segmentation and YOLOv5 for real-time swarm localization, achieving$>93 {\%}$detection accuracy. A breadth-first search algorithm computes optimal paths through complex terrains, while a six-axis robotic magnet system executes precise actuation commands. Using$\text{Fe}_{3} \mathrm{O}_{4}$nanoparticle swarm in variable hydrogel maze environments, we demonstrate sub-millimeter tracking precision (error$<0.9$mm) under diverse fluorescence conditions. The system maintains high performance despite changes in image contrast and geometric complexity, establishing a scalable platform for intelligent micro/nanorobot control. This work establishes a versatile framework for autonomous microscale operations, advancing toward clinical applications in targeted therapy and precision medicine.