液体环境中的微纳机器人普遍具有高效的运动能力，已在光学和生物医学等领域展现出广阔应用前景，如衍射光学成像、体内药物递送和微创医疗操作等。然而在陆地环境中，微尺度表面粘附力和摩擦力的显著波动使得微纳机器人的定向可控运动面临巨大挑战。
近日，复旦大学智能机器人与先进制造创新学院/智慧纳米机器人与纳米系统国际研究院的梅永丰教授、崔继斋青年研究员团队报道了一种新型陆上微纳机器人，该机器人依靠三维薄膜的非互易变形产生稳定可控的非对称力，并结合AI强化学习算法实现了精准运动控制。相关成果以研究论文在《美国科学院院刊》（Proceedings of the National Academy of Sciences of the United States of America）上发表，题为《基于非互易变形的三维纳米薄膜作为微纳机器人实现陆上运动》（“Terrestrial locomotion of microscopic robots enabled by 3D nanomembranes with nonreciprocal shape morphing”）。
研究团队首先通过理论证明，非互易变形能够克服表面力的波动，实现高效的定向运动。在此基础上，团队采用自卷曲工艺将二氧化钒纳米薄膜构造成三维微纳机器人，并利用激光驱动与二氧化钒固有的相变滞后特性协同实现非互易变形。实验表明，该机器人可在硅片、纸张、砂纸、树叶乃至垂直墙面等多种陆地表面实现可靠的定向运动。此外，微纳机器人的外形可按需设计为方形、圆形以及仿生蜘蛛、瓢虫等多种形态（图1），并均可在高频激光的精准驱动下实现全向运动，例如微机器人顺利穿过迷宫环境。
图1 自卷曲二氧化钒陆上微纳机器人基于非互易变形在陆地表面实现运动。图中展示了仿生形状为蜘蛛与瓢虫形的微纳机器人实现定向可控运动。
团队进一步展示了该陆上微纳机器人的潜在应用，并构建了基于人工智能算法的微纳机器人操纵系统。通过对微纳机器人集群的控制自组装，形成预设图形，可能用于显示与芯片实验室(Lab on a chip)应用中；通过利用二氧化钒相变过程中的电阻变化，令微纳机器人导通电路从而点亮灯泡，为柔性电子提供新思路。此外，研究团队基于强化学习的人工智能控制算法，通过配合振镜系统对激光位置进行高速切换，可控制不同形状的微纳机器人朝目标方向运动(图2A-B)。利用训练后的AI控制系统，微纳机器人可自主遵循复杂的轨迹运动，例如书写“hello world”等字样（图2C）。
图2 陆上微纳机器人的人工智能控制。(A)人工智能控制系统。(B)基于强化学习算法使圆形机器人朝目标移动。 (C) 机器人自主书写“hello world”轨迹。
综上所述，该研究奠定了陆地微纳机器人的理论和技术基础，为其在纳米光子、微电子和生物医学等前沿应用的发展铺平了道路。
论文第一作者为汪洋博士，通讯作者为梅永丰教授、崔继斋青年研究员。该工作得到了国家重点研发计划、国家自然科学基金、上海市科委等项目的资助和支持，部分实验在复旦大学微纳加工与器件公共实验室开展，部分理论计算工作基于复旦大学CFFF平台开展。
文章信息：
Yang Wang, Xing Li, Chang Liu, Yunqi Wang, Chunyu You, Hong Zhu, Zhi Zheng, Ziyu Zhang, Guobang Jiang, Xiang Dong, Tianjun Cai, Ziao Tian, Zengfeng Di, Gaoshan Huang, Xiangzhong Chen, Enming Song, Jizhai Cui*, Yongfeng Mei*, Terrestrial locomotion of microscopic robots enabled by 3D nanomembranes with nonreciprocal shape morphing, Proceedings of the National Academy of Sciences of the United States of America, 2025, 122(25), e2500680122.
文章链接：
http://www.pnas.org/doi/10.1073/pnas.2500680122
Terrestrial locomotion of microscopic robots enabled by 3D nanomembranes with nonreciprocal shape morphing
Microscopic robots exhibit efficient locomotion in liquids by leveraging fluid dynamics and chemical reactions to generate force asymmetry, thereby enabling critical applications in photonics and biomedicine. However, achieving controllable locomotion of such robots on terrestrial surfaces remains challenging because fluctuating adhesion on nonideal surfaces disrupts the necessary asymmetry for propulsion. Here, we present a microscopic robot composed of three-dimensional nanomembranes, which navigate diverse terrestrial surfaces with omnidirectional motion. We propose a general mechanism employing nonreciprocal shape morphing to generate stable asymmetric forces on surfaces. This nonreciprocal shape morphing is realized through a laser-actuated vanadium dioxide nanomembrane, leveraging the material's inherent hysteresis properties. We demonstrate that these robots can be fabricated in various shapes, ranging from simple square structures to bioinspired bipedal helical designs, enabling them to directionally navigate challenging surfaces such as paper, leaves, sand, and vertical walls. Furthermore, their omnidirectional motion facilitates applications in microassembly and microelectronic circuit integration. Additionally, we developed an artificial intelligence control algorithm based on reinforcement learning, enabling these robots to autonomously follow complex trajectories, such as tracing the phrase hello world. Our study lays a theoretical and technological foundation for microscopic robots with terrestrial locomotion and paves a way for microscopic robots capable of operating on surfaces for advanced nanophotonic, microelectronic, and biomedical applications.
Article information:
Yang Wang, Xing Li, Chang Liu, Yunqi Wang, Chunyu You, Hong Zhu, Zhi Zheng, Ziyu Zhang, Guobang Jiang, Xiang Dong, Tianjun Cai, Ziao Tian, Zengfeng Di, Gaoshan Huang, Xiangzhong Chen, Enming Song, Jizhai Cui*, Yongfeng Mei*, Terrestrial locomotion of microscopic robots enabled by 3D nanomembranes with nonreciprocal shape morphing, Proceedings of the National Academy of Sciences of the United States of America, 2025, 122(25), e2500680122.
Article link:
http://www.pnas.org/doi/10.1073/pnas.2500680122