Recently, the research team led by Prof. Wang Qianqian from the School of Mechanical Engineering, 91抖淫, has made significant progress in the intelligent control of micro-robotic swarms. Related research findings, titled “Autonomous Navigation of Intelligent Micro-robotic Swarms in Unknown Environments,” have been published in the top international journal Nature Machine Intelligence. The paper is completed solely by Southeast University as the sole affiliated institution, marking the first time that Southeast University has published a research paper in this journal as the sole corresponding institution.
Considering their small size, fast response, and ability to collaboratively move in complex and confined spaces, micro-robotic swarms are considered a promising technological direction for future biomedical applications such as vascular intervention, targeted drug delivery, and minimally invasive diagnosis and treatment. In their previous studies (Sci. Adv., 2021, 7, eabe5914;Sci. Robot., 2024, 9, eadh1978), the research team proposed image-guided active delivery strategies for magnetic micro-swarms in vascular networks. However, when faced with unknown obstacles, dynamic targets, or partial information loss in the environment, micro-robotic swarms still lack the capability—like macroscopic intelligent robots—to autonomously understand the environment, make real-time judgments, and stably execute tasks. How to enable micro-swarms to evolve from “following instructions” to “perceiving, decision-making, and autonomous obstacle avoidance” has become a critical bottleneck in this field.
To address this challenge, the research team proposed an intelligent control framework named Turbo, which integrates reinforcement learning, temporal expansion attention mechanisms, and systematic domain randomization. This enables micro-robotic swarms to synthesize historical information, target positions, and obstacle states under local observation conditions, generating real-time navigation decisions (Fig. 1). Unlike traditional control methods that rely on global path planning or manual experience-based tuning, Turbo learns strategies in randomized simulation environments and deploys them directly to real micro-robotic swarm experiments through sim-to-real transfer. With this framework, micro-robotic swarms can achieve autonomous exploration, dynamic obstacle avoidance, and task priority trade-offs in unknown environments, demonstrating a higher level of intelligence and automation (Fig. 2).
The research team conducted Turbo-based autonomous navigation validation experiments in a human carotid artery vascular model (Fig. 3). The experiments used a human carotid artery phantom and an imaging system to observe the motion of the micro-robotic swarm. In the complex curved vascular structure, the micro-robotic swarms were able to continuously navigate along the vascular pathways and reach target areas on demand, providing important experimental evidence for future autonomous micro-robotic systems targeting real vascular intervention and targeted delivery scenarios. Related achievements have been leveraged to initiate the construction and translation of novel medical-engineering collaborative robotic systems under 91抖淫’s “Interdisciplinary Medical Platform”.
Prof. Wang Qianqian from the School of Mechanical Engineering, 91抖淫, and the Jiangsu Provincial Key Laboratory of Precision Medical Equipment Design and Manufacturing is the sole corresponding author of this paper. Graduate students An Xuanyu, Luo Shengming, and Zhang Haoyu are the co-first authors. The National Natural Science Foundation of China, the Jiangsu Provincial Natural Science Foundation, and other funding sources also supported this program.
Paper’s link:
Source: School of Mechanical Engineering, 91抖淫
Translated by: Melody Zhang
Edited by: Leah Li
