摘要
磁性微型机器人在生物医学中具有较小的侵入性,且易于通过微窄的环境,近些年来成为一个新兴的领域。然而,设计可以应用在复杂环境中的微型机器人自主导航控制系统仍旧是一个挑战。提出了一个基于显微视觉反馈的微型机器人自主导航控制方案,设计了语义分割模型以构建显微视野下的全局语义地图,然后改进了RRT-connect路径规划算法,并基于强化学习设计了轨迹跟踪控制模型以跟随规划路径。最后,搭建实验平台进行了微型机器人自主导航的实验,结果证明:所提方案能够在复杂的环境下构建环境语义地图,并实现路径的规划与导航控制。
Magnetic microrobots have become an emerging field because they are less invasive in biomedicine and easy to pass through a narrow environment.However, it is still a challenge to design autonomous navigation control systems for microrobots in complex environments.An autonomous navigation control scheme for microrobots based on microscopic visual feedback is proposed.Firstly, a semantic segmentation model is designed to build a global semantic map under the microscope.Then, RRT-connect path planning algorithm is improved and a trajectory tracking control model based on reinforcement learning are designed to follow the planned path.Finally, an experimental platform is setup to conduct an experiment of autonomous navigation of a magnetic microrobot.The result proves that the solution can construct a global semantic map in a complex environment, and realize path planning and navigation control.
作者
张鹏松
樊启高
于振中
ZHANG Pengsong;FAN Qigao;YU Zhenzhong(School of IOT Engineering,Jiangnan University,Wuxi 214122,China;Harbin Institute of Technology Robotics(Hefei)International Innovation Research Institute,Hefei 230011,China)
出处
《传感器与微系统》
CSCD
北大核心
2021年第6期11-15,共5页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(51405198)
江苏省“六大人才高峰”高层次人才项目(GDZB-138)
江苏省研究生科研与实践创新计划项目(KYCX201935)。
关键词
磁性微型机器人
地图构建
路径规划
强化学习
magnetic microrobot
map construction
path planning
reinforcement learning
