复杂环境下磁弹性微型游泳机器人的路径规划与识别跟踪

磁弹性材料设计的毫米级微型游泳机器人可通过外部均匀磁场实现连续形变来完成游泳动作，通过视觉反馈能提高机器人的路径跟随精度，但在复杂背景下机器人的视觉闭环控制易发生识别错误、跟踪失败等现象。针对上述问题，首先获取复杂环境下的障碍物信息，提出改进RRT*算法(IIC-RRT*)进行路径规划，同时基于环形平滑标签的YOLOv5识别跟踪算法(CSL-YOLOv5)，在复杂背景下对微型游泳机器人的中心位置与旋转角度进行实时更新。在此基础上进行微型游泳机器人在障碍物复杂背景下的位置与角度的双闭环伺服控制，试验结果表明，提出的路径规划算法改善了路径生成的效率与平滑性，识别跟踪算法提高了微型游泳机器人的识别稳定性与精确性，为磁控微型游泳机器人在复杂背景下的精确控制提供了新思路。

Path Planning and Recognition Tracking of a Magnetoelastic Miniature Swimmer in Complex Environment

The miniature swimmer made of magnetic elastic material can achieve continuous deformation through the external uniform magnetic field to complete the swimming action. Despite visual feedback improving path tracking accuracy, visual closed-loop robot control is prone to recognition errors and tracking failures under complex backgrounds. Aiming at the above problems, firstly, the obstacle information in complex environments is obtained, and the improved RRT* algorithm (IIC-RRT*) is proposed for path planning. In addition, YOLOv5 based on circular smooth label (CSL-YOLOv5) recognition and tracking algorithm is used to update the central position and rotation angle of miniature swimmers in real time under complex conditions. Using this method, the miniature swimmer is controlled via dual closed loop servos with position and angle against a complex obstacle background. According to the experimental results, the proposed path planning algorithm improves the efficiency and smoothness of the path generation process, while the recognition and tracking algorithm improves the recognition stability and accuracy of the miniature swimmer. This work may provide an innovative idea for the precision control of the magnetic miniature swimmer under complex backgrounds.