医护场景下移动机器人障碍物检测和安全运行-《中国医学物理学杂志》

文章信息/Info

Title:: Obstacle detection and safe operation of mobile robots in healthcare scenarios

作者:: 付韬1; 刘元浩2; 张吟龙2; 1.中国医科大学附属盛京医院计算机中心，辽宁沈阳 110004； 2.中国科学院沈阳自动化研究所/机器人学国家重点实验室，辽宁沈阳 110016

Author(s):: FU Tao1; LIU Yuanhao2; ZHANG Yinlong2; 1. Computer Center, Shengjing Hospital of China Medical University, Shenyang 110004, China 2. State Key Laboratory of Robotics/Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang 110016, China

Keywords:: Keywords: healthcare scenario mobile robot obstacle segmentation network safe operation

摘要:: 为实现在动态医护场景下基于障碍物检测的移动机器人速度调节与安全间距保持，提出一种新型医护机器人实时障碍物分割与安全运行框架。采用紧耦合方式利用轮式编码器和单目RGB图像来计算机器人的线速度；提出了HS-EOSN模型对障碍物进行精准分割，并利用噪声滤波方法对障碍物距离进行精确识别；最后设计了一种自适应控制策略动态调整机器人速度以保证安全。该算法已在开发的移动机器人平台上进行了测试，实验结果表明：本文方法的速度估计误差小于10%，障碍物分割的全类平均精度和平均交并比分别达到96.87%和94.35%，平均深度误差为0.24 m。【关键词】医护场景；移动机器人；障碍物分割网络；安全运行

Abstract:: Abstract: To realize speed regulation and safe distance maintenance of mobile robots based on obstacle detection in dynamic healthcare scenarios, a novel real-time obstacle segmentation and safe operation framework for healthcare robots is proposed. A tightly-coupled approach utilizing wheel encoders and monocular RGB images to calculate the robots linear velocity is adopted. Concurrently, a healthcare scenario end-to-end obstacle segmentation network (HS-EOSN) model is established for precise obstacle segmentation, supplemented by a noise filtering method to accurately estimate obstacle distance. Finally, an adaptive control strategy is designed to dynamically regulate the robots speed, thereby ensuring safety. The algorithm is tested on a developed mobile robot platform. Experimental results show that the proposed method achieves a velocity estimation error below 10%. For obstacle segmentation task, the mean average precision and mean intersection over union reach 96.87% and 94.35%, respectively and the mean depth error is within 0.24 m.