基于示教学习和视线跟踪的机器人遥操作系统设计  

Design of Robot Teleoperation System Based on Learning from Demonstration and Gaze Tracking

作　　者：胡皓 柴馨雪 HU Hao;CHAI Xinxue(School of Mechanical Engineering,Zhejiang Sci-Tech University,Hangzhou 310018,China)

机构地区：[1]浙江理工大学机械工程学院,浙江杭州1310018

出　　处：《轻工机械》2024年第6期1-10,22,共11页Light Industry Machinery

基　　金：浙江省重点研发计划项目(2021C04017)。

摘　　要：针对传统的机器人遥操作系统缺乏对任务的认知和操作者主观性判断的问题,课题组提出了一种基于示教学习与视线追踪的遥操作系统设计策略。首先运用高斯混合模型(Gaussian mixture model,GMM)和高斯混合回归(Gaussian mixture regression,GMR)对机器人的技能进行学习,实现人机技能转移;然后基于示教学习设计了一种人工势场法的虚拟夹具(virtual fixtures,VF),并设计了力反馈算法;最后使用网络摄像头对操作者的视觉意图进行判别,将意图识别结果和人工势场法应用于虚拟夹具力反馈算法中,并进行实验。研究结果表明:基于视觉意图判别的虚拟夹具力反馈遥操作系统能够对操作者的意图进行判别,有效解决了操作者的主观意图受传统虚拟夹具产生的反馈力限制的问题,增强了操作者在遥操作过程中的主观性,是遥操作控制系统中一种有效且实用的力反馈控制方案。研究结果可为高危场景下具有力觉临场感的机器人遥操作系统的实际应用提供参考。Aiming at the problem of traditional teleoperation system for medical robots,which lacks the cognition of task and the judgment of operator's subjectivity,a teleoperation system design strategy based on learning from demonstration and gaze tracking was proposed.Firstly,Gaussian mixture model(GMM)and Gaussian mixture regression(GMR)were applied to learn the robot's skills and realize the transfer of human-robot skills.Then,a virtual fixtures(VF)with artificial potential field method based on learning from demonstration was designed,and a force feedback algorithm was designed.Finally,a webcam was used to discriminate the operator's visual intention,and the intention recognition results and the artificial potential field method were applied to the virtual fixtures force feedback algorithm,and experiments were conducted.The research results show that the virtual fixtures force feedback teleoperation system based on visual intention discrimination can discriminate the operator's intention,effectively solve the problem that the operator's subjective intention was limited by the feedback force generated by the traditional virtual fixture,and enhance the operator's subjectivity in the process of teleoperation.It is an effective and practical force feedback control scheme in the teleoperation control system.The research results can provide a theoretical basis for the practical application of robot teleoperation system with the force-aware presence sensing motion in high-risk scenarios.

关 键 词：遥操作系统 示教学习 高斯混合模型 力反馈 意图识别 

分 类 号：TH113.2[机械工程—机械设计及理论] TP274[自动化与计算机技术—检测技术与自动化装置]