SIA OpenIR  > 机器人学研究室
极端环境遥操作机器人多模人机交互
其他题名Multimodal Human-Robot Interaction for Extreme Environmental Telerobot
魏青1,2
导师赵忆文
分类号TP242
关键词极端环境遥操作 时变时延 多模交互信息增强重建 人机交互
索取号TP242/W55/2018
页数125页
学位专业模式识别与智能系统
学位名称博士
2018-05-19
学位授予单位中国科学院沈阳自动化研究所
学位授予地点沈阳
作者部门机器人学研究室
摘要本文从人机交互过程出发,提出了一种基于多模交互信息增强重建的遥操作机器人人机交互方法。该方法通过对反馈信息补全、交互信息预测和预测信息表达这三个方面的研究来补偿通讯链路对主从交互过程的影响,从而提高了遥操作机器人对复杂任务和精细操作的适应能力。视觉和力觉是操作者感知环境的主要手段,因此本文首先对环境在线建模方法和虚拟机器人接触力计算方法进行了研究,以解决受传感器和通信链路限制所导致的关键反馈信息缺失问题,并为下一步多模交互信息的预测奠定基础。在环境几何模型构建方面,本文利用结构光点云对从端环境进行在线建模,从而保证了虚实环境的一致性,并提出一种自适应随机抽取一致性算法对点云数据进行去噪,而且还对点云的配准和网格化方法进行了研究。在接触力计算方面,提出了一种基于机器人动力学模型的主动式接触力计算方法,不仅解决了刚性接触任务中机器人接触力的计算问题,而且配合环境动力学参数还可以很方便的推广到非刚性接触任务场合。实验表明所提算法具有较高的实时性,且可通过主手运动滤波和反馈力变刚度渲染在人机协作中实现机器人与环境的稳定接触。其次,本文对大时变时延遥操作系统中多模交互信息的预测方法进行了研究,并提出一种对主从交互信息的畸变现象具有补偿功能的预测显示算法。文中首先对基于预测显示的遥操作机器人系统进行动力学分析,并把时延抖动看作是系统的未知扰动;然后分别对系统通讯时延和操作者的前臂运动进行了分析,并根据二者的产生机理和统计规律设计了相应的预测方法;最后利用二者的预测结果在整个遥操作系统中对时变时延引起的系统扰动进行前馈补偿,并在仿真平台上对所提算法进行了实验验证。实验结果表明,在大时变时延遥操作系统中本文提出的预测显示方法可显著提高对从端机器人速度和接触力的预测精度。再次,为了提高人机协作的效率和质量,本文对遥操作系统中多模交互信息的表达方法进行了研究。在力觉和运动信息表达上,本文采用了自行设计的通用型6自由度力反馈主手,并根据主手的物理结构特点和运动学约束提出一种基于简化模型的动力学补偿算法对其自身的动力学特性进行补偿,实验表明该算法可有效减轻操作者的体力负担,且能够保证力觉交互的实时性;在视觉信息表达方面,本文选择了具有深度临场感的头盔显示器来向操作者提供预测图形,为了最小化运动画面延时对操作者的影响,提出了一种对运动画面延时抖动具有补偿功能的卡尔曼滤波算法对用户头部位姿进行预测,文中先对运动画面延时进行分析和预测,然后根据头部运动的相关性分析结果选择合适的位姿预测方式,并在头部运动建模和位姿预测时分别利用延时预测结果对延时抖动进行了补偿,实验结果表明所提算法在多种运动模式下均可显著提高用户头部位姿的预测精度。最后,为了验证本文提出的遥操作机器人人机交互方法的有效性,搭建了具有多模交互信息增强重建功能的遥操作平台框架,并对其软件平台实现过程中要解决的关键问题进行了研究。然后分别在空间遥操作仿真系统和核聚变堆维护遥操作仿真系统中对所提方法进行了验证。实验结果表明:在大时变时延遥操作系统中,本文方法可显著提高接触任务的执行效率和操作安全性;即便是在时延较小的地面遥操作系统中,本文方法也可有效提高精细操作任务的执行效率。
其他摘要In this research, the incompletion, delay and distortion of the interaction information is compensated by certain interactive information completion, prediction and expression methods. Therefore, the efficiency and operation safety of fine and complex teleoperation tasks are significantly improved. Visual and force perception are the main means for the operator to perceive the environment, so the environment online modeling method and the virtual contact force calculation method are firstly studied in this dissertation, which not only compensate the incompletion of the feedback information caused by the limitation of sensor and communication link, but also lay the foundation for the multimode interactive information prediction. When constructing the environment geometric models, the structure light is used to modeling the slave environment online, and an adaptive sample consensus is proposed to denoise the structure light point cloud, and the registration and grid method of point cloud have also been studied. In the calculation of contact force, an active contact force calculation method based on robot dynamics model is proposed, which not only solves the problem of contact force calculation in the rigid contact task, but also can be easily extended to the non-rigid contact task with the environment dynamic parameters. The experiments show that the algorithm has high real-time performance, and the stable contact between the robot and the environment can be achieved through the master hand motion filtering and feedback force variable stiffness rendering in human-telerobot collaboration. Secondly, the prediction method for multimode interactive information in large time-varying delay teleoperation system is studied, and a predictive display algorithm with compensation function for the distortion of master-slave interaction information is presented. Firstly, the dynamic model of the predictive display based teleoperation system is presented, and the delay jitter is regarded as the unknown disturbance of the system. Then the communication delay and the forearm motion of the operator are analyzed respectively, and the corresponding prediction method is designed according to their mechanism and statistic law. Finally, the feedforward compensation for the disturbances is carried out in the whole teleoperation system by using the predicted results. The experimental results show that the proposed method can significantly improve the prediction accuracy of the speed and contact force of the robot in the large time-varying delay teleoperating system. Thirdly, in order to improve the efficiency and quality of the human-telerobot collaboration system, the expression method of multimode interactive information is studied. In the expression of force and motion information, a general-purpose 6-DOF force feedback handle is designed, and a dynamic compensation algorithm based on the simplified model is proposed to compensate the dynamic characteristics of the manipulator according to its structural characteristics and kinematics. In the aspect of visual information expression, the helmet-mounted display is selected to provide the operator with predictive graphics. And in order to minimize the impact of motion to photon latency, a kalman filter algorithm with motion to photon latency jitter compensation function is proposed to predict the user's head pose in this dissertation. Finally, in order to verify the effectiveness of the proposed human-telerobot interaction method, a universal teleoperation platform with multimode interactive information enhancement function is built. And the algorithm is verified in the space teleoperation simulation system and the nuclear fusion reactor maintenance teleoperation simulation system respectively. And the experimental results show that the proposed method can significantly improve the execution efficiency and operation security of the contact task in the large time-varying delay teleoperation system. And even in the ground teleoperation system with small time delay, the proposed method can also effectively improve the execution efficiency of the fine operation task.
语种中文
产权排序1
文献类型学位论文
条目标识符http://ir.sia.cn/handle/173321/21815
专题机器人学研究室
作者单位1.中国科学院沈阳自动化研究所
2.中国科学院大学
推荐引用方式
GB/T 7714
魏青. 极端环境遥操作机器人多模人机交互[D]. 沈阳. 中国科学院沈阳自动化研究所,2018.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
极端环境遥操作机器人多模人机交互.pdf(6201KB)学位论文 开放获取CC BY-NC-SA请求全文
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[魏青]的文章
百度学术
百度学术中相似的文章
[魏青]的文章
必应学术
必应学术中相似的文章
[魏青]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。