【摘要】 机器人遥操作，即人类操作者通过远程控制设备操纵机器人完成任务，实现了机器人的执行能力和人类操作者在非结构化环境中的决策能力的融合，已被广泛应用于各种工作要求复杂且对人类有潜在危险的领域.高效遥操作的关键因素之一在于能否为操作者提供良好的临场感，临场感取决于操作者对远程环境的身临其境的感知(视觉、触觉等感官反馈)以及对远程机器人精准、直观的控制.本文提出了一种基于视觉定位与虚拟现实技术的机器人遥操作方法，重建并传输远程场景的静态点云模型，利用Gen6D位姿估计算法根据RGB数据自动检测远程场景中的物体，然后将检测到的物体位姿信息传输到Unity虚拟现实场景中，定位虚拟物体，从而取代物体的点云模型.充分发挥虚拟现实的固有优势，沉浸式可视化重建后的远程场景，利用Unity物理引擎驱动虚拟物体，提高操作者的临场感.其次，提出一种人机解耦的信息物理(cyber-physical)映射模型，将操作者与远程机器人的控制回路解耦，消除了时延对遥操作的影响.最后，提出一种基于虚拟现实空间的自适应位置增量映射算法，根据虚拟机器人末端抓手与目标物体的距离改变位姿映射的比例，提高了遥操作的灵活性，并赋予虚拟现实遥操作避障的功能，保障了虚拟现实遥操作的安全性.更多还原

【Abstract】 Robot teleoperation,where a human operator remotely controls a robot to perform tasks,combines the decision-making capabilities of humans with the execution capabilities of robots in unstructured environments. This approach has been widely used in various fields that involve complex tasks and potential hazards for humans. Crucial for efficient teleoperation is providing the operator with a strong sense of presence,which depends on the operator’s immersive perception of the remote environment,including visual,tactile,and other sensory feedback,as well as precise and intuitive control of the robot. In this paper,an innovative robot teleoperation method that leverages visual localization and virtual reality technology is proposed. The proposed method begins by reconstructing and transmitting a static point-cloud model of the remote scene. Using the Gen6D pose estimation algorithm,objects in the scene are automatically detected based on RGB data. The pose information of these detected objects is then transmitted to a Unity virtual reality scene,where the point-cloud models of the objects are replaced with virtual equivalents. We fully leverage the inherent advantages of virtual reality to visualize the reconstructed remote scene immersively. The Unity physics engine drives the virtual objects,improving the operator’s sense of presence. Moreover,we propose a cyberphysical mapping model with human-machine decoupling,which separates the control loops of the operator and the robot,effectively mitigating the impact of time delays on remote operations. We also propose an adaptive position incremental mapping algorithm within the virtual reality space. This algorithm adjusts the scale of pose mapping according to the distance between the virtual robot’s end gripper and the target object,thereby improving teleoperation flexibility. This also incorporates obstacle avoidance features,ensuring the safety of teleoperation within the virtual reality environment.更多还原