面向危险作业的人形机器人遥操作系统设计

Design of a humanoid robot teleoperation control system for hazardous manufacturing operations

  • 摘要: 针对高压电柜按钮人工作业风险高、操作环境危险等问题,文章面向高压电柜按钮作业场景,在已有视觉遥操作方法基础上,设计了一种人形机器人遥操作控制系统。为实现操作者手部动作到人形机器人双臂及灵巧手的实时协同控制,进一步提出了一种基于头戴式视觉设备的非接触式手—臂一体遥操作方法。该方法采用视觉设备获取操作者手腕位姿和手指运动信息,通过双臂逆运动学求解、灵巧手运动重映射和手势触发的多状态切换机制,实现操作者上肢动作到机器人双臂末端及灵巧手关节的控制转换。实验结果表明,该系统主循环频率达到90 Hz,控制延迟小于30 ms,末端操作误差小于3 mm,能够满足高压电柜单按钮按压任务对实时性和定位精度的要求。与已有面向人形机器人上半身遥操作的研究达到的50 Hz实时控制频率相比,该系统在主循环频率方面达到90 Hz,并进一步结合灵巧手末端精度测试完成了按钮按压任务验证,为人形机器人上半身遥操作应用于高压电柜按钮作业提供了一种可行方案。

     

    Abstract: For the high risks and hazardous operating environment associated with manual operation of buttons on high-voltage electrical cabinets, based on the existing visual remote operation method, a teleoperation control system for humanoid robots was developed for high-voltage electrical cabinet button operation. To achieve real-time coordinated control from the operator's hand motions to the dual arms and dexterous hands of the humanoid robot, a non-contact hand-arm integrated teleoperation method based on a head-mounted vision device was proposed. In this method, the wrist pose and finger motion information of the operator were acquired by the vision device. The control transformation from the operator's upper-limb motions to the robot dual-arm end-effectors and dexterous hand joints was realized through dual-arm inverse kinematics, dexterous hand motion remapping, and a gesture-triggered multi-state switching mechanism. The experimental results showed that the main control loop frequency reached 90 Hz, and the control latency was less than 30 ms. The robot was able to complete the button-pressing task in a high-voltage electrical cabinet. Compared with the current system, the proposed system has certain advantages in non-contact motion acquisition, hand-arm coordinated mapping, and high-frequency real-time control. It can improve the teleoperation adaptability and safe operation capability of humanoid robots in hazardous electrical operation scenarios.

     

/

返回文章
返回