搜维尔科技提供完整的人形机器人解决方案以及训练系统

问题：从灵巧手收集的数据是否也会在大脑大模型中训练，或是在专门用于手部控制的单独模型中训练？

Q: If the data collected from dexterous hands will be trained as well in the brain large model, or in a separate model dedicated for hands control？

A：Manus glove finger data can be transported. 

答：可以传输Manus手套手指数据。

问题：在整个应用链路中除了人形机器人设备外，Xsens是否只提供一套动捕设备？是否拥有一套完整的解决方案以及训练系统？

Q: Does Xsens only provide one set of motion capture devices in the entire application chain, apart from humanoid robot devices? Do you have a complete solution and training system?

A：We supply a motion capture system, which data can be transferred to your own training system.

答：我们提供运动捕捉系统，数据可以传输到您自己的训练系统。

问题：遥操作与动作捕捉之间是何关系，两者是否是同一件事？

Q: What is the relationship between teleoperation and motion capture, and are they the same thing?

A：Motion Capture is capturing movement (motion) data, when live streamed you can control a robot with it. This is called teleoperation.

答：运动捕捉是捕捉运动（运动）数据，当（数据）实时流传输时，你可以用它来控制机器人。这被称为遥操作。

问题：请介绍一下关于XSENS 的用户隐私保护政策？

Q: Could you please introduce the user privacy protection policy of XSENS?

A： see file, sent by email

答：请查看通过电子邮件发送的文件

问题：在生产工位将真人操作的动捕关节数据用于人形机器人训练，除了手臂和灵巧手的训练，能否可以把对零件的视觉识别结合起来，例如将头显数据，形成视觉识别输入到关节操作输出的闭环。上述输入、输出关系是否需要通过SDK自己编程实现？

Q: Can we combine visual recognition of parts, such as using headset data, to form a closed loop of visual recognition input to joint operation output, in addition to training arms and dexterous hands, by using human operated motion capture joint data for humanoid robot training at the production station. Do the above input and output relationships need to be implemented through SDK programming by myself?

A：YES

答：是的

问题：真人和机器人运动结构有较大差别。如何保证真人的关节数据可直接用于机器人关节控制？能否只能针对单机械臂进行运用，VR应用中因延时导致的用户眩晕问题如何解决？

Q: There is a significant difference in the motion structure between humans and robots. How to ensure that human joint data can be directly used for robot joint control? Can it only be applied to a single robotic arm? How to solve the problem of user dizziness caused by delay in VR applications?

A：the latency of our system is 20ms, this ensures fast control of the robot. This can be a single arm but also a full humanoid robot. By switching off the head sensor of the motion capture system and use the internal IMU data of the head mounted display.

答：我们系统的延迟为20ms，这确保了机器人的快速控制。可以是单臂机器人，也可以是全人形机器人。通过关闭运动捕捉系统的头部传感器并使用头戴式显示器的内部IMU数据。