· Research & technology· 日本語で読む
RoboParty Publishes Instructions for Bimanual Teleoperation Software Using PICO VR
RoboParty’s GitHub repository describes bimanual teleoperation of RPO/Roboto robots using PICO VR. Based on Unitree’s xr_teleoperate, it recommends an environment with Python 3.10 and ROS 2. VR control covers the upper body, while separate robot-side software handles standing and balance control.
Key facts
- The software is based on Unitree’s xr_teleoperate and is designed for bimanual teleoperation of RPO/Roboto robots using PICO VR.
- The recommended environment is Ubuntu 20.04 or 22.04, Python 3.10 and a ROS 2 environment with a matching Python minor version.
- The default control frequency is 30 Hz, and the --headless and --profile-loop options are documented.
- The VR side handles upper-body control, while roboparty_deploy is required for standing and balance control on the physical robot.
- The repository is released under GPL-3.0-only and says it retains the Apache-2.0 designation for portions derived from Unitree.
What happened
RoboParty’s GitHub repository provides bimanual teleoperation software and setup instructions for controlling RPO/Roboto robots from PICO VR, based on Unitree’s xr_teleoperate. It lists Ubuntu 20.04 or 22.04, Python 3.10 and a ROS 2 Python environment as supported requirements, and describes a setup that converts XR data into joint commands through inverse kinematics (IK) and sends them via ROS. The system controls the upper body; separate roboparty_deploy software is required for standing and balance control on the physical robot.
Why it matters
The README’s setup means that PICO VR upper-body control alone cannot operate the physical robot: standing and balance control by roboparty_deploy, as well as authorization on both sides, are required. For developers considering deployment, the documentation shows that checking compatibility across the VR input system, ROS 2 environment and robot-side policy is part of the overall task.
Analysis
The key point about this repository is that it does not use VR to control the robot’s entire body as one system. Instead, it separates upper-body teleoperation from the robot-side policy that maintains standing and balance. The README says VR motion input is converted into joint commands through IK and sent via ROS, while roboparty_deploy handles posture maintenance on the physical robot. In other words, this implementation example depends not only on the control interface but also on integration with robot-side inference and control software.
The repository also divides the conditions for sending commands to the physical robot between an LB operation on the robot side and an A operation on the teleoperation side. It further instructs users to bring the simulated posture close to that of the physical robot before starting to send commands. The two-step process for starting control and the posture check can be understood as operational gates designed to prevent XR input from immediately reaching the physical robot. However, the README documents procedures; it does not present results evaluating the system’s accident rate or safety performance.
On the technical side, the README emphasizes matching Python 3.10 with the Python minor version used by ROS 2. It says rclpy may fail to load with different versions, indicating that software environment compatibility, as well as connectivity with the XR device, is a requirement for operating the physical robot. The requirement for an HTTPS certificate and port 8012 for the PICO browser connection also shows that network configuration is part of setting up the control system.
A key consideration is the range of supported robots and the differences in models and settings required for each platform. The README lists several policy configuration files, but does not provide information on robot-by-robot compatibility, measured control cycle times, latency, continuous operating time or physical-robot safety validation. The listed 30 Hz is a default setting and does not mean that the rate is consistently achieved on a physical robot. Areas to verify include support by model, measured control cycle times and latency, behavior when switching policies, and validation of safety functions, including emergency stops.
Details
The README recommends Python 3.10 and a compatible ROS 2 Python environment. The PICO browser connection requires an HTTPS certificate and uses port 8012. The default control frequency at startup is 30 Hz. The --headless option disables IK visualization, while --profile-loop can be used to check the time taken for XR data processing, IK, ROS transmission and other steps.
On the physical robot, the policy in roboparty_deploy handles standing and balance, while the VR side sends upper-body joint commands. The README describes a procedure in which upper-body control is enabled through both an LB operation on the robot side and an A operation on the teleoperation side. It also instructs users to bring the simulated posture close to that of the physical robot before sending commands. The repository is licensed under GPL-3.0-only and separately notes the Apache-2.0 designation for portions derived from Unitree, as well as rights information for the Atom01 URDF and other materials.
Sources
Translated from the Japanese original: /news/roboparty-pico-vr-wo-tsukau-souwan-enkakusousa-sofuto-no-tej-dd4a7d. The Japanese version is authoritative if they differ.