· Product launches· 日本語で読む
RoboParty releases ROS 2 deployment framework for RPO and RP1
RoboParty has published a ROS 2 deployment framework for RPO and RP1 on GitHub. It uses a modular architecture for motor and IMU connectivity, inference and depth perception. The guide includes deployment instructions for RPO, while instructions for RP1 are marked “Coming soon.”
Key facts
- roboparty_deploy has been released as a ROS 2 deployment framework for RPO and RP1.
- The repository’s motor-related directories include drivers for DM, EVO, LRO, RP and XYN.
- The inference component handles ONNX inference, assembling observations and loading motion data.
- The camera component integrates RealSense and publishes perception information after depth processing to ROS topics.
- A deployment guide for RPO is available, while the guide for RP1 is marked “Coming soon.”
What happened
RoboParty has published “roboparty_deploy,” a ROS 2 deployment framework for RPO and RP1, on GitHub. It includes motor drivers, an IMU interface, policy inference and depth perception, in a modular architecture that supports custom packages. The guide includes deployment instructions for RPO, while instructions for RP1 are marked “Coming soon.”
Why it matters
RoboParty’s repository lays out components for reading motor and IMU status, converting policy outputs into joint targets, and publishing depth information to ROS topics. Developers considering code and configurations that connect control and perception on physical robots can review how the functions are divided and how deployment works. However, the guide for RP1 is still marked “Coming soon,” so the scope of the documented instructions differs by model.
Analysis
A key feature of the repository is that it not only brings robot control, perception and inference together in one execution process, but also separates the functions into ROS 2 modules. It describes reading motor and IMU status, converting policy outputs into joint targets and sending them to drivers, while depth-camera processing separately provides perception information through ROS topics. The guide is structured to cover the connections, configuration, zero-point calibration and startup required for deployment on a physical robot. In practical terms, a focus of the release appears to be how far it can standardize the connection between hardware and inference, beyond simply providing a model.
However, while there is a deployment guide for RPO, instructions for RP1 are marked “Coming soon.” The fact that both models are listed as targets should therefore be distinguished from having the same scope of deployment instructions available for both. For RP1, we reported in “RoboParty launches RP1 humanoid robot, opening up hardware and software” that it is designed to open up everything from mechanical design to control, simulation and training tools. This repository can be viewed as a resource for reviewing code and instructions related to physical-robot connections, inference and sensor processing within that open offering. The repository alone, however, does not show the extent to which the two offerings are compatible.
The architecture appears to treat robot-specific differences separately through motor and IMU drivers, as well as model-specific hardware configurations, policies, models and motion data. The repository says it accepts custom packages, leaving room for users to add functions, but does not specify the scope of compatibility or the amount of modification required. The GPL-3.0-only license applies to the repository’s original works; submodules, bundled binaries and third-party materials remain subject to their own terms. Use and redistribution therefore cannot be treated as subject to a single set of license terms for the entire repository.
The next points to check are which functions will actually be available once the RP1 guide is published, the scope of supported hardware configurations and policies, and the operating conditions for the camera, inference and motor control on physical robots. The repository does not provide operational track records or performance figures for these. How far the code release provides a shared development and deployment environment for RPO and RP1 will need to be assessed by examining model-specific configurations and implementation updates.
Details
The repository includes motor-control drivers for DM, EVO, LRO, RP and XYN, and IMU drivers for HIPNUC and MCT7123. The inference component handles ONNX inference, assembling observations and loading motion data. It receives motor and IMU status through a robot interface, converts policy outputs into joint targets and sends them to motor drivers.
The camera component supports RealSense and depth processing, and publishes perception information to ROS topics. The guide covers connecting the RPO controller, environment setup, a real-time kernel, wireless LAN access point configuration, wiring, zero-point calibration, and starting the policy and camera. It also describes YAML configuration, building, logs, gamepads, services and the Python SDK. The repository’s original works are licensed under GPL-3.0-only unless otherwise noted.
Sources
Translated from the Japanese original: /news/roboparty-rpo-to-rp1-muke-ros-2-tenkai-fureemuwaaku-wo-kouka-dcab80. The Japanese version is authoritative if they differ.