· Research & technology· 日本語で読む
RoboParty publishes RP1 mechanical design through training on GitHub
RoboParty has published the hardware, software and training-related repositories for its humanoid robot, RP1, on GitHub. The listed supported configurations include models with 24, 28, 30 and 42 degrees of freedom. The development platform connects mechanical design with deployment and training, but RP1 support varies by project.
Key facts
- The listed RP1 configurations include variants with 24, 28, 30 and 42 degrees of freedom.
- Published materials include mechanical design files, schematics, board layouts, Gerber files, a BOM and assembly instructions.
- The deployment framework for RPO and RP1 provides modules for IMUs, motors, inference and cameras accessible from C++ and Python.
- The training work environment uses Isaac Lab and covers workflows for AMP, BeyondMimic and Parkour.
- IBFM, PredActor and INTACT-JEPA are listed as “coming soon” for RP1 support.
What happened
RoboParty has published an open-source development platform for its humanoid robot, RP1, on GitHub. It covers mechanical and electrical design, firmware, control, robot models, deployment and training as an integrated development stack, and lists supported configurations with 24, 28, 30 and 42 degrees of freedom.
Why it matters
RoboParty's published materials include RP1 mechanical designs, a bill of materials (BOM) and assembly instructions, as well as a low-level SDK, models and deployment modules. The setup gives developers assembling the robot and users building control and training software a way to examine the connection between design information and implementation. However, RP1 support for training projects needs to be checked individually.
Analysis
The key point is that the release includes not only RP1 hardware data but also assembly, low-level control, model descriptions, a deployment SDK and training environments. These are in separate repositories but connected as a development stack. The BOM and assembly instructions relate to reproducing and modifying the robot, while the Linux BSP and USB-CAN bridge provide interfaces to hardware control. The inclusion of URDF and MJCF models, deployment modules and training workflows also makes public a path from robot design data to control and training.
We reported on product configurations including the standard 24-degree-of-freedom version and the 30-degree-of-freedom Pro version in RoboParty launches humanoid robot RP1 with hardware and software made public. The repository lists configurations with 24, 28, 30 and 42 degrees of freedom, covering more model configurations in the development platform than the two included in the product announcement. However, the list alone does not establish the shipping status or specification differences of each configuration. The deployment framework covered in RoboParty releases ROS 2 deployment framework for RPO and RP1 is presented here as SDKs for IMUs, motors, inference and cameras.
For developers, an important feature is the ability to examine not only mechanical designs and electronic circuits but also to compare robot descriptions with control interfaces and training projects. However, the software is not uniformly available for RP1. IBFM, PredActor and INTACT-JEPA are described as planned for RP1 support, while training environments such as AMP are described as being for RPO. Listing a project does not in itself mean it has been verified to run on RP1.
The next points to check are differences in supported parts and configurations by degrees of freedom, the contents of releases in each repository, test results on the physical robot, and the extent to which trained policies can be deployed on RP1. The published page does not give prices, production volumes or operating performance for each configuration. The overview page lists GPL-3.0, but asks users to check the license for each individual repository. Users redistributing or modifying the robot or software need to confirm licensing and support status for each component.
Details
Hardware materials include mechanical design files, schematics, board layouts, Gerber files, a bill of materials (BOM) and assembly instructions. Software materials include a Linux BSP, USB-CAN bridge, low-level SDK, and URDF and MJCF models. The deployment framework targets RPO and RP1, with modules for IMUs, motors, inference and cameras accessible from C++ and Python.
For training, the repository points to a work environment for RPO walking policies using Isaac Lab and lists workflows including AMP, BeyondMimic and Parkour. IBFM, PredActor and INTACT-JEPA are marked as “coming soon” for RP1 support. The overview repository lists GPL-3.0, but directs users to check the license in each individual repository.
Sources
Translated from the Japanese original: /news/roboparty-rp1-no-kikai-sekkei-kara-gakushuu-madewo-github-de-1f743d. The Japanese version is authoritative if they differ.