RoboParty RP1. Fully open-source humanoid platform.

RP1: An OpenSource Full-Stack Platform for Agentic Humanoids

From mechanical design and electronics to firmware, control and learning.
Built to be understood. Open to be built upon.

Open by Design

RP1 is RoboParty's fully open-source humanoid platform: a shared foundation for building, understanding and extending a real robot.

The platform brings mechanical and electrical design, joint modules, firmware, robot models, deployment and learning into one connected development stack. Explore a single subsystem, reproduce a build, or develop your own control and learning workflows.

Hardware you can inspect	Software you can extend	Intelligence you can explore
Mechanical CAD, schematics, PCB layouts, Gerber files, BOM and assembly documentation.	Linux BSP, communication, device interfaces, robot descriptions and deployment modules.	Locomotion training, motion imitation, tracking and behavior-model projects.

_{RP1, photographed in the real world.}

One Robot. A Connected Stack.

RP1 control architecture: input / perception → motion control base (motion retargeting, motion policies) → real-time control (action decoding + serial-link solving at 500 Hz) → whole-body execution (ROMOMO joint modules, RP-hand), with an E-stop and a high-frequency feedback loop.

No black boxes, nothing withheld — full control handed back to the developer

Explore the Repositories

Release contents, supported configurations and setup instructions are maintained in each repository.

Contributions belong in the relevant sub-repository. This overview repository does not accept pull requests; please open issues and pull requests in the project you are working on.

Hardware & Tools

Repository	Description
rp1_hardware	Mechanical source files, electronics, testing and assembly SOPs.
ROMOMO	Joint modules, firmware and host tools.
RP Hand	Dexterous hand product documentation, wiring, companion applications, CAD and robot models. Includes the C++ / Python / ROS 2 SDK entry point.
RP1_TEST	RP1 full-machine testing.
roboparty_joystick	RoboParty controller and app.

Robot configurations

See the description repository's current release for supported robot configurations.

Supported configurations include 24, 28, 30 and 42 DoF variants.

Software & Systems

Repository	Description
roboparty_deploy	Deployment framework for RPO and RP1, with easy-to-use SDKs for `imu`, `motor`, `inference` and `camera` modules, callable from C++ and Python.
RP1_description	URDF and MJCF models, 3D meshes, and robot and RP Hand configurations.
roboparty_train	Isaac Lab training workspace for RPO locomotion policies, covering AMP, BeyondMimic and Parkour workflows with RSL-RL training and motion retargeting.
RP1_firmware	Linux BSP, USB-CAN bridge, low-level SDK and packages.

Demo

Project	Description
AMP	Learns motion priors from reference motions via adversarial training, enabling natural-looking locomotion and control.
BeyondMimic	Reproduces expressive whole-body movements by tracking reference motions, including dance and motion-imitation sequences.
Parkour	Learns perceptive locomotion policies to traverse challenging terrain, including stairs, gaps and obstacles.
MimicLite	Provides an efficient, general-purpose humanoid motion-tracking system for training deployable PPO and PPO-ROA policies.
UFO	Provides a general unsupervised reinforcement learning framework for humanoid control, supporting FB/TeCH training, motion tracking and reward and goal inference.
PBFM	Maps motion references, target body states and rewards into a shared behavioral intent space, combining intent with depth perception for terrain-adaptive control.
IBFM (RP1 support coming soon)	Learns reward-conditioned robust humanoid interaction via unsupervised reinforcement learning, demonstrated on carry, push and kick box tasks.
INTACT-JEPA (RP1 support coming soon)	Learns an intent-to-action mapping with JEPA for goal-conditioned robot control without broad action search.