allworld.mediaNews from all over the world
RoboParty releases source code for its humanoid robot RP1, including mechanical design and control software
Read more
Pandaily
pandaily.com

RoboParty releases source code for its humanoid robot RP1, including mechanical design and control software

RoboParty has released the RP1, a humanoid robot positioned as a fully open, fully functional platform. Instead of providing only a Software Development Kit (SDK) or a few Application Programming Interfaces (APIs), the company is publishing materials across the entire development chain: from mechanical and electrical design and proprietary joint modules to firmware, robot models, deployment tools, and training. The description of this project was presented in technical documentation on October 25th.

The RP1 robot stands 135 cm tall. The open version is equipped with 24 active degrees of freedom, while the Professional version offers 30 degrees of freedom, including two in the head and seven in each arm, suitable for more complex manipulation research. The head and hand modules can be replaced for use in various applications such as education, scientific research, data collection, tours, or entertainment. RoboParty demonstrated the robot's ability to walk continuously, track human movements, recover from pushes, and stand up after falls.

The hardware is based on ROMOMO, RoboParty's integrated rotary joint module. According to published specifications, the nominal peak torque ranges from 25 to 160 Newton-meters when powered by 48V and using CAN-FD communication. These modules combine inner rotor motors with planetary gearboxes, allowing for the sacrifice of low rotor inertia for fast response, stiffness, overload resistance, and reduced cost. RP1 uses an off-axis hip joint and a parallel ankle, where two motors jointly control pitch and roll, and the hollow cable routing attachment simplifies joint disassembly.

On the GitHub platform, the RP1 overview repository contains links to subprojects dedicated to mechanical source files, electronics, testing and assembly procedures, ROMOMO joint firmware, and host tools, as well as the RP Hand dexterity with SDKs in C++, Python, and ROS 2, full machine testing, and controller application. The repository lists mechanical CAD models, schematics, PCB layouts, Gerber files, material specifications, and assembly documentation for the hardware, while URDF, MJCF, and mesh models are available for simulation. The original article notes that the exact scope of each release should be checked in individual repositories.

Regarding software, the embedded stack connects motors, inertial sensors, and the battery management system through a unified device interface, enabling control at a frequency of 500 Hz and policy updates at 50 Hz. An open-source motion remapping tool called hhtools maps human movements onto the robot while respecting joint limits and ground contact. Within the basic development platform, RoboParty's PartyOS opens parts of the base motion layer, including hhtools, MimicLite, and UFO—a recovery project for disturbances and falls. Higher levels for intent representation and agent skill orchestration remain under research.

RoboParty has opened applications for the RP1 beta program, intended for developers and teams with real projects ready to report issues and provide feedback on this platform.

Similar stories

Amap releases Lambda-0, a humanoid model trained exclusively on first-person human data
Read more
pandaily.com

Amap releases Lambda-0, a humanoid model trained exclusively on first-person human data

Amap, a division of Alibaba Group focused on mapping and geolocation, has introduced Lambda-0. This model is positioned as the first foundational model for humanoid robots, created based on observed human behavior. The research was conducted in collaboration with Alibaba's ATH business group, and the code is available on GitHub under the Apache 2.0 license alongside a scientific paper on arXiv.

The core idea is that a humanoid can learn walking, posture, and two-handed manipulation simultaneously, even if the robot is absent during data collection. The team asserts that many vision, language, and first-person action-based models focus only on hands and limbs, ignoring crucial information about the torso, body, and posture needed for long-term tasks in open environments. Using teleoperation of real robots to fill this gap is costly and slow to scale.

Instead, Lambda-0 utilizes a whole-body data collection method without the robot's involvement, which Amap calls 'robot-free whole-body data collection.' Specialists wear consumer devices and record movements in everyday settings. For a single demonstration task, collecting one segment took about 29 seconds, which is significantly more efficient compared to the approximately 100 seconds required when using a physical robot, demonstrating a 3.4 times efficiency gain. Through this process, the team created the HumanVerse-500 dataset, containing 500 hours of human behavior data in open environments, synchronized across vision, body movement, and hand actions.

The training process occurs in three stages: hand interaction, full-body coordination, and finally, embodiment in a humanoid body. To account for differences between human and robotic bodies, as well as end-effector errors, the model employs a viewpoint alignment scheme and a coarse-to-fine action alignment method while maintaining full-body coordination and improving contact accuracy.

According to the presented paper, Lambda-0 demonstrated state-of-the-art results in the SIMPLE simulation environment, achieving about 90 percent success rate, as well as in real-world robot tests. In seven long-horizon, full-body mobile manipulation tasks, the model successfully transferred its learning to the Unitree G1 using only human data. The team reports a real-robot success rate of 62.5 percent, which is 22.5 percentage points higher than the strongest baseline. With a constant number of robot demonstrations, increasing the volume of human behavior data generally continued to improve performance on the real robot, which the team presents as evidence of the scalability of humanoid training based on human data. These figures are taken from the team's paper and have not yet been independently reproduced. For Amap, the release of this model complements their recent work on 3D city world models.

Popular