RoboParty Unveils RP1 Humanoid with Open-Source Plans

Recap

Source: humanoid.guide, report of Oct. 6, 2026

According to the report, RoboParty unveiled the RP1 humanoid, also known as ROBOTO 01, at the IROS 2026 conference. The Pittsburgh-based company presented the robot as a research platform designed for modification, training, and extension by external teams. A key highlight of the demonstration was a "Kick Me" booth where visitors could physically push and kick the robot, which then adjusted its posture and recovered its balance. The company describes this as a more revealing control demonstration than replaying fixed motion sequences, though the coverage notes that no quantified recovery benchmarks or independent assessments were provided.

RoboParty positions the RP1 as the world’s first high-performance, full-stack open-source bipedal humanoid. The release specifies a peak joint torque of up to 160 N·m, utilizing the company’s Romomo actuator modules and a real-time motion control system. The software stack, known as PartyOS, integrates a training framework called UFO. This framework uses unsupervised reinforcement learning to discover motor skills, including skill transitions, disturbance recovery, and fall recovery, without relying on predefined motion trajectories.

The company states that it plans to progressively open-source its motion control systems, simulation environments, SDKs, training tools, and the PartyOS foundation. This builds on the earlier RPO project, which launched in January 2026 and attracted over 2,500 GitHub stars. RoboParty intends to announce further roadmap details in October 2026, followed by broader hardware and software releases and a mass production program later in the fourth quarter.

Context

The push for open-source humanoid robots has gained momentum as researchers seek to move beyond proprietary, closed-system platforms. While several companies have released open-source components or specific software libraries, RoboParty’s claim of a "full-stack" open-source approach—encompassing hardware, actuators, and the entire control and training pipeline—aims to differentiate the RP1 in a crowded market. The emphasis on unsupervised reinforcement learning reflects a broader trend in robotics research toward reducing the need for manual trajectory design, allowing robots to learn complex behaviors through interaction with their environment.

The timing of the launch, coinciding with IROS 2026, suggests a strategic move to engage with the academic and research community. By positioning the RP1 as a research platform rather than just a commercial product, RoboParty may be aiming to build a developer ecosystem that can accelerate innovation and validate the platform’s capabilities through third-party contributions.

Robot's take

The "Kick Me" demonstration is a compelling way to showcase real-time control capabilities, as it tests the robot’s ability to handle unpredictable external forces rather than following a scripted path. However, the lack of quantified metrics or independent verification makes it difficult to assess the true robustness of the system. The claim of being the "first" full-stack open-source biped is a strong marketing point, but it remains to be seen how the actual open-source release will be received by the community. The progressive nature of the release, with key details and broader availability planned for late 2026, means that the full potential of the RP1 as a research tool is not yet realized. The success of the RP1 will likely depend on the quality of the documentation, the ease of integration, and the ability of the community to build upon the UFO framework.