RoboParty debuts RP1, world’s first full-stack open-source humanoid

The global debut of the RP1 (ROBOTO 01) took place at IROS 2026, where Pittsburgh-based RoboParty introduced its new bipedal humanoid. The company describes the machine as the world’s first high-performance, full-stack open-source system in this category. It targets researchers, educators, robotics developers, and embodied AI teams seeking a platform that supports modification, training, and continuous extension within real development workflows.

Founder advocates for open access

Yi Huang, founder of RoboParty, emphasized that humanoid robots should not remain closed systems accessible only to a select few while others merely observe. He stated that the goal is for RP1 to serve as an open foundation that developers can access, modify, train, and continuously improve. According to Huang, full-stack open source has been the technical direction since the RPO project began.

Interactive stability demonstration

At the RoboParty booth, the “Kick Me” interactive demonstration allowed visitors to directly apply external disturbances to the robot. When pushed or kicked, RP1 rapidly adjusted its posture and recovered its balance. This open interaction differed from demonstrations based solely on predefined motions, enabling observers to assess dynamic stability and control performance under disturbance, according to a press release. Repeated high-intensity interactions also tested the robot’s body performance and system-level robustness.

These capabilities were supported by PartyOS, RoboParty’s open R&D foundation for humanoid robotics, which integrates the UFO framework.

Technical specifications and design

RoboParty highlighted that RP1 delivers peak joint torque of up to 160 N•m. The design incorporates in-house-developed Romomo actuator modules and mechanical structures alongside a real-time motion-control system. These components are engineered to meet requirements for dynamic motion, disturbance recovery, and long-term development experiments.

The platform is designed to support embodied AI model training, validation, deployment, data collection, and continuous development. During the “Kick Me” interaction, the robot demonstrated continuous motion, disturbance recovery, and fall recovery. These behaviors reflected the coordination of the robot body, actuator modules, and motion-control system, rather than algorithmic performance alone.