RoboParty Debuts RP1 Humanoid to Challenge Closed Robotics Systems

The RP1 made its public entrance in Pittsburgh with a live demonstration of its dynamic stability. Visitors at the RoboParty booth were invited to physically push and kick the robot, which consistently regained its balance and posture without relying on pre-programmed motion paths. This performance is governed by PartyOS and the UFO framework, which utilizes unsupervised reinforcement learning to train low-level motor skills, including fall recovery and complex transitions.

Unlike current market leaders that often operate as closed systems, RoboParty intends to provide researchers and developers with full access to the robot's architecture. The company plans to release mechanical files, software stacks, simulation environments, and SDKs throughout late 2026. Founder Yi Huang stated that the goal is to prevent humanoid robotics from remaining an exclusive domain for a few select teams, instead fostering a collaborative environment where developers can modify and refine both hardware and AI algorithms.

Technically, the RP1 features Romomo actuator modules capable of 160 N•m peak joint torque, designed specifically for long-term experimental use. The company is building upon the success of its earlier project, RPO, which garnered over 2,500 GitHub stars and established a foundation for community-led robotics development. With a core team drawn from institutions like Stanford and Tsinghua University, RoboParty is now focusing on transitioning from its IROS validation milestone to a broader release and mass-production program scheduled for the fourth quarter of 2026.