RoboParty has unveiled RP1 (ROBOTO 01) at IROS 2026, positioning the bipedal humanoid as an open-source platform for robotics researchers, developers, educators and embodied AI teams. The robot combines in-house actuators, real-time motion control and reinforcement-learning software, with RoboParty planning to progressively open its hardware and software stack.
RoboParty Wants Humanoid Robotics to Be an Open Platform
Humanoid robotics is increasingly becoming a software problem as much as a hardware one. RoboParty is betting that opening both sides of that equation could accelerate development, and its newly unveiled RP1 is designed around that idea.
The company publicly debuted RP1 (ROBOTO 01) on September 28 at IROS 2026, describing it as a full-stack open-source bipedal humanoid aimed at researchers, educators, robotics developers and embodied AI teams.
RP1 follows RPO (ROBOTO ORIGIN), RoboParty’s open-source humanoid project launched in January 2026. The company says RPO has attracted more than 2,500 GitHub stars, while its broader developer community has grown to more than 5,000 people.
The more consequential part of RP1’s launch is the scope of the planned openness. RoboParty says it intends to progressively release not only mechanical designs, but also motion-control systems, simulation environments, software development kits, training tools and its PartyOS robotics development foundation.
That approach puts the robot in the growing intersection between open-source robotics, embodied AI and physical AI.
Testing Stability With Real-World Disturbances
At IROS, RoboParty demonstrated RP1’s dynamic stability using a deliberately simple test: visitors could push and kick the robot.
The “Kick Me” demonstration was designed to expose how the humanoid responds when its balance is disrupted. RP1 adjusted its posture and recovered its balance after external disturbances, providing a visible test of the interaction between its body, actuators and control software.
Dynamic recovery is an important capability for bipedal robots. Unlike wheeled machines, humanoids must continuously manage their center of mass while walking, standing and interacting with uneven or unpredictable environments.
RoboParty attributes RP1’s capabilities partly to PartyOS, its open R&D foundation for humanoid robotics.
One component is UFO, a training framework that uses unsupervised reinforcement learning to discover motor skills. The framework covers areas including skill transitions, disturbance recovery and fall recovery without relying on predefined motion trajectories.
That distinction is significant for embodied AI. Rather than programming every movement in advance, reinforcement-learning systems can allow robots to develop control policies through training and interaction.
Hardware and AI Developed as One Stack
RoboParty is also attempting to avoid the hardware-software separation that can slow robotics development.
RP1 integrates its robot body, actuator modules, low-level control and embodied AI algorithms so the components can be validated together. The robot’s joints can deliver peak torque of up to 160 N·m, according to the company, using its internally developed Romomo actuator modules and real-time motion-control system.
The platform also builds on components introduced with earlier RoboParty projects, including the RPO robot body, Romomo actuators and the RP Hand dexterous end effector.
The company’s stated goal is to make the complete development workflow accessible to outside developers. That could allow researchers to modify hardware, experiment with low-level controllers, train models and deploy new behaviors without being locked into a proprietary robotics stack.
This is an important distinction in the increasingly crowded humanoid robotics market. Companies such as Tesla, Figure AI, Apptronik and others are pursuing highly integrated humanoid systems, but most commercial platforms remain substantially closed.
An open-source approach shifts the competitive model from selling only robots to building a developer ecosystem around them.
The Open-Source Bet on Embodied AI
RoboParty’s strategy mirrors the role that open-source software has played in conventional AI development. Frameworks, models, datasets and developer tools have helped researchers build on one another’s work rather than starting from scratch.
Robotics is harder to open because physical systems introduce manufacturing, safety, calibration and hardware reliability challenges. Releasing a mechanical design does not automatically make a robot reproducible or useful.
That is why RoboParty’s emphasis on a full-stack approach matters. The company says it plans to open the simulation environment, SDKs, training infrastructure and motion-control stack alongside hardware.
PartyOS is intended to evolve across locomotion, perceptual interaction, whole-body manipulation and autonomous humanoid behavior. Those capabilities represent some of the core technical problems that embodied AI researchers are attempting to solve: getting machines to perceive an environment, reason about it and physically execute actions.
The open-source community could become particularly important for that development. Different research teams can experiment with control policies, manipulation strategies and training approaches without waiting for a vendor’s software roadmap.
From Research Platform to Production Robot
RoboParty is not stopping at a research prototype. The company says it plans to announce more details of RP1’s open-source roadmap in October 2026, followed by broader hardware and software releases and a mass-production program later in the fourth quarter.
That next phase will be a more meaningful test of the platform’s ambitions.
Open-source robotics succeeds only if developers can actually reproduce, modify and deploy the technology. Hardware availability, documentation, manufacturing consistency, simulation quality and software tooling will matter as much as the robot’s headline specifications.
For now, RP1 represents a notable attempt to apply open-source principles to the complete humanoid robotics stack. If RoboParty delivers on its roadmap, the platform could give embodied AI developers access to a system spanning physical hardware, control software and machine-learning infrastructure.
The larger question is whether that openness can produce a sufficiently large community to accelerate humanoid development faster than proprietary platforms can.
Market Landscape
The humanoid robotics market is moving toward increasingly integrated systems combining robotic hardware, real-time control, computer vision, reinforcement learning and foundation models. Companies including Tesla, Figure AI, Apptronik and others are pursuing commercial humanoids, while research institutions continue to develop open robotics platforms and simulation environments.
RoboParty’s differentiator is its emphasis on full-stack openness. Rather than limiting open-source access to mechanical designs, the company says it intends to release software, simulation, SDKs, control systems and training tools.
This aligns with the broader embodied AI and physical AI trend, where researchers are attempting to transfer increasingly capable AI models from digital environments into robots that can perceive and act in the physical world.
Top Insights
- RoboParty’s RP1 combines open-source hardware ambitions with motion control, simulation, SDKs and reinforcement-learning tools for embodied AI development.
- A 160 N·m peak joint-torque specification positions RP1 around dynamic humanoid robotics rather than basic educational platforms.
- The company’s “Kick Me” demonstration highlights disturbance recovery, an important requirement for stable bipedal robots operating outside controlled environments.
- PartyOS and its UFO reinforcement-learning framework are designed to support motor-skill discovery without relying exclusively on predefined trajectories.
- RoboParty plans broader hardware and software releases and a mass-production program later in Q4 2026.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
