ROBOTIS AI Sapiens integrates NVIDIA Kimodo for text-to-motion control
Open-source humanoid executes full-body motions from text prompts using RL
ROBOTIS has announced a significant milestone for its open-source humanoid platform, AI Sapiens, by integrating NVIDIA Kimodo to generate humanoid motions directly from text prompts. These full-body motions are executed on the physical AI platform using reinforcement learning (RL) for dynamic control. The platform is built on ROBOTIS' in-house DYNAMIXEL-Q (QDD) actuators, which provide high backdrivability and precise torque control essential for stable real-world operation. This integration allows researchers and developers to command complex humanoid behaviors simply by inputting natural language descriptions, bridging the gap between language understanding and physical action.
As a fully open-source project, ROBOTIS is preparing to release all development resources by the end of June 2026, including 3D CAD drawings, schematics, firmware, control source code, simulation assets, and RL configuration files. The goal is to provide a robust, reproducible starting point for the Physical AI community. The demonstration video showcases AI Sapiens performing various motions generated from text cues, highlighting the potential for rapid prototyping of humanoid behaviors. With ROS 2 as the backbone, this platform aims to accelerate research in embodied AI, human-robot interaction, and reinforcement learning for manipulation and locomotion.
- NVIDIA Kimodo generates full-body humanoid motions from text prompts, executed via RL control on AI Sapiens.
- DYNAMIXEL-Q (QDD) actuators provide high backdrivability and precise torque control for stable real-world operation.
- Complete open-source release by end of June 2026: 3D CAD, schematics, firmware, control code, simulation assets, and RL configs.
Why It Matters
Democratizes humanoid robotics research by providing a fully reproducible, open-source platform with text-driven motion generation.