Pollen Robotics: Microduck, a desktop biped for reinforcement learning and sim2real teaching
Pollen Robotics is a robot maker founded in Bordeaux in 2016 by former Inria researchers; it joined Hugging Face in April 2025 and is its robotics team today. Its first consumer robot, Reachy Mini, is a desktop robot built for human-robot interaction; according to the manufacturer more than 10,000 of them have reached people around the world. The Reachy 2 humanoid research platform is also in its portfolio. Microduck is the team's second consumer robot and its focus is movement: a 25 cm biped weighing under 800 g, designed for training behaviours in simulation with reinforcement learning and transferring them to the real robot. The software stack is open source under Apache-2.0; the mechanical and electronic design files are not open. Robotlar.org offers Microduck to universities, schools and R&D teams in Turkey through enterprise sourcing.
The Robotlar.org catalogue lists 1 model: Microduck, a biped robot for reinforcement learning and sim2real teaching. The manufacturer's portfolio also includes the Reachy Mini desktop interaction robot and the Reachy 2 humanoid research platform.
Pollen Robotics was founded in Bordeaux in 2016 by former Inria researchers and joined Hugging Face in April 2025. It works as Hugging Face's robotics team today, and Microduck's training tools integrate with the Hugging Face Hub and Hugging Face Jobs.
The software stack is open source under Apache-2.0; the robot software, SDK, simulation and reinforcement learning training tools are on GitHub. The mechanical and electronic design files are not open, so the robot is not open-source hardware. The 3D model files in the RL repository are released under the non-commercial CC BY-NC-SA licence.
| Model | Category | Motors | Height | Weight | Policy Loop | |
|---|---|---|---|---|---|---|
| Microduck | Desktop Biped Robot for Reinforcement Learning | 15Legs, neck and head | 25 cm14 cm wide | < 800 gFits in a hand | 50 HzOnboard | Detail |
Choosing by brand name alone is not enough. Payload, operating environment, shift pattern, bidirectional traffic, and integration requirements should all shape the decision. We can map the right architecture with you.
