Back to Search
M

Microduck

RobotActive

Microduck is a tiny open-source biped robot developed by Pollen Robotics, the Bordeaux-based robotics team at Hugging Face. The official repository describes it as "a tiny biped robot that moves using reinforcement learning policies". It is about 25 cm tall and weighs about 800 g, with 15 motors driven by a 50 Hz onboard policy loop running on a Rockchip RK3566 computer, and ships with a camera, LiDAR, and two IMUs. Pollen Robotics opened pre-orders for Microduck at US$399 on August 27, 2026. The official product page presents it as a robot you train yourself with reinforcement learning: every behavior is a policy that can be retrained on your own machine, with seven trained moves shipped out of the box, including walking, sitting and standing, kicking a ball, beak-grabbing objects, roller skating with an accessory, and getting back up by itself. The SDK, the simulation, and the full reinforcement-learning training stack are published on GitHub: the pollen-robotics/microduck repository hosts the robot software, and the pollen-robotics/microduck_rl repository hosts the training stack, built on MuJoCo and PPO with ONNX export for the sim2real pipeline that produces the shipped policies. The software is licensed under Apache-2.0. Policies are shared on Hugging Face, where community fine-tunes appeared within days of the announcement. Microduck targets developers, makers, and AI learners, complementing Pollen Robotics' research-oriented Reachy platforms with a low-cost, playful robot for learning and experimenting with learned locomotion and behaviors.

Details

Updated:9/4/2026
weight kg0.80
height cm25.00
robot typebiped

Tags

sim2realopen-sourcehugging-faceeducationbipedreinforcement-learning

Relationships

Sources

pollen-robotics/microduck - Microduck robot software
github
Visit
pollen-robotics/microduck_rl - Microduck reinforcement-learning training stack
github
Visit
Microduck - A tiny biped robot you can teach new tricks (official product page)
website
Visit

Appears In

No related landscapes.