Pollen Robotics presents the Microduck biped robot
Pollen Robotics’ Microduck repository describes a roughly 25 cm, 800 g biped duck robot whose locomotion uses reinforcement-learning policies. The vendor says its Rockchip RK3566 runs a 50 Hz control loop for 15 servos. Policies are trained in the companion microduck_rl project using MuJoCo, PPO, domain randomization and a sim-to-real process, then exported to ONNX. The repository includes robot code, documentation and a simulator, and provides a purchase link.
The project documents a complete reinforcement-learning workflow for a small biped robot, spanning simulation, training, model export and robot control. The hardware specifications and claimed capabilities are vendor-reported.
A Tiny biped duck robot 🦆
https://pollen-robotics.com/microduck
Microduck
A tiny biped robot that moves using reinforcement learning policies.
#Pollen Robotics#Microduck#microduck_rl#MuJoCo#PPO#ONNX#Rockchip RK3566