Observed arrival · 2026-09-08
Microduck — an open-source biped for retraining
Pollen Robotics presents a 25 cm, 800 g robot that ships with seven trained behaviours and lets owners retrain policies in simulation before deploying them on the real machine.
Field notes
The page describes a reinforcement-learning workflow built around a physical 25 cm biped: behaviours are trained in simulation, deployed to the robot, refined, and shared as policies. It lists seven existing moves, including self-righting and roller skating, and identifies MuJoCo as the physics simulator. The robot runs a 50 Hz onboard policy loop, while the linked software stack includes SDK tooling, simulation, and training code presented under an Apache-2.0 license.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue