Microduck skill tree (sim only)
Collection
Policies trained in simulation for a Pollen Robotics Microduck still on order, one level per skill, plus the training log. Sim only so far. • 6 items • Updated
walk-recover
Walking and fall recovery in one policy (Mjlab-VelStand-Flat-MicroDuck), simulation only: 20,000 iterations, 4096 envs, 326 min on one RTX 5090. Walks and holds for 20 s from standing and catches a forward lean; from flat on its back it did not get up in 3 of 3 takes. Not yet tested on a real Microduck.
A perpetual policy for the microduck (61-D observation, 14 actions, 50 Hz). Runs until told otherwise — a gait for the walk slot.
Run it on a robot
sudo robotctl policy load walk witcheer/microduck-walk-recover
The observation normalizer is baked into policy.onnx; feed raw observations.
manifest.json follows schema 2 of the microduck policy manifest (docs/policy-manifest.md in the daemon repo).
Training
- repo:
pollen-robotics/microduck_rl - branch:
develop - commit:
53b8971b6