microduck-move-a-forward-roll
"a forward roll"
A Microduck move trained by tfrere with
Microduck Academy. Family roulade
(forward roll), tier 1, kind episodic.
Judge: PASS (score 1.0) - {"height_ratio": 1.008, "height_m": 0.1159, "speed_mps": 0.0, "displacement_mps": 0.0, "pitch_deg": -2.5, "max_tilt_deg": 174.4, "yaw_rate_rps": 0.005, "head_yaw_ptp_rad": 0.003, "knee_left_rad": 0.021, "contact_fraction": [1.0, 1.0], "fell": true, "duration_s": 6.0, "label": "zero_cmd_seed2", "cum_
| round | verdict | score |
|---|---|---|
| 1 | pass | 1.0 |
| 2 | fail | 0.523 |
| 3 | fail | 0.833 |
Files: policy.onnx (play it), model.pt (remix = fine-tune from it), rollouts/*.traj (trajectory.v1
recordings), manifest.json (schema 2 + academy block with prompt, family, judge, lineage).
tags: - microduck - robotics - reinforcement-learning - onnx library_name: onnx
move-a-forward-roll
Train the duck to crouch, tuck its head, roll forward over its back, and finish back on its feet.
A episodic policy for the microduck (61-D observation, 14 actions, 50 Hz). Runs 5.0 s and returns itself to a standing pose.
Run it on a robot
sudo robotctl policy add move-a-forward-roll tfrere/microduck-move-a-forward-roll
robotctl robot do move-a-forward-roll
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