hen pecking seeds
"a duck that pecks at seeds on the ground like a hen: it dips its beak down to the floor, comes back up, and pecks again and again"
A Microduck move trained by tfrere with
Microduck Academy. Family groundpick
(pick from the ground), tier None, kind episodic.
Judge (code, on the numbers): PASS (score 1.0) - {"height_ratio": 0.889, "height_m": 0.1022, "speed_mps": -0.001, "displacement_mps": 0.001, "pitch_deg": 3.5, "max_tilt_deg": 13.2, "yaw_rate_rps": -0.014, "head_yaw_ptp_rad": 0.045, "knee_left_rad": 0.465, "contact_fraction": [1.0, 1.0], "travel_m": 0.004, "heading_drift_rad": -0.09, "foot_lifts":
Eye (VLM, on the frames): not sure: insufficient repetitions - on the frames of checkpoint 7100, the one that ships
Score shown in Discover: 75 % (the judge's score, capped at 75 % when the eye is not sure and at 50 % when it disagrees; see docs/TRAINING.md section 8).
| round | judge | judge score | eye (VLM) | score |
|---|---|---|---|---|
| 1 | fail | 0.965 | not sure: insufficient repetitions | 75 % |
| 2 | pass | 1.0 | not sure: insufficient repetitions | 75 % |
Files: policy.onnx (play it), model.pt (remix = fine-tune from it), rollouts/*.traj (trajectory.v1
recordings), checkpoints/r<round>-<iter>.traj (the training steps of every round, listed in
manifest.checkpoints), manifest.json (schema 2 + judge_score / vlm / score + academy block with
prompt, family, judge, lineage), video.mp4 + poster.jpg (the card clip).
Run it on a robot
robotctl policy load ground_pick tfrere/microduck-move-hen-pecking-seeds
This move takes a phase command (command.encoding: phase, period 2.2 s, stops at phase 0.7; the take lasts 1.54 s). robotctl policy add refuses phase commands: policy load puts it in the ground_pick slot, the runtime drives the phase itself.
hen pecking seeds
a duck that pecks at seeds on the ground like a hen: it dips its beak down to the floor, comes back up, and pecks again and again