flat out sprint
"a duck that sprints as fast as it possibly can, flat out but still on its feet"
A Microduck move trained by tfrere with
Microduck Academy. Family velocity
(walking styles (gaits)), tier 1, kind perpetual.
Judge (code, on the numbers): PASS (score 1.0) - {"height_ratio": 1.07, "height_m": 0.123, "speed_mps": 0.531, "displacement_mps": 0.515, "pitch_deg": 8.6, "max_tilt_deg": 14.8, "yaw_rate_rps": 0.147, "head_yaw_ptp_rad": 1.239, "knee_left_rad": 0.185, "contact_fraction": [0.5, 0.5], "fell": false, "duration_s": 8.0, "label": "forward", "mouth_min_
Eye (VLM, on the frames): agrees (The duck robot sprints forward steadily, staying upright on its feet without falling.)
Score shown in Discover: 100 % (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 | pass | 1.0 | agrees (The duck robot is moving forward quickly on its feet without falling.) | 100 % |
| final | pass | 1.0 | agrees (The duck robot sprints forward steadily, staying upright on its feet without falling.) | 100 % |
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
sudo robotctl policy add flat out sprint tfrere/microduck-move-flat-out-sprint
robotctl robot do flat out sprint
flat out sprint
Train the duck to walk at its maximum speed, around 0.35 m/s, with fast driving strides while staying balanced on its feet.