publish walk-rough: perpetual, velrough-2026-09-19-it8000.onnx
Browse files- README.md +29 -0
- manifest.json +36 -0
- policy.onnx +3 -0
README.md
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
tags:
|
| 3 |
+
- microduck
|
| 4 |
+
- robotics
|
| 5 |
+
- reinforcement-learning
|
| 6 |
+
- onnx
|
| 7 |
+
library_name: onnx
|
| 8 |
+
---
|
| 9 |
+
|
| 10 |
+
# walk-rough
|
| 11 |
+
|
| 12 |
+
Walking gait trained on rough terrain (Mjlab-Velocity-Rough-MicroDuck), simulation only: 8,000 iterations, 4096 envs, 324 min on one RTX 5090. Upright for a 12 s flat-ground take; no rough-ground take recorded yet. Not yet tested on a real Microduck.
|
| 13 |
+
|
| 14 |
+
A **perpetual** policy for the [microduck](https://github.com/pollen-robotics/microduck) (61-D observation, 14 actions, 50 Hz). Runs until told otherwise — a gait for the `walk` slot.
|
| 15 |
+
|
| 16 |
+
## Run it on a robot
|
| 17 |
+
|
| 18 |
+
```bash
|
| 19 |
+
sudo robotctl policy load walk witcheer/microduck-walk-rough
|
| 20 |
+
```
|
| 21 |
+
|
| 22 |
+
The observation normalizer is baked into `policy.onnx`; feed raw observations.
|
| 23 |
+
`manifest.json` follows schema 2 of the microduck policy manifest (`docs/policy-manifest.md` in the daemon repo).
|
| 24 |
+
|
| 25 |
+
## Training
|
| 26 |
+
|
| 27 |
+
- **repo**: `pollen-robotics/microduck_rl`
|
| 28 |
+
- **branch**: `develop`
|
| 29 |
+
- **commit**: `53b8971b6`
|
manifest.json
ADDED
|
@@ -0,0 +1,36 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"schema_version": 2,
|
| 3 |
+
"model_api": 1,
|
| 4 |
+
"obs_len": 61,
|
| 5 |
+
"action_len": 14,
|
| 6 |
+
"robot": {
|
| 7 |
+
"model": "microduck",
|
| 8 |
+
"hw_rev": 1,
|
| 9 |
+
"servos": "xl330",
|
| 10 |
+
"control_hz": 50
|
| 11 |
+
},
|
| 12 |
+
"name": "walk-rough",
|
| 13 |
+
"kind": "perpetual",
|
| 14 |
+
"entry_pose": "standing",
|
| 15 |
+
"description": "Walking gait trained on rough terrain (Mjlab-Velocity-Rough-MicroDuck), simulation only: 8,000 iterations, 4096 envs, 324 min on one RTX 5090. Upright for a 12 s flat-ground take; no rough-ground take recorded yet. Not yet tested on a real Microduck.",
|
| 16 |
+
"command": {
|
| 17 |
+
"encoding": "constant",
|
| 18 |
+
"idle": [
|
| 19 |
+
0.0,
|
| 20 |
+
0.0,
|
| 21 |
+
0.0
|
| 22 |
+
],
|
| 23 |
+
"twist": "unused (zeros)",
|
| 24 |
+
"head": "unused (zeros)",
|
| 25 |
+
"body": "unused (zeros)"
|
| 26 |
+
},
|
| 27 |
+
"duration_s": null,
|
| 28 |
+
"slot": "walk",
|
| 29 |
+
"training": {
|
| 30 |
+
"repo": "pollen-robotics/microduck_rl",
|
| 31 |
+
"commit": "53b8971b6",
|
| 32 |
+
"branch": "develop",
|
| 33 |
+
"dirty": false,
|
| 34 |
+
"source_file": "velrough-2026-09-19-it8000.onnx"
|
| 35 |
+
}
|
| 36 |
+
}
|
policy.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9c9f14af4b60e9587db8a26241476ba31586f013667950bd8d75c080a42fd533
|
| 3 |
+
size 793807
|