---
tags:
- microduck-academy-policy
- microduck-policy
- academy
- official
- family:groundpick
- kind:episodic
- tier:1
- microduck
- robotics
- reinforcement-learning
- onnx
license: apache-2.0
library_name: onnx
---

# Ground pick

<video src="video.mp4" poster="poster.jpg" width="480" muted loop autoplay playsinline controls></video>

![Ground pick (poster)](poster.jpg)


An **official** Microduck move: [Pollen Robotics](https://huggingface.co/pollen-robotics)' **ground pick**,
retrained from scratch by [Microduck Academy](https://huggingface.co/spaces/tfrere/microduck) with the upstream recipe
as-is (task `Mjlab-GroundPick-Flat-MicroDuck` of [pollen-robotics/microduck_rl](https://github.com/pollen-robotics/microduck_rl) at `2b25a48`,
5999 iterations, 4096 envs, no reward edit), so that it carries the
`base/model.pt` checkpoint the Academy needs to **remix** it. The policy Pollen ships is
[`alpha_ground_pick.onnx`](https://huggingface.co/pollen-robotics/microduck-policies/blob/main/alpha_ground_pick.onnx).

Pollen Robotics' ground pick: the beak goes down to the floor and comes back up, driven by a phase command over 2.8 s.

## Fidelity check (this move vs Pollen's policy, same scene, same command, take `phase`)

| number | this move | Pollen's `alpha_ground_pick.onnx` |
|---|---|---|
| final_upright | True | True |
| max_tilt | 0.502 | 0.638 |
| base_z_mean | 0.0914 | 0.0979 |
| base_z_min | 0.0805 | 0.0839 |
| mouth_z_min | 0.0212 | 0.0204 |
| t_mouth_min_s | 0.26 | 1.08 |
| mouth_z_start | 0.188 | 0.1883 |
| mouth_z_end | 0.2162 | 0.2166 |
| base_z_end | 0.1166 | 0.1164 |

Recorded with `space/payload/render/rollout.py` on the upstream robot MJCF (`spikes/base_moves/fidelity.py`).

## Run it on a robot

```sh
robotctl policy load ground_pick tfrere/microduck-move-base-ground-pick
```

This move takes a **phase** command (`command.encoding: phase`, period 4.0 s, stops at phase 0.7; the take lasts 2.8 s). `robotctl policy add` refuses phase commands: `policy load` puts it in the `ground_pick` slot, the runtime drives the phase itself.

## Files

`policy.onnx` (play it, normalizer baked in), `base/model.pt` + `base/agent.yaml` + `base/env.yaml` (the rsl_rl
checkpoint and configs: **remix** = fine-tune from it), `rollouts/0.traj` (trajectory.v1 showcase take, with the
commands), `video.mp4` + `poster.jpg` (the card clip), `manifest.json`
(schema 2 + `academy` block: origin, source task and policy, training facts, fidelity numbers), `fidelity.json`,
`train.json`.

## Credits and license

The task, the reward design, the robot model and the training recipe are Pollen Robotics' work
([pollen-robotics/microduck_rl](https://github.com/pollen-robotics/microduck_rl), Apache-2.0). This repo republishes a retrained policy of
that recipe under the same Apache-2.0 license, with the checkpoint, so the Academy's users can start from it.