---
tags:
- microduck
- robotics
- reinforcement-learning
- onnx
library_name: microduck
pipeline_tag: robotics
base_model: Nico-robot/microduck-tug-of-war
base_model_relation: finetune
---

# tug-puller1

Tug of War: Nico-robot/microduck-tug-of-war warm-started and trained 50 PPO iterations against that same duck as the frozen opponent (microduck-challenges 4b675eb, Mjlab-TugOfWar-MicroDuck). No seed tables or fingerprints: one plain network. Fresh-seed head-to-head (20 bouts each): beat greecree tug-leader-v2, Nico-robot tug-of-war and the reference 20/20. Credit to Nico-robot.

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, loaded into a policy slot.

## Run it on a robot

```bash
sudo robotctl policy load <slot> juicenv/microduck-tug-puller1
```

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).

A remix of [Nico-robot/microduck-tug-of-war](https://huggingface.co/Nico-robot/microduck-tug-of-war).

## Training

- **repo**: `pollen-robotics/microduck_rl`
- **branch**: `main`
- **commit**: `4b675ebf4`