> ## Documentation Index
> Fetch the complete documentation index at: https://docs.wuji.tech/llms.txt
> Use this file to discover all available pages before exploring further.

# Introduction

> In-hand cube reorientation on the Wuji Hand — PPO policies trained in mjlab that cover the full SO(3) goal space, with a sim-to-real bridge for closed-loop deployment on the physical hand.

wuji-mjlab is an in-hand cube reorientation reinforcement learning project for the Wuji Hand. It trains PPO policies in [mjlab](https://github.com/mujocolab/mjlab) that cover the full SO(3) goal space, then deploys them on the physical hand through a sim-to-real bridge. The project ships pretrained checkpoints, deployment scripts, and hardware guides so you can reproduce the demo end to end.

<Info>
  This release targets the Wuji Hand. The deploy bridge wraps the Wuji Hand SDK and expects the matching firmware revision.
</Info>

<div style={{ display: 'flex', gap: '2%', justifyContent: 'center', flexWrap: 'nowrap' }}>
  <img src="https://mintcdn.com/wuji-0bbde051/GsT0bEnPZ_18e5HL/assets/source/images/en/wuji-mjlab/sim.gif?s=e7d467b96fb6a56739f4eece74ba3d00" style={{ width: '49%', height: 'auto' }} alt="Simulation reorient demo" width="480" height="360" data-path="assets/source/images/en/wuji-mjlab/sim.gif" />

  <img src="https://mintcdn.com/wuji-0bbde051/KO3tT6qNwl3EeRut/assets/source/images/en/wuji-mjlab/real.gif?s=7bb8be7fa34cd574688199013603b69b" style={{ width: '49%', height: 'auto' }} alt="Real-hand reorient demo" width="640" height="480" data-path="assets/source/images/en/wuji-mjlab/real.gif" />
</div>

## Task

The project centers on one task, with two training configurations.

| Robot | Task ID | Pretrained checkpoint | Demo |
| - | - | - | - |
| Wuji Hand | `WujiHand_Reorient` | [Latest release assets](https://github.com/wuji-technology/wuji-mjlab/releases/latest) | Sim and real demos above |

The policy holds a cube in a downward-facing hand, receives a target orientation in the palm frame, and rotates the cube in place until its orientation matches the goal within a hold window, without dropping it.

## What You Get

* A calibrated pipeline that runs the trained ONNX policy on real Wuji Hand hardware
* Pretrained checkpoints and released policy assets, so you can deploy without training
* A 3D-printed ArUco-tagged cube, wrist AprilTag world frame, and camera calibration workflow
* Two training configurations: a release config that reproduces the demo, and a lower-VRAM variant

## Get the Released Assets

Pull the pretrained checkpoint and CAD bundle from the latest release:

```bash theme={null}
# Requires the GitHub CLI (https://cli.github.com). The glob keeps this
# command working unchanged across future release tags.
gh release download --repo wuji-technology/wuji-mjlab --pattern '*-assets.zip'
unzip wuji-mjlab-*-assets.zip
mv wuji-mjlab-*-assets release-assets
```

Without the `gh` CLI, open the [latest release page](https://github.com/wuji-technology/wuji-mjlab/releases/latest), download the `wuji-mjlab-v*-assets.zip` attachment by hand, then run the same `unzip` and `mv` pair.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.