---
title: oli-rl-policies
canonical_url: "https://www.modelscope.cn/models/LimXDynamics/oli-rl-policies"
md_url: "https://www.modelscope.cn/models/LimXDynamics/oli-rl-policies.md"
repository: LimXDynamics/oli-rl-policies
last_updated: 2026-11-06
license: apache-2.0
pipeline_tag: robotics
tasks:
  - robotics
library_name:
  - onnx
  - pytorch
frameworks:
  - pytorch
language:
  - en
  - zh
downloads: 0
stars: 0
tags:
  - robotics
  - reinforcement-learning
  - humanoid
  - locomotion
  - motion-imitation
  - onnx
  - mujoco
  - limx-dynamics
  - oli
---

# oli-rl-policies

> oli-rl-policies - LimXDynamics 在 ModelScope 开源的模型。LimX Oli 全尺寸人形机器人（HUD0401，31 个关节）的强化学习策略，ONNX 格式，包括行走、动作模仿（挥手打招呼）、站立和阻尼控制器，可在 MuJoCo 仿真或真机上运行，配套部署代码为 humanoid-rl-deploy-python。文件原样同步自 GitHub，以 GitHub 仓库为准。

LimXDynamics/oli-rl-policies 是 ModelScope 魔搭社区上的robotics模型，采用 apache-2.0 许可。

- **Repository**: LimXDynamics/oli-rl-policies
- **License**: apache-2.0
- **Tasks**: robotics
- **Tags**: robotics, reinforcement-learning, humanoid, locomotion, motion-imitation, onnx, mujoco, limx-dynamics, oli
- **Downloads**: 0
- **Stars**: 0
- **Last updated**: 2026-11-06

Source: https://www.modelscope.cn/models/LimXDynamics/oli-rl-policies

---

## 中文说明

LimX Oli 全尺寸人形机器人（`HU_D04_01`，31 个关节）的强化学习策略，ONNX 格式，包括行走、动作模仿（挥手打招呼）、站立和阻尼控制器，可在 MuJoCo 仿真或真机上运行，配套部署代码为 [humanoid-rl-deploy-python](https://github.com/limxdynamics/humanoid-rl-deploy-python)。文件原样同步自 GitHub，以 GitHub 仓库为准。

### 快速上手（仿真）

1. 将 [humanoid-mujoco-sim](https://github.com/limxdynamics/humanoid-mujoco-sim) 和 [humanoid-rl-deploy-python](https://github.com/limxdynamics/humanoid-rl-deploy-python) 克隆到同一目录（部署代码已包含这些策略）
2. 安装 limxsdk-lowlevel 和 click
3. 设置环境变量 `ROBOT_TYPE=HU_D04_01`
4. 依次启动仿真器 `simulator.py`、部署程序 `main.py` 和手柄程序 `robot-joystick`

遥控器按键：行走 R1 + X，动作模仿 L1 + B，站立 L1 + Y，阻尼 L1 + A。

下载全部策略：

```bash
hf download limxdynamics/oli-rl-policies --local-dir ./oli-rl-policies
```

国内用户也可以从魔搭下载：[LimXDynamics/oli-rl-policies](https://www.modelscope.cn/models/LimXDynamics/oli-rl-policies)。

真机部署前请先在仿真中测试，确保有经过培训的操作人员在场，急停随手可及。

采购咨询与合作：**opensource@limxdynamics.com**

---

# LimX Oli RL Policies

![LimX Oli humanoid](https://raw.githubusercontent.com/limxdynamics/humanoid-description/master/docs/images/banner.png)

Reinforcement-learning policies in ONNX for the **LimX Oli** full-size humanoid (`HU_D04_01`, 31 joints), ready to run in MuJoCo or on the real robot with [humanoid-rl-deploy-python](https://github.com/limxdynamics/humanoid-rl-deploy-python).

These files are mirrored as-is from [humanoid-rl-deploy-python](https://github.com/limxdynamics/humanoid-rl-deploy-python/tree/main/controllers/HU_D04_01). The GitHub repository is the source of truth.

## What's included

| Controller | Remote buttons | Files |
| --- | --- | --- |
| Walk | R1 + X | `walk_controller/policy/default/policy.onnx`, `walk_controller/walk_param.yaml` |
| Mimic (greeting motion) | L1 + B | `mimic_controller/policy/default/policy.onnx`, `lin_encoder.onnx`, `priv_encoder.onnx`, `mimic_controller/mimic_param.yaml` |
| Stand | L1 + Y | `stand_controller/joint_params.yaml` |
| Damping | L1 + A | `damping_controller/joint_params.yaml` |

`HU_D04_01/controllers.yaml` lists the controllers loaded by the deploy code.

## Quick start (simulation)

Run on Linux x86_64 with [humanoid-mujoco-sim](https://github.com/limxdynamics/humanoid-mujoco-sim) and [humanoid-rl-deploy-python](https://github.com/limxdynamics/humanoid-rl-deploy-python) in the same folder. The deploy repository already contains these policies.

```bash
pip install humanoid-mujoco-sim/limxsdk-lowlevel/python3/amd64/limxsdk-*.whl click
export ROBOT_TYPE=HU_D04_01
python humanoid-mujoco-sim/simulator.py            # terminal 1
python humanoid-rl-deploy-python/main.py           # terminal 2
humanoid-mujoco-sim/robot-joystick/robot-joystick  # terminal 3
```

To download the policies from here instead:

```bash
hf download limxdynamics/oli-rl-policies --local-dir ./oli-rl-policies
```

To deploy your own walking policy, replace `walk_controller/policy/default/policy.onnx`. See the deploy repository for details.

## Safety

These policies drive a physical robot. Test in simulation first, run with a trained operator present, and keep the emergency stop within reach.

## Related

- Robot models (URDF / MJCF / USD): [humanoid-description](https://github.com/limxdynamics/humanoid-description)
- Simulator: [humanoid-mujoco-sim](https://github.com/limxdynamics/humanoid-mujoco-sim)
- ROS 1 / ROS 2 / C++ deployment: [humanoid-rl-deploy-ros](https://github.com/limxdynamics/humanoid-rl-deploy-ros) · [humanoid-rl-deploy-ros2](https://github.com/limxdynamics/humanoid-rl-deploy-ros2) · [humanoid-rl-deploy-cpp](https://github.com/limxdynamics/humanoid-rl-deploy-cpp)

## License

Apache 2.0, the same as [humanoid-rl-deploy-python](https://github.com/limxdynamics/humanoid-rl-deploy-python).

## Contact

- Business inquiries and collaboration: **opensource@limxdynamics.com**
- Questions: open a discussion in the Community tab or an issue on GitHub
- More from LimX Dynamics: [Hugging Face](https://huggingface.co/limxdynamics) · [GitHub](https://github.com/limxdynamics) · [Website](https://www.limxdynamics.com/en)