+
+
For partnership and procurement inquiries, contact opensource@limxdynamics.com
+ +# LimX Humanoid Description + +> **Part of the Oli humanoid open-source ecosystem.** Where to start and the other Oli repositories: [Oli quick start](https://github.com/limxdynamics/limxdynamics/blob/main/doc/quickstart.md#oli). + +**Simulation-ready models of a real, full-size humanoid robot — in URDF, MuJoCo MJCF and USD.** + +Drop the LimX Oli humanoid (HU_D04) into ROS, MuJoCo or NVIDIA Isaac Sim in minutes, with real actuator parameters and the true parallel-linkage ankles. + +[Models](#robot-models) · [Quick Start](#quick-start) · [Specs](#specifications) · [Ecosystem](#ecosystem) · [Citation](#citation) + +
| **HU_D04** full-size humanoid | `HU_D04_01` | 31 | URDF · MJCF · USD |
+|
| **HU_D04 + Gripper** | `HU_D04_01_with_gripper` | 39
| **HU_D04 + Dexterous Hand** | `HU_D04_01_with_hand` | 53
| HU_D03 *(discontinued)* | `HU_D03_03` | 31 | URDF · MJCF · USD |
+
+> **Note:** MuJoCo MJCF is currently provided for the base HU_D04 only. HU_D03 is no longer maintained — please use HU_D04 for new projects.
+
+## Quick Start
+
+```bash
+git clone --depth 1 https://github.com/limxdynamics/humanoid-description.git
+cd humanoid-description
+```
+
+> `--depth 1` is recommended — the meshes and USD files make the full history large.
+
+### MuJoCo (30 seconds)
+
+```bash
+pip install mujoco
+python -m mujoco.viewer --mjcf=HU_D04_description/xml/HU_D04_01.xml
+```
+
+Or load it from Python:
+
+```python
+import mujoco
+
+model = mujoco.MjModel.from_xml_path("HU_D04_description/xml/HU_D04_01.xml")
+data = mujoco.MjData(model)
+print(f"{model.nu} actuators, {model.nsensor} sensors") # 31 actuators, 65 sensors
+
+for _ in range(1000):
+ data.ctrl[:] = 0.0 # joint torques (N·m) — plug in your controller here
+ mujoco.mj_step(model, data)
+```
+
+> `HU_D04_01.xml` is the full simulation model (actuators, sensors, closed-loop linkages). `HU_D04_01_vis.xml` is a lightweight visualization-only version.
+
+### ROS 1 / ROS 2
+
+Clone into your workspace and build — each `HU_D0x_description` folder is a ROS package that works with both catkin and colcon:
+
+```bash
+cd ~/ros_ws/src
+git clone --depth 1 https://github.com/limxdynamics/humanoid-description.git
+cd ..
+catkin_make # ROS 1
+colcon build # ROS 2
+```
+
+View in RViz (ROS 1, requires `urdf_tutorial`):
+
+```bash
+roslaunch urdf_tutorial display.launch model:=$(rospack find HU_D04_description)/urdf/HU_D04_01.urdf
+```
+
+View in RViz (ROS 2):
+
+```bash
+URDF=$(ros2 pkg prefix HU_D04_description)/share/HU_D04_description/urdf/HU_D04_01.urdf
+ros2 run robot_state_publisher robot_state_publisher --ros-args -p robot_description:="$(cat $URDF)" &
+ros2 run joint_state_publisher_gui joint_state_publisher_gui &
+rviz2 # Add → RobotModel, set Fixed Frame to base_link
+```
+
+### Isaac Sim / Isaac Lab
+
+Open `HU_D04_description/usd/HU_D04_01.usd` directly in Isaac Sim (**File → Open**), or add it to your stage as a reference. The top-level USD composes the layers in `usd/configuration/` automatically (`sensor → physics → base`), so keep that folder next to it.
+
+For RL training, see [humanoid-rl-isaaclab](https://github.com/limxdynamics/humanoid-rl-isaaclab), which uses these models out of the box.
+
+## Specifications
+
+Values below are read from the HU_D04 model files.
+
+| | HU_D04 |
+| --- | --- |
+| Height (model) | ≈ 1.65 m |
+| Mass (URDF, base model) | ≈ 52.5 kg |
+| Legs | 2 × 6 DoF (hip pitch / roll / yaw, knee, ankle pitch / roll) |
+| Waist | 3 DoF (yaw / roll / pitch) |
+| Arms | 2 × 7 DoF (shoulder pitch / roll / yaw, elbow, wrist yaw / pitch / roll) |
+| Head | 2 DoF (yaw / pitch) |
+| Dexterous hand | 11 joints / 6 independent per hand |
+| Parallel linkages | Ankles (A/B push-rods) and waist roll / pitch |
+
+
+
+
合作采购请联系 opensource@limxdynamics.com
+ +# LimX 人形机器人描述文件 + +> **Oli 人形机器人开源生态的一部分。** 上手路线和其他 Oli 仓库:[Oli 快速上手](https://github.com/limxdynamics/limxdynamics/blob/main/doc/quickstart_cn.md#oli)。 + +**真实全尺寸人形机器人的开箱即用仿真模型 —— URDF、MuJoCo MJCF、USD 三种格式齐全。** + +几分钟内把逐际动力 Oli 人形机器人(HU_D04)接入 ROS、MuJoCo 或 NVIDIA Isaac Sim,自带真实执行器参数与并联踝关节结构。 + +[模型一览](#模型一览) · [快速开始](#快速开始) · [规格参数](#规格参数) · [生态](#生态) · [引用](#引用) + +
| **HU_D04** 全尺寸人形 | `HU_D04_01` | 31 | URDF · MJCF · USD |
+|
| **HU_D04 + 夹爪** | `HU_D04_01_with_gripper` | 39
| **HU_D04 + 灵巧手** | `HU_D04_01_with_hand` | 53
| HU_D03 *(已停止维护)* | `HU_D03_03` | 31 | URDF · MJCF · USD |
+
+> **注意:** 目前仅基础款 HU_D04 提供 MuJoCo MJCF。HU_D03 已停止维护,新项目请使用 HU_D04。
+
+## 快速开始
+
+```bash
+git clone --depth 1 https://github.com/limxdynamics/humanoid-description.git
+cd humanoid-description
+```
+
+> 建议加 `--depth 1`:网格和 USD 文件较大,完整历史体积不小。
+
+### MuJoCo(30 秒上手)
+
+```bash
+pip install mujoco
+python -m mujoco.viewer --mjcf=HU_D04_description/xml/HU_D04_01.xml
+```
+
+或在 Python 中加载:
+
+```python
+import mujoco
+
+model = mujoco.MjModel.from_xml_path("HU_D04_description/xml/HU_D04_01.xml")
+data = mujoco.MjData(model)
+print(f"{model.nu} actuators, {model.nsensor} sensors") # 31 actuators, 65 sensors
+
+for _ in range(1000):
+ data.ctrl[:] = 0.0 # 关节力矩(N·m)—— 在这里接入你的控制器
+ mujoco.mj_step(model, data)
+```
+
+> `HU_D04_01.xml` 是完整仿真模型(执行器、传感器、闭链连杆);`HU_D04_01_vis.xml` 是仅用于可视化的轻量版本。
+
+### ROS 1 / ROS 2
+
+克隆到工作空间后直接编译 —— 每个 `HU_D0x_description` 目录都是同时兼容 catkin 和 colcon 的 ROS 包:
+
+```bash
+cd ~/ros_ws/src
+git clone --depth 1 https://github.com/limxdynamics/humanoid-description.git
+cd ..
+catkin_make # ROS 1
+colcon build # ROS 2
+```
+
+在 RViz 中查看(ROS 1,需安装 `urdf_tutorial`):
+
+```bash
+roslaunch urdf_tutorial display.launch model:=$(rospack find HU_D04_description)/urdf/HU_D04_01.urdf
+```
+
+在 RViz 中查看(ROS 2):
+
+```bash
+URDF=$(ros2 pkg prefix HU_D04_description)/share/HU_D04_description/urdf/HU_D04_01.urdf
+ros2 run robot_state_publisher robot_state_publisher --ros-args -p robot_description:="$(cat $URDF)" &
+ros2 run joint_state_publisher_gui joint_state_publisher_gui &
+rviz2 # Add → RobotModel,Fixed Frame 设为 base_link
+```
+
+### Isaac Sim / Isaac Lab
+
+在 Isaac Sim 中直接打开 `HU_D04_description/usd/HU_D04_01.usd`(**File → Open**),或作为 reference 添加到场景中。顶层 USD 会自动组合 `usd/configuration/` 中的各层(`sensor → physics → base`),请保持该目录与其在同一位置。
+
+强化学习训练请参考 [humanoid-rl-isaaclab](https://github.com/limxdynamics/humanoid-rl-isaaclab),可直接使用本仓库模型。
+
+## 规格参数
+
+以下数值读取自 HU_D04 模型文件。
+
+| | HU_D04 |
+| --- | --- |
+| 身高(模型) | ≈ 1.65 m |
+| 质量(URDF,基础款) | ≈ 52.5 kg |
+| 腿部 | 2 × 6 自由度(髋 pitch / roll / yaw、膝、踝 pitch / roll) |
+| 腰部 | 3 自由度(yaw / roll / pitch) |
+| 手臂 | 2 × 7 自由度(肩 pitch / roll / yaw、肘、腕 yaw / pitch / roll) |
+| 头部 | 2 自由度(yaw / pitch) |
+| 灵巧手 | 每只手 11 个关节,6 个独立 |
+| 并联连杆 | 踝关节(A/B 推杆)、腰部 roll / pitch |
+
+