docs: expand org card with stats, sim engines, correct email and benchmark count
Browse files
README.md
CHANGED
|
@@ -11,6 +11,64 @@ thumbnail: >-
|
|
| 11 |
https://cdn-uploads.huggingface.co/production/uploads/679bec571de528e679aae409/KR7hrL_kADf-oPybj7Q0C.png
|
| 12 |
---
|
| 13 |
|
| 14 |
-
|
| 15 |
|
| 16 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
https://cdn-uploads.huggingface.co/production/uploads/679bec571de528e679aae409/KR7hrL_kADf-oPybj7Q0C.png
|
| 12 |
---
|
| 13 |
|
| 14 |
+
<div align="center">
|
| 15 |
|
| 16 |
+
# OpenRAL
|
| 17 |
+
|
| 18 |
+
**An open-source operating layer for embodied AI** — OpenRAL unifies fast policies, slow reasoning, and classical control into one typed, traceable, safety-first runtime for deployable robot agents.
|
| 19 |
+
|
| 20 |
+
[](https://github.com/OpenRAL/openral)
|
| 21 |
+
[](https://github.com/OpenRAL/openral/blob/master/LICENSE)
|
| 22 |
+
[](https://github.com/OpenRAL/openral)
|
| 23 |
+
[](https://docs.ros.org/en/jazzy/)
|
| 24 |
+
[](https://discord.gg/3paXT2bVyB)
|
| 25 |
+
|
| 26 |
+
</div>
|
| 27 |
+
|
| 28 |
+
---
|
| 29 |
+
|
| 30 |
+
## What is OpenRAL?
|
| 31 |
+
|
| 32 |
+
OpenRAL is a typed, layered runtime that sits between a robot's motor API and a task planner:
|
| 33 |
+
|
| 34 |
+
- **Typed runtime** — eight well-defined layers (HAL → Sensors → World State → rSkill → Reasoning → WAM → Safety → Observability) with Pydantic v2 contracts at every boundary.
|
| 35 |
+
- **rSkill packaging** — HuggingFace Hub artifacts with weights, a `rskill.yaml` manifest, quantization hints, latency budgets, and reproducible eval. Install like a model: `openral rskill install OpenRAL/rskill-smolvla-libero`.
|
| 36 |
+
- **Planning kernel** — a slow LLM-based reasoner (S2) emitting typed tool-calls, and a fast visuomotor policy (S1, 30–200 Hz) executing dispatched skills.
|
| 37 |
+
- **Safety kernel** — C++ separate process, deny-by-default. Python proposes; C++ disposes.
|
| 38 |
+
|
| 39 |
+
---
|
| 40 |
+
|
| 41 |
+
## What's shipped
|
| 42 |
+
|
| 43 |
+
| | |
|
| 44 |
+
|---|---|
|
| 45 |
+
| 🤖 **16 robot platforms** | SO-100/SO-101, Franka Panda, UR5e/UR10e, ALOHA bimanual, OpenArm v2, Unitree H1/G1, Rethink Sawyer, Fourier GR1 |
|
| 46 |
+
| 🧠 **31 rSkills** | SmolVLA, π0.5, xVLA, MolmoAct2, ACT, Diffusion Policy, RLDX-1, GR00T N1.7, RT-DETR, LocateAnything, Qwen3.5-4B VLM |
|
| 47 |
+
| 🎮 **12 benchmark scenes** | LIBERO, MetaWorld, ManiSkill3, SimplerEnv, RoboCasa, gym-aloha, gym-pusht, Isaac Sim |
|
| 48 |
+
| 🔬 **Sim engines** | MuJoCo, RoboCasa, Isaac Sim |
|
| 49 |
+
| 📡 **Sensors** | RealSense D435/D435i/D415, Luxonis OAK-D Pro, USB UVC, Robotiq FT-300 |
|
| 50 |
+
|
| 51 |
+
---
|
| 52 |
+
|
| 53 |
+
## Quick start
|
| 54 |
+
|
| 55 |
+
```bash
|
| 56 |
+
curl -fsSL https://raw.githubusercontent.com/OpenRAL/openral/master/scripts/install.sh | bash
|
| 57 |
+
openral doctor
|
| 58 |
+
openral rskill install OpenRAL/rskill-smolvla-libero
|
| 59 |
+
openral sim run scenes/benchmark/libero_spatial.yaml
|
| 60 |
+
```
|
| 61 |
+
|
| 62 |
+
---
|
| 63 |
+
|
| 64 |
+
## Community & support
|
| 65 |
+
|
| 66 |
+
- 💬 **[Discord](https://discord.gg/3paXT2bVyB)** — questions, help, design chat
|
| 67 |
+
- 📧 **[hello@openral.com](mailto:hello@openral.com)** — general enquiries
|
| 68 |
+
- 🔒 **Security disclosures:** [security@openral.com](mailto:security@openral.com) — never a public issue
|
| 69 |
+
|
| 70 |
+
---
|
| 71 |
+
|
| 72 |
+
## License
|
| 73 |
+
|
| 74 |
+
OpenRAL is **Apache-2.0**. Third-party model weights keep their own upstream licenses; the loader surfaces each weight's license at install time.
|