Add README
Browse files
README.md
CHANGED
|
@@ -1,3 +1,46 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
+
library_name: lerobot
|
| 4 |
+
tags:
|
| 5 |
+
- robot
|
| 6 |
+
- imitation-learning
|
| 7 |
+
- act
|
| 8 |
+
- nexarm
|
| 9 |
+
- real-robot
|
| 10 |
---
|
| 11 |
+
|
| 12 |
+
# NexArm ACT: Pick and Place
|
| 13 |
+
|
| 14 |
+
This model is trained with [ACT (Action Chunking with Transformers)](https://arxiv.org/abs/2304.13705) on a **Hiwonder NexArm** 6-DOF robotic arm for a pick-and-place task.
|
| 15 |
+
|
| 16 |
+
## Training Details
|
| 17 |
+
|
| 18 |
+
| Item | Value |
|
| 19 |
+
|------|-------|
|
| 20 |
+
| **Policy** | ACT (Action Chunking with Transformers) |
|
| 21 |
+
| **Dataset** | 31 episodes, ~27K frames |
|
| 22 |
+
| **Cameras** | 2× USB cameras (front + wrist), 640×480 |
|
| 23 |
+
| **Training Time** | ~100K steps on NVIDIA RTX 5070 Ti |
|
| 24 |
+
| **Final Loss** | ~0.057 |
|
| 25 |
+
|
| 26 |
+
## Usage
|
| 27 |
+
|
| 28 |
+
```python
|
| 29 |
+
from lerobot.policies import ACTPolicy
|
| 30 |
+
from lerobot.rollout import rollout
|
| 31 |
+
|
| 32 |
+
policy = ACTPolicy.from_pretrained("CypherChen/NexarmControlModel")
|
| 33 |
+
# Run inference on your NexArm robot
|
| 34 |
+
rollout(policy, robot_type="nexarm_follower")
|
| 35 |
+
```
|
| 36 |
+
|
| 37 |
+
## Hardware
|
| 38 |
+
|
| 39 |
+
- **Robot**: Hiwonder NexArm (ESP32 + AT32 dual-chip)
|
| 40 |
+
- **Servos**: HX-series serial bus servos
|
| 41 |
+
- **Leader arm**: Drag-to-teach teleoperation
|
| 42 |
+
- **Follower arm**: Executes learned policy
|
| 43 |
+
|
| 44 |
+
## License
|
| 45 |
+
|
| 46 |
+
Apache 2.0
|