|
|
--- |
|
|
license: mit |
|
|
pipeline_tag: robotics |
|
|
library_name: lerobot |
|
|
tags: |
|
|
- action-chunking-with-transformers |
|
|
--- |
|
|
|
|
|
# Robot Learning: A Tutorial - ACT Model |
|
|
|
|
|
This repository contains a model checkpoint (`act-resnet18-upside-down-side-v0.1`) associated with the paper "[Robot Learning: A Tutorial](https://huggingface.co/papers/2510.12403)". This tutorial navigates the landscape of modern robot learning, providing practical examples using the `lerobot` library developed by Hugging Face. The model provided here is an instance of the Action Chunking with Transformers (ACT) framework. |
|
|
|
|
|
- **Paper:** [Robot Learning: A Tutorial](https://huggingface.co/papers/2510.12403) |
|
|
- **GitHub Repository:** [https://github.com/fracapuano/robot-learning-tutorial](https://github.com/fracapuano/robot-learning-tutorial) |
|
|
- **Project Page (Hugging Face Space):** [https://huggingface.co/spaces/lerobot/robot-learning-tutorial](https://huggingface.co/spaces/lerobot/robot-learning-tutorial) |
|
|
|
|
|
## Abstract |
|
|
Robot learning is at an inflection point, driven by rapid advancements in machine learning and the growing availability of large-scale robotics data. This shift from classical, model-based methods to data-driven, learning-based paradigms is unlocking unprecedented capabilities in autonomous systems. This tutorial navigates the landscape of modern robot learning, charting a course from the foundational principles of Reinforcement Learning and Behavioral Cloning to generalist, language-conditioned models capable of operating across diverse tasks and even robot embodiments. This work is intended as a guide for researchers and practitioners, and our goal is to equip the reader with the conceptual understanding and practical tools necessary to contribute to developments in robot learning, with ready-to-use examples implemented in $\texttt{lerobot}$. |
|
|
|
|
|
## Model Details |
|
|
This model checkpoint is configured as an Action Chunking with Transformers (ACT) policy, as defined in its `config.json`. It utilizes a `resnet18` vision backbone for processing visual observations and is designed for robotic action prediction. |
|
|
|
|
|
- **Model Type:** `act` (Action Chunking with Transformers) |
|
|
- **Vision Backbone:** `resnet18` |
|
|
- **Input Features:** |
|
|
- `observation.state`: State vector of shape `[6]` |
|
|
- `observation.images.up`: Visual input from an 'up' camera, shape `[3, 480, 640]` |
|
|
- `observation.images.side`: Visual input from a 'side' camera, shape `[3, 480, 640]` |
|
|
- **Output Features:** |
|
|
- `action`: Action vector of shape `[6]` |
|
|
- **Uses VAE:** `true` |
|
|
|
|
|
## Usage |
|
|
For detailed instructions on installation, training, and utilizing robot learning models within the `lerobot` ecosystem, please refer to the comprehensive [GitHub repository](https://github.com/fracapuano/robot-learning-tutorial) of the "Robot Learning: A Tutorial". The repository provides various code examples and guidance for implementing and experimenting with such models. |
|
|
|
|
|
## License |
|
|
The code examples within the "Robot Learning: A Tutorial" GitHub repository are licensed under the [MIT License](https://opensource.org/licenses/MIT). This model artifact is associated with that codebase. |
