This repository contains the source code for the "Robot Learning: A Tutorial" report. This tutorial covers many of the most pressing aspects in modern robot learning, and provides practice examples using lerobot, the robot-learning library developed by Hugging Face.

You’re more than welcome to contribute to the next edition of the tutorial! Simply open an issue, tag @fracapuano, and start a discussion about the scope and content you’d like to add. Check out CONTRIBUTING.md for more details 😊 All merged pull requests will receive public acknowledgment in the main body of the tutorial. Items marked with an empty [ ] in the following Table of Contents are open for community contribution!

• 1.1 lerobot Dataset
  • 1.1.1 The dataset class design
• 1.2 Code Example: Batching a (Streaming) Dataset
• 1.3 Code Example: Collecting Data

• 2.1 Explicit and Implicit Models
• 2.2 Different Types of Motion
• 2.3 Example: Planar Manipulation
  • 2.3.1 Adding Feedback Loops
• 2.4 Limitations of Dynamics-based Robotics

• 3.1 A (Concise) Introduction to RL
• 3.2 Real-world RL for Robotics
• 3.3 Code Example: Real-world RL
• 3.4 Limitations of RL in Real-World Robotics: Simulators and Reward Design

• 4.1 A (Concise) Introduction to Generative Models
  • 4.1.1 Variational Auto-Encoders
  • 4.1.2 Diffusion Models
  • 4.1.3 Flow Matching
• 4.2 Action Chunking with Transformers
  • 4.2.1 Code Example: Training and Using ACT in Practice
• 4.3 Diffusion Policy
  • 4.3.1 Code Example: Training and Using Diffusion Policies in Practice
• 4.4 Optimized Inference
  • 4.4.1 Code Example: Using Async Inference

• 5.1 Preliminaries: Models and Data
• 5.2 Modern VLAs
  • 5.2.1 VLMs for VLAs
• 5.3 PI0
  • 5.3.1 Code Example: Using PI0
• 5.4 SmolVLA
  • 5.4.1 Code Example: Using SmolVLA
• 5.5 GR00T (1/2)
  • 5.5.1 Code Example: Using GR00T
• 5.6 PI05
  • 5.6.1 Code Example: Using PI05
• Large-scale datasets
  • Open-X
  • DROID
  • BEHAVIOR

• 6.1 Post training VLAs
  • 6.1.1 From Imitation to Refinement
  • 6.1.2 EXPO
• 6.2 World Models for robotics
  • 6.2.1 Cosmos
  • 6.2.2 World Models (1X)
  • 6.2.3 Sima and Genie 1

• 7.1 Conclusions

The written content of this book is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

All source code examples in the snippets/ directory are licensed under the MIT License.