Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Robot Learning: A Tutorial

Francesco Capuano, Caroline Pascal, Adil Zouitine, Thomas Wolf, Michel Aractingi

TL;DR

Robot Learning: A Tutorial surveys the shift from classical dynamics-based robotics to data-driven methods, emphasizing how large-scale robotics data and learning algorithms enable robust perception-to-action control across embodiments. It couples foundational RL and imitation-learning techniques with modern generalist, language-conditioned Vision-Language-Action models, and highlights practical open-source tools like LeRobotDataset and lerobot for end-to-end experimentation. The work appraises strengths and limitations of RL, BC, and generative modeling (VAEs, diffusion, flow matching), and introduces scalable architectures (ACT, DP, π0, SmolVLA) backed by large-scale datasets and open pipelines. Through code examples and open data, it argues for a democratized, modular approach to building generalist robot policies that can adapt across tasks and robots, accelerating progress in real-world robotics.

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 $\texttt{lerobot}$.

Robot Learning: A Tutorial

TL;DR

Robot Learning: A Tutorial surveys the shift from classical dynamics-based robotics to data-driven methods, emphasizing how large-scale robotics data and learning algorithms enable robust perception-to-action control across embodiments. It couples foundational RL and imitation-learning techniques with modern generalist, language-conditioned Vision-Language-Action models, and highlights practical open-source tools like LeRobotDataset and lerobot for end-to-end experimentation. The work appraises strengths and limitations of RL, BC, and generative modeling (VAEs, diffusion, flow matching), and introduces scalable architectures (ACT, DP, π0, SmolVLA) backed by large-scale datasets and open pipelines. Through code examples and open data, it argues for a democratized, modular approach to building generalist robot policies that can adapt across tasks and robots, accelerating progress in real-world robotics.

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 .
Paper Structure (39 sections, 39 equations, 39 figures, 1 algorithm)

This paper contains 39 sections, 39 equations, 39 figures, 1 algorithm.

Table of Contents

  1. Introduction
  2. LeRobotDataset
  3. The dataset class design
  4. Code Example: Batching a (Streaming) Dataset
  5. Code Example: Collecting Data
  6. Classical Robotics
  7. Explicit and Implicit Models
  8. Different Types of Motion
  9. Example: Planar Manipulation
  10. Adding Feedback Loops
  11. Limitations of Dynamics-based Robotics
  12. Robot (Reinforcement) Learning
  13. A (Concise) Introduction to RL
  14. Real-world RL for Robotics
  15. Sample-efficient RL
  16. ...and 24 more sections

Figures (39)

  • Figure 1: lerobot is the open-source library for end-to-end robotics developed by Hugging Face. The library is vertically integrated on the entire robotics stack, supporting low-level control of real-world robot devices, advanced data and inference optimizations, as well as SOTA robot learning methods with simple implementations in pure Pytorch.
  • Figure 2: Overview of methods to generate motion (clearly non-exhausitve, see bekrisStateRobotMotion2024). The different methods can be grouped based on whether they explicitly (dynamics-based) or implicitly (learning-based) model robot-environment interactions.
  • Figure 3: Different kinds of motions are achieved with potentially very different robotic platforms. From left to right, top to bottom: ViperX, SO-100, Boston Dynamics' Spot, Open-Duck, 1X's NEO, Boston Dynamics' Atlas. This is an example list of robotic platforms and is (very) far from being exhaustive.
  • Figure 4: Cheaper, more accessible robots are starting to rival traditional platforms like the Panda arm platforms in adoption in resource-constrained scenarios. The SO-100, in particular, has a cost in the 100s of Euros, and can be entirely 3D-printed in hours, while the industrially-manufactured Panda arm costs tens of thousands of Euros and is not openly available.
  • Figure 5: The SO-100 arm is a 6-dof manipulator arm. Preventing some of its joints (shoulder pane, wrist flex and wrist roll) from actuating, it can be represented as a traditional 2-dof planar manipulator (the gripper joint in the end-effector is not considered towards the count of the degrees of freedom used to produce motion).
  • ...and 34 more figures