Robots Atlas>ROBOTS ATLAS
LeRobot

Development · Control & Planning

LeRobot

LeRobot 1.5 + Unitree G1 support·Hugging Face

Active Open source API available
CATEGORYDevelopment · Control & Planning
READINESSTRL 6
ADOPTION SCALEGrowing Community
LICENSESApache-2.0
FIRST RELEASE2024

**LeRobot** is Hugging Face's flagship AI-robotics project, announced in May 2024 (lead: Remi Cadene, ex-Tesla Autopilot). Mission: 'democratize AI robotics' by providing a complete, open pipeline — from affordable hardware (~USD 250 SO-100 Standard Open Arm gates designed by The Robot Studio + Hugging Face) to pretrained models and datasets on the HuggingFace Hub.

The library contains SOTA policy implementations in PyTorch: **ACT** (Action Chunking with Transformers, the original from Mobile Aloha), **Diffusion Policy** (Columbia, actions as conditional diffusion), **VQ-BeT** (Vector-Quantized Behavior Transformer), **TDMPC** (Temporal Difference Model Predictive Control), **π0** (since April 2025 — official integration with Physical Intelligence). Each model has a standard interface: `Policy(observation) -> action`, compatible with gym and LeRobot datasets.

Datasets: LeRobot introduces the **LeRobotDataset** — a standard demonstration format based on Parquet + Hugging Face Datasets, with metadata (camera intrinsics, action space spec, robot model). 50+ public datasets (PushT, ALOHA Cube Transfer, Mobile Aloha Towel Folding, SO-100 demos). Hardware: official support for the **SO-100** (a 5-DOF leader + follower arm pair for ~USD 250, 3D printing + standard DYNAMIXEL servos), **Koch Arm**, **ALOHA**, **Stretch RE2**, **Unitree G1**, plus prototype support for custom robots.

Ecosystem: native integration with HuggingFace Spaces (interactive demos), a Trainer API (fine-tune a policy in one line), TensorBoard logging, video recording. Active development — ~30 commits/week, ~80 active contributors (Q1 2026). Targets: academic research, education, hobbyist robotics, startups testing VLAs on their own task domains.

Type & Roles
Software types
VLA / Foundation Model
SDK

An SDK (Software Development Kit) is a curated set of libraries, interfaces, tools, sample code, and documentation intended for building applications and integrating with a specific hardware device, platform, or service. In robotics, an SDK typically exposes device control, telemetry, sensor access, configuration, and execution functions, significantly reducing the time-to-first-integration for developers targeting a specific robot or platform.

Developer Tool

A Developer Tool is software designed to support the development workflow, including configuration, debugging, testing, monitoring, validation, and integration of robotic and embedded systems. Examples include IDE plugins, visual debuggers, log analysers, hardware-in-the-loop (HIL) test harnesses, and code-generation utilities specific to robotics platforms.

Runtime

A Runtime is the environment or execution layer used to run code, load libraries, manage dependencies, and operate applications or services — either in real time or during normal system operation. In robotics this includes real-time operating system (RTOS) runtimes, ROS 2 executor runtimes, containerised execution environments (Docker, podman), and embedded C++ runtimes on microcontrollers.

Select an item to see its description.
Main category
Control & PlanningSDKsDeveloper ToolsSimulation & Digital Twins
Roles in robotics ecosystem
Motion PlanningPerceptionComputer VisionDeveloper EnablementAPI Access
Robot Control

Robot Control denotes the role of software responsible for motion control, command execution, coordination of actuating elements and the direct operational logic of the robot.

Select an item to see its description.
Software family
Family

A family of open Vision-Language-Action (VLA) and foundation models for robotics: OpenVLA (Stanford/Berkeley), LeRobot (Hugging Face), RoboAgent (CMU), RT-2 (Google DeepMind, publication). Trained on datasets such as Open X-Embodiment, BridgeData V2, and RoboNet.

Maturity & Adoption
6 / 9
Demonstration phase
ResearchPrototypeProduction
Adoption scaleGrowing Community
Maintenance statusActively Maintained
First release2024
Last update20 May 2026
Deployments

Education: courses at Stanford CS 224R, MIT 6.4200, EPFL Robot Learning use LeRobot as a baseline. The SO-100 has become a popular hobbyist arm (~5,000 units printed by Q1 2026 according to community reports). Startups: Skild AI, Nimble Robotics, Realbotix use LeRobot for prototyping. HuggingFace Spaces: 200+ public demos with LeRobot policies. Showcased at CES 2025 (Hugging Face booth with 6 SO-100 robots performing tasks).

Community

github.com/huggingface/lerobot ~9.5k★, ~1,100 forks (Q1 2026). PyPI `lerobot` ~30k downloads/month. HuggingFace Hub has ~50 LeRobot datasets + 80 policy checkpoints. LeRobot Discord ~6k active members. Active development: ~30 commits/week, ~80 active contributors.

ROS supportCompatibility with ROS / ROS 2 ecosystem
Community ROS 2 WrapperWrapper ROS 2 tworzony i utrzymywany przez społeczność, nie przez producenta
System capabilities
Open source
Source code is publicly available under an open-source license — enables security audits, custom modifications, and integration without licensing barriers.
Real-time capable
Designed with timing-determinism guarantees — meets the requirements of control loops, safety systems, and tasks demanding low, predictable latency.
×
⟨/⟩
API available
The software exposes a programmable interface (REST, gRPC, SDK, or language bindings) that enables automation and integration with other systems.
📦
Pre-built / binary
Distributed as ready-to-use binary packages, container images, or installers — no need to build from source.
Programming languages
PythonCUDA
Operating systems
Ubuntu 22.04Ubuntu 20.04DebianmacOS
Ubuntu 24.04

Ubuntu 24.04 LTS 'Noble Numbat' — supported until April 2029. The host for ROS 2 Jazzy.

Select an item to see its description.
Minimum hardware requirements
Minimum hardware requirements
CPUInference of small policies: dual-core x86-64 ≥ 2.4 GHz. ACT/Diffusion Policy training: quad-core ≥ 3 GHz with a decent GPU.
RAM (GB)16
GPUInference: 1× RTX 3060 12 GB or Jetson Orin Nano for ACT/Diffusion. π0 fine-tuning in LeRobot: 1× A100 40 GB. ACT training from scratch: 1× RTX 4090 (~2-6 hours per task).
Disk (GB)50

PyPI `lerobot` (`pip install lerobot`). SO-100 hardware can be 3D printed — STEP files + BOM are provided in the repo. Requires DYNAMIXEL servos (~USD 150 per set) + Raspberry Pi 5 or a mini-PC.

Packaging & distribution
Package managers
pip / PyPIGitHub Releases / GitHub Actions ArtifactsDocker / Docker Hub
CPU architectures
x86_64 (AMD64)NVIDIA GPU (CUDA – x86_64)ARM64 / AArch64NVIDIA Jetson – AArch64 (JetPack)Apple Silicon – AArch64 (macOS)
Installation difficulty
LevelModerate
Protocols and interfaces
Communication protocols
gRPCREST API (HTTP/HTTPS)WebSocketUSB (Universal Serial Bus)ROS 2 Topics
Hardware interfaces
USB 2.0USB 3.0 / 3.1 Gen 1Ethernet 1000BASE-T (Gigabit Ethernet)
Latency classes
Soft Real-Time (20–100 ms)Soft Real-Time (100–500 ms)
Deployment types
Local WorkstationEdgeContainerizedOn Robot
Supported simulators
MuJoCo
PyBullet / Bullet3
Gazebo Harmonic
NVIDIA Isaac Sim
Licenses
Apache-2.0Apache License 2.0v2.0

License family: Permissive

ModificationDistributionCommercial useSublicensingPrivate useROS-compatibleOSI approvedFSF Free/LibreRequires attributionPatent grant
Version history
LeRobot 1.5 + Unitree G1 supportNov 2025

Official support for the Unitree G1 humanoid (datasets + policy fine-tuning recipes).

LeRobot 1.0 (stable API)Jun 2025

API stabilization, backward compatibility guarantee, official release notes per minor version.

LeRobot + π0 integrationApr 2025

Official integration of π0 (Physical Intelligence) with LeRobot — fine-tune π0 in a single line of code.

LeRobot 0.5 + VQ-BeTNov 2024

Introduction of VQ-BeT (Vector-Quantized Behavior Transformer) as a native policy. Reformulates action discretization.

SO-100 Arm releaseAug 2024

Launch of the affordable open 5-DOF SO-100 arm (~USD 250), 3D-printed, with BOM and assembly instructions.

LeRobot 0.1 (announcement)May 2024

First public announcement of the Hugging Face library. Support for ACT, Diffusion Policy, TDMPC. PushT baseline.