← AI GitHub

AI GitHub

LeRobot for robot demonstrations and policy training

Record robot demonstrations and train policies using reusable data and control tools.

Science and ResearchPublished 2 min read

Key Takeaways

  • LeRobot provides dataset conventions, policy implementations and integrations for supported hardware.
  • Start with a supported dataset and policy and match observation and action definitions.
  • Begin with a supported dataset or simulation before moving hardware.
  • A policy can behave poorly when camera placement or joint calibration changes.

How it works

LeRobot provides dataset conventions, policy implementations and integrations for supported hardware. Recorded observations and actions become inputs to imitation learning.

What you need

Start with a supported dataset and policy and match observation and action definitions. For hardware, verify calibration and workspace limits before enabling a policy.

What to try first

Begin with a supported dataset or simulation before moving hardware. Inspect timestamps, action units and camera views, then evaluate a policy on a short controlled task.

Inspect one episode after dropping a camera frame. Check whether the loader preserves time alignment between images, state and action labels.

Where to look for failures

A policy can behave poorly when camera placement or joint calibration changes. Use workspace limits and a stop mechanism during physical evaluation.

Related Franklin pages

Start with saved data or a simulator before deploying a policy on hardware. Verify observation layout, action conventions and task success using a known reference episode.

Franklin’s robotics coverage gives examples of where simulation, training and hardware meet. Franklin’s robotics coverage

Project documentation

Try it for your workflow

Explore the project’s documentation for setup and usage, or save this page to a collection for later.

Read the project documentation ↗