Skip to content

Learning to Fly

Train a quadrotor policy in seconds

Visit official site

What learning-to-fly is

The project is an open source framework for training end-to-end control policies with deep reinforcement learning on a consumer laptop.

What you can do with it

  • Train a quadrotor policy from sensor input to control output
  • Move a trained policy onto a real Crazyflie
  • Watch training progress through the web UI

Getting started

  • Clone the repository with git clone
  • Install the required libraries for your system
  • Configure with cmake, build, then run training_headless

Pros & cons

✓ What we like

  • The full pipeline runs from sensors to control output
  • A web UI lets users monitor training

! What to watch out for

  • A cmake build step is part of the setup
  • Dependencies differ between Ubuntu and macOS

FAQ

How fast is training?

The material states training finishes in 18 seconds on a laptop.

Does it transfer to hardware?

The material says trained policies can move to real-world quadrotors.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review