---

---

CrazyPlayGround - Collection of CrazyFlie Environments

A collection of Crazyflie reinforcement learning environments built on Isaac Lab. Trained policies can be deployed on real Crazyflie 2.1 drones via the scripts present in execution/.

⚠️ This repository is under construction. If you find a bug or a problem, do not hesitate to open an issue! ⚠️

Quick start

1. Install Isaac Lab

Install Isaac Lab 2.3.2 (Isaac Sim 4.5+).

2. Install CrazyPlayGround

Install with the extras you need:

pip install -e "source/CrazyPlayGround[rl]"      # Simulation / Training
pip install -e "source/CrazyPlayGround[deploy]"  # Deployment Hardware
pip install -e "source/CrazyPlayGround[all]"    

3. Verify and train

python scripts/list_envs.py
python scripts/skrl/train.py --task=Vel-Hovering --num_envs=4096

Docker

A Dockerfile is provided for reproducible training (Linux + NVIDIA GPU required):

docker login nvcr.io                    
docker build -t crazyplayground -f docker/Dockerfile .
docker run --gpus all -it --rm -v $(pwd)/logs:/workspace/crazyplayground/logs crazyplayground

Once inside the container, use the following command to verify and train (inside the container the --headless is mandatory):

/workspace/isaaclab/isaaclab.sh -p scripts/list_envs.py
/workspace/isaaclab/isaaclab.sh -p scripts/skrl/train.py --task=Vel-Hovering --num_envs=4096 --headless

Documentation

Full documentation (environments, controller architecture, configuration, real-drone deployment) lives in docs/ and is built with MkDocs Material, an online version of the documentation will soon be available