JuggleRL: Mastering Ball Juggling with a Quadrotor via Deep Reinforcement Learning

Overview

Highlights

Zero-shot sim-to-real deployment (no real data for training)
Calibrated dynamics + domain randomization reduce sim-to-real gap
Lightweight Communication Protocol (LCP) for low-latency streaming
Real-world performance: up to 462 hits (avg 311 across 10 trials)

This page hosts figures, demo videos, and links to paper & code.

Project Links

Paper: arXiv:2509.24892
Training: GitHub
ROS Pack: GitHub
NatNet SDK: GitHub

Key Metrics

462

Max real-world hits

311

Avg hits (10 trials)

Real data for training

Figures

Real-world Video (Bilibili)

BibTeX

@article{JuggleRL2025,
  title   = {JuggleRL: Mastering Ball Juggling with a Quadrotor via Deep Reinforcement Learning},
  author  = {Your Name and Coauthors},
  journal = {arXiv preprint arXiv:2509.24892},
  year    = {2025}
}