← Back ICRA 2026

How to Model Your Crazyflie Brushless

Alexander Gräfe, Christoph Scherer, Wolfgang Hoenig, Sebastian Trimpe

PDF

AI summary

Key figure (auto-extracted from paper)

An accurate, open-source dynamics model and MJX-based simulator for the Crazyflie Brushless enables successful sim-to-real transfer of reinforcement learning controllers for agile nano-quadcopter maneuvers.

Crazyflie Brushless quadcopter dynamics reinforcement learning sim-to-real transfer MJX simulator parameter identification

Problem

The newly released Crazyflie Brushless lacks an accurate, open-source dynamics model and simulator, creating a barrier to developing agile control policies and bridging the sim-to-real gap for nano-quadcopter research.

Approach

The authors derive a first-principles dynamics model, identify key parameters through physical experiments, and implement it in a highly parallelized MJX simulator. They validate the model by training end-to-end neural network controllers via reinforcement learning and testing sim-to-real transfer.

Key results

Derived a first-principles dynamics model with experimentally identified thrust, torque, and inertia parameters
Model accurately predicts hardware trajectories over 200ms horizons, outperforming existing CF 2.1 simulators
Successfully trained RL controllers for position control and acrobatic backflips that transfer to real hardware
Quantified domain randomization requirements for sim-to-real transfer, revealing key model accuracy metrics

Why it matters

Provides researchers with a validated simulation platform to rapidly develop and test agile control algorithms for the new Crazyflie Brushless without costly hardware trial-and-error.

Abstract

The Crazyflie quadcopter is widely recognized as a leading platform for nano-quadcopter research. In early 2025, the Crazyflie Brushless was introduced, featuring brushless motors that provide around 50% more thrust compared to the brushed motors of its predecessor, the Crazyflie 2.1. This advancement has opened new opportunities for research in agile nano-quadcopter control. To support researchers utilizing this new platform, this work presents a dynamics model of the Crazyflie Brushless and identifies its key parameters. Through simulations and hardware analyses, we assess the accuracy of our model. We furthermore demonstrate its suitability for reinforcement learning applications by training an end-to-end neural network position controller and learning a backflip controller capable of executing two complete rotations with a vertical movement of just 1.8 meters. This showcases the model’s ability to facilitate the learning of controllers and acrobatic maneuvers that successfully transfer from simulation to hardware. Utilizing this application, we investigate the impact of domain randomization on control performance, offering valuable insights into bridging the sim-to-real gap with the presented model. We have open-sourced the entire project, enabling users of the Crazyflie Brushless to swiftly implement and test their own controllers on an accurate simulation platform.

Index terms

Aerial Systems: Mechanics and Control Dynamics Reinforcement Learning