How to Model Your Crazyflie Brushless

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.

Figures (8)

• Figure 1: Overview of the different components of the cfb bitcraze2025crazyfliecrazyflie_bl_schematics. A Kalman filter estimates position, velocities and attitude from the data of different sensors. The controller uses these in combination with rotation rates of the imu to give motor inputs to the esc, which perform the electronic commutation of the brushless motors using the motor angle measured via the back electromotive force (EMF).
• Figure 2: Comparison between measured and model-predicted thrust and torque characteristics. The propeller model accurately captures both thrust and torque behavior across the operating range.
• Figure 3: Comparison of hardware measurements (dotted lines) versus simulation (solid lines). The simulation incorporated 4096 quadcopters with different mass, moment of inertia and thrust/torque scalings (differing at maximum 2% from the nominal ones). The line in the middle denotes the mean, whereas the shaded areas the minimum and maximum. We excluded position in the plots as it is almost constant.
• Figure 4: Comparison of hardware measurements (dotted lines) versus simulation (solid lines) of angular rates for the cf 2.1. We use the model parameters used in PyBullet-drones panerati2021learning for baselines comparisons.
• Figure 5: Target controller flying back and forth between two points 6m apart. The nn controller guides the cfb to its target with an accuracy of a few cm (cf. \ref{['tab:compare_controllers']}), achieving velocities of approximately 5ms during the maneuver.