Modeling and Identification of Multirotor Drone Dynamics for Onboard MPC Motion Planning

This paper presents a nonlinear model of multirotor drone dynamics selected specifically to be suited for onboard motion planning using Model Predictive Control (MPC), and presents a generally applicable procedure to identify the model parameters solely based on outdoor flight data. The model stems from a trade-off between prediction accuracy and computational complexity, approximating attitude and thrust control loops as low order linear time-invariant subsystems but without linearizing the thrust vector orientation, and including linear aerodynamic drag and a simplified battery voltage dependency. An open loop simulation compared to recorded data from a free flight maneuver motivates the proposed model complexity in contrast to further simplifications of the proposed model.