Non-Linear Identification and Virtual Validation of Multi-Rotor Dynamics
Year: 2026
Type: article
Abstract: High-fidelity virtual testing can reduce the cost, time, and risk of multirotor controller development; however, conventional physics-based simulators often suffer from sim-to-real mismatch due to unmodeled actuator/sensor nonlinearities and coupled dynamics, which is particularly challenging for hexarotors with higher input dimensionality. This paper proposes a real-flight-data-driven pipeline to identify and virtually validate hexarotor dynamics using a MIMO Hammerstein–Wiener model. Flight input–output data from a DJI F550 hexarotor were collected using onboard sensors and an OptiTrack motion-capture system. The Hammerstein–Wiener structure captures static input/output nonlinearities and linear time-invariant dynamics, and the linear-block order was selected via the Akaike Information Criterion. The identified model was integrated into a ROS–MATLAB/Simulink–Gazebo co-simulation environment for closed-loop evaluation using a PX4-style cascaded PID controller. In a zig-zag mission, fine-tuned gains reduce roll attitude Root-mean-square error (RMSE) by 24.6% ( $0.0362^{\circ }\to 0.0273^{\circ }$ ) and Y-position RMSE by 17.8% (0.1937 m $\to 0.1593$ m) versus the PX4 default, while aggressive gains induce oscillations and increase pitch and yaw RMSE by $8.38\times $ and $10.33\times $ . These results support the Hammerstein–Wiener model as a repeatable virtual testbed for rapid controller screening before real-flight trials.
Source: IEEE Access
Authors: Hsin-Lin Chiu, Cheng-Jung Yang, Po-Tuan Chen
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,075
OpenAlex ID: W7138928882
FWCI:
Cites:
Cited by:
Related to: