System Identification Study for A Small-Scale Hybrid Unmanned Aerial Vehicle Using Differential Evolution Approach

Abstract

This research develops a dynamic model for a small-scale hybrid unmanned aerial vehicle (UAV) using the system identification approach, optimized by the Differential Evolution (DE) method, aiming to enhance the performance, stability, and controllability characteristics of this complex type of UAV. The methodology involved gathering data from simulated tests using LabVIEW and X-Plane 9, establishing the model structure, and then refining and training the model parameters. The differential evolution algorithm was applied in the tuning phase, comprising initialization, mutation, crossover, and selection, to efficiently optimize parameters for precision and performance. Flight data was collected from simulated hover and forward flight tests. The training dataset employed frequency-swept excitation, while the test dataset utilized doublet excitation for validation. Model validation was conducted through statistical analysis, demonstrating high prediction accuracy. The dynamic models demonstrated low mean square error with minimal standard deviation, and the predicted responses aligned closely with measured data, confirming robustness and accuracy. This study successfully applied the DE algorithm based on the system identification approach to develop and validate accurate dynamic models for a small hybrid UAV. The results establish a valuable framework for improving UAV performance and inspiring further advancements in system identification.