Simulated System Identification of Morphing Trailing Edge Fixed Wing UAV

Abstract

This report outlines a grey-box approach to controller training and simulation of a Morphing Trailing Edge (TE) fixed-wing Unmanned Aerial Vehicle (UAV). The method combines physical modelling through a Flight Dynamics Model (FDM) with synthetic flight data generation for non-linear system identification. This enables the development of an analytical model that captures the systems dynamic behaviour across a wide flight envelope. The project is part of the morphAIR initiative, aiming to enable safe flight and control of a morphing UAV demonstrator. The final goal is to establish a model-based design pipeline, capable of transitioning between conventional and morphing wing configurations while supporting accurate, simulation-based controller development. The thesis developed a working UAV simulation and excitation pipeline but struggles to accurately identify non-linear MIMO dynamics in Closed-Loop simulations. Although the SINDy-based non-linear system identification overfitted in both open- and closed-loop flights further experiments show that a mixed excitation model can still capture Open-Loop responses under simulation-based excitations.