Identification of time-varying models for flapping-wing micro aerial vehicles

Abstract

The demand for always smaller, more manoeuvrable and versatile unmanned aerial vehicles cannot be met with conventional manned flight approaches. This has led engineers to seek inspiration in nature, giving rise to the bio-inspired flapping-wing micro aerial vehicle (FWMAV). FWMAVs achieve a remarkable flight performance at small scales, however their flight mechanics are extremely complex. This hinders the development of effective dynamic models, which are essential for simulation, design and advanced controller development, and would enhance the performance and autonomy of such vehicles. This thesis addresses the challenge of modelling flapping-wing dynamics, using free-flight and wind tunnel data, with the aim of devising new models that are both accurate and computationally simple enough for control and simulation applications. The research is based on a test vehicle, i.e. the DelFly, developed at TU Delft. To meet the stated objectives, two modelling approaches are developed. The first approach is based on free-flight system identification and yields time-varying grey-box state-space models of the full vehicle dynamics, covering different flight conditions. The second approach results in physically meaningful phenomenological models of the aerodynamics specifically, accounting for complex effects such as the clap-and-fling mechanism and the interaction between the unsteady wing wake and tail. In addition to the modelling, recommendations for effective FWMAV flight testing are put forth, and a sensor fusion method is developed to advantageously combine on-board sensor data with off-board motion tracking data. All the developed models are accurate and computationally inexpensive, and the approaches can be generalised to comparable FWMAVs. While each model is best suited for different applications, thanks to its specific properties, all the developed models pave the way for new work in design, simulation, and control of FWMAVs.