Data-Driven Controls of a Flapping Wing Unmanned Aerial Vehicle Inspired by Monarch Butterfly
Open AccessThis dissertation studies dynamical modeling, stability analysis, optimal controls, and data-driven control policies for a flapping wing unmanned aerial vehicle, inspired by Monarch butterflies. The dynamics and control of flapping wing aerial vehicles are challenging, as they are represented by infinite-dimensional nonlinear time-varying systems, where unsteady aerodynamics is coupled with structural deformation of wings and body dynamics in a sophisticated manner. To address this, we present a global formulation of geometric dynamic model, to which various control strategies are developed, including Lyapunov-Floquet stability, optimal control, imitation learning, and vision-based controls.First, inspired by flight characteristics observed in live Monarch butterflies, a new dynamic model is presented to account the effects of low-frequency flapping and abdomen undulation. The dynamics are developed according to an intrinsic formulation of Lagrangian mechanics on an abstract Lie group. This provides an elegant, global formulation of the dynamics for flapping-wing aerial vehicles, avoiding complexities and singularities associated with local coordinates.Second, an optimal periodic motion that minimizes the energy variations is constructed, and a feedback control system is proposed to asymptotically stabilize it according to the Floquet stability theory without averaging. Furthermore, the beneficial effects of abdomen undulation in the flight of Monarch butterflies are investigated in terms of energy consumption and stability properties, which are compared to those of other insects. Third, a new framework for constrained imitation learning is proposed to transform a set of optimal trajectories into data-driven feedback control. The proposed structured learning enhances the stability properties of the controlled dynamics without the need for generating additional optimal trajectories online, thereby improving the data efficiency significantly.Finally, a vision-based control scheme is proposed to avoid estimating the state of the flapping wing aerial vehicle completely in real time. Instead of training a monolithic network, we propose a modular construction where a pose estimation network and a control network are concatenated, which are trained alternatingly to achieve the complex tasks of end-to-end perception and control efficiently. All of the dynamic models and the control schemes proposed in this dissertation overcome the common restrictions of flapping wing aerial vehicles, namely high frequency flapping to justify the linearization and averaging assumptions. The proposed geometric formulation establishes the foundation of the nonlinear controls for the sophisticated dynamics of flapping wing flights without such simplification.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.
| Thumbnail | Title | Date Uploaded | Visibility | Actions |
|---|---|---|---|---|
|
|
KodihalliChandrappa_gwu_0075A_16626.pdf | 2024-01-11 | Open Access |
|