Intelligent Control for Fixed-Wing eVTOL Aircraft

Urban Air Mobility promises revolution to transportation by having "flying cars" over cities. Electric fixed-wing vertical take-off and landing aircraft gain popularity as they can swiftly traverse through cluttered metropolitan areas, while efficiently cover longer distances. Such modes of operation call for elevated level of precision, safety, and intelligence for flight control.

In the first part of this work, we begin by studying the theoretical benefits of these aircraft, followed with proposal of a novel unified control framework. It consists of nonlinear position and attitude controllers using forces and moments as inputs; and allocation modules that determines desired attitude and thruster input. Next, we present a composite adaptation scheme for linear-in-parameter dynamics models, which provides accurate estimation for wing and rotor forces in real-time based on information from a three-dimensional airflow sensor. Then, we introduce a design method for multirotor configuration that ensures robustness against failures.

In the second part, we use deep neural networks to learn unmodeled dynamics and incorporate them in control design. Spectral normalization that regulates Lipschitz constant of the network is applied for better generalization ability. The resultant network is utilized in a nonlinear feedback controller with contraction mapping, solving non-affine-in-control issues that arise. Finally, a delay compensation method that transforms controllers for an undelayed system into a sample-based predictive controller with numerical integration is proposed. It handles both first order and transport delays in actuators and balances between numerical accuracy and computation time to guarantee stability under hardware limitations.

Live Zoom Event: <https://caltech.zoom.us/j/84852988290>

Box Recordings for Caltech: <https://caltech.box.com/s/yscs22k3xqk05s9moxzay7sd8enw5qq4>