Safety Critical Control for UAVs: Integrating Reconfigurable Control Barrier Functions with Sliding Mode Control

In this thesis, we propose a safety-critical controller design for nonlinear affine systems under actuator cyberattacks and model (i.e., system) uncertainties. To achieve this, our approach addresses two primary challenges: First, we propose a robust controller that employs a sliding mode-based control barrier function (SM-CBF) to adeptly handle model uncertainties. Second, we devise an LSTM (Long Short Term Memory)-based attack detection mechanism to promptly alert the presence of cyberattacks within the actuator channel. The natural drawback of a conventional CBF-based controller is that its performance (i.e., safety guarantees) is highly sensitive to the model dynamics. These model uncertainties and cyberattacks may cause unwanted safety breaches/violations. To overcome the technical challenge, our proposed controller with SM-CBF approach enables us to effectively enforce the safety conditions of the system even in the presence of model uncertainties. Furthermore, we propose a novel LSTM-based attack detector such that it can swiftly determine which control input channels (i.e., motors) are compromised. By synthesizing the robust controller with the attack detector, our proposed safety-critical controller can greatly enhance system safety while concurrently achieving control performance. Finally, an illustrative example of the stabilization of quadrotor unmanned aerial vehicles (UAVs) using a high-fidelity simulator is provided to demonstrate the effectiveness of the proposed methodology.