"The ongoing evolution of the aerospace industry has seen significant developments of advanced aerial vehicles, such as reusable launch vehicles (RLVs) and hypersonic flight vehicles (HFVs). Despite the potential benefits of these vehicles, their design presents substantial challenges, including complicated aerodynamics, strong nonlinearities, large flight envelopes, and demanding flight environments. This dissertation aims to enhance robust adaptive control methodologies to render them more suitable for practical implementation in advanced aerial vehicles. " Read more...

"Visual localization aims to estimate the spatial relationship between a camera and its environment based on visual media captures. However, the increasing need for robustness in challenging environments and the requirement to solve semantic-centric problems has presented challenges for geometric localization workflows. Conventional methods rely on hand-crafted heuristics that struggle to meet the growing demands of such environments. On the other hand, emerging deep learning methods face issues with generalization and interpretability due to their geometryagnostic nature. To address these challenges, this dissertation presents a hybrid localization workflow that leverages both geometric and data-driven priors. " Read more...

"Machine learning is a data-driven process that heavily relies on the quality of the data being used. Trivial approaches trade data for performance resulting in heavy, noisy, and erroneous results. In particular, when the real-world test domain is different from the training, the laboratory experiments will not find practical applications. To bridge this gap, previous work has explored selecting data from a specific target domain to address this discrepancy. " Read more...

"The call for trustworthy AI has been a long pursuit, yet still far to reach. My thesis joins the initiative that brings three trustworthiness aspects, Accuracy, Interpretability, and Resilience (AI-R), under one umbrella. Unlike conventional approaches striving to balance AI-R aspects by sacrificing model performance for interpretability or compromising transparency for resilience, we take a unique direction and show that model interpretation helps to improve prediction quality and model resilience. Our methods advance attention mechanisms, LIME, SHAP, self-learn post-hoc explainability models, and incorporate model-intrinsic techniques into a multi-agent reinforcement learning (MARL) framework." Read more...

"As the world faces ongoing rapid changes and unforeseen crises, it is crucial to develop a comprehensive understanding of the interconnectedness of systems and the critical role of human-technology-governance interactions in shaping the resilience of complex systems. In addressing these challenges, this dissertation adopts systems thinking as its foundational approach, guiding the examination of intricate relationships and their implications to expedite digital transformation in complex healthcare systems." Read more...