Richard Agbeyibor
(Advisor: Prof. Karen Feigh]

will defend a doctoral thesis entitled,

Human-AI Collaboration Aboard Crewed Autonomous Intelligence, Surveillance, and Reconnaissance Aircraft 

On

Friday, November 21 at 1:30 p.m. 
 Montgomery Knight Building 317

and

Microsoft Teams

Abstract

The rapid emergence of capable autonomous systems presents both opportunities and challenges for human–autonomy teaming (HAT) in specialized aviation missions. This dissertation investigates how AI-enabled autonomous pilots (APs) can collaborate effectively with humans aboard crewed autonomous aircraft during complex mission scenarios such as Intelligence, Surveillance, and Reconnaissance (ISR) and Medical Evacuation (MedEvac). The research centers on the design and experimental use of the Crewed Autonomous Aircraft Flight Simulator (CAAFS), a high-fidelity testbed enabling controlled studies of adaptive autonomy and transparency in realistic flight contexts. Physiological sensing using heart rate, respiration rate, and eye-tracking is employed to estimate operator workload and attention. Nonlinear dynamics and machine-learning analyses demonstrate that multimodal physiological features can significantly differentiate high- and low-workload states, establishing a cognitive state estimator that enables autonomy to adjust its behavior based on operator state. A single-agent ISR study evaluates static AP behaviors to benchmark performance, workload, and trust across fixed allocations of authority, transparency, and function, revealing the limitations of rigid control modes in dynamic mission environments. Building on these results, a human-subjects study introduces Attention-Tunable Control Barrier Functions (AT-CBFs), a formal methods approach that adapts safety margins to operator attention, enhancing safety while maintaining mission efficiency. Finally, a multi-agent ISR experiment examines robot-to-human transparency using the Situation Awareness-Based Agent Transparency (SAT) model, showing that increased transparency improves performance and team fluency. Collectively, the findings demonstrate that effective HAT requires bidirectional  (robot-of-human and robot-to-human) transparency---autonomy must sense and adapt to human cognitive state just as humans must understand autonomy’s actions, plans, reasoning process, predicted outcomes and uncertainties---and provide a framework for designing next-generation crewed autonomous aircraft.

Committee

• Prof. Karen M. Feigh – School of Aerospace Engineering (Advisor)
• Prof. Samuel Coogan – School of Electrical and Computer Engineering
• Prof. Kyriakos Vamvoudakis– School of Aerospace Engineering
• Prof. Mengyao Li – School of Psychology
• Dr. Joseph B. Lyons – Air Force Research Laboratory