Human Behavior and Cognitive States Assistance (Post-Event)

To enable users to interact safely with Physical AI systems, these systems must identify user needs in real time and deliver tailored assistance. Unlike conventional approaches that rely on pre-defined, rigid interventions, my work leverages real-time computing to dynamically provide context-aware information and support. This personalized assistance ensures that users can utilize Physical AI systems more safely and effectively.

• My research processes stream-level driver data, such as control inputs and gaze position, in real time to determine optimal control strategies and visual attention targets, providing adaptive feedback on how drivers can ideally operate the vehicle and where they should allocate their attention [Working Paper].