Research
Embedded Files
Research Vision: Contributing foundational research that enables physical AI systems to work with humans, not simply around them.
Main Research Question: How can we model humans-physical AI system interactions under unreliable and safety-critical situations?
Methodology & Research Goal: By computationally modeling human behavior and cognitive states, I aim to enhance safety and optimize human–physical AI system performance.
Research Framework
My research framework spans three temporal phases of human–physical AI interaction during safety-critical events (Figure).
- Pre-safety-critical event: Predicting driver behavior and cognitive states to prevent safety-critical situations.
- In the moment: Computationally intervening to support human decision-making as a safety-critical situation unfolds.
- Post-safety-critical event: Characterizing human behavior and cognitive states after encountering a safety-critical situation.
My research framework spans three temporal phases of human–physical AI interaction during safety-critical events (Figure).
- Pre-safety-critical event: Predicting driver behavior and cognitive states to prevent safety-critical situations.
- In the moment: Computationally intervening to support human decision-making as a safety-critical situation unfolds.
- Post-safety-critical event: Characterizing human behavior and cognitive states after encountering a safety-critical situation.
1. Human Behavior and Cognitive States Characterization (Post-Event)
General RQ: How do humans behave in specific scenarios following a critical event?
Methodology: Traditional statistical modeling and inferential analysis
Application: Understanding human characteristics and limitations; suggesting system design guidelines
Related Publications: J1, J2, J5, J6, J7, J9, W2
2. Real-Time Human Assistance & Intervention (In the Moment)
General RQ: How can adaptive assistance be provided by analyzing humans’ real-time multimodal data, including driving performance, eye-tracking, and physiological signals?
Methodology: Real-time data communication (TCP/IP, UDP), online computation (Python, MATLAB), and multimodal intervention delivery (visual, auditory, or haptic feedback)
Application: Enhancing (driving) safety; developing real-time intervention strategies
Related Publications: J4, W3
3. Human Behavior and Cognitive States Prediction (Pre-Event)
General RQ: Can human behavior be predicted n seconds into the future using current multimodal data?
Methodology: Machine learning and deep learning approaches
Application: Enhancing (driving) safety; enabling proactive collision avoidance / hazard mitigation
Related Publications: J3, J8, W4
Research Methodology
Methodologically, I utilize driving simulators and physiological sensors, including eye-tracking, heart rate, galvanic skin response, and 3D skeleton tracking, to assess drivers' behavioral and cognitive responses.
Methodologically, I utilize driving simulators and physiological sensors, including eye-tracking, heart rate, galvanic skin response, and 3D skeleton tracking, to assess drivers' behavioral and cognitive responses.
Using Python and MATLAB, I model these dynamics through real-time interventions, machine learning, and statistical approaches.
Using Python and MATLAB, I model these dynamics through real-time interventions, machine learning, and statistical approaches.
Page updated
Google Sites
Report abuse