Balancing temporal dynamics with measurement noise in real-time situation awareness prediction

Area of Science:

• Human-computer interaction
• Cognitive science
• Autonomous systems

Background:

• Situation awareness (SA) is crucial for operator safety and performance.
• Autonomous systems can decrease operator SA, necessitating real-time SA indicators.
• Current SA measurement methods have limitations for real-time predictive modeling.

Purpose of the Study:

• To develop a real-time, nondisruptive SA measurement for adaptive human-autonomy teams.
• To address limitations of single-trial SA measures and temporal dynamics.
• To validate a moving average SA score for improved prediction.

Main Methods:

• Implemented a 3-trial moving average SA score.
• Compared moving average SA scores with single-trial SA scores.
• Utilized single-trial physiological signals to predict SA scores.

Main Results:

• The 3-trial moving average SA score reduces noise while preserving temporal information.
• Moving average SA scores showed stronger correlation with performance (r=0.36, p<0.01) than single-trial scores.
• Physiological signals predicted moving average SA scores with greater accuracy (SAE=0.61, Q2=0.36) than single-trial SA scores.

Conclusions:

• A 3-trial moving average SA score is a viable method for real-time SA assessment.
• This approach enhances the predictive accuracy of SA from physiological signals.
• Improved SA measurement supports adaptive cooperation in human-autonomy systems.