Development of a Method to Estimate Shortage of Situation Awareness in Takeover based on Reference Model of Cognitive Behavior

Abstract

Highly automated vehicles require takeover (TO) to shift the control authority to drivers in an emergency, but the driver only have about 4–10 seconds for TO. Driver’s situation awareness should be thus monitored for safe TO. In this study, we develop a situation awareness estimation method based on cognitive behavior. We made TO situations with different conditions in the driving simulator, and corrected the driver gaze data and car position, and labels (safe or dangerous). We classified obtained dataset into safe and dangerous by using a support vector machine, and used the classifier as a reference model for cognitive behavior. We found that the reference model could classify the correctness of situation awareness with high accuracy.