Abstract
Current automated driving systems are not capable of responding to all situations, require driver intervention when necessary, and require time for appropriate driver intervention. This study proposes a gradual alert that combines a reliable alert issued when there is no time to spare and an advisory alert issued when there is time to spare, and aims to clarify its effectiveness by analyzing the human adaptation process. Experiment 1 was conducted using a driving simulator, and eye tracking data confirmed that subjects responded differently to each alarm, even though they were given the same instructional content. The cause may be the influence of personal characteristics in responding to alarms. Experiment 2 was conducted using a driving simulator, as well as a questionnaire on the level of trust in automated driving systems and alarms. After the experiment, Balloon Analogue Risk Task (BART) was performed to measure risk-taking behavior and assess risk perception characteristics. Comparison of eye tracking data with indicators of personal characteristics indicates that driver's risk perception and confidence in alarms may influence response to alarms.
