To design better fallback procedures and enhance road safety for conditionally automated vehicles (SAE Level 3), it is important to understand the factors that affect driver’s takeover performance (i.e., driving performance while resuming manual control). This study investigates the impacts of driver’s pre-warning cognitive state (i.e., before the issuance of a takeover warning) on takeover performance. Most existing studies assess takeover performance by independently analyzing driving performance indicators (e.g., minimum time-to-collision and maximum deceleration), and thereby ignore their associated interdependencies. This study proposes a novel comprehensive takeover performance metric, Takeover Performance Index (TOPI), that combines multiple driving performance indicators representing three aspects of takeover performance: risk of collision, the intensity of the driver’s response, and trajectory quality. Further, the driver’s pre-warning cognitive state is estimated by analyzing neurophysiological data (i.e., brain electrical activity) measured using an electroencephalogram (EEG) for 118 participants in driving simulator experiments. Linear mixed models are estimated for takeover performance to analyze its linkages to the driver’s pre-warning cognitive state, novelty in takeover experience (i.e., prior experience with a takeover situation), type of takeover warning (i.e., non-mandatory takeover vs. mandatory takeover), age, and driving experience. In this study, most drivers intervened in non-mandatory takeover scenarios and exhibited poor takeover performance. We observed three crashes across 287 runs. The study results show that takeover performance decreases with age but increases with driving experience when the driver is under certain pre-warning cognitive states, including fatigue, drowsiness, passive attention, and low level of alertness. They also illustrate that the novelty in takeover experience and mandatory takeover warning negatively affects takeover performance. The study findings provide insights for developing operator training and licensing strategies, designing regulations for the use of automated vehicles, and factoring driver cognition in designing fallback procedures in automated vehicles.
Read full abstract