Visual gaze and engagement with non-driving related tasks: a driving simulator study with automated driving systems.

Abstract

Objective: Automated Driving Systems (ADS), classified as Level 3 automated systems (SAE 2021), can potentially reduce risks by conditionally taking control of the driving task. However, drivers must remain alert and be ready to take back control if necessary. This may introduce risks, especially if drivers are distracted. Observing driver behaviors as they engage in different types of NDRTs could help understand how behaviors differ while driving with Level 3 automation. To that end, in this study, we observed drivers when driving with Level 3 automation. Specifically, we analyzed eye movements, non-driving-related task (NDRT) engagement, and responses to takeover requests (TOR) to understand behaviors during automation and transitions to manual driving.

Methods: We conducted a simulator study with 24 fully licensed drivers. Participants drove in a simulator equipped with Level 3 automation and performed two NDRTs: a Surrogate Reference Task and a cellphone task. Drivers were notified visually and verbally about automation status and TORs. Participants' gaze behavior and takeover times were measured during the drive, and post-drive surveys assessed trust and usability scores.

Results: NDRT type had a significant impact on takeover time, with drivers taking longer to take over during cellphone tasks. Drivers tended to focus more on non-driving related areas right until a TOR. After TORs, drivers tended to shift focus to the Instrument Cluster, underlining the criticality of displaying information about the TOR. Trust and usability scores were comparable across groups, suggesting that drivers generally found the system easy to use and exhibited a reasonable level of trust in it.

Conclusions: Findings reveal that regardless of the NDRT, drivers continued engaging in NDRTs right up till the TOR. Designing intuitive, context-specific interfaces that guide drivers' attention to driving-related areas and provide information can improve drivers' awareness of the TOR and, consequently, their takeover performance. The findings provide significant insights on the potential methods to keep drivers aware of their surroundings while using automation, and while transitioning to manual control. These insights provide information on driving behaviors with Level 3 automation, specifically how fully licensed drivers engage with distraction while driving with Level 3.