Supervising the self-driving car: Situation awareness and fatigue during highly automated driving. (July 2023)

• Record Type: Journal Article
• Authors: McKerral, Angus; Pammer, Kristen; Gauld, Cassandra
• Is Part Of: Accident analysis and prevention. Volume 187(2023)
• Publication Date: 2023-07
• Subjects: Situation awareness -- Automated vehicles -- Non-driving related tasks -- Attention; Accidents -- Prevention -- Periodicals; Accident Prevention -- Periodicals; Accidents -- Prévention -- Périodiques; 363.106
• DOI: 10.1016/j.aap.2023.107068
• Languages: English
• ISSNs: 0001-4575

Abstract

Highlights: The present study uses a PCM-based approach to assess AV operator SA over an extended fatiguing automated drive. Operator SA is assessed at two critical takeover events near the beginning and end of the simulated drive. Operator SA is shown to decline in response to passive monitoring of the simulated AV. Self-selected NDRT engagement counteracts SA decline. Impaired schema-related SA appears to explain poorer outcomes in the AV monitoring condition compared to the NDRT-permitted condition. Abstract: Vehicle automation promises to reduce the demands of the driving task, making driving less fatiguing, more convenient, and safer. Nevertheless, Level 3 automated vehicles rely on a human driver to be ready to resume control, requiring the driver to reconstruct situation awareness (SA) and resume the driving task. Understanding the interaction between non-driving-related task (NDRT) use, SA, and takeover capacity is important because an effective takeover is entirely dependent on, and scaffolds from, effectively reconstructed SA. While a number of studies have looked at the behavioural impact of being 'in- or on-the-loop', fewer consider the cognitive impact, particularly the consequences for SA. The present study exposed participants to an extended simulated automated drive involving two critical takeover scenarios (early- and late-drive). We compared automated vehicle (AV) operators who were required to passively monitor the vehicle to those engaging with self-selected NDRTs. Monitoring operators demonstrated lower total- and schema-specific SA count scores following a fatiguing drive compared to those engaging with self-selected NDRTs. NDRT engagement resulted in no significant difference in SA count scores early- and late-drive. Assessment of differences in the type and sensory modality of NDRTs indicated operators make fundamentally different selections about the NDRTs they engage with in an automated driving environment compared to a manual environment. The present study provides further evidence linking SA and AV operator behaviour and underscores the need to understand the role of SA in takeover capacity. Our findings suggest that although SA declines over time regardless of driving task requirements (Monitoring versus NDRT engagement), NDRT use may facilitate better SA construction, with implications for the regulation of NDRT use in AVs as the technology progresses.