Abstract
This study develops a conceptual driver-system performance model within the context of Human Reliability Analysis (HRA), using data from controlled driving simulator experiments involving takeover scenarios in vehicles equipped with Level 3 Automated Driving Systems (ADSs). Driving simulation-based experiments, collecting both objective and subjective data, informing the model’s development and quantification, with a focus on operation and takeover performance. These experiments examined how warning availability and traffic complexity affect driver takeover performance during automated driving failures. Fifty participants experienced ADS-initiated tailgating events across four warning/traffic conditions in a driving simulator. Warnings significantly reduced collision probability and improved driver reaction times and time-to-collision. Traffic complexity increased hard-braking, tailgating, and steering variability, and elevated workload and post-exposure distrust of the automation. A significant interaction between warning availability and traffic level was observed for collision probability. These initial findings indicate warnings provide measurable safety benefits, and will inform the calibration of the model, including warning design and timing, in future work.
