Driver Behaviour in Level 3 Automated Vehicles: The Effects of Distraction, Fatigue and driver Age

Abstract

As Level 3 autonomous vehicles (AVs) require drivers to resume control upon a Takeover Request (TOR), understanding the factors influencing takeover performance (TOP) is critical for safety. This thesis examined cognitive, demographic, and situational variables including age, time budget (TB), Non-Driving Related Tasks (NDRTs), and fatigue to determine their impact on TOP. This thesis employed a multi-method approach, including a systematic literature review, cognitive experiments, and driving simulations using hi-fidelity simulator, to assess how these variables impact TOP metrics like reaction time, steering behaviour, and lane-keeping performance. The experimental design used a mixed-factorial approach, with TB, NDRTs, and fatigue as within-subject conditions and age as between-subject factors. Objective driving performance metrics including response takeover time (RTOT), steering wheel angle (SWA), steering wheel speed (SWS), lateral and longitudinal deceleration, and standard deviation of lane position (SDLnP) evaluated takeover quality and timeliness. Subjective workload measures (NASA-TLX, Karolinska Sleepiness Scale) assessed cognitive demand and fatigue. Key findings revealed that while older adults exhibited slower cognitive responses in lab tasks, their real-world TOP remained comparable to that of younger drivers, likely due to compensatory behaviours like driving at lower speed. TB influenced steering precision but not RTOT, suggesting reflexive responses remained stable regardless of warning duration. NDRTs disrupted vehicle control, with passive tasks (e.g., monitoring) leading to greater variability in steering response, while active tasks (e.g., gaming) had a more pronounced effect on lane-keeping performance. Notably, moderate NDRT engagement partially offset fatigue-induced lateral control degradation, though fatigue still adversely impacted steering stability without affecting RTOT. However, active NDRTs (e.g., gaming) disrupted vehicle control (in terms of lane keeping) more severely than fatigue alone, underscoring the need for balanced workload management in automated driving systems. These results challenge assumptions about age-related declines in TOP, emphasizing the role of cognitive functions, driving experience, and adaptive strategies. Future research should develop personalized automation strategies that adapt TOR parameters such as TB duration and modality based on individual driver state, ensuring both safety and efficiency in semi-autonomous driving.