Alertness-related differences in CF strategy among truck drivers based on naturalistic driving data

Driver drowsiness is a major safety concern in heavy-duty truck operation, particularly in freeway car-following (CF) scenarios that require prolonged vigilance and continuous longitudinal control. However, how drowsiness-induced reduced alertness influences truck drivers' CF behavior in real-world settings remains insufficiently understood. To address this gap, this study established a naturalistic heavy-truck CF dataset from long-haul freeway operations, with driver alertness labels obtained from a Smart Eye system. A three-level analytical framework was adopted to examine the effects of alertness on CF behavior from the perspectives of high-level behavioral indicators, model-based mechanism characterization, and primitive-level temporal structure. Specifically, overall CF performance was first described using commonly used indicators, then parameterized using the Intelligent Driver Model (IDM), and further interpreted at a finer temporal scale through behavioral primitives identified by a disentangled sticky hierarchical Dirichlet process hidden Markov model (DS-HDP-HMM). The results consistently showed that reduced alertness was associated with changes in drivers' longitudinal adjustment patterns, which may partly reflect compensatory responses. Compared with alert driving, non-alert driving was associated with a more conservative spacing strategy, milder and less aggressive longitudinal regulation, and more near-constant and weak-regulation behavioral primitives, while active acceleration and deceleration corrections were less frequent. These findings provide a multi-level understanding of the longitudinal control mechanisms underlying non-alert truck driving in freeway environments and offer behavioral evidence for the development of intelligent driving assistance systems that can better respond to non-alert road truck drivers.