A systematic review of abnormal behaviour detection and analysis in driving simulators

Abstract

Driving safety is increasingly recognised as a critical global issue, addressed extensively through both naturalistic and simulator-based research. Driving simulators, in particular, offer valuable practical and theoretical contributions to the field, with numerous studies affirming their effectiveness. This review sought to examine relevant simulator-based research, focusing specifically on the detection and analysis of unsafe driving behaviours. While previous studies predominantly focused on individual behaviours, this review encompasses a broader spectrum. Initially, a comprehensive search from 2013 to 2023 yielded 759 research articles from Scopus, Web of Science, and ScienceDirect. Employing established search strategies and adhering to specific inclusion and exclusion criteria, 70 papers were ultimately selected for detailed review. This analysis examined the methodological approaches of these studies, including the types of unsafe behaviours investigated, the parameters measured, the equipment utilised, and the classification and analysis techniques employed. This review provides an extensive overview of the field, detailing how various simulators detect a range of unsafe driving behaviours and analysing the algorithms used to assess each driving parameter. It also guides researchers in selecting simulator hardware and choosing appropriate detection algorithms. The review highlights the importance of incorporating both vehicle-based and driver-based parameters in driving behaviour studies and advocates for the use of simulators with high levels of freedom and fidelity in experiments. This comprehensive synthesis serves as a valuable resource for regulators and stakeholders, offering foundational insights for developing strategies to reduce unsafe driving behaviours and enhance road safety.