Systematic review of driving simulator validation studies

摘要

Driving simulators are a common tool for researching driver behaviour, providing practical, safe, and controlled environments. Despite their frequent use in research, there is relatively little evidence confirming their validity (i.e., how accurately they represent or reproduce real-world driving). Moreover, there is inconsistency in both the types of simulators used, and the operationalisation of "real-world" driving in validations. This systematic review was undertaken to evaluate the evidence regarding driving simulator accuracy when compared with real-world driving. The review included 44 studies reporting a direct comparison between simulated driving and on-road driving in a vehicle. Measures reported for comparison varied but included mean speed, speed variability, lateral position, overall driving performance, and number of driving errors. Simulators in approximately half of the studies achieved absolute or relative validity, whereas one third produced non-valid results. To understand this further, the fidelity of simulators was considered, however this further clouded our understanding as the relationship between simulator fidelity and validity was not straightforward. The findings suggest that the reporting of driving simulator studies requires improvement, particularly around the validation evidence associated with the simulator, the specific details of the simulated driving environment, and the outputs of statistical analyses. Guidelines are proposed for future research to ensure consistency in the conduct, and reporting, of simulator-based research.