Drivers’ response to attentional demand in automated driving
Vehicle automation can make driving safer; it can compensate for human impairments that are recognized as the leading cause of crashes. Vehicle automation has become a central topic in transportation and human factors research. This thesis addresses some unresolved challenges on how to guide attention for safe use of automation and on how to improve the design of automation to account for humans' abilities and limitations. Specifically, this thesis investigated how driver attention changed with automation and the driving situation. The objective was to inform the design of vehicle systems and develop design knowledge to support safe driving. A novelty of this thesis was in the use of real-world driving data and Bayesian methods (improved statistical modeling techniques). The analysis of driver behavior was based on data collected in naturalistic driving studies (to study the effect of assistive automation) and in a simulator experiment (to study the effect of unsupervised automation). Driver behavior was examined with measures of visual and motor response, together with contextual information, on the driving situation. The results show that assistive automation affected driver attention in real-world driving. In general, drivers devoted less attention at the forward path with automation than without. However, driver attention was sensitive to the presence of other traffic and changes in illumination---variations in the surrounding environment that increased the uncertainty of the driving situation---and it was elicited by visual, audio, and vestibular-kinesthetic-somatosensory information (perceptual cues) that alerted to an impending conflict. Driver response to a critical situation with unsupervised automation had a reflexive component (glance on-path, hands on wheel, and feet on pedals) and a planned component (decision and execution of evasive maneuver). Warnings primarily alerted attention rather than triggering an intervention. Expectation, which changed over time depending on experience, affected driver response substantially. This thesis found that the safety implications of diverting attention away from the driving situation need to be interpreted in relation to the characteristics and criticality of the driving situation (driving context) and need to consider the reduction of risk exposure due to automation (e.g., headway maintenance and collision warnings). Drivers were, for example, successful at changing their behavior in the presence of other vehicles and in different light conditions independently of automation. If drivers are not attentive at critical points, warnings are effective for triggering a quick shift of attention to the driving task in preparation to an evasive action. The results improved on those of earlier studies by providing a comprehensive assessment of driver attentional response in routine driving and critical situations. The results can support evidence-based recommendations (inattention guidelines) and be used as a reference for driver modeling and vehicle systems development.
driving simulator data
Campus Lindholmen, Saga building, Hörselgången 4, room Alfa
Opponent: Prof. John D. Lee, Department of Industrial and Systems Engineering, University of Wisconsin-Madison