Three experiments were conducted to assess a test of distraction potential for in-vehicle information systems (IVIS) and portable devices used while driving. The test used a low-fidelity (PC-based) driving simulator; sensors record drivers’ control inputs in stationary production vehicles. Participants performed car-following and target detection together with secondary tasks. Experiment 1 examined the effects of two levels of driving task (car-following) difficulty and two detection tasks on test sensitivity. Detection tasks included a head-mounted task (HDT) and a computer-generated multiple-target task (MDT), which incorporated simple targets into the simulated roadway display. Secondary tasks included simple (Circles) and complex (navigation destination entry) visual-manual tasks and a hands-free auditory-vocal task (N-back). The MDT was more sensitive to task load differences, while the HDT created problems for the eye tracker. Increasing car-following task difficulty had no effect on metric sensitivity. The complex visual-manual task was more disruptive than the simple visual-manual task or the auditory-vocal task. The second experiment compared metrics provided by two occlusion paradigms: (1) traditional occlusion, which involves intermittent masking of the task to simulate the visual demands of driving, and (2) Enhanced Occlusion Task (EOT), which added auditory tracking to more realistically simulate driving task demands. Task duration estimates with EOT were closer to static completion times than those obtained with traditional occlusion; however their usefulness in estimating task duration requires a stronger connection to comparable estimates obtained from a driving protocol. EOT did not improve the R (task resumability) metric, although there was no independent evidence to support the expectation of differences in this metric. The R metric was not related to driving performance degradation. Auditory tracking performance metrics obtained in the EOT paradigm revealed performance degradation consistent with effects observed with simulator metrics. Experiment 3 used the simulator-based test to assess the distraction potential of navigation tasks performed with three systems with comparable functionality, including one original equipment manufacturer (OEM) and two portable systems, which had been independently rated as having different levels of usability. Metrics revealed strong and consistent differences between driving alone and driving with a secondary task. Three metrics (car-following coherence, detection task response time, and proportion of long glances) revealed differences between the (simple and complex) navigation tasks across all systems. Two metrics (standard deviation of lane position [SDLP] and detection task proportion correct) exhibited significant Systems x Task interactions; differences between navigation tasks were not consistent across systems. It was concluded that developing a simulator-based distraction potential test is feasible. Core metrics include those sensitive to visual-manual task conditions (SDLP, car-following delay, and detection task response time) and those sensitive to auditory-vocal task conditions (car-following delay, detection task response time, and detection task proportion of correct responses). Measures based on eye position data, primarily the proportion of long glances away from the forward roadway, revealed significant promise. Estimates of distraction potential can be combined with task duration estimates provided by the EOT to compute estimates of drivers’ exposure to risk (Author/publisher)
Abstract