Predicting Driver Distraction Using Computed Occlusion Task Times: Estimation of Task Element Times and Distributions.

摘要

To determine conformance with NHTSA’s visual-manual interface distraction guidelines and to reduce the associated number of crashes, NHTSA recommends a visual-occlusion test procedure. As an alternative to testing subjects following that procedure, this report provides experimentally based estimated times for in-vehicle task elements (e.g., flick, press button). Those estimated times can be summed and then adjusted using Pettitt’s method (which assumes that visual tasks progress only when the test goggles are open) to estimate total task occlusion time. The estimated times were determined from a frame-by-frame analysis of data from an occlusion experiment evaluating a next-generation Hyundai navigation radio. That analysis revealed the mean element time for middle-aged subjects (45-55) was only about 16% longer than young (25-35) subjects, whereas the mean task time was 44% greater, primarily because there were 32% more occurrences of elements to complete tasks. The elements and their mean times were flick (0.50 s), flick/scroll return (0.38 s), press button (0.64 s), quick flick (0.35 s), reach for button (0.42 s), reach for center console (0.75 s), read instructions (0.53 s), scroll (0.66 s), search (0.54 s), stop screen (0.24 s), turn knob (0.43 s), reposition hand on knob (0.33 s), wait-loading (0.90 s), wait after loading (0.92 s), wait for goggles-known location (1.34 s), and wait for goggles-unknown location (0.92 s). Most element distributions were lognormal. Interestingly, 45% of all element occurrences were when the goggles were closed or in open-closed or closed-open periods. Given this, the assumptions of Pettitt’s method need further thought