Abstract
We studied the monocular visual cues (as opposed to ttc, d theta or r dot) that the drivers of following vehicles may utilize to determine their relation to a lead vehicle. The purpose is to improve collision intervention signals. We focused on increases in lead vehicle width (horizontal size change), height (vertical change) and area (concurrent horizontal and vertical change). Six normal subjects were more sensitive to brief (0.1 sec) width and area increases than to height increases. The stimulus was a grey rectangle (∼6.5 deg wide × 6.5 deg high) that increased symmetrically in the direction of the parameter being tested. Three different degrees of border increase, 0.6, 1.2 and 4.2 minutes of visual angle, were used (e.g. 0.6 min increase means a 1.2 min overall increase and a “rubbery” transformation of the stimulus). With a 0.6 min change, we observed a 27% advantage in detection performance (measured in terms of average % error of detection in the yes-no task) for width increases and 36% advantage for area increases over height (p=0.086 and 0.041 respectively). For 1.2 min change, the improvement was 175% (width) and 59% (area) (p<0.0005 and 0.058). There was no significant difference between the performance on the width and area tasks. This result is not consistent with prior studies suggesting greater sensitivity to rigid (area) vs. rubbery (width and height) transformations. We plan to test other parameters such as texture and distance to a fixed landmark. These results imply that a warning system would only need to draw the attention of the following driver to the changing width or area of the lead vehicle, rather than to the height. If costs are a concern, focusing on just the width change would be sufficient.
Funding: CA-PATH TO-4221, USDOT-FTA RICWS (AATA & Veridian Int.), UCB-SUPERB (NSF), UCB-GUIDANT.