Video simulation of pedestrian crossings at signalized intersections

This paper compares users' accommodation-based pedestrian level-of-service (LOS) ratings of signalized intersections in a video simulation environment and the physical roadway environment. Video simulation and walking course responses were compared with the use of data obtained from Walk for Science 2004, in which the participants reported their perceived level of safety and comfort as pedestrians. These data were used to test the hypothesis that subjects rate intersections' LOSs differently in a controlled (video simulation) environment than they do by walking along a real-world course and experiencing real-time traffic conditions. The results for intersections common to both experiences (549 observations) indicated a statistically significant difference in the respective ratings (t = 2.34). Accordingly, a calibration factor was developed to adjust the participants' scores from the video event to what they would have scored in the real-time field event. The adjusted video simulation data were then used to refine a field-calibrated model for pedestrian LOS at intersections (which is discussed in detail in another paper). Although further hypothesis testing may be conducted, the resulting general model for the pedestrian LOS at intersections is highly reliable, has a high correlation coefficient (R-2 = .73) with the average observations, and is transferable to the vast majority of metropolitan areas in the United States. This video simulation research design can serve as a model for similar data-gathering initiatives in the future as the effort to calibrate the modeling of video simulation data to field-based perception data continues.