A digital video camera was used to record left-turning vehicles and through vehicles at an urban intersection. A total of 39 left-turn events, with a total of 195 gap decisions, were identified and vehicle trajectories corresponding to those were extracted from the video and transformed into real coordinates using photogrammetry. Bayes estimates of each opposing vehicle's distance, speed, and time-to-arrival were then computed from the trajectories and used as predictors in logit models of acceptance/rejection decisions. It was found, when models are penalized for the numbers of their parameters, that arrival time, the ratio of initial distance to initial speed, was best predictor. This contrasts with an earlier study that found distance clearly superior to arrival time. This may be due to the fact that in the current study, speeds and initial distances were substantially higher than in the earlier study. (Author/publisher)
Abstract