G. Guido, A. Vitale, F. Saccomanno, V. Gallelli

Pages: 65-74

Abstract
The main objective of this study is to extract experimentally from a videotaping of traffic operations vehicle tracking profiles for a case study and compare these profiles to the simulated ones for a given set of parameter inputs. In particular, this study seeks to investigate a) how simulated vehicle profiles are positioned with respect to experimental values for a sample of vehicles, b) how simulated profiles are affected by parameter inputs and c) which parameters have the greatest effect on the accuracy of the simulated results. The purpose is to establish, at the aggregate level, if it can be assumed that the simulated values come from the same population as the observed values for a sample of trajectories. In this study tracking profiles are categorized into two classes of attributes: 1) Operational and 2) Safety-related. The operational profiles include information on instantaneous vehicle speed and acceleration. The safety-related profiles, which are derived from the operational variables, include vehicle pair time-to-collision (TTC). The transferability of the model inputs is assessed for different traffic conditions through a validation process performed on an independent sample from the case study. This demonstrate how the input values suggested in the calibration stage can identify potentially unsafe vehicle interactions for vehicle movements based on car-following behavior protocol (potential rear-end crashes). In this way it is possible to provide a link between simulated safety performance indicators and observed high risk vehicular interactions.

Keywords: video image processing; microsimulation; safety performances; genetic algorithms