Book Description

Safety of traffic facilities is most often measured by counting the number (and severity) of crashes that occur. It is not possible to apply such a measurement technique to traffic facility designs that have not yet been built or deployed in the real world. This project has resulted in the development of a software tool for deriving surrogate safety measures for traffic facilities from data output by traffic simulation models. This software is referred to as SSAM, an acronym for the Surrogate Safety Assessment Model. The surrogate measures developed in this project are based on the identification, classification, and evaluation of traffic conflicts that occur in the simulation model. By comparing one simulated design case with another, this software allows an analyst to make statistical judgments about the relative safety of the two designs. The method is based on processing detailed trajectories of all vehicles in the simulation environment and identifying conflicts that occur between any two vehicles. An open-standard vehicle trajectory data format was designed, and support for this format has been added as an output option by four simulation model vendors/developers PTV (VISSIM), TSS (AIMSUN), Quadstone (Paramics), and Rioux Engineering (TEXAS). Eleven 'theoretical' validation tests were performed to compare the surrogate safety assessment results of pairs of simulated design alternatives. In some cases, these comparative tests showed clear agreement with traditional results from crash prediction models; and, in other cases, the results from SSAM utilizing the traffic simulation models were counterindicative. Typically, one design would exhibit a statistically significant lower number of conflicts, but the severity of conflicts that did occur would be higher. Nevertheless, SSAM does provide metrics for use by analysts in comparing traffic facility designs that do not have established crash statistics. More research work is necessary to refine the metrics for broader consumption by a general audience. A field validation exercise was completed to compare the output from SSAM with real-world crash records. Eighty-three intersections from British Columbia, Canada were modeled in VISSM and simulated under AM-peak traffic conditions. The processed conflict results were then compared with the crash records in a number of different statistical validation tests. In general, the correlation results between the outputs from the simulation models and the crash records were significant, with the regression model predicting crashes from conflict data exhibiting a correlation (R-squared) value of = 0.41. However, the simulation modeling effort did present a number of issues. Sensitivity analysis was performed to identify differences between the SSAM-related outputs of each simulation model vendors system on the same traffic facility designs. These comparative analyses provide some guidance to the relative use of surrogate measures data from each simulation system. For more information regarding the SSAM tool and user manual (FHWA-HRT-08-050) see companion report, PB2008-111196.