5. Summary of Simulation Model Features Review

The results of the simulation model review do not indicate a clear superiority of any particular model when considering all of the elements required for modeling surrogate measures. From the evaluation done to produce this report, it does not appear possible that surrogate measures could be obtained from any of the simulation models that were reviewed without some internal modifications to either the API(s), if provided, or the source-code modules themselves. The selection of a simulation model or models is tightly coupled to both:

• Selection of a recommended approach to obtaining surrogate safety measures from simulation models.
• Determination of the surrogates that are to be extracted or computed.

The literature indicates a clear preference for surrogate measures that include specific vehicle-to-vehicle proximity measures, including GT, TTC, and PET. Our evaluation and the results of the Surrogate Safety Measures Workshop held at the beginning of this project confirm this. These event-based measures hold the most promise for evaluating the relative safety of traffic facilities. Aggregated measures such as gap-acceptance distributions, link or lane speeds, etc. have a less credible connection with crash probability and thus are viewed as secondary surrogates of safety. The other most significant finding of the Surrogate Safety Measures Workshop was that an SSAM must be flexible in the aggregation of the measures, i.e., across approaches, movements, conflict types, etc., rather than a priori assuming a particular aggregation methodology.

The recommended approach to collecting surrogates from a number of different simulation models in a uniform way is to develop a post-processor surrogate safety assessment module. This requires internal enhancements to a specific model, or models, for calculation of surrogate measures and output of those calculations in a formatted output data file that is post-processed by the stand-alone surrogate safety assessment module to aggregate the internally calculated measures in a variety of ways. This approach has many advantages:

1. Allows specification of a standard SSAM that can be implemented in any simulation model that wishes to support the output format.
2. Maintains an external aggregation and statistics calculation module so the SSAM itself remains decoupled from the simulation.
3. Demonstrates the necessity of distribution-based statistical evaluation and assessment of simulation output (i.e., measurement and display of variance information) rather than the typical average-value analysis that is conducted in practice.
4. Supports the future goal of common standards for traffic simulation input and output specifications.

These recommendations are captured in the next section on requirements for an SSAM software tool.

Previous | Table of Contents | Next

TFHRC Home | FHWA Home | Feedback United States Department of Transportation - Federal Highway Administration