U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590

Skip to content U.S. Department of Transportation/Federal Highway AdministrationU.S. Department of Transportation/Federal Highway Administration

Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations

This report is an archived publication and may contain dated technical, contact, and link information
Publication Number: FHWA-RD-03-050

Surrogate Safety Measures From Traffic Simulation Models

PDF Version (903 KB)

PDF files can be viewed with the Acrobat® Reader®

4. Discussion of Microscopic Simulation Model Comparisons

As the elements in each of the tables in the previous section indicate, each simulation model has its own strengths and weaknesses with respect to both traffic modeling in general and simulation of surrogate safety measures. Previous European research has indicated similar results (39, 65). Summaries of each reviewed model are presented below. All of the models reviewed would require some level of modification, upgrade, or enhancement to support the derivation of surrogate measures of safety—both internal enhancements to the source code and external enhancements for additional output file(s), statistics, and possibly new input value(s).


CORSIM provides the most natural choice for further FHWA development of a surrogate safety assessment tool. The primary disadvantage of using CORSIM for surrogate safety analysis is the use of the fixed 1-s time-step update interval. The accuracy of the calculations of surrogate measures will be limited by this fundamental issue (i.e., surrogate measure calculations can only be made at certain position and velocity initial conditions). It is possible that certain surrogate measures could be neglected because the update interval was too coarse. The second key issue affecting the use of CORSIM for surrogate safety analysis is the lack of appropriate modeling of vehicle movements in the intersection box. Although the current model includes "micronode" modeling of intersection box movements, the calculation of conflicts and output statistics for conflicts using the micronode logic is not consistent with the definition of conflict used in the safety community. In addition, the model used to evaluate the "real-estate blocks" used to identify situations where conflict opportunities occur is not complete for intersections with more than four legs or irregular intersection geometries (61).

Analysis of the micronode logic also indicates two additional difficulties for the current implementation:

  • The micronode logic is not implemented in CORSIM as a stand-alone module, i.e., it could not be easily replaced with a new logic that would facilitate both surrogate safety analysis and simulation of vehicle movements in the intersection box.
  • Several modules used for simulating vehicle movements and interactions with other vehicles (e.g., car following) have assumptions of 1-s time steps embedded into the equations.


SIMTRAFFIC is a relatively easy-to-use traffic simulation tool that is designed for use by field traffic engineers primarily as an adjunct to the SYNCHRO signal-timing optimization software. A significant disadvantage of SIMTRAFFIC is the lack of API functions or supporting detailed output of vehicle-state variable information and automated statistical analysis capabilities of other codes. On the other hand, SIMTRAFFIC has the most resolute state variable standard update intervals of all models surveyed (0.1 s) and claims many improvements over the CORSIM models for representing real-world traffic conditions, although the validity of those improvements is not known. SIMTRAFFIC appears to model most of the behaviors necessary for collecting surrogate measures, but at a less resolute level than AIMSUN, VISSIM, or Paramics.


VISSIM appears to be a full-featured microscopic simulation model with the ability to obtain detailed state variable information on each vehicle on time scales with better than second-by-second accuracy. VISSIM has been interfaced to other external codes before, including hardware signal controllers, thus the developers have experience in development collaboration. The priority rules feature of VISSIM appears to allow complex modeling of junction behavior, including friendly merging (situations where following vehicles will slow for merging vehicles to create a gap), as it occurs in the real world. It is not apparent that other simulation models are able to represent such behavior (AIMSUN supports such an effect at freeway ramp junctions). Another advantage of VISSIM is the representation of on-street parking behavior and double parking. VISSIM has NEMA controller models available (i.e., using the VAP macro language), and adaptive algorithms and real controllers can be integrated and evaluated rather easily with the real-time interface. There are complexity issues involved with setting up the multitude of priority rules at each junction, although, again, this flexibility allows for very detailed modeling of location- and vehicle-specific interactions. However these affect the usability of the software more so than the ability to obtain surrogate measures. VISSIM appears to support most of the modeling features required for obtaining surrogate measures at a reasonable level of fidelity.


HUTSIM is currently being modified by Helsinki University to evaluate the use of a nanoscopic driver behavior model to produce delays in driver reaction time that lead to surrogate safety measures (59). They have tentatively selected TTC as their primary surrogate measure. Some of the details of driver behavior modeling in the HUTSIM simulation were unavailable given the scope of this project. The demo software available indicates a less sophisticated visualization and model-building GUI than other tools, although the software contains an add-on analyzer module for post-processing output data into graphs/charts. It appears that all modifications to HUTSIM are being made internally since no API is available. Sight-distance limitation modeling is a significant advantage of the HUTSIM simulation model.


Paramics appears to be a full-featured microscopic simulation model with the ability to obtain detailed state variable information on each vehicle on time scales with better than second-by-second accuracy. Paramics has been interfaced to other codes before, and the API continues to be refined and extended by researchers around the world, including "extensibility" of the input processor(s) and output processor(s). A significant disadvantage of the Paramics model is the use and reliance on origin-destination matrices to derive traffic volumes. Paramics appears to support most of the modeling features required for obtaining surrogate measures at a reasonable level of fidelity, although some modeling elements are described only at a functional level.


Integration is a simulation model developed primarily for research use that has recently been distributed on a commercial basis. Integration does not have an API or access to vehicle state variables on a time step-by-time step basis. Integration appears to be weaker at explicit simulation of detailed vehicle-to-vehicle interactions than other simulation models, given that it originated from a hybrid "mesoscopic" macro/micro modeling base. Integration does not appear to explicitly model movements in the intersection box. Integration has been modified to output TTC distributions and predict crash rate statistics using previously developed nonlinear regression models (based on link mean speed).


AIMSUN is a full-featured simulation model with the ability to obtain detailed state variable information on each vehicle on time scales with better than second-by-second accuracy. AIMSUN has a set of APIs and has been interfaced to external codes in the past, such as EMME/2 and SCATS. AIMSUN's car-following logic has been shown to be realistic in tests during the SMARTEST study (Algers, et al., 1997). A significant advantage of AIMSUN is that the gap-acceptance behavior of drivers is modified based on their delay time. Most other models do not represent such phenomena. AIMSUN also has a model for vehicle-actuated NEMA controllers and allows for a look-ahead distance restriction at junctions. Most of the necessary elements are modeled in AIMSUN to support the collection of surrogate measures at a reasonable level of fidelity.


WATSIM is an enhancement of the NETSIM model by one of the original developers of NETSIM. As such, WATSIM inherits many of the limitations of the CORSIM model, including fixed 1-s time steps. WATSIM has many additional features over CORSIM, including light-rail modeling. WATSIM lacks many of the features of general-distribution tools for supporting this type of surrogate safety research, such as configurable output files, post-processing tools, and APIs.


Texas allows microscopic simulation modeling of only a single, isolated intersection. Texas includes no significant built-in post-processing tools, configurable outputs, graphical input editors, or APIs (Texas can create a file of the traditional measures of effectiveness (MOEs) that are processed to determine replicate-run MOEs). Texas can output vehicle-state variables for each time step, including data that are along the lines of output of the relevant variables needed for calculation of surrogate measures. FHWA is currently considering making modifications to Texas to allow red-light running in the behavioral modeling in order to study the safety implications.

A significant advantage of Texas is explicit modeling of North American semi- and fully-actuated signal controllers, including control of diamond interchanges (i.e., two intersections operated as a single controlled-intersection entity). Texas does allow simulation of vehicle movements at time steps down to 0.1 s. The inclusion of sight-distance limitation modeling is also a significant advantage of the Texas model, as well as the explicit modeling of U-turns. The main drawback of Texas is modeling of just a single intersection junction; however, model appears to support the primary modeling elements necessary for obtaining surrogate measures at a reasonable level of fidelity.

Previous | Table of Contents | Next

United States Department of Transportation - Federal Highway Administration