How to use simulation

Microscopic simulation models are numerous and new ones are being developed, while existing models are upgraded frequently. Each model may have particular strengths and weaknesses. Therefore, when selecting a model, analysts should consider the following:

• Should a simulation model be used, or is an isolated analytic roundabout model sufficient?
• What are the model input requirements, are they sufficient, and how can they be provided or estimated?
• What outputs does the model provide in animated, graphical, or tabular form?
• What special features of the model are pertinent to the problem being addressed?
• Does the user manual for the simulation model specifically address modeling a roundabout?
• How sensitive is the model to various geometric parameters?
• Is there literature on the validation of this model for evaluating roundabouts?
• Is there sufficient information available on the microscopic processes being used by the model such as car following, gap acceptance, lane changing, or steering? (The availability of animation can assist in exposing model logic.)
• Are relevant past project examples available?

When a simulation model is used, the analyst is advised to use the results to make relative comparisons of the differences between results from changing conditions, and not to conclude that the absolute values found from the model are equivalent to field results. It is also advisable to perform a sensitivity analysis by changing selected parameters over a range and comparing the results. If a particular parameter is found to affect the outcomes significantly, then more attention should be paid to accurate representation and calibration of this parameter. Finally, the analyst should check differences in results from using different random number seeds. If the differences are large, then the simulation time should be increased substantially.