The use of common random numbers to reduce the variance in network simulation of traffic

Computer simulation is now used routinely as a decision support tool in many facets of transportation engineering. Simulation models are often used to compare system alternatives, i.e. alternative system designs and/or operating policies. The presence of high variance of the simulation output variables is a critical problem in such comparative analyses because the models must either be run longer or executed several times to achieve reasonably accurate point and interval estimates of the parameters of interest. The high variability in output measures can also lead to concern about the validity of the model when they are used by the practitioners who are not intimately familiar with the stochastic nature of the model processes. This paper describes and illustrates the effectiveness of variance reduction based on the concept of common random numbers (CRN) for the TRAF-NETSIM simulation model.