Correlated Random Number Generation for Simulation Experiments

The design of adequate input models is crucial for the validity of simulation experiments. Nowadays it is common to analyse trace data and to determine appropriate probability distributions for the generation of identically and independently distributed random numbers. Although it is well known that ignoring the correlation structure might give false results, modelling and generation of correlated random numbers is not supported by most simulation tools. In this paper we present an approach for generating correlated random numbers within the wellknown simulator Arena. Fitting of input processes is done using the free available toolkit ProFiDo. Resultant process definitions are loaded into an Arena module with specific Visual Basic code for the generation of correlated random numbers. The effect of correlated versus uncorrelated input data is shown for the simulation results of an Arena assembly line model.