A New Approach to Compute Performance Sensitivities of Stochastic Discrete Event Dynamic Systems (DEDS)

Abstract

To analyse a complex system, e.g. flexible manufacturing systems (FMS), Hospitals, or Airports, viewed as stochastic discrete event systems (DEDS, e.g. [2],[5],[6]), as close to reality as required, there are some alternatives. For the queueing theory, which has obtained a fair amount of success, no analytical (product-form) solution is yet available, as soon as serious restrictions are involved (e.g. blocking only under given additional restrictions, see [1 ],[5]). Therefore, it is not an appropriate approach for the application to a large category of man-made systems. Thus, one may use, the natural alternative to the exact analysis of an approximated model, which is the approximated analysis of a more exact model, e.g. the Queueing Network Analyser described by Whitt [9]. Finally, experimentation (on a simulation model or the actual system) could be preferred to optimise system performance with respect to given decision parameters. Hence, simulation tools on computers are frequently used, e.g. GASP, SIMULA, or SIMAN, to perform the brute force analysis (e.g. [5], [6]).