On the efficient generation of discrete event sample paths under different system parameter values (Q1103283)

The techniques of perturbation analysis (PA) are generalized from the viewpoint of efficiently generating sample paths of discrete event dynamic systems (DEDS) operating under different parameter values. The central idea, denoted as `cut-and-paste', is simply to utilize one segment of a sample path generated under one parameter as legitimate sample paths of as many different parameter values of the DEDS as possible. Viewed in this light, many new ways to perform discrete event simulation for parametric study appear possible. Both analytical and experimental results are offered to substantiate our case.