Comparsion of perturbation analysis with conventional sensitivity estimates for regenerative stochastic systems

Perturbation Analysis (P/A) of Discrete Event Dynamical Systems is a recently introduced method for estimating the sensitivity of the performance of such systems to adjustable parameters. We examine the Mean Squared Error (MSE) of these estimates and compare it to that of estimates obtained by conventional methods. We assume that the P/A estimate is unbiased (a question dealt with in other papers). The results obtained establish the asymptotic superiority of P/A over conventional methods, for any stochastic system with regenerative structure. Here we give an overview of the results: details are available in [1].