Efficient estimation of arc criticalities in stochastic activity networks

An algorithm is described for estimating arc and path criticalities in stochastic activity networks by combining Monte Carlo simulation with exact analysis conditioned on node release times. These estimators are proved to be unbiased and to have lower variance than the corresponding standard Monte Carlo estimators. The algorithm is applied to a variety of standard and randomly generated test networks to establish that the estimators are significantly and robustly more efficient then the standard estimators when run time and statistical efficiency are properly combined.