Conditional Monte Carlo: A Simulation Technique for Stochastic Network Analysis

Abstract

This paper is concerned with a simulation procedure for estimating the distribution functions of the time to complete stochastic networks. The procedure, called conditional Monte Carlo, is shown to be substantially more efficient (in terms of the computational effort required) than traditional simulation methods. The efficacy of conditional Monte Carlo and its use in conjunction with other Monte Carlo methods is illustrated for the Wheatstone bridge network. The applicability of the procedure to larger networks, as well as other stochastic systems, is discussed, and a general method is given for its implementation.