Randomized Estimators for Time Integrals.

Abstract : This document considers a stochastic system for which one needs an estimate of the expected discounted cost over the infinite horizon. Standard Monte Carlo procedures do not apply since the parameter to be estimated involves values of the process over an infinitely long time interval. This paper, presents Monte Carlo estimation techniques, based on randomization, that can be used in the above setting. The techniques developed turn out to be more efficient than the standard approach, even when the parameter to be estimated cumulates costs over a finite time interval. (Author)