Surveillance Problems: Poisson Process under Costly Surveillance

摘要

This paper is concerned with optimal strategies for controlling a one-dimensional Poisson production process under costly surveillance. Observations may be taken and repairs may be made at any time at positive cost. A form of the long run average income is used as the criterion for comparing strategies. The class of non-randomized stationary strategies which allow production only when the process is observed in one of a finite number of states is shown to contain an optimal strategy under a certain assumption of monotonicity of the income function. (Author)