Research on Adaptive Sampling and Stochastic Scheduling

点击翻译按钮获取中文摘要

Two dynamic allocation problems are considered in this research. The first is the so-called "Multi-Armed Bandit Problem" (MABP), which has applications to scheduling in job shops and in optimal replacement of parts. It has been shown that the general N dimensional MABP can be reduced to N one-dimensional problems involving a single project, the solution of which is the dynamic allocation index value for the current state of the project. The principal investigator proposes a new approach for calculating the dynamic allocation index value and will investigate how this approach can be applied to several problems in sequential statistics and stochastic scheduling. The second part of this research deals with the relationship between closed (no arrivals) and open (arrivals) stochastic scheduling problems. The aim is to understand the role played by the arrival process on the optimal policies.