Distributionally robust multistage inventory models with moment constraints

摘要: In this paper, we consider a minimax approach to managing an inventory under distributional uncertainty. particular, study the associated multistage distributionally robust optimization problem, when only mean, variance, and distribution support are known for demand at each stage. It is that if policy maker allowed recompute her choice after stage (i.e. dynamic formulation), thus taking prior realizations of into consideration performing relevant calculations later stages, basestock optimal. contrast, not static far less known. If these two formulations have common optimal policy, i.e. would be content with given whether or she has power stage, say \emph{time consistent}, problem \emph{weakly time consistent}. every formulation consistent, \emph{strongly give sufficient conditions weak strong consistency. We also provide several examples demonstrating consistent in general. Furthermore, show question consistency can quite subtle setting. that: (i) fail weakly (ii) but strongly (iii) even different values. Interestingly, stands contrast analogous setting which mean it such inconsistency cannot occur \cite{S-12}.