Sample Path Derivatives for (s, S) Inventory Systems

摘要: for (s, S) inventory systems, we derive sample path derivatives of performance measures with respect to the two parameters s and S. These yield derivative estimators which can be estimated from a single or simulation system, in some cases not even requiring actual knowledge underlying demand distribution. Such estimates would useful sensitivity analysis gradient-based optimization techniques. We consider nondiscounted periodic review system general independent identically distributed (i.i.d.) continuous demands, full backlogging, holding shortage costs. For infinite horizon model, consistency proofs are given special cases, although argue why should correct more case.