|
|
Value of Demand Observability for Inventory Control with
Markovian Demand and Lost Sales
submitted, 2006.
(w/ O. Ortiz and C.C. White, III)
This paper considers a periodic-review, single-product
inventory system serving serially-correlated customer demands,
where demad is modeled using an exogenous Markov chain, such
that the (discrete) demand distribution is dependent on the
level of demand in the prior period. As in many real-world systems,
unmet demand is lost and not directly observed; sales provide a screened
observation of demand. The paper provides: (1) the development of a
partially-observed Markov decision process (POMDP) model for
determining optimal inventory replenishment policies for systems
where demand is partially-observed via sales data; (2) the development
of a parametric family of heuristic algorithms for determining
near-optimal replenishment policies for large-scale problems; and (3)
the development of an approach for bounding the maximum value of improving demand
observability in such settings.
|
|
An Inventory Control Model with Possible Border Disruptions
submitted, 2005.
(w/ B.M. Lewis and C.C. White, III)
This paper develops a quantitative framework for analyzing the impact
of single-bottleneck disruptions on simple one-to-one supply chains, focusing
on the problem of determining optimal inventory control policies. Disruptions
at the bottleneck lead to freight queues, which require multiple periods to
dissipate. A methodology is developed to determine an appropriate distribution
for the stochastic lead time induced by such systems, and then to determine
optimal average cost inventory control policies and their costs.
|
|
|
|
