Details

Autor(en) / Beteiligte

• Truong, Van Anh

Titel

Approximation methods for supply-chain problems

Erscheinungsjahr

2007

Link zum Volltext

Quelle

Alma/SFX Local Collection

Beschreibungen/Notizen

• We develop solution approaches to two classical supply-chain problems: the capacity expansion problem and the joint inventory-control and pricing problem. The capacity expansion problem suffers from computational intractability due to its having a high-dimensional state space and an embedded NP-hard, subset selection problem. The joint inventory-control and pricing problem suffers from model dependence: existing solutions rely on an explicitly specified, and difficult to estimate, price-demand relationship. For both problems, we propose computational approaches yielding approximate solutions. For the capacity-expansion problem, we develop multi-dimensional balancing algorithms to compute provably near-optimal policies. These are the first approximation algorithms for multi-machine, multi-product systems facing non-stationary, stochastic and correlated demands. Our approach is computationally efficient and guaranteed to produce a policy with total expected cost no more than twice that of an optimal policy. We overcome the problem's high-dimensionality by introducing novel quasi-separable resolution schemes to decompose the system's lost-sales cost by machine types. We make assumptions of minimal inventory and lost sales and treat two different models for making production decisions: Monotone Production and Revenue-Maximizing Production. For the joint inventory-control and pricing problem, We propose a new data-driven algorithm called ICP for computing a joint inventory and pricing policy for a multi-stage, single-product system under independent demands. We obtain a polynomial bound for the number of samples required to compute the optimal policy to any degree of accuracy with high certainty. Our approach is truly parameter-free, in the sense that we make neither parametric assumptions about the form of the objective function, nor distributional assumptions about the underlying randomness. In particular, we do not use an a priori model of the demand as a function of the price. We allow capacity and minimum-order-quantity constraints on the inventory ordered in each period under stronger assumptions.