Supply Chain Asset Configuration via Approximation Algorithms

Cornell University, Ithaca NY

Investigators

Abstract

We address the problem of fixed asset configuration in supply chains. Specifically, we look at where manufacturing and distribution centers should be located, which products should be produced or stocked at those facilities, and which clients should be served from which facilities. Our research will be based on approximation algorithms, for the following reason: both solution quality and run times are strong, are predictable, and scale very well as the problem sizes grow. Competing approaches are less robust. The major challenge we need to address is to develop approximation algorithms that incorporate key aspects of real-world supply chains, including more accurate models of capacity, probabilistic-scenario-based modeling, stochastic demand for inventory, and more flexible topologies for distribution networks. The algorithms we will develop are useful both as stand-alone algorithms, and in conjunction with branch-and-bound. The main activities to be carried out are algorithm development, and computational testing with generated and corporate data sets. The computational aspects of our research provide us with an excellent opportunity to involve undergraduate students, and to reach out to underrepresented groups. Finally, we intend to interact closely with software vendors, the primary implementers of research of this type, to be sure that our view of the needs of supply-chain-driven organizations is correct, and as a source of industrial data. Asset configuration is one of the most important and expensive aspects of supply chain design. These issues are faced by a vast array of companies, of all sizes, both nationally and globally. The financial implications are huge. Commercial supply chains have strong similarities to other problems domains, such as global and national relief networks (e.g., The Red Cross), networks that distribute humanitarian aid (e.g., the United Nations), telecommunications systems, etc. This research has a strong potential to have an impact on those domains as well.