Abstract |
How can logistics, telecommunications, and other network service providers meet customer requirements for speedy and reliable service when economies of scale favor consolidating traffic, possibly degrading performance? Also, how can service designers manage the extremely large size of the underlying network while effectively designing and planning operations on it? In this talk, I will present a series of models and algorithms that my coauthors and I have been developing to address these very practical but challenging questions. Our models, developed on large-scale integer programming, incorporate the complex trade-offs involved in designing a cost-effective, but responsive, network that meets end-to-end service requirements. Using insights about the structure of this network design problem, we develop polyhedral results and methods to strengthen the model and accelerate the solution procedure. We also provide powerful tools to help construct a smaller or reduced network in place of an existing large network so that service performances in the small network are equal or only slightly degraded to those in the original network. The theoretical results and methods developed in our research provide a valuable foundation to design contemporary service networks. And real word applications of our methods have resulted in significant cost saving and performance improvements in network design and operations. |