Using column generation and Rosen`s partitioning method to solve large multi-commodity network flow problems
This talk will discuss a specialization of Rosen`s Partitioning Method designed to exploit the structure of multi-commodity network flow problems. The multi-commodity problem we address, prevalent in the telecommunications industry, defines each commodity as the flow from an origin to a destination. The solution method uses shortest path column generation and constraint relaxation to reduce the number of rows of the master problem to the number of capacitated edges in the network. Each stage of the algorithm consists of a column generation phase followed by a dual ascent step which results in a dual feasible basis. Computational results comparing variants of this method to standard shortest path formulations for providing solutions to the LP relaxation will be shown. The specialized branch-and-bound algorithm for generating minimum-cost unique path solutions will also be discussed.
- OSTI ID:
- 36110
- Report Number(s):
- CONF-9408161--
- Country of Publication:
- United States
- Language:
- English
Similar Records
Network design with column generation
Branch-and-cut: Column generation for solving huge integer programs