Interior Point Methods for Linear Multicommodity Flow Problems
Luis Guardia, Gílson Brito Alves Lima
This article studies the linear multicommodity network flow problem. This kind of problem arises in a wide variety of contexts. A numerical implementation of the primal-dual interior-point method is designed to solve the problem. In the interior-point method, at each iteration, the corresponding linear problem, expressed as an augmented indefinite system, is solved by using the AINV algorithm combined with an indefinite preconditioned conjugate gradient algorithm. Numerical experiments are conducted for networks of different dimensions and numbers of products and for some nonlinear costs. The computational results show the effectiveness of the interior-point method for this class of network problems.