Relaxation methods for the strictly convex multicommodity flow problem with capacity constraints on individual commodities

We study the multicommodity flow problem that minimizes a strictly convex cost objective function subject to the capacity constraints on individual commodities as well as the total flows in each arc. By making use of its dual, we formulate the problem as a nonlinear unconstrained optimization problem and propose relaxation methods. Computational results show that the proposed methods can practically solve problem instances, for example, with up to 100 nodes, 1000 arcs, and seven commodities.